Split (3)

train · 4.35k rows

article stringlengths 1.98k 169k	summary stringlengths 1.01k 4.15k	section_headings listlengths 2 38	keywords listlengths 0 12	year stringclasses 11 values	title stringlengths 30 189
Myopathies decrease muscle functionality . Mutations in ryanodine receptor 1 ( RyR1 ) are often associated with myopathies with microscopic core-like structures in the muscle fiber . In this study , we identify a mouse RyR1 model in which heterozygous animals display clinical and pathological hallmarks of myopathy with core-like structures . The RyR1 mutation decreases sensitivity to activated calcium release and myoplasmic calcium levels , subsequently affecting mitochondrial calcium and ATP production . Mutant muscle shows a persistent potassium leak and disrupted expression of regulators of potassium homeostasis . Inhibition of KATP channels or increasing interstitial potassium by diet or FDA-approved drugs can reverse the muscle weakness , fatigue-like physiology and pathology . We identify regulators of potassium homeostasis as biomarkers of disease that may reveal therapeutic targets in human patients with myopathy of central core disease ( CCD ) . Altogether , our results suggest that amelioration of potassium leaks through potassium homeostasis mechanisms may minimize muscle damage of myopathies due to certain RyR1 mutations . Myopathies due to mutations in RyR1 are inherited and currently incurable . Congenital myopathies are characterized by hypotonia and delay of motor development with weakness in skeletal muscles . In normal muscle , nuclei are located at the periphery of the myofibers and the mitochondria reside throughout the myofiber . Muscle biopsies of patients with myopathies and mutations in ryanodine receptor type 1 ( RyR1 ) show centralized nuclei as well as disorganized areas in the center of the myofiber , called cores , that lack mitochondria and are devoid of metabolic activity , reflective of cellular damage . The most common RyR1-associated myopathy is central core disease ( CCD ) ( Wu et al . , 2006; Jungbluth , 2007 ) . In addition , RyR1 mutations in muscle cause diseases such as multi-mini core disease ( MMD ) , hypokalemic periodic paralysis , and malignant hyperthermia ( MH ) , although the phenotype/genotype relationship is unclear ( Fujii et al . , 1991; Marchant et al . , 2004; Treves et al . , 2008; Wilmshurst et al . , 2010 ) . Muscle movement is produced through excitation–contraction ( E–C ) coupling elicited by t-tubule conductance of the muscle action potential . Depolarization stimulates the dihydropyridine receptor ( DHPR , CaV1 . 1 ) to mechanically open RyR1 channels in the sarcoplasmic reticulum ( SR ) , which release intracellular calcium stores to flood the myofiber with calcium and initiate muscle contraction ( Numa et al . , 1990 ) . During contractions , muscle cells influx Na+ and efflux K+ , leading to depolarization of the membrane , a rise in extracellular K+ , and loss of excitability . The myoplasmic loss of potassium is stabilized via increased potassium conductance from the opening of voltage-gated potassium channels ( Lindinger et al . , 2001; Nielsen et al . , 2004 ) and the activation of sodium–potassium ATPase ( Na+ , K+-ATPase ) transport ( Clausen , 2013b ) . During muscle contractions , the enzymes Na+/K+-ATPase , Ca2+-ATPase , and myosin ATPase consume most of the stored ATP . Myosin ATPase is considered to account for the majority of ATP consumption at 60–70% ( Smith et al . , 2005 ) . Ca2+-ATPase is assumed to account for the majority of remaining ATP , while Na+–K+-ATPase can consume up to 10% of available ATP ( Homsher , 1987; Clausen et al . , 1991; Barclay et al . , 2007 ) . As ATP levels decrease during repeated muscle contractions and fatigue , ATP-sensitive KATP6 . 2 channels open causing potassium efflux , membrane depolarization , and inhibition of the action potential ( Gong et al . , 2003 ) . Thus , potassium homeostasis mechanisms preserve E–C coupled release of internal calcium stores , protect against muscle fatigue , and prevent excessive K+ loss ( Clausen , 1986; Cifelli et al . , 2008 ) and may protect the muscle during ATP loss ( Gong et al . , 2003 ) . In this study , we explore the function of RyR1 as it pertains to mechanisms of muscle weakness and potassium homeostasis . Through a forward genetic screen , we identified a mutation in Ryr1 that causes muscle weakness and myopathic pathological hallmarks in heterozygous mice . Furthermore , we examined the potential of this animal model for clinical and therapeutic studies . Mutant Ryr1 muscle showed decreased concentrations of Ca2+ and K+ in the myoplasm and an increased permeability for K+ in muscles at rest . In vitro , the potassium leak can be rectified through the application of higher concentrations of K+ or inhibition of KATP channels . In vivo , the muscle weakness and myopathic pathology can be reversed through increased interstitial potassium and pharmacological inhibition of KATP channels either by diet or FDA-approved drugs , suggesting defects in K+ homeostasis mechanisms as possible myopathological hallmarks due to mutations in RyR1 . KATP6 . 2 channels and other regulators of potassium transport are misexpressed in mutant muscle and may partly underlie the disease phenotype in this myopathic animal model . Finally , we find that human patients with CCD show dysregulation of genes involved in the control of potassium homeostasis and suggest that these may serve as important disease biomarkers . Through a forward genetic screen in mice , we identified a mutation in which heterozygous mutant mice display muscle weakness . Sequencing identified an E4242G change in exon 93 of Ryr1 ( Figure 1A ) outside of the channel region of RyR1 . This allele is named RyR1m1Nisw , but here we will refer to it as Ryr1AG . Gene identification was confirmed by a lack of complementation in a genetic cross between RyR1AG and RyR1tm1TAlle allele ( see ‘Materials and methods’ ) . RyR1 mutations in humans are associated with several congenital myopathies including CCD , multi-mini core disease , hypokalemic periodic paralysis , and malignant hyperthermia ( MH ) ( Fujii et al . , 1991; Marchant et al . , 2004; Jungbluth , 2007; Treves et al . , 2008; Wilmshurst et al . , 2010 ) . 1-month , 2-month , and 1-year old Ryr1AG/+ mice showed a significant decrease in grip strength and considerable deficits on a wire hanging task compared to Ryr1+/+ mice ( Figure 1B , C ) . Malignant hyperthermia test was performed by placing Ryr1+/+ and Ryr1AG/+ mice in a 41°C humidified incubator for 30 min . There was no sign of muscle rigidity or spasticity or seizure activity for any of the animals . To test for sensitivity to halogenated anesthetics , anesthetic isoflurane was administered to Ryr1AG/+ mice at ∼5 . 5 × 10−5 ml/cm3 for a maximum of 30 min exposure . Ryr1AG/+ mice showed no MH response after exposure to isoflurane . The rectal temperature in Ryr1AG/+ mice after isoflurane exposure was similar to wild-type littermates [Ryr1AG/+ was 35 . 6 ± 0 . 9°C at 2 months ( n = 5 ) and 35 . 9 ± 0 . 6°C at 6 months ( n = 5 ) compared with Ryr1+/+ 35 . 4 ± 0 . 8°C and 35 . 8 ± 0 . 7°C at 2 and 6 months respectively ( n = 5 for both ) ] . Thus , heterozygous mutant mice show muscle weakness , the clinical definition of a myopathy . 10 . 7554/eLife . 02923 . 003Figure 1 . ENU-induced Ryr1AG mutation mimics clinical and pathological features of CCD in heterozygous mice . ( A ) Missense mutation in exon 93 of Ryr1 changes A to G , resulting in substitution of glutamic acid with glycine ( E4242G ) . ( B ) Average grip strength assayed using vertical digital push–pull strain gauge on Ryr1+/+ and Ryr1AG/+ mice at different ages raised on control 0 . 6% potassium diet . ( C ) In vivo hanging task determination of upper-body strength of Ryr1+/+ and Ryr1AG/+ mice at different ages ( B , C , 10 trials/mouse , n = 5 per set ) . ( D , E ) H&E staining indicate central nuclei ( arrows ) in Ryr1AG/+ ( E ) but not Ryr1+/+ ( D ) vastus lateralis muscle from 1-year old mice raised on control 0 . 6% potassium diet . ( F , G ) NADH-TR staining indicates cores ( white arrows ) in 1-year old vastus lateralis of RyR1AG/+ but not Ryr1+/+ . ( H , I ) Cytochrome oxidase ( COX ) staining denotes a decrease in mitochondrial function ( white arrows ) in vastus lateralis muscle of Ryr1AG/+ ( I ) compared to Ryr1+/+ ( H ) . Scale bar = 50 μm . ( J–L ) Transmission electron microscopy of 2-month old Ryr1AG/+ soleus muscle . ( J , K ) Regions of Z line streaming and associated sarcoplasmic disruption and cores ( white arrows ) of myofibrils . ( L ) Enlarged T-tubules are present in type I fibers ( scale bars: J = 1 μm; K = 2 μm; L = 0 . 2 μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 003 In typical muscle , nuclei are located at the periphery of the myofiber and mitochondria are present throughout . Muscle biopsies of CCD patients show internalized , centrally located nuclei , disorganized cores that lack mitochondria , and reduced metabolic activity ( visualized with cytochrome oxidase ( COX ) and nicotinamide adenine dinucleotide hydride-tetrazolium reductase ( NADH-TR ) staining ) ( Wu et al . , 2006 ) . However , the number of cores present in muscle biopsies is not reflective of the degree of muscle weakness ( Wu et al . , 2006 ) . Because proximal muscles of limbs are typically taken for muscle biopsies in patients , we examined the vastus lateralis and adductor magnus . Additionally , we examined the soleus muscle as it is commonly used in vitro studies of fatigue and RyR1-associated myopathy in mouse models . Analysis of these three muscles in 1-year old Ryr1AG/+ mice showed centrally located nuclei ( 13 . 2 ± 3 . 8 per 100 myofibers vs none detected in wild-type ) , core-like structures ( Table 1 ) , and decreased COX and NADH-TR staining , whereas no cores were observed in Ryr1+/+ muscle ( Figure 1D–I ) . Focusing on type 1 fibers , core-like structures appear in 8 . 9 ± 1 . 8 per 100 NADH-TR stained type 1 fibers in 12-month old Ryr1AG/+ soleus muscle , which is comparable to the 12 cores per 100 NADH-TR stained fibers of 18-month old Ryr1IT ( RyR1I4895T ) soleus muscle , which has a congenital myopathy with cores ( Zvaritch et al . , 2009 ) . In 2-month old Ryr1AG/+ muscle , we observed ultrastructural hallmarks of RyR1-associated myopathies ( Boncompagni et al . , 2009 , 2010 ) , including z-line streaming ( white arrows in Figure 1J: Table 2 ) , sarcomeric degeneration , irregularly shaped mitochondria , core-like structures ( white arrow in Figure 1K ) , and swelling of the t-tubules ( Figure 1L ) as well as internalized nuclei ( 9 . 8 ± 4 . 9 per 100 myofibers vs none detected in wild-type ) . The onset and extent of the pathological and histological changes in multiple muscle types are similar to a mouse CCD model , RyR1I4895T ( Boncompagni et al . , 2009; Zvaritch et al . , 2009; Boncompagni et al . , 2010 ) , altogether suggesting Ryr1AG mice as an additional model for myopathies with core-like structures . 10 . 7554/eLife . 02923 . 004Table 1 . Core-like structures in type 1 muscle fibers of 1-year old RyR1AG/+ mice as revealed by NADH-TR and COX labelingDOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 004NADH-TR negative coresCOX negative coresVastus lateralis6 . 2 ± 2 . 47 . 8 ± 2 . 1Adductor magnus muscle9 . 3 ± 2 . 99 . 7 ± 2 . 2Soleus muscle8 . 9 ± 1 . 88 . 1 ± 1 . 4Average values from 100 fibers with 10 slices per muscle and three muscles per group . 10 . 7554/eLife . 02923 . 005Table 2 . Number of Z line streaming sites within 100 μm2 grid in 2-month old wild-type and RyR1AG/+ soleus muscleDOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 005Genotype12-month old2-month old0 . 6% Diet5 . 2% DietEnalaprilGlibenclamideRyr1+/+0 . 48 ± 0 . 210 . 04 ± 0 . 010 . 03 ± 0 . 010 . 05 ± 0 . 030 . 09 ± 0 . 08Ryr1AG/+10 . 67 ± 0 . 823 . 75 ± 0 . 050 . 34 ± 0 . 161 . 13 ± 0 . 141 . 12 ± 0 . 02Average values from five grids per mice and three mice per group . Asterisks indicate significant value <0 . 05 , <0 . 005 . Ryr1 functions as a muscle isotype of ryanodine receptor that is mechanically activated by the dihydropyridine receptor to allow Ca2+ release from the sarcoplasmic reticulum , thus evoking E–C coupling ( Fill and Copello , 2002 ) . To determine whether RyR1AG mutant protein affects calcium release , we examined the effects of treatment with 4-chloro-m-cresol ( 4-CmC ) , a potent activator of RyR1 in skeletal muscle and RyR2 in cardiac muscle ( Zorzato et al . , 1993; Choisy et al . , 1999 ) . The domain , which is activated by 4-CmC , is located between residues 4007 and 4180 of RyR1 ( Fessenden et al . , 2003 ) , near the Ryr1AG mutation ( E4242G ) . An increased sensitivity to 4-CmC has been linked to patients with malignant hyperthermia ( Herrmann-Frank et al . , 1996a; Baur et al . , 2000 ) and ratiometric calcium imaging has been used to show hypersensitivity of 4-CmC in muscle cells from a patient with a mutation in RyR1 associated with malignant hyperthermia susceptibility ( Roesl et al . , 2014 ) . Thus , we performed ratiometric Ca2+ imaging using the calcium indicator , Fura-2 AM , and recorded the ratio of 340/380 nm fluorescence over a range of 4-CmC concentrations up to 1000 μM ( Zorzato et al . , 1993; Herrmann-Frank et al . , 1996b ) . 4-CmC was applied to explanted soleus muscle from 2-month old Ryr1AG/+ and littermate controls and ratiometric data of individual muscle fibers were normalized to the fluorescent ratiometric intensity at 1000 μM 4-CmC for each group . In wild-type mice , application of 200 μM 4-CmC induced measurable changes in Fura-2 signal ( Figure 2A; 20 individual fibers from 4 mice ) . However , in littermate Ryr1AG/+ muscle fibers , measurable changes in Fura-2 signal were not observed until the application of 600 μM 4-CmC ( Figure 2A; 27 individual fibers from 5 mice ) . These data suggest that the Ryr1AG/+ mutation results in decreased sensitivity to 4-CmC . 10 . 7554/eLife . 02923 . 006Figure 2 . Calcium homeostasis is disrupted in Ryr1AG/+ muscle . Fura-2 ratiometric imaging of myoplasmic Ca2+ in 2-month old muscle . ( A ) Ratiometric analysis of 4-CmC sensitivity in soleus muscles of wild-type ( black ) and Ryr1AG/+ littermates ( grey ) . ( B ) Average ratiometric analysis of myoplasmic Fura-2 signal normalized to wild-type signal ( n above bar represents number of analyzed fibers ) . ( C ) Representative fura-2 ratiometric signal of muscle fibers showing percent change after addition of 20 μM thapsigargin ( TG ) from wild-type ( black ) and Ryr1AG/+ ( grey ) littermate . ( D ) TG induced maximum response in muscle fibers ( n above bar represents number of analyzed fibers ) . ( E , F ) Representation and quantification of Mn2+ quenching of Fura-2 fluorescence ratiometric signal illustrating increased influx of store operated calcium entry in soleus muscle fibers of Ryr1AG/+ ( grey ) muscle fibers compared to wild-type ( black ) muscle fibers ( dash lines represent slope ) . In ( E ) , arrows indicate when media was introduced with 0 Ca2+ followed by 0 . 5 mM Mn2+ . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 006 The reduced sensitivity of Ryr1AG to 4-CmC may suggest that less Ca2+ is released into the myoplasm upon RyR1 activation . In myotubes expressing malignant hyperthermia RyR1 cDNAs , myoplasmic Ca2+ is elevated due to a significant passive Ca2+ leak from the SR ( Yang et al . , 2007 ) . However , in Ryr1AG/+ muscle fibers , we found that myoplasmic Ca2+ levels were significantly decreased compared to wild-type muscle fibers ( Figure 2B; 85 . 3 ± 2 . 2% in Ryr1AG/+ muscle fibers normalized to wild-type concentration ) . This suggests that a Ca2+ leak that significantly elevates resting Ca2+ is not present in Ryr1AG/+ muscle fibers . In addition , the reduced sensitivity of Ryr1AG/+ muscle fibers to 4-CmC may suggest an increase in SR Ca2+ . To examine SR calcium levels , 20 μM thapsigargin ( TG ) , which inhibits SR Ca2+ ATPases , was applied to Fura-2 labeled muscle fibers . Ratiometric analysis of Ca2+ intensities provide evidence that Ryr1AG/+ muscle fibers have an increase in TG-dependent SR Ca2+ release ( Figure 2C , D; 0 . 23 ± 0 . 05; 37 muscle fibers from 6 mice; p < 0 . 001 ) compared to wild-type muscle fibers ( 0 . 15 ± 0 . 03; 24 muscle fibers from 4 mice ) . This suggests increased levels of Ca2+ in the SR of Ryr1AG/+ muscle fibers and is consistent with the lack of a Ca2+ leak and decreased RyR1-mediated Ca2+ release from the SR in the mutant myofibers . For muscle to function properly throughout the process of E–C coupling , the muscle fibers must maintain functional levels of Ca2+ within the SR . In response to reduced myoplasmic Ca2+ , external Ca2+ enters the myofiber to replenish internal stores through store-operated calcium entry ( SOCE ) . To measure unidirectional ion flux through SOCE , we used Mn2+ quenching of Fura-2 fluorescence in muscle fibers from Ryr1AG/+ and wild-type mice as Mn2+ has an increased affinity to Fura-2 compared to Ca2+ , thus decreasing Ca2+-dependent myoplasmic fluorescence . To examine Mn2+ quenching , Fura-2 fluorescence was measured in muscle fibers first washed in 0 mM Ca2+ , followed by Mn2+ , and finally Triton-X 100 with EGTA to provide a baseline measurement . Ryr1AG/+ muscle fibers showed a substantially increased Mn2+ entry rate ( −0 . 08 ± 0 . 02 rate in wild-type; n = 16 muscle fibers in 4 mice ) compared to that of wild-type ( Figure 2E , F; −0 . 2 ± 0 . 02 rate in wild-type; n = 18 muscle fibers in 4 mice ) . These data suggest that Ca2+ homeostasis is disrupted in Ryr1AG/+ muscle . During excitation–contraction coupling , RyR1 floods Ca2+ into the myoplasm for ATP-dependent contraction of muscle fibers . Immediately following contraction , Ca2+ is pumped back into the SR and taken up by adjacent mitochondria for the purpose of ATP production . The decrease in myoplasmic calcium could be due to decreased release of Ca2+ from the SR or increased uptake of Ca2+ by the mitochondria , which prompted us to evaluate mitochondrial function . Local calcium release mechanisms or calcium ‘sparks’ communicate with adjacent mitochondria to spatially confine Ca2+ release ( Rizzuto et al . , 1993; Hajnoczky et al . , 1995; Rizzuto et al . , 1998 ) . In skeletal myotubes , RyR1 and RyR3 generate calcium sparks ( Ward et al . , 2001; Weisleder et al . , 2012 ) , with RyR1 being the predominant ryanodine receptor expressed in limb skeletal muscles ( Marks et al . , 1989; Takeshima et al . , 1989; Otsu et al . , 1990; Zorzato et al . , 1990 ) . In cardiac muscle , the in vivo depletion of RyR2 is sufficient to reduce mitochondrial Ca2+ influx , oxidative metabolism , and ATP levels ( Bround et al . , 2013 ) . Moreover , the activation of cytosolic Ca2+ sparks can lead to the influx of mitochondrial calcium termed Rhod-2 labeled calcium ‘marks’ ( Pacher et al . , 2002 ) . Rhod-2 distributes to mitochondria when applied at low temperature and can be used as a measure of changes in fluorescence intensity ( Trollinger et al . , 1997 ) . Influxes in mitochondrial calcium induce increased fluorescence intensities in Rhod-2 labeled mitochondria . In Ringer's solution , the number of mitochondria with changes in fluorescence intensity over 10 min of imaging were significantly decreased in Ryr1AG/+ compared to Ryr1+/+ littermate isolated muscle fibers ( 6 . 1 ± 1 . 1 compared to 10 . 2 ± 1 . 4 per 1000 μm2; Figure 3D ) , but the Rhod-2 intensity changes in mitochondria were increased after the application of 4-CmC to the bath in both Ryr1AG/+ and Ryr1+/+ isolated fibers ( Figure 3A–D ) . Recent evidence in a mouse model with an induced mutation in RyR1 illustrates that leakage of SR Ca2+ induces ROS activation and mitochondrial dysfunction ( Andersson et al . , 2011 ) . However , no difference in MitoSOX , an indicator of superoxides , was detected in Ryr1AG/+ compared to Ryr1+/+ littermate 2-month old soleus muscle fibers ( Figure 3E ) . This is consistent with our data showing that Ryr1AG/+ muscle does not have an abnormal Ca2+ leak . 10 . 7554/eLife . 02923 . 007Figure 3 . Defects in mitochondrial function in Ryr1AG/+ mice . ( A–C ) Representative isolated fiber with Rhod2 fluorescence of 2-month old Ryr1AG/+ in 3 mM KCl ( A ) and 1 min ( B ) and 10 min ( C ) after application of 1 mM 4-CmC ( punctate fluorescence are calcium marks ) . ( D ) Quantification of mitochondrial calcium marks visualized with Rhod2-AM from isolated soleus muscle fibers from Ryr1+/+ ( black bars ) and Ryr1AG/+ ( gray bars ) mice . ( E ) Superoxide labeling of isolated muscle fibers from Ryr1+/+ and Ryr1AG/+ mice . ( F ) Intensity of TMRE labeling of isolated soleus muscle fibers from Ryr1+/+ and Ryr1AG/+ littermates with saponin . The numbers on top of bars in graphs represent number of fibers examined . ( G ) ATP content from Ryr1+/+ and Ryr1AG/+ isolated soleus muscle from age-matched mice ( n = 4 for each age and group ) . Scale bar = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 007 The decrease in myoplasmic Ca2+ and Rhod-2 labeled mitochondria Ca2+ intensities , and the hyposensitivity in RyR1AG/+ channels may lead to a reduction in mitochondrial generation of ATP . Mitochondrial membrane potential drives the generation of ATP . To analyze mitochondrial membrane potential , we used the mitochondrial membrane potential indicator TMRE along with the amphipathic glycoside , saponin , to avoid the potentially confounding influence of variations in plasma membrane potential . Ryr1AG/+ isolated muscle fibers showed no significant difference relative to Ryr1+/+ in overall TMRE labeling of mitochondria ( Figure 3F ) . Therefore , muscle was examined for total muscle ATP generation . ATP determination showed ∼30% less ATP in Ryr1AG/+ soleus muscle compared to age-matched Ryr1+/+ ( 1 , 2 , and 6-month old mice; n = 4 mice per group; Figure 3G ) . In human subjects that undergo fatigue , muscle fiber ATP levels can be reduced by up to 20% ( Sahlin et al . , 1997; Karatzaferi et al . , 2001a , 2001b; Jones et al . , 2009 ) , suggesting that the reduced ATP levels in Ryr1AG/+ muscle might impair ATP-dependent enzyme activity during and after contraction ( Homsher , 1987; Clausen et al . , 1991; Smith et al . , 2005; Barclay et al . , 2007; Clausen , 2013b ) . These data suggest that dysfunctional Ca2+ release in RyR1AG muscle may influence mitochondrial function and ATP production . Together the SR Ca2+ release deficiencies and decreased mitochondrial ATP production may be sufficient to induce the myopathic phenotype in RyR1AG/+ mice . The significant depletion of ATP in RyR1AG/+ muscle could consequently affect ATP-dependent mechanisms that normally act within the muscle , such as the influx of potassium to enhance membrane excitability . To determine whether potassium homeostasis is disrupted in Ryr1AG/+ muscle , we used the potassium indicator PBFI-AM ( see ‘Materials and methods’ ) in combination with ratiometric potassium imaging of muscle fibers . Experiments using wild-type soleus in normal Ringer's solution ( 3 mM K+ ) showed that intracellular potassium concentration was 137 ± 7 mM ( n = 23 fibers from 4 mice; Figure 4A , C ) . However , Ryr1AG/+ muscle showed lower intracellular potassium concentrations of 106 ± 8 mM ( n = 29 fibers from 4 mice; p < 0 . 005; Figure 4B , C ) , a similar decline as that observed for wild-type muscles undergoing fatigue ( McKenna et al . , 2008 ) . Moreover , Ryr1AG/+ muscle fibers showed a slow decrease in potassium fluorescent ratio intensity over the course of a minute suggesting an increase in K+ ion permeability ( Figure 4D , E; n = 23 fibers in 4 Ryr1AG/+ mice and n = 29 fibers in 4 wild-type mice ) , which may be indicative of a potassium leak . We hypothesized that if Ryr1AG/+ muscles have a defect in potassium homeostasis due to an increase in K+ ion permeability , it might be possible to decrease the semi-permeability of the K+ ion with the addition of external K+ . However , we did not observe a significant difference in serum K+ levels in 2-month old Ryr1AG/+ mice , but there was a 12 ± 3% increase in serum K+ in 6-month old Ryr1AG/+ mice relative to wild-type ( Figure 4—figure supplement 1 , n = 6 samples per group per age for Ryr1+/+ and Ryr1AG/+ ) . The acute addition of 7 mM KCl to the muscle increased PBFI-AM fluorescence intensity , suggesting inhibition of a potassium leak , while 0 mM KCl decreased PBFI-AM fluorescence intensity ( Figure 4D , E; n = 23 fibers in 4 Ryr1AG/+ mice and n = 29 fibers in 4 wild-type mice ) . Glibenclamide ( 2 μM ) selectively binds to and inhibits SUR1/2 ( Porat et al . , 2011 ) , subunits of the KATP6 . 1 and KATP6 . 2 channels in muscle ( Pedersen et al . , 2009 ) . In wild-type adult mouse skeletal muscle , glibenclamide protects against fatigue caused by tetanic force ( Duty and Allen , 1995 ) . Additionally , in chick muscle fibers , glibenclamide increases the twitch and tetanus tension induced by application of caffeine , an agonist of RyR1 ( Andrade et al . , 2011 ) . The acute addition of 2 μM glibenclamide to the 3 mM KCl bath resulted in increased intracellular K+ in wild-type ( n = 17 fibers in 4 mice ) and Ryr1AG/+ ( n = 16 fibers in 4 mice ) soleus muscle fibers ( Figure 4D , F ) . Upon bath application of 7 mM KCl for 1 . 5 hr to examine a continuous counterbalancing of the K+ leak , the intracellular potassium concentration in the mutant soleus increased to 132 ± 8 mM , similar to wild-type ( Figure 4G , n = 25 fibers in 4 Ryr1AG/+ mice and n = 26 fibers in 4 wild-type mice ) . Upon bath application of 2 μM glibenclamide for 1 . 5 hr , the intracellular potassium concentration in the mutant soleus increased to 138 ±12 mM , similar to wild-type ( Figure 4H , n = 28 fibers in 4 Ryr1AG/+ mice and n = 28 fibers in 4 wild-type mice ) . The decline in intracellular K+ fluorescence in Ryr1AG/+ muscle compared to wild-type muscle ( Figure 4A–E ) and the increase in intracellular K+ upon KATP channel inhibition ( Figure 4G , H ) suggest an increase in KATP channel activity in Ryr1AG/+ muscle . These data suggest that increased extracellular K+ ion and inhibition of KATP channels might prevent the excessive K+ leak and protect the mutant muscle from decreased intracellular concentrations of K+ . 10 . 7554/eLife . 02923 . 008Figure 4 . Detection and compensation of an internal potassium leak in RyR1AG/+ muscle . Fluorescence imaging of PBFI at 340 nm in Ryr1+/+ ( A ) and Ryr1AG/+ ( B ) soleus muscle in 3 mM Ringer's solutions . ( C ) Ratiometric potassium imaging obtained at 340 and 380 nm wavelengths provided a ratio of fluorescence in Ryr1+/+ ( black ) and Ryr1AG/+ ( grey ) soleus muscle ( normalized to Ryr1+/+ ) . ( D ) Representation of the ratiometric imaging experimental paradigms used in Figure 4 showing bath applications of 3 mM KCl , 7 mM KCl , 0 mM KCl , and 3 mM KCl with 2 μM glibenclamide in Ringer's solutions . ( E ) Slope of intracellular K+ fluorescence intensities in experimental conditions . ( F , H ) Normalized intracellular K+ concentration in Ryr1+/+ ( black ) and Ryr1AG/+ ( grey ) soleus muscle in 3 mM KCL ( F , H ) compared to soleus from contralateral limb in 7 mM KCl Ringer's solutions ( F ) or 3 mM KCL plus 2 μM Glibenclamide ( H ) ( muscle was bathed in solutions for 1 . 5 hr before imaging , n is the number of fibers examined from four mice ) . ( G ) Rate of change in PBFI fluorescence after acute bath application of 2 μM Glibenclamide . Scale bar = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 00810 . 7554/eLife . 02923 . 009Figure 4—figure supplement 1 . Serum level measurements from Ryr1+/+ ( black bars ) and Ryr1AG/+ ( grey bars ) in 2- and 6-month old mice . Measurements of serum levels were performed on Ryr1+/+ ( black bars ) and Ryr1AG/+ ( grey bars ) in 2- and 6-month old mice ( n = 6 samples per group per age for Ryr1+/+ and Ryr1AG/+ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 009 The decreased ATP concentrations and ratiometric potassium data in Ryr1AG/+ muscle suggest that there may be an aberrant activation of mechanisms that regulate potassium efflux . Therefore , we examined several proteins involved in K+ homeostasis in muscle of 2-month old Ryr1AG/+ mice . As mitochondrial dysfunction may confound the analysis of whole muscle , we prepared membrane-enriched lysates in which the plasma membrane and t-tubules were isolated from the fraction containing mitochondria , peroxisomes , and lysosomes ( see ‘Materials and methods’ ) . The membrane-enriched preparations showed a 33 ± 6% increase in KATP6 . 2 protein in 2-month old Ryr1AG/+ soleus muscle compared to Ryr1+/+ littermates ( Figure 5A , B; p < 0 . 005 , the RNA encoding KATP6 . 2 ( Kcnj11 ) was not altered in expression , Table 3 , 0 . 6% diet ) , but no significant change in the protein expression of KATP6 . 1 or KIR2 . 1 . KATP6 . 2 channels localize to skeletal muscle membranes and t-tubules ( Banas et al . , 2011 ) , suggesting that KATP6 . 2 channel function may be increased , leading to increased potassium efflux and disrupted potassium homeostasis , and contributing to the weakness in Ryr1AG/+ muscle . 10 . 7554/eLife . 02923 . 010Figure 5 . Altered activity of KATP channels involved in potassium transport . ( A ) Western blots of membrane enriched lysate ( that lacks mitochondria , peroxisomes , and lysosomes ) from 2-month old wild-type and heterozygous Ryr1AG/+ soleus muscle analyzed for KATP6 . 1 , KATP6 . 2 , and KIR2 . 1 . ( B ) Quantification of protein levels from Western blots of membrane enriched lysates . Each sample was first normalized to its own loading control and then the values from mutant and wild-type on the same blot were compared . Statistical analyses were determined from a minimum of 3 blots and at least two independent samples ( n = 4 for each Western ) . ( C ) KATP current densities of isolated soleus muscle fibers from 2-month old wild-type and Ryr1AG/+ littermates with and without glibenclamide . Numbers on top of bars are number of fiber recordings . ( D–E ) Representative current density recordings with the same tip resistance from Ryr1+/+ and Ryr1AG/+ soleus muscle fibers ( E ) with and ( D ) without 2 μM glibenclamide . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 01010 . 7554/eLife . 02923 . 011Table 3 . Quantitative RT-PCR of muscles from 2-month old wild-type and heterozygous Ryr1AG/+ mice fed for 4 weeks on the indicated dietsDOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 011Gene0 . 6% K Diet5 . 2% K DietRyr1+/+Ryr1AG/+Ryr1+/+Ryr1AG/+Abcc8 ( SUR1 ) Vastus lateralis1 . 00 ± 0 . 071 . 30 ± 0 . 211 . 05 ± 0 . 301 . 51 ± 0 . 33* Tibialis anterior1 . 00 ± 0 . 041 . 20 ± 0 . 17**1 . 09 ± 0 . 231 . 31 ± 0 . 27 Adductor magnus1 . 00 ± 0 . 121 . 28 ± 0 . 220 . 98 ± 0 . 231 . 61 ± 0 . 42Abcc9 ( SUR2 ) Vastus lateralis1 . 00 ± 0 . 190 . 81 ± 0 . 241 . 03 ± 0 . 201 . 09 ± 0 . 18 Tibialis anterior1 . 00 ± 0 . 230 . 86 ± 0 . 210 . 99 ± 0 . 131 . 18 ± 0 . 20 Adductor magnus1 . 00 ± 0 . 180 . 88 ± 0 . 191 . 01 ± 0 . 191 . 11 ± 0 . 11Atpa1 ( NKAα1 ) Vastus lateralis1 . 00 ± 0 . 151 . 23 ± 0 . 210 . 94 ± 0 . 320 . 18 ± 0 . 06** Tibialis anterior1 . 00 ± 0 . 211 . 22 ± 0 . 130 . 75 ± 0 . 390 . 23 ± 0 . 06** Adductor magnus1 . 00 ± 0 . 261 . 25 ± 0 . 170 . 93 ± 0 . 290 . 24 ± 0 . 07Clc1 ( CLC1 ) Vastus lateralis1 . 00 ± 0 . 091 . 05 ± 0 . 190 . 90 ± 0 . 210 . 84 ± 0 . 25 Tibialis anterior1 . 00 ± 0 . 101 . 10 ± 0 . 160 . 93 ± 0 . 110 . 82 ± 0 . 27 Adductor magnus1 . 00 ± 0 . 121 . 03 ± 0 . 210 . 89 ± 0 . 150 . 88 ± 0 . 17Kcnj2 ( KIR2 . 1 ) Vastus lateralis1 . 00 ± 0 . 090 . 77 ± 0 . 121 . 75 ± 0 . 171 . 62 ± 0 . 13* Tibialis anterior1 . 00 ± 0 . 170 . 68 ± 0 . 17*1 . 70 ± 0 . 261 . 55 ± 0 . 15 Adductor magnus1 . 00 ± 0 . 180 . 59 ± 0 . 131 . 67 ± 0 . 331 . 52 ± 0 . 18Kcnj8 ( KATP6 . 1 ) Vastus lateralis1 . 00 ± 0 . 200 . 56 ± 0 . 120 . 97 ± 0 . 200 . 97 ± 0 . 23 Tibialis anterior1 . 00 ± 0 . 130 . 74 ± 0 . 190 . 87 ± 0 . 110 . 91 ± 0 . 12 Adductor magnus1 . 00 ± 0 . 240 . 69 ± 0 . 161 . 16 ± 0 . 201 . 08 ± 0 . 17Kcnj11 ( KATP6 . 2 ) Vastus lateralis1 . 00 ± 0 . 150 . 96 ± 0 . 171 . 10 ± 0 . 131 . 42 ± 0 . 14* Tibialis anterior1 . 00 ± 0 . 181 . 01 ± 0 . 161 . 05 ± 0 . 081 . 34 ± 0 . 09* Adductor magnus1 . 00 ± 0 . 170 . 94 ± 0 . 091 . 0 9± 0 . 061 . 38 ± 0 . 21*Prkaa1 ( AMPK ) Vastus lateralis1 . 00 ± 0 . 141 . 23 ± 0 . 151 . 11 ± 0 . 101 . 07 ± 0 . 14 Tibialis anterior1 . 00 ± 0 . 041 . 11 ± 0 . 061 . 08 ± 0 . 051 . 07 ± 0 . 10 Adductor magnus1 . 00 ± 0 . 111 . 20 ± 0 . 111 . 01 ± 0 . 111 . 10 ± 0 . 06Values normalized to GAPDH before normalization to Ryr1+/+ control on 0 . 6% diet . Asterisks indicates significant value = 0 . 05 , = 0 . 001 , * = 0 . 005 . An increased number of KATP channels in the membrane ( or a change in activity ) would be expected to alter current densities . In rat skeletal muscle , KATP6 . 2 pores have a single-channel conductance of around 80 pS , although the current densities ( pA ) of KATP channels can vary between muscles ( Tricarico et al . , 2006 ) . To examine current densities of KATP channels in soleus muscle fibers , inside-out patch excision in ATP-free solutions was performed . Currents from excised patches of 2-month old wild-type soleus muscle fibers were −106 ± 12 pA with an average pipette resistance of 2 . 2 ± 0 . 4 MΩ after −60 mV voltage injection ( n = 10 fibers from 3 mice ) . In 2-month old Ryr1AG/+ soleus muscle fibers , the currents increased to −142 ± 24 pA with an average pipette resistance of 2 . 3 ± 0 . 3 MΩ ( n = 10 fibers from 3 mice ) . Thus , Ryr1AG/+ soleus muscle fibers showed a 34 ± 9% increase in normalized current densities compared to wild-type littermates ( Figure 5C , D ) . Glibenclamide addition removed the current differences between Ryr1+/+ and Ryr1AG/+ muscle , suggesting that the channels opened in the inside-out patch were KATP channels ( Figure 5C , E , n = 8 fibers from three mice for both Ryr1+/+ and Ryr1AG/+ ) . These data indicate an increase in KATP channel activity in the membrane of Ryr1AG/+ muscle and this could explain the increased potassium efflux in spite of the decreased ATP levels in mutant muscle . Potassium transport activity of KATP channels is positively regulated by the AMP-activated protein kinase AMPK ( Prkkaa1 ) ( Wang et al . , 2005 ) , and this transcript was upregulated in 2-month old Ryr1AG/+ muscles ( Table 3 , 0 . 6% diet ) . The sulfonylurea receptors ( SUR ) are ATP-binding cassette ( ABC ) transporters and subunits of KATP channels . SUR1 ( Abcc8 ) , which binds to KATP6 . 2 , was significantly upregulated in Ryr1AG/+ muscles at the RNA level ( Table 3 ) . Thus , the increased expression of genes that encode proteins upstream of KATP6 . 2 channels or that interact with KATP6 . 2 channels might contribute to the increased activity of KATP6 . 2 channels in membrane and t-tubules . ClC-1-dependent chloride conductance is increased during muscle fatigue ( Pedersen et al . , 2009 ) , yet ClC-1 transcripts were not significantly altered in Ryr1AG/+ muscle compared to Ryr1+/+ littermates . Ryr1AG/+ muscle showed a rapid increase in intracellular potassium when KATP channels were blocked ( Figure 4C , D , F ) , suggesting that mechanisms of K+ transport into the cell are active in these muscles but may not be able to overcome the potassium leak and compensate for decreased intracellular K+ . Na+ , K+-ATPase activity maintains Na+ and K+ equilibrium at the membrane in the muscle ( Kristensen and Juel , 2010 ) . Transcripts of the pump subunit Na+ , K+-ATPase α1 were increased in Ryr1AG/+ muscle ( Table 3 ) . Taken together , these data suggest that expression levels of genes and proteins in the potassium transport pathway are dysregulated when RyR1-dependent calcium release is disrupted . As our in vitro studies showed that increased extracellular K+ leads to a normalization of intracellular K+ in Ryr1AG/+ muscles , we then asked whether increased serum K+ could alter the in vivo phenotype of Ryr1AG/+ mice . Serum potassium levels can be increased through diet ( Christensen et al . , 2010 ) or with the angiotensin-converting enzyme ( ACE ) inhibitor , enalapril ( Cleland et al . , 1985 ) . Therefore , we asked whether these treatments might decrease K+ ion permeability in muscle and whether this could ameliorate the physiological and pathological symptoms of the myopathy in Ryr1AG/+ mice . To test these hypotheses , Ryr1AG/+ and Ryr1+/+ mice were weaned onto 0 . 6% K+ ( level of potassium in standard mouse chow ) , and four weeks later , when muscle weakness is already observed ( Figure 1B , C ) , they were given either 5 . 2% K+ diet or enalapril ( 0 . 02 mg/ml in drinking water ) for an additional four weeks ( Figure 6—figure supplement 1 , for blood pressure readings ) . Examination of soleus muscle fibers from these mice showed that internal potassium concentrations were increased in Ryr1AG/+ soleus fibers , similar to wild-type , following 5 . 2% K+ diet ( 145 ± 22 mM compared to 147 ± 20 mM; Figure 6A ) or with enalapril ( 147 ± 18 mM compared to 148 ± 25 mM; Figure 6A ) . As noted in Figure 1 , weakness in 2-month old Ryr1AG/+ mice on 0 . 6% K+ diets was clearly evident with an ∼50% reduction in grip strength and inability to hang onto the wire , but in line with our hypothesis , Ryr1AG/+ mice treated with enalapril showed a significant improvement in the wire hanging task ( Figure 6B , C ) . Most dramatically , the 5 . 2% K+ diet rescued muscle strength in Ryr1AG/+ mice ( Figure 6B , C ) . Thus , potassium supplementation and an FDA-approved drug can rescue muscle weakness in mice carrying the Ryr1AG mutation . 10 . 7554/eLife . 02923 . 012Figure 6 . Increased potassium diet can rescue muscle strength and reverse the CCD histology and myopathy . ( A ) Normalized internal potassium concentrations of soleus muscle from 2-month old Ryr1+/+ and Ryr1AG/+ mice fed 0 . 6% K+ diet for 4 weeks , then placed on 5 . 2% K+ diet or 0 . 6% diet + enalapril for 4 weeks , and then bath exposed to different extracellular potassium concentrations . ( B , C ) Average grip strength ( B , five trials/mouse and 5 mice/set; p values <0 . 001 ) and in vivo hanging task ( C , 10 trials/mouse , n = 5 per set; p values <0 . 001 ) assayed from 2-month old Ryr1+/+ ( black bar ) and Ryr1AG/+ ( grey bar ) mice maintained for 4 weeks on control 0 . 6% K+ diet , 5 . 2% K+ diet , or 0 . 6% K+ diet supplemented with enalapril . ( D ) Quantification of number of internalized nuclei per 100 myofibers in 12-month old mice or in 2-month old mice maintained for 4 weeks on control 0 . 6% K+ diet , 5 . 2% K+ diet , or 0 . 6% K+ diet supplemented with enalapril ( n = 10 per muscle , n = 3 per set of muscles for total of n = 30; p values <0 . 001 ) . ( E–P ) Vastus lateralis myofibers from 2-month old Ryr1+/+ and Ryr1AG/+ mice maintained for 4 weeks on control 0 . 6% K+ diet , 5 . 2% K+ diet , or 0 . 6% K+ diet supplemented with enalapril . Cross-sections stained with H&E ( left panels ) and COX ( right panels ) . Ryr1AG/+ mice on 5 . 2% K diet show increased COX staining and no internalized nuclei , similar to Ryr1+/+ . These pathological features are still observed in Ryr1AG/+ mice on 0 . 6% K+ diets . Enalapril increases COX staining but some internalized nuclei are observed , even in Ryr1+/+ . Scale bar = 50 μm; error bars as standard error of the mean ( SEM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 01210 . 7554/eLife . 02923 . 013Figure 6—figure supplement 1 . Blood pressure measurements from Ryr1+/+ ( black bars ) and Ryr1AG/+ ( grey bars ) in 2-month old mice . Non-invasive measurements of systolic ( A ) and diastolic ( B ) blood pressure were performed using a Visitec-2000 tail-cuff apparatus . Measurements were taken for 3 days to acclimatize the animals for a more reproducible blood pressure measurement and then the data were collected on day 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 013 Histological examination is the hallmark diagnosis of RyR1-related myopathies . On control 0 . 6% K+ diets , internally located nuclei and decreased mitochondrial activity were detected in Ryr1AG/+ muscle ( Figure 6D , E–H ) . Strikingly , muscle from Ryr1AG/+ mice on high-potassium diet resembled RyR1+/+ muscle ( Figure 6D , I–L ) , while enalapril treatment showed a less dramatic improvement in muscle histology ( Figure 6D , M–P ) . The number of Z line streaming areas was also significantly reduced by these therapies ( Table 2 ) . Thus , even after observable muscle dysfunction , myopathic pathology can be rescued with increased extracellular potassium . Chronic treatment of Ryr1AG/+ animals for 1 month with a 5 . 2% potassium diet returned KATP6 . 2 protein expression to levels similar to those seen in diet-treated Ryr1+/+ littermates ( Figure 7A ) . We also noted that protein levels of the inwardly rectifying potassium channel KIR2 . 1 were increased by 64 ± 23% in Ryr1AG/+ muscle following 5 . 2% K+ diet ( Figure 7A , p < 0 . 01 ) , suggesting that increased KIR2 . 1 channel activity may also assist in rectifying the K+ ion leak . This physiological rescue and the normalization of KATP expression suggest a decrease in KATP channel activity . KATP current density recordings in Ryr1AG/+ soleus muscle fibers following 5 . 2% potassium diet were returned to levels similar to wild-type on normal diet ( −107 ± 16 pA in 2 . 0 ± 0 . 3 MΩ ) , whereas currents in Ryr1+/+ on 5 . 2% diet were substantially decreased ( −74 ± 15 pA in 2 . 1 ± 0 . 4 MΩ; Figure 7B ) . ATP levels were also increased by 5 . 2% potassium diet in Ryr1AG/+ soleus muscle , similar to Ryr1+/+ ( 38 . 5 ± 6 . 2 and 41 . 1 ±8 . 3 nmol ATP/mg soleus , respectively on 5 . 2% diet; Figure 7C ) . Moreover , the 5 . 2% K+ diet increased mitochondrial calcium marks in isolated fibers from Ryr1AG/+ soleus muscles to a rate similar to wild-type littermates , with no effect on ROS activity ( Figure 7D , E ) . These data indicate a rescue of the myopathic muscle physiology by diet , indicating potential therapeutic avenues for RyR1-related myopathies in human . 10 . 7554/eLife . 02923 . 014Figure 7 . Increased potassium diet influences KATP channel activity and mitochondrial function . ( A ) Membrane enriched lysate ( lacking mitochondria , peroxisomes , and lysosomes ) from soleus muscle of 2-month old Ryr1+/+ and Ryr1AG/+ mice fed 5 . 2% potassium supplemented diet for 4 weeks analyzed by Western blot for KATP6 . 1 , KATP6 . 2 , and KIR2 . 1 ( * is p < 0 . 001 ) . ( B ) KATP current densities of isolated soleus muscle fibers from Ryr1+/+ and Ryr1AG/+ littermates after 4 weeks on 5 . 2% K diet compared to 0 . 6% K diet . ( C ) ATP content from Ryr1+/+ and Ryr1AG/+ isolated soleus muscle from mice on 0 . 6% or 5 . 2% potassium diet . ( D ) Quantification of calcium marks visualized with Rhod2-AM from isolated soleus muscle from Ryr1+/+ and Ryr1AG/+ mice with and without diet therapy . ( E ) Superoxide labeling of isolated soleus muscle from Ryr1+/+ and Ryr1AG/+ mice with and without therapy . Error bars as standard error of the mean ( SEM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 014 Glibenclamide can inhibit KATP channels , which are overexpressed in the mutant muscle ( Figure 5A ) , and rectify the K+ ion leak when acutely applied in vitro ( Figure 4C ) . To extend these findings in vivo , we asked whether this FDA-approved drug could ameliorate the myopathic pathology in mice . To test this , glibenclamide ( 15 mg/kg/day ) was orally administered to 1-month old mice for 30 days . Ryr1AG/+ mice treated with glibenclamide showed significantly increased grip strength ( Figure 8A , B ) . ATP content and mitochondrial calcium changes in intensities in glibenclamide treated Ryr1AG/+ mice were increased to similar levels as Ryr1+/+ mice ( 36 . 4 ± 3 . 9 and 39 . 3 ± 4 . 3 nmol ATP/mg soleus muscle , respectively; n = 4 mice per condition , Figure 8C; 11 . 8 ± 1 . 5 and 11 . 1 ± 1 . 8 mitochondrial changes in intensities , respectively; n = 8 muscle fibers for each group , Figure 8D ) with no alteration in MitoSOX labeling ( data not shown ) . Ryr1+/+ and Ryr1AG/+ mice on glibenclamide therapy showed no significant difference in internal potassium concentrations ( 140 ± 9 mM in Ryr1+/+ soleus fibers compared to 145 ± 28 mM in Ryr1AG/+ soleus fibers; n = 17 and 20 muscle fibers , respectively; normalized in Figure 8E ) . Current densities from Ryr1AG/+ soleus muscle fibers from mice treated with glibenclamide were −82 ± 13 pA with an average pipette resistance of 2 . 1 ± 0 . 3 MΩ ( n = 8 fibers from 4 muscles; Figure 8F ) and showed no significant difference from wild-type littermates on glibenclamide ( −93 ± 18 pA with 2 . 2 ± 4 MΩ tip resistance , n = 8 fibers from 4 muscles; Figure 8F ) . Furthermore , a significant decrease in internal nuclei was detected in Ryr1AG/+ muscles ( 9 . 8 ± 5 . 2 in 0 . 6% K+ diet compared to 4 . 8 ± 2 . 7 in glibenclamide diet; n = 4 mice per condition; Figure 8G ) . Administration of glibenclamide also rescued the histological manifestations of the myopathy in Ryr1AG/+ mice ( Figure 8H–K , Table 2 ) , suggesting an alternative treatment with an FDA-approved drug that does not involve modulation of serum potassium levels . Together our in vivo data suggest that the mitochondrial dysfunction in Ryr1AG/+ soleus muscle is significantly improved through 5 . 2% potassium diets and glibenclamide administration . These studies suggest that rescue of the weakened state of muscles in this mouse myopathic model are , in part , due to a potassium-dependent mechanism . 10 . 7554/eLife . 02923 . 015Figure 8 . Inhibition of KATP channels can reverse the histological and myopathic phenotypes . ( A , B ) Average grip strength ( A , five trials/mouse and 5 mice/set; p values <0 . 001 ) and in vivo hanging task ( B , 10 trials/mouse , n = 5 per set; p values <0 . 001 ) assayed from 2-month old Ryr1+/+ ( black bar ) and Ryr1AG/+ ( grey bar ) mice maintained for 4 weeks on 0 . 6% K diet and glibenclamide ( 15 mg/kg/day ) . ( C ) ATP content from soleus muscle of Ryr1+/+ and Ryr1AG/+ mice after 4 weeks on 0 . 6% K diet without or with glibenclamide ( n = 4 muscles per condition ) . ( D ) Quantification of calcium marks visualized with Rhod2-AM from isolated soleus muscle from Ryr1+/+ and Ryr1AG/+ mice with and without glibenclamide therapy . ( E ) Normalized internal potassium concentrations after glibenclamide therapy . ( F ) KATP current densities of isolated soleus muscle fibers from Ryr1+/+ and Ryr1AG/+ littermates after 4 weeks on glibenclamide therapy compared to 0 . 6% K diet . ( G ) Quantification of number of internalized nuclei per 100 myofibers in all conditions ( n = 10 per muscle , n = 3 per set of muscles for total of n = 30; p values <0 . 001 ) . ( H–K ) Vastus lateralis myofiber from Ryr1+/+ and Ryr1AG/+ mice maintained for 4 weeks on glibenclamide . Arrow in Ryr1+/+ myofiber ( F ) shows the rare occurrence of internal nuclei . Cross-sections stained with H&E ( left panels ) and COX ( right panels ) . Ryr1AG/+ mice on glibenclamide show increased COX staining , similar to Ryr1+/+ . Scale bar = 50 μm; error bars as standard error of the mean ( SEM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 015 The levels of protein and gene expression might serve as measurements of recovery in Ryr1AG/+ mice following administration of a high-potassium diet . Indeed , as shown in Figure 7A and Table 3 , we found that expression of proteins or transcripts that encode KATP6 . 1 , KATP6 . 2 , AMPK , and SUR2 were returned to normal levels on high-potassium diet in Ryr1AG/+ adult muscle , whereas Na+ , K+-ATPase α1 transcripts were considerably decreased and KIR2 . 1 protein was increased in high-potassium diet in Ryr1AG/+ muscle . Thus , potassium supplementation rescues the expression of key molecules required for potassium homeostasis in RyR1AG/+ muscle , suggesting that KIR2 . 1 , KATP6 . 1 , KATP6 . 2 , Prkaa1 , Abbc9 , and Na+ , K+-ATPase α1 are potential biomarkers for myopathies associated with mutations in RyR1 and for evaluation of future potential therapies . Central core disease is the most common diagnosed myopathy due to mutations in RyR1 ( Jungbluth , 2007 ) . Using the molecular insights gained from our myopathic mouse model , we examined these biomarkers in RNA isolated from human skeletal muscles collected from CCD patients and controls without muscle disease ( normalized to a pooled control muscle sample , see methods ) . Expression levels of potential biomarkers from human muscle samples of healthy individuals , patients with RyR1 mutations causing CCD , and patients with CCD without identified RyR1 mutations were compared ( Table 4; Figure 9 ) . While PRKAA1 was increased in all muscle biopsies compared to the pooled human RNA sample , excluding it as a useful biomarker , all CCD patients showed a significant increase in expression of ABCC8 ( SUR1 ) , KCNJ2 ( KIR2 . 1 ) , KCNJ8 ( KATP6 . 1 ) , and KCNJ11 ( KATP6 . 2 ) , indicating that these four transcripts may be useful biomarkers for CCD in humans . Together , these data suggest specific biomarkers that could aid in diagnosis and treatment of CCD . 10 . 7554/eLife . 02923 . 016Table 4 . Relative variation of quantitative RT-PCR of control RNA and RNA from human muscle biopsiesDOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 016PatientRYR1 MutationCongenital MyopathyABCC8 ( SUR1 ) ATPA1 ( NKAα1 ) KCNJ8 ( KATP6 . 1 ) KCNJ11 ( KATP6 . 2 ) PRKAA1 ( AMPK ) 1NoNo0 . 98 ± 0 . 021 . 35 ± 0 . 052 . 15 ± 0 . 191 . 59 ± 0 . 393 . 88 ± 0 . 062NoNo0 . 94 ± 0 . 080 . 85 ± 011 . 06 ± 0 . 060 . 78 ± 0 . 164 . 35 ± 0 . 023YesYes13 . 61 ± 0 . 074 . 35 ± 0 . 114 . 95 ± 0 . 053 . 28 ± 0 . 162 . 90 ± 0 . 114YesYes30 . 68 ± 0 . 129 . 40 ± 0 . 087 . 72 ± 0/043 . 12 ± 0 . 1815 . 45 ± 0 . 135YesYes19 . 37 ± 0 . 3419 . 98 ± 0 . 546 . 48 ± 0 . 073 . 92 ± 0 . 0917 . 85 ± 0 . 096YesYes108 . 61 ± 1 . 8841 . 43 ± 0 . 319 . 74 ± 0 . 0816 . 24 ± 0 . 8917 . 32 ± 0 . 117YesYes4 . 69 ± 0 . 0737 . 72 ± 0 . 3617 . 45 ± 0 . 145 . 21 ± 0 . 354 . 04 ± 0 . 088YesYes46 . 10 ± 0 . 223 . 64 ± 0 . 0422 . 96 ± 117 . 77 ± 1710 . 73 ± 0 . 099YesYes159 . 20 ± 1 . 079 . 61 ± 0 . 1667 . 76 ± 0 . 248 . 33 ± 0 . 4235 . 59 ± 0 . 2910NoYes34 . 18 ± 0 . 191 . 09 ± 0 . 119 . 64 ± 0 . 134 . 43 ± 0 . 298 . 23 ± 0 . 0211NoYes22 . 30 ± 0 . 661 . 09 ± 0 . 1045 . 14 ± 0 . 1113 . 82 ± 0 . 8121 . 03 ± 0 . 04Values normalized to GAPDH before normalization to pooled control human muscle RNA . Red indicates significant value 0 . 005; Blue indicates significant value 0 . 05 . 10 . 7554/eLife . 02923 . 017Figure 9 . Relative variation of quantitative qRT-PCR of human control RNA and RNA from human muscle biopsies . Whisker plots derived from the data in Table 2 of control muscle ( yellow ) , congenital myopathy ( CCD ) with a mutation in RyR1 ( green ) , and without a mutation in RyR1 ( white ) . ( A ) ABCC8 , ( B ) ATPA1 , ( C ) KCNJ2 , ( D ) KCNJ8 , ( E ) KCNJ11 , ( F ) PRKAA1 . Values normalized to GAPDH before normalization to pooled control human muscle RNA . DOI: http://dx . doi . org/10 . 7554/eLife . 02923 . 017 In this study , we present a myopathic mouse model with core-like pathology that illuminates mechanisms underlying muscle weakness and mitochondria dysfunction . First , we identified a mutation in RyR1 that induces a myopathy with ultrastructural hallmarks of cores similar to the Ryr1I4895T/+ mutant mouse , a known mouse model of CCD ( Zvaritch et al . 2007 , 2009; Boncompagni et al . , 2010 ) . Second , we used this myopathic mouse model to identify a novel link between decreased internal calcium ( Figure 2 ) , decreased mitochondrial ATP production ( Figure 3 ) , and altered potassium homeostasis ( Figure 4 ) , mechanisms that are important in controlling excitability in muscle . Third , we provide evidence that KATP channels are elevated in the membrane and overactive , leading to increased potassium efflux in this myopathic mouse model , which could have direct effects on membrane excitability . Fourth , we provide clinically relevant findings that suggest a potassium-enriched diet , perhaps in combination with the sulfonylurea class of anti-diabetic drugs , which are potential therapies for myopathies with defects in potassium homeostasis . Using multiple assays and different muscles with different fiber type compositions , our data are consistent and show muscle weakness , pathological changes , and gene expression changes . The consistency of our data between muscle types further strengthens our hypotheses and conclusions . It should be noted that malignant hyperthermia , a disease associated with RyR1 mutations , is sensitive to potassium chloride ( Moulds and Denborough , 1974 ) , and therefore a testing method in patients or biopsy samples need to be initiated prior to potassium supplementation . Importantly , our work identifies potential diagnostic biomarkers of disease in patients with CCD and potentially non-RyR1 myopathies , which could be utilized to identify patients who may be candidates for potassium modulation-based therapies . The physiological mechanisms underlying RyR1-related myopathy are not well defined . One histological hallmark of RyR1-related myopathy is the presence of muscle cell cores , which indicate focal mitochondrial loss and decreased ATP production ( Boncompagni et al . , 2009 , 2010; Jungbluth et al . , 2011 ) . The Ryr1AG/+ mutant mouse muscles have decreased intracellular calcium , core-like structures and decreased ATP content , as well as other ultrastructural hallmarks of core-like diseases . Data from our model suggest that proper calcium release from internal stores is required for mitochondrial production of ATP and increased ATP levels , which in a feedback loop can then stimulate RyR1-dependent calcium release ( Meissner et al . , 1986; Laver et al . , 2001 ) . As muscle ATP decreases , KATP channels open to efflux K+ from the myoplasm to the interstitial space , which may protect against excessive muscle contraction . Efflux of K+ into the t-tubules depolarizes the tubular membranes and inhibits the opening of sodium channels ( Fraser et al . , 2011 ) . Sodium channel inhibition decreases the probability that a nerve-derived action potential will activate DHPR receptors that mechanically activate Ryanodine receptors , and thus myoplasmic release of stored Ca2+ is reduced . Therefore , in the Ryr1AG/+ model , RyR1 dysfunction , decreased internal Ca2+ release , potassium leakage and increased KATP-dependent efflux may potentiate muscle weakness and muscle deterioration associated with core-like structures . The chronic low-level calcium uptake into the mitochondria may then induce long-term mitochondrial dysfunction . The histological appearance of cores in the center of the muscle but more normal staining near the plasma membrane may be explained by the possibility that more centralized mitochondria receive less calcium , whereas SOCE and calcium influx from external sources at the plasma membrane may provide buffering to mitochondria that are adjacent to the plasma membrane . Our data suggest that the SR calcium release deficiencies and decreased mitochondrial ATP production might be sufficient to induce the myopathic phenotype in RyR1AG/+ mice . ATP provides the energy to cause muscle contraction , and relative ATP/ADP levels are important in providing information about muscle fatigue and ion homeostasis to the cell . In skeletal muscle , 90% of ATP is used for muscle contractions by the myosin ATPase , Ca2+ ATPase , and Na+/K+ ATPase ( Homsher , 1987; Clausen et al . , 1991; Smith et al . , 2005; Barclay et al . , 2007; Clausen , 2013b ) . Additional ATP-dependent enzymes may be particularly sensitive to a 30% loss of ATP in RyR1AG/+ muscle and this might contribute to a significant effect on excitability and membrane homeostasis after muscle contraction and fatigue . During skeletal muscle fatigue and fluctuations in potassium homeostasis , KATP channels open and efflux potassium from the muscle ( Pedersen et al . , 2009 ) . Moreover , plasma membrane and t-tubule KATP channels become more active as ATP is utilized and ADP accumulates ( Baczko et al . , 2005 ) . KATP channels are mechanistically linked to the DHPR-RyR1 complex ( Tricarico et al . , 1999 ) but the ATP binding sites within RyR1 are at a considerable distance from the point mutation in the RyR1AG protein ( Popova et al . , 2012 ) . Finally , KATP channel activity increases potassium conductance within the t-tubules after an action potential ( Gramolini and Renaud , 1997; Cifelli et al . , 2008; Banas et al . , 2011 ) . These data suggest a mechanistic link between E–C coupling , RyR1 function , ATP levels , KATP channel function , and potassium homeostasis . In our mouse model of a myopathy with core-like structures , increased intake of dietary potassium increases interstitial K+ , which leads to an increase in myoplasmic K+ . The resulting increase in interstitial K+ might counterbalance the K+ leak permeability and increase membrane excitability . A more excitable membrane increases the likelihood of voltage-dependent DHPR-RyR1 release of calcium from the SR during an action potential . Increased calcium in the myoplasm will enhance mitochondrial uptake of calcium . We propose that the waves of increased calcium into the mitochondria enhance ATP production ( Griffiths and Rutter , 2009 ) , which inhibits KATP channel opening and reverses the K+ leak , and stimulates RyR1-dependent calcium release from internal stores ( Meissner et al . , 1986; Laver et al . , 2001 ) , further perpetuating mitochondrial ATP production in this model of myopathy in which the phenotype can be ameliorated by increased potassium . However , we cannot rule out the possibility that the potassium supplementation may work indirectly at the muscle by depolarizing the locomotor circuit to increase neurotransmitter release at the neuromuscular junction and potentiate the voltage-dependent DHPR-RyR1 release of calcium . Defective potassium mechanisms appear to be the hallmark of Ryr1AG/+ mice as evidenced by 1 ) increased KATP channel current that would drive K+ efflux , 2 ) decreased internal K+ concentrations , and 3 ) a potassium leak . Altogether this may suggest an increase in interstitial K+ , but serum K+ levels may not correlate with interstitial K+ levels ( Clausen , 2013a ) and increased interstitial K+ would be expected to be rapidly removed by diffusion and through the capillary network of blood vessels . We did not observe a significant difference in serum K+ levels in 2-month old Ryr1AG/+ mice , but there was an increase in serum K+ in 6-month old Ryr1AG/+ mice relative to wild-type ( Figure 4—figure supplement 1 ) . In vitro , the rapid rise in intracellular K+ in Ryr1AG/+ muscle bathed in 3 mM K+ and treated with glibenclamide , an inhibitor of KATP channels , is in part due to the transport of K+ into the myoplasm through KIR2 . 1 and Na+ , K+-ATPases . These mechanisms that transport K+ into the myoplasm in Ryr1AG/+ muscles may not be sufficient in vivo to compensate for the K+ leakage or the increased activation of KATP channels located on the plasma membrane . We suggest that Ryr1AG/+ myopathic muscles have a K+ leak that persistently decreases intracellular K+ and leads to a loss of membrane excitability . Amelioration of muscle weakness through low-cost potassium supplementation could have profound impact on patients suffering from core myopathies and these data may have important implications for other disease states . Recent guidelines from the World Health Organization suggest that the world population is consuming inadequate levels of dietary potassium , which can lead to heart disease , stroke , respiratory distress , and increased fatigue ( Hartl , 2013 ) . Our studies suggest that Ryr1AG/+ muscle is in a myopathic state due to decreased intracellular Ca2+ levels , decreased K+ levels , decreased ATP levels and increased activation of KATP6 . 2 channels , which may mimic a fatigue-like state . Currently , myopathies with mutations in RyR1 are not considered metabolic myopathies , even though RyR1-S2844D mice with an intracellular calcium leak and muscle weakness exhibit decreased levels of mitochondrial activity and metabolic enzymes ( Andersson et al . , 2011 ) . Considering that our data suggest mitochondrial dysregulation of metabolic pathways and potassium signaling in Ryr1AG/+ muscle and muscles from patients with CCD , we were intrigued by the findings that some fatigue-inducing metabolic myopathies can be rescued through dietary potassium supplementation ( Villabona et al . , 1987; Caradonna et al . , 1992; Ismail et al . , 2001 ) . In the future , it will be of interest to determine whether the positive response to increased potassium levels is applicable to myopathies due to other RyR1 mutations and to other genetic causes . Our data also suggest that muscles of human CCD patients are sensitive to pathways that mediate potassium homeostasis . ATPA1 ( Na+ , K+-ATPase α1 ) showed increased expression only in CCD patients with mutations in RYR1 , suggesting a selective sensitivity to this ATP-sensitive transporter . With our suggestion that KATP6 . 2 channel inhibition may help protect skeletal muscles from the formation of cores , one might predict that the human patient data would show consistently higher levels of KCNJ11 ( KATP6 . 2 ) transcripts in the muscle biopsies . This is the case in CCD patients , although the level is variable . This variability could be due to the location of the muscle biopsy , as KATP6 . 2 channels in mice are more highly expressed in glycolytic muscle than oxidative skeletal muscle fibers ( Banas et al . , 2011 ) . These biomarkers may therefore provide a less invasive approach to identify patients with CCD myopathies and may eventually identify patients who are suitable candidates for treatment of CCD through modulation of potassium signaling pathways . ENU mutagenesis was performed as described ( Kasarskis et al . , 1998 ) on males of C57BL/6J background and then outcrossed onto 129S1/Svlmj background to score G3 embryos at embryonic day 18 . 5 for recessive mutations that affect embryonic locomotion . The Ryr1m1Nisw allele showed muscle weakness when heterozygous ( described here ) and a non-motile embryonic phenotype when homozygous ( unpublished results ) . The gene mutation was mapped using data from homozygous mutants , starting with a panel of 96 MIT and SKI SSLP markers , which mapped the Line1–4 mutation to the proximal third of chromosome 7 . The genetic region containing the mutation was narrowed to 3 Mb by the use of additional MIT SSLP markers on chromosome 7 and further meiotic mapping which followed linkage between the phenotype and C57BL/6J markers . As Ryr1 was a strong candidate gene in the critical interval , a complementation cross was performed between Ryr1AG ( Ryr1m1Nisw ) and Ryr1 tm1TAlle ( dyspedic null allele generated by Richard Allen , [Buck et al . , 1997] ) . These alleles did not complement as E18 . 5 trans-heterozygotes showed a similar non-motile phenotype as homozygotes of either allele . To identify the mutation in Line1-4 , genomic DNA in overlapping segments of the Ryr1 gene was amplified by PCR from phenotypic E18 . 5 embryos and compared with control E18 . 5 C57Bl/6J DNA . Subsequently , embryos were genotyped as follows: tissue was placed in tail lysis buffer ( 100 mM Tris–Cl , pH 8 . 0 , 5 mM EDTA , 0 . 2% SDS , 200 mM NaCl ) overnight . DNA was amplified using Taqman Gold ( Applied Biosystems ) with primer set to RyR1 mutation ( forward primer: GTGGAGTGTGGTGCTGTATATC forward and CCGTCACTGTCACCTTCTTG reverse primers ) . PCR amplification was performed for 35 cycles at 55°C . PCR product was sent to Barbara Davis Center Molecular Biology Service Center at the University of Colorado Denver for sequencing . For non-sequencing genotyping , the MIT SSLP markers , D7MIT114 and D7MIT267 , were amplified to examine linkage between phenotypic background ( C57BL/6J ) and control background ( 129/SvJ ) . Transcript accession for RyR1 was NM_0091009 and the ENU-induced change was A12864G ( protein NP_033135 , resulting in E4242G ) . All of the data presented here were obtained after outcrossing >8 generations onto 129S1/Svlmj background . Soleus muscles of 2-month old mice were fixed with 3 . 5% glutaraldehyde in 0 . 1 M cacodylate buffer ( pH 7 . 2 ) at 37°C . The central region of muscle was dissected to maximize the endplate regions , followed by a secondary post-fixation in 2% osmium tetroxide + 1 . 5% potassium ferrocyanide in 0 . 1 M cacodylate buffer for 1 hr at room temperature . Samples were dehydrated though an ethanol series ( 50 , 70 , 90 , 95 , and 100% ETOH ) and embedded in Epon/Araldite resin polymerized overnight at 60°C . Ultrathin sections were cut , placed on copper mesh grids , and double contrasted with 2% aqueous uranyl acetate and Reynold's lead citrate . Grids were photographed using a FEI Technai G2 BioTwin transmission electron microscope . COX and NADH staining protocols are available at html://neuromuscular . wustl . edu/pathol/histol . Histostained samples were examined on a Zeiss LSM 510 META laser scanning microscope . Soleus muscle was dissected from limb with tendons attached and bifurcated at the proximal tendon . Bifurcated muscle was affixed to glass bottom culture dishes ( MatTeK ) with Vetbond ( 3M , St . Paul , MN ) and cleaned of debris and adipose tissue . Tyrode buffer: 140 mM NaCl , 5 mM KCl , 10 mM HEPES , 2 mM MgCl2 , 2 . 5 mM CaCl2 , pH 7 . 2 , 290 mosm . The ratiometric cell-permeant calcium indicator Fura-2 AM ( Molecular Probes , Eugene , OR , USA ) was administered to the preparation for 45 min at 33°C then washed two times for 20 min at room temperature . Muscles were analyzed at 33°C using the 340/380 filters in a Zeiss Axio Observer Z1 with a Photometrics CoolSNAP HQ2 camera for wide-field imaging , then performing ratio intensity analysis . To deplete SR Ca2+ store , soleus fibers were treated with 10 μm thapsigargin for 2–4 min to induce Ca2+ release . The perfusion solution was then switched to 0 . 5 mM Mn2+ for 2–4 min to observe the extent of Mn2+ entry and establish the rate of SOCE . For all measurements of SOCE by Mn2+ quenching , the fluorescence signal was normalized using values determined by lysis of the cells with a solution containing 0 mM Ca2+/EGTA and 0 . 1% Triton X-100 at the end of the experiment . All experiments were conducted at ∼33°C . For [K+]i measurements , soleus muscle was affixed to glass bottom culture dishes ( MatTeK , Ashland , MA ) with Vetbond ( 3 M ) . The ratiometric cell-permeant potassium indicator PBFI-AM ( 5 μM; Life Technologies , Grand Island , NY ) together with 0 . 2% Pluronic F-127 for enhanced dye loading was loaded for 30 min in Ringer's solution ( 3 mM K+ ) . After loading , muscles were washed for 30 min to remove excess dye and to allow de-esterification of the AM dye . Data were collected from 10 regions per muscle with three muscles per condition . Muscles were analyzed at 33°C using the 340/380 filters in a Zeiss Axio Observer Z1 with a Photometrics CoolSNAP HQ2 camera for wide-field imaging , then performing ratio intensity analysis . To calibrate intracellular potassium concentrations , PBFI ratio intensities were fitted to PBFI intensities due to the variable K+ solutions with the addition of 10 μM gramicidin in extracellular solutions and using the intensity of ∼10 fibers of interest per muscle in >3 muscles from each condition . Calibration solutions were prepared with appropriate volumes of a high [K+] solution with potassium gluconate . Muscle tissue from the limbs was dissected and washed in PBS and resuspended in RIPA buffer supplemented with Complete Mini Protease Inhibitor Cocktail ( Roche Basel , Switzerland ) and Phosphatase Inhibitor Cocktail Set II ( EMD Millipore , Billerica , MA ) . Samples were sonicated for 1 min ( 15 s on , 15 s off , 25% power ) and incubated at 4°C for 1 hr on a rocking platform . Protein was quantified with the Bio-Rad Protein Assay ( BioRad , Hercules , CA ) . All samples were incubated at 37°C for 30 min prior to loading . Protein was separated on 4–12% Bis-Tris gels ( Invitrogen ) and transferred to Immobilon-FL PVDF membranes ( Millipore ) . Membranes were blocked in 5% dry milk in TBST for 1 hr . Primary antibody incubations were performed overnight at 4°C while rocking . Antibodies were diluted in TBST supplemented with 5% wt/vol BSA ( Sigma , St Louis , MO ) . Secondary antibody incubations were performed for 1 hr at room temperature in TBST supplemented with 5% dry milk . Western blots were imaged using the Odyssey Infrared Imaging System ( Li-Cor , Lincoln , NE ) . Quantitation was performed using Odyssey v3 . 0 software ( Li-Cor ) . Primary and secondary antibodies were used as follows: KIR6 . 1 ( KATP6 . 1; Santa Cruz , goat , 1:200 ) , KIR6 . 2 ( KATP6 . 2; Abcam , goat , 1:1000 ) , KIR2 . 1 ( Abcam , rabbit , 1:1000 ) , Goat anti-Rabbit IRDye 680 ( Li-Cor , 1:10 , 000 ) , Goat anti-Mouse IRDye 800CW ( Li-Cor , 1:10 , 000 ) , Donkey anti-Goat IRDye 680LT ( Li-Cor , 1:30 , 000 ) , Donkey anti-Mouse 800CW ( Li-Cor , 1:10 , 000 ) . Mitochondrial and cytosolic protein fractions were isolated using differential centrifugation . Immediately after sacrifice , soleus skeletal muscles from RyR1+/+ and RyR1AG/+ mice were extracted , rinsed in saline solution , weighed , then cut into four fragments , and homogenized in 1:5 wt/vol ice-cold isolation buffer containing 0 . 21 M mannitol , 0 . 07 M sucrose , 0 . 005 M HEPES , 0 . 001 M EDTA , 0 . 2% fatty acid-free BSA , pH 7 . 4 , using glass homogenizer . The homogenate was centrifuged at 600×g for 10 min at 4°C . The supernatant was then centrifuged at 12 , 000×g for 20 min . After the second spin , the supernatant ( membrane , t-tubule and cytosol ) was separated from the pellet ( peroxisome , lysome , and mitochondria ) and was stored at −80°C until used for Western blot assays . Soleus skeletal muscles were assayed immediately after homogenization using buffer described in membrane extract lysate to determine ATP content with a luciferin–luciferase based bioluminescence assay . The methodology of the ATP determination kit is provided in the experimental protocol by Molecular Probes ( A-22066 , Eugene , OR , USA ) . In order to determine ATP content , freshly extracted homogenate was added to a cuvette containing reaction buffer , D-luciferin , luciferase , and DTT and placed in a Sirius luminometer v . 2 . 2 ( Berthold Detection Systems , Pforzheim , Germany ) . Known concentrations of ATP standards were used to establish a standard curve . RNA was extracted from muscles using Trizol followed by DNaseI digestion and clean-up ( Qiagen RNAeasy Minikit , Venlo , Limburg ) and reverse transcribed using random hexamer primers and SuperScript III Reverse Transcriptase ( Invitrogen ) and amplified using TaqMan Universal PCR Master Mix ( Applied Biosystems ) . Quantitative PCR was performed on a Roche LightCycler 480 Real-Time PCR System . Calculations were performed by a relative standard curve method . Probes for target genes were from TaqMan Assay-on-Demand kits ( Applied Biosystems ) . Samples were adjusted for total RNA content by GAPDH in mice on diets and for the human samples . Control human muscle RNA was purchased from Amsbio ( R1234171-50 ) and Invitrogen ( AM7982 ) . The RNA samples were compared to a five sample human RNA skeletal muscle pool ( R1234171-P ) . RNA samples of myopathies were generously provided by Telethon Biobank ( Italy ) and University of Colorado Neurology Department following University of Colorado COMIRB approval for human subject research . Diets were based on studies examining hypertension in mice ( Grimm et al . , 2009 ) . Diets were purchased through Harlan Laboratories , Inc . and are TD . 10942 ( 0 . 6% K+ ) and TD . 94121 ( 5 . 2% K+ ) . These diets have similar ratios of three K+ sources . Following weaning ( 1 month of age ) , animals were placed on potassium diets ad libitum for 4 weeks followed by determination of muscle strength and molecular and histological examination . For enalapril and glibenclamide studies , following weaning , mice were placed on 0 . 6% K+ diet and given enalapril in drinking water ( 0 . 02 mg/ml ) or glibenclamide via daily oral gavage ( 15 mg/kg/day ) for 4 weeks . Skeletal muscle fibers were isolated from the soleus muscles by a classical enzymatic dissociation process; muscles were incubated for 1 hr at 37°C in Tyrode's solution containing collagenase ( 2 . 0 mg/ml , Sigma , Type 1 ) for patch-clamp experiments or 1 . 5 hr at 37°C in Tyrode's solution containing collagenase ( 2 . 2 mg/ml , Sigma , Type 1 ) for calcium imaging and mitochondrial function experiments . After enzyme treatment , muscles were rinsed with Tyrode's solution and kept in low calcium Ringer's solution at 4°C until analysis . Ringer's solution used during collection of isolated muscle fibers contained 145 mM NaCl , 5 mM KCl , 1 mM MgCl2 , 0 . 5 mM CaCl2 , 5 mM glucose , and 5 mM HEPES and was adjusted to pH 7 . 2 . The patch-pipette solutions contained 150 mM KCl , 2 mM CaCl2 , and 1 mM HEPES ( pH 7 . 2 ) . The bath solution contained 150 mM KCl , 5 mM EGTA , and 1 mM HEPES ( pH 7 . 2 ) . Intact skeletal muscle fibers were separated from the muscle mass by gently triturating the muscle with a plastic 1000 μl pipette . Experiments were performed in inside-out configurations by using the standard patch-clamp technique . Initial pipette resistance was between 1 . 3–3 . 0 MΩ . Channel currents were recorded during voltage steps going from 0 mV of holding potential to −60 mV voltage membrane ( Vm ) immediately after excision , at 20–22°C , in the presence of KCl on both sides of membrane patches . Useful patches required minimal seal resistance of 1 . 1 GΩ . The mean currents were calculated by subtracting the baseline level from the open-channel level of each current trace . A minimum of three traces was collected for each seal . Digital averaging of traces was performed using CLAMPFIT ( Molecular Devices , Sunnyvale , CA ) . ATP ( 5 mM ) or glibenclamide ( 1 mM ) was used to examine the pharmacological response of the channel . To evaluate muscle weakness in vivo , we compared RyR1AG/+ and RyR1+/+ mice on the 5 . 2% K+ or 0 . 6% K+ diet without or with glibenclamide or enalapril using grip strength tests ( Costa et al . , 2010 ) and wire hanging task ( Ogura et al . , 2001 ) . Similar tests have been used to evaluate muscle weakness in CCD mice ( Loy et al . , 2011 ) . Performance in Hanging Wire Task was scored on a 0 to 5 scale according to Loy et al . ( 2011 ) : 0 , immediately fell off the bar; 1 , hung onto bar with two forepaws; 2 , hung onto bar with two forepaws and attempted to climb onto the bar; 3 , hung onto the bar with two forepaws and one or both hind paws; 4 , hung onto the bar with all four paws and tail wrapped around the bar; 5 , hung onto the bar with all four paws and tail wrapped around the bar and escaped onto one of the supports . Body weight was measured before and after therapies with no significant difference in weight ( data not shown ) . Malignant hyperthermia test was performed by placing Ryr1+/+ and Ryr1AG/+ mice in a 41°C humidified incubator for 30 min . To test for sensitivity to halogenated anesthetics , stage 3 anesthetic isoflurane ( ∼5 . 5 × 10−5 ml/cm3 ) was administered through a nasal cone to Ryr1AG/+ mice at a maximum of 30 min exposure . Isolated soleus muscle fibers from 2-month old Ryr1+/+ and Ryr1AG/+ were examined for mitochondrial function . MitoTracker Deep Red FM ( Invitrogen ) was added to all fibers to identify the mitochondria and Powerload ( Invitrogen ) was added to calcium indicators for 20 min . To examine mitochondria calcium changes in intensity , Rhod2-AM ( Invitrogen ) was added to the fibers in Normal Ringer's solution for 30 min at room temperature and then the cultures were washed with normal Ringers . To examine mitochondrial membrane potential , TMRE ( Invitrogen ) with saponin ( Sigma ) were added at the manufacturer's recommended dose for 30 min at room temperature . FCCP ( carbonyl cyanide 4- ( trifluoromethoxy ) phenylhydrazone; ABCAM , Cambridge , England ) was added at the end of the experiment to eliminate the mitochondrial membrane potential and the value was used to normalized to the experimental fluorescent TMRE value . To examine superoxides , MitoSOX ( Invitrogen ) indicator was added at the manufacturer's recommended dose for 15 min at 37°C . Antimycin A ( Sigma ) was used as a positive control at the end of the experiment to normalize to the experimental signal . We placed the fiber cultures into a Zeiss Environmental incubation system ( 5% humidity , 95% O2 ) at 22°C . Images of mitochondrial regions ( 3 per fiber and 5 fibers per muscle in 3 muscles ) were taken on a Zeiss LSM510 laser-scanning microscope of 1000 μm2 fields . Scans were taken every 2 min . All data were normalized to baseline per mitochondrial region to give a Δf/f value ( Image J ) . The unpaired two-tailed Student's t test was used to compare means for statistical differences . Data in the manuscript are represented as mean ± SEM unless otherwise indicated . p < 0 . 05 was considered significant . Normal methods to collect serum do not provide an accurate representation of interstitial ( extracellular ) potassium levels ( Mazzaccara et al . , 2008 ) . However , to determine whether any overt discrepancies of serum potassium , sodium , and chloride can be detected , serum was collected from 2-month and 6-month old wild-type and Ryr1AG/+ mice . Analysis was performed using a Starlyte ISE Analyzer ( Alfa Wassermann , West Caldwell , NJ ) . All experiments were conducted in accordance with the protocols described in the Guide for the Care and Use of Laboratory Animals ( NIH . Revised , 2011 ) and we received University of Colorado COMIRB approval for human subject research .	Skeletal muscle covers our skeleton and allows us to move around . One disorder that leads to weakness in skeletal muscle—known as central core disease—can leave affected infants ‘floppy’ and delay the development of motor skills such as sitting , crawling , and walking . While no cure or treatment currently exists for the disease , researchers have found that most cases are connected to a mutation in the gene that makes a protein called ryanodine receptor type 1 ( RyR1 ) . RyR1 belongs to a family of proteins that create channels for the controlled release of calcium ions from stores within cells . For muscle cells to contract , calcium ions must be released from these internal stores at the same time as potassium ions leave the cells . To relax the muscle cells , calcium ions are pumped back into the internal stores and potassium ions are taken back into the cell . Previous studies have established a role for RyR1 in the contraction of skeletal muscle , but the precise molecular details are not known . Here , Hanson et al . studied mice that had symptoms of central core disease due to a mutation in the gene that makes RyR1 . The muscle weakness in these mice was caused by defects that hindered the release of calcium ions from internal stores and leakage of potassium ions from the muscle cells . The experiments reveal that a high-potassium diet alleviates the symptoms of disease in the mice by increasing the amount of potassium surrounding the muscle cells . Treatment with an existing drug called glibenclamide also reversed the disease symptoms by reducing the leakage of potassium ions from the cells . Hanson et al . also found several genes involved in controlling potassium ion levels in cells that could act as indicators of the presence of the disease . These findings suggest that therapies targeting the control of potassium ion levels in muscle cells could minimize muscle damage in patients with central core disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2015	Potassium dependent rescue of a myopathy with core-like structures in mouse
The extracellular matrix ( ECM ) plays an instrumental role in determining the spatial orientation of epithelial polarity and the formation of lumens in glandular tissues during morphogenesis . Here , we show that the Endoplasmic Reticulum ( ER ) -resident protein anterior gradient-2 ( AGR2 ) , a soluble protein-disulfide isomerase involved in ER protein folding and quality control , is secreted and interacts with the ECM . Extracellular AGR2 ( eAGR2 ) is a microenvironmental regulator of epithelial tissue architecture , which plays a role in the preneoplastic phenotype and contributes to epithelial tumorigenicity . Indeed , eAGR2 , is secreted as a functionally active protein independently of its thioredoxin-like domain ( CXXS ) and of its ER-retention domain ( KTEL ) , and is sufficient , by itself , to promote the acquisition of invasive and metastatic features . Therefore , we conclude that eAGR2 plays an extracellular role independent of its ER function and we elucidate this gain-of-function as a novel and unexpected critical ECM microenvironmental pro-oncogenic regulator of epithelial morphogenesis and tumorigenesis . It is well established that proper organization , maintenance , and function of the epithelial organs are largely ascribed to cell-cell adhesion and polarization of epithelium and its interaction with extracellular matrix ( ECM ) . Apical-basal polarity is a fundamental feature of epithelial cells required for cell division , differentiation and morphogenesis ( Arumugam et al . , 2008; Nelson , 2009 ) . Polarization is initiated by signaling from cell-to-cell and cell-to-ECM contacts ( Nelson and Bissell , 2006 ) . Polarization of epithelial cells is implicit for the development of lumens , which are essential for glandular tissues to carry out their normal functions . Conversely , loss of tissue organization is among the earliest hallmarks of epithelial cancer . Therefore , understanding mechanisms of lumen and apical surface formation will improve future prospects for therapy . Anterior gradient 2 ( AGR2 ) encodes an endoplasmic reticulum ( ER ) -resident protein mainly expressed in epithelial cells in human . Enhanced intracellular AGR2 ( iAGR2 ) expression is observed in many cancers ( reviewed in Ref [Chevet et al . , 2013] ) . Previously , we have demonstrated that iAGR2 overexpression could represent a mechanistic intermediate between endoplasmic reticulum quality control ( ERQC ) and tumor development ( Higa et al . , 2011; Chevet et al . , 2013 ) . In such model , increased iAGR2 expression could enhance ER protein homeostasis/proteostasis thereby allowing tumor cells to cope with abnormal protein production and secretion and contributing to the aggressiveness of cancer ( Higa et al . , 2011 ) . The latter was demonstrated using both in vitro and in vivo approaches ( Chevet et al . , 2013 ) . Although the iAGR2-mediated ER proteostasis control model is appealing , it was also observed that in cancer , AGR2 was present in the extracellular space , serum , and urine ( Shi et al . , 2014; Park et al . , 2011 ) , thereby opening other avenues for its role on tumor microenvironment . Despite the detailed characterization of its intracellular function , the physiological role of extracellular AGR2 ( eAGR2 ) remains unknown . AGR2 is a Protein-Disulfide Isomerase ( PDI ) , PDIA17 ( Persson et al . , 2005 ) , and although the intracellular roles of PDIs have been well documented , some of these proteins were also found in the extracellular milieu , with unclear functions . For instance , we have previously shown that PDIA2 is secreted into the lumen of the thyroid follicles by thyrocytes to control extracellular thyroglobulin folding and multimerisation ( Delom et al . , 1999; Delom et al . , 2001 ) . Further , PDIA3 was found to be secreted and to interact with ECM proteins ( Dihazi et al . , 2013 ) and QSOX1 was reported to participate in laminin assembly thereby controlling ECM functionality ( Ilani et al . , 2013 ) . We and others , have recently demonstrated that epithelial organization and many physiological cell-cell and cell-ECM contacts , cellular polarity , and secretory functions are preserved in epithelial organoids ( Fessart et al . , 2013; Kimlin et al . , 2013 ) . Therefore , to address whether eAGR2 could act as a pro-oncogenic molecule in the ECM , we have used our human epithelial organoid model ( Fessart et al . , 2013 ) . We demonstrate , for the first time , that eAGR2 plays an extracellular role independent of its ER function and we elucidate this gain-of-function as a novel and unexpected critical ECM microenvironmental pro-oncogenic regulator of epithelial morphogenesis and tumorigenesis . To evaluate the correlation between AGR2 expression levels and lung cancer , we monitored AGR2 endogenous expression in a panel of human lung bronchial epithelial cell lines . High AGR2 expression was only observed in lung tumor cell lines ( A549 , H23 , H1838 ) compared to a non-tumorigenic human bronchial epithelial cell ( HBEC ) ( Figure 1A–C ) . Moreover , the expression pattern of AGR2 in tumor and non-tumor bronchial organoids ( Figure 1D ) was similar to that observed in 2D culture ( Figure 1A ) . Immunohistochemistry of AGR2 in a cohort of 34 non-small cell lung cancer ( NSCLC ) patients ( Supplementary file 1A ) revealed that AGR2 was overexpressed in tumors compared to adjacent non-tumor tissue ( Figure 1E ) . Consequently , AGR2 expression was increased in NSCLC tissues ( Figure 1E ) , and was essentially restricted to type II pneumocytes ( Figure 1F ) . We then used a log-rank test with Kaplan–Meier estimates to analyze the cohort in order to stratify patient samples as having high , low/intermediate AGR2 expression status ( Supplementary file 1A ) . High AGR2 expression correlated with low survival rate and the low/intermediate AGR2 expression with high survival rate in NSCLCs patients ( Figure 1G ) . Hence NSCLC patients can be sorted into poor and good prognosis groups as a function of high or low/intermediate AGR2 expression levels , respectively . Taken together , these results demonstrate in vitro and in vivo correlations between AGR2 expression levels and lung cancer , suggesting a function for this protein in tumor development , progression and aggressiveness . 10 . 7554/eLife . 13887 . 003Figure 1 . AGR2 is overexpressed in lung cancer cell lines and tumor tissues . ( A ) Analysis by immunofluorescence of AGR2 expression in normal human bronchial epithelial cells ( HBEC ) and three lung cancer cell lines ( A549 , H23 and H1838 ) grown in 2D culture . Scale bars , 50 μm . ( B ) Quantification of AGR2 protein expression in cell lines according to immunofluorescence . The stacked bars show the percent contribution of high and low AGR2-positive cells relative to the total number of cells per field . ( C ) Expression of AGR2 protein detected by Western blot in a panel of human lung epithelial cell lines . Values correspond to three independent experiments . Data are mean ± SEM . ( D ) Confocal cross-sections of organoids stained with AGR2 antibody ( red ) and DAPI ( blue ) for nucleus , in normal HBECs and three lung cancer cell lines ( A549 , H23 and H1838 ) . Scale bars , 50 μm . ( E ) AGR2 expression determined by immunohistochemistry in sections of formalin-fixed paraffin-embedded normal human lung samples and in the different lung adenocarcinoma subtypes . II = squamous cell carcinoma , IV = adenocarcinoma , VI = large cell carcinoma as compared to normal tissues ( I , III , V ) ( x200 ) . ( F ) Pulmonary lung carcinoma showing brown , nuclear immunostaining for TTF-1 expression and cytoplasmic immunostaining for AGR2 expression ( dual color Multiplex TTF-1 + AGR2 immunostain; x200 ) . ( G ) Kaplan-Meier survival curves of lung cancer patients . The cumulative survival was related to different levels of AGR2 expression: Group 1 , low to moderately positive stains ( n=17 ) ; and Group 2 , strongly positive AGR2 stains ( n=16 ) , as defined in Materials and methods and Supplementary file 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 003 To test the relevance of AGR2 overexpression in tumorigenesis , we silenced its expression in lung adenocarcinoma cell lines ( A549 , H23 and H1838 ) using lentivirus-mediated infection with AGR2 shRNA ( Sh-AGR2 ) . First , we monitored the efficacy of AGR2 depletion using immunofluorescence ( Figure 2A , D , G ) and Western blotting ( Figure 2B , E , H ) . Cell growth was strongly inhibited in Sh-AGR2 cells compared to control cells ( Sh-ctl ) ( Figure 2C , F , I–J and Figure 2—figure supplement 1 ) with no cell death detected ( Figure 2K ) . Interestingly , soft-agar colony formation ( Figure 2L–N ) ( used as indicator of malignant transformation ) significantly reduced upon AGR2 silencing in adenocarcinoma cell lines , thereby indicating that AGR2 may favor tumor progression by increasing anchorage-independent growth . To further characterize the role of AGR2 in cancer , we assessed the impact of AGR2 depletion using the 3D human bronchial organoids system which we previously established ( Fessart et al . , 2013 ) . AGR2 silencing dramatically decreased the formation of tumor organoids ( Figure 3A–C ) . To determine whether AGR2 is required for tumor progression in vivo , we then evaluated the in vivo tumorigenic potential of Sh-ctl and Sh-AGR2 cells using two different adenocarcinoma cell lines ( A549 and H1838 ) , following tail vein injection in immunodeficient mice . Sh-ctl mice presented large metastases that were easily detected at the surface of the lungs compared to mice injected with Sh-AGR2 cells ( Figure 3D , F ) . Moreover , AGR2 depletion significantly reduced total metastasis number ( by ~50% ) in both A549-Sh-AGR2 and H1838-Sh-AGR2 injected mice ( Figure 3E , G ) . Taken together , these results show that overexpression of AGR2 in cancer cells increases their tumorigenic and metastatic potential . 10 . 7554/eLife . 13887 . 004Figure 2 . AGR2 knock down in lung cancer cells decreases cell proliferation and malignant transformation . ( A , D , G ) Analysis by immunofluorescence of AGR2 depletion in our three lung cancer cell lines ( A549 , H23 , and H1838 ) . Scale bars , 50 μm . ( B , E , H ) Western blot analysis showing the down-regulation of AGR2 protein levels in control and Sh-AGR2 transfected cells . Calnexin ( CANX ) concentrations are shown as the loading control . One representative experiment ( n = 3 ) is shown . ( C , F , I ) Growth of cells harboring Sh-ctl or Sh-AGR2 ( three independent experiments ) . Data are mean ± SEM . ( J ) Quantification of the percentage of BrdU positive cells . Data are presented as mean ± SEM of at least three independent experiments . P < 0 . 05 . ( K ) Quantitation of cell death in AGR2 depleted cells ( Sh-AGR2 ) . Results are representative of three independent experiments . The cell death rate was determined by Trypan blue dye-exclusion assay . n . s: not significant . ( L–N ) Soft agar colony formation of cells expressing Sh-ctl or Sh-AGR2 . The graph shows the number of colonies ( mean ± SEM . ) after 3 weeks of three independent experiments . The p values ( determined by Student's t test ) are relative to Sh-ctl cells . p≤0 . 05 . p≤0 . 01 and p≤0 . 001 . Shown at the left are representative images of the colonies formed by each cell type . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 00410 . 7554/eLife . 13887 . 005Figure 2—figure supplement 1 . Cell proliferation on A549-Sh-AGR2 depleted cells . Representative images of A549 control cell ( Sh-ctl ) and A549-Sh-AGR2 depleted cells ( Sh-AGR2 ) stained for BrdU ( red ) administrated 12 hr before , and with DAPI ( blue ) . Images were subjected to high throughput imaging ( bottom panels , analysis data ) . After acquisition of the dataset , images were segmented by a watershed transformation to identify individual cells ( bottom right ) . Scale bars: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 00510 . 7554/eLife . 13887 . 006Figure 3 . AGR2 knockdown reduces tumorigenicity and metastatic potential . ( A–C ) Representative brightfield pictures of tumor organoids grown in 3D from cells harboring Sh-ctl or Sh-AGR2 ( three independent experiments ) . The bar graph shows the mean of organoids per well ( mean ± SEM . ) after 10 days of culture from three independent experiments . The p values ( determined by Student's t test ) are relative to Sh-ctl cells . p≤0 . 05 . p≤0 . 01 and p≤0 . 001 . ( D , F ) Representative pictures of the lungs in mice injected with Sh-ctl cells or Sh-AGR2 cells allowed to developed tumors for 2 weeks . Note the extensive lesions ( white areas ) on the surface of lungs from Sh-ctl mice and the relatively normal appearance of the lungs from Sh-AGR2 mice . ( E , G ) Representative H&E-stained sections for lungs injected with Sh-ctl cells or Sh-AGR2 cells ( montage of x5 magnification of lung sections ) . Scale bars , 1 mm . Box plot showing the mean number of tumors per mouse in Sh-ctl ( n=6 ) vs Sh-AGR2 ( n=6 ) groups . The p values ( determined by Student's t test ) are relative to Sh-ctl cells . p≤0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 006 To further confirm that AGR2 is essential for cancer tumorigenesis , we next investigated the effects of overexpressing AGR2 in non-tumorigenic HBECs using lentivirus-mediated infection with either empty vector ( HBEC-EV ) ( Figure 4A , top panels ) or AGR2 containing vector ( HBEC-AGR2 ) ( Figure 4A , bottom panels ) . In HBEC-AGR2 cells , AGR2 co-localized with calnexin ( CANX ) , an ER resident type I integral membrane protein ( Figure 4B , top panels ) and BiP ( GRP78 ) , a soluble HSP70 molecular chaperone located in the lumen of the ER ( Figure 4B , middle panels ) . In contrast and as expected , AGR2 did not co-localize with GM130 , a cis-Golgi marker ( Figure 4B , bottom panels ) . These results show that AGR2 localized to the ER in HBEC-AGR2 cells ( Figure 4B ) . Increased AGR2 expression in these cells ( HBEC-AGR2 ) was also confirmed using Western blotting ( Figure 4C ) . Remarkably , AGR2 overexpression enhanced cell proliferation ( Figure 4D–G ) and increased the formation of organoids ( Figure 4H ) , thereby indicating that the expression of AGR2 dictates organoid-initiating frequency . 10 . 7554/eLife . 13887 . 007Figure 4 . Enhanced cell proliferation and transformation after AGR2 overexpression in HBECs . ( A ) Analysis by immunofluorescence of AGR2 overexpression in HBECs . Scale bars , 50 μm . ( B ) Analysis of AGR2 sub-cellular localization in infected HEBCs by immunofluorescence . Calnexin ( CANX ) and Bip ( GRP78 ) are used as an ER localization control and GM130 is used as a Golgi localization control . Scale bars , 50 μm . ( C ) Up-regulation of AGR2 protein concentrations in control cells ( HBEC-EV ) and in cells infected with AGR2 ( HBEC-AGR2 ) , as analyzed by western blot . Calnexin ( CANX ) concentrations are shown as the loading control . One representative experiment ( n = 3 ) is shown . ( D ) Growth of cells stably expressing AGR2 or empty vector ( EV ) ( three independent experiments ) . Data are mean ± SEM . ( E ) Quantification of the percentage of BrdU positive cells . Data are presented as mean ± SEM of at least three independent experiments . p<0 . 001 . ( F ) Quantitation of cell death in AGR2 overexpression cells . Results are representative of three independent experiments . The cell death rate was determined by Trypan blue dye-exclusion assay . n . s: not significant . ( G ) Cell proliferation on HBEC overexpressing AGR2 cells . Representative images of control HBECs ( HBEC-EV ) and HBEC-AGR2 stained for BrdU ( red ) administrated 12 hr before , and with DAPI ( blue ) . Images were subjected to high throughput imaging ( bottom panels , analysis data ) . After acquisition of the dataset , images were segmented by a watershed transformation to identify individual cells ( bottom right ) . Scale bars: 100 μm . ( H ) The bar graph shows the mean of organoids per well ( mean ± SEM . ) after 10 days of culture from three independent experiments . Shown at the left are representative images of the organoids formed by each type of cell HBEC-AGR2 and HBEC-empty vector . The p values ( determined by Student's t test ) are relative to empty vector infected cells . p≤0 . 01 and *p≤0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 007 Since AGR2 has been detected in the body fluids from cancer patients ( Chevet et al . , 2013 ) as well as in the conditioned media from pancreatic and prostate cancer cells ( Arumugam et al . , 2008; Zhang et al . , 2005 ) , we hypothesized that tumor organoids may also secrete AGR2 ( extracellular AGR2 / eAGR2 ) . Indeed , eAGR2 was detected in the extracellular medium of adenocarcinoma organoids ( Sh-ctl ) , but not in adenocarcinoma organoids silenced for AGR2 ( Sh-AGR2 ) ( Figure 5A ) . To ensure that the presence of eAGR2 in the extracellular medium correlated to its overall level of expression , we also assessed the levels of eAGR2 in non-tumor organoids overexpressing AGR2 ( HBEC-AGR2 ) ( Figure 5B ) . Similarly , we found that eAGR2 was secreted in the extracellular medium from HEBC-AGR2 organoids ( Figure 5B ) . We also tested the presence of two other proteins of the secretory pathway: Calnexin ( CANX ) and BiP/GRP78 ( Figure 5B ) . None of these ER resident proteins were detected in the extracellular medium ( Figure 5B ) . This demonstrates the specific presence of ER-resident AGR2 in extracellular medium from non-tumor organoids overexpressing AGR2 ( HBEC-AGR2 ) . 10 . 7554/eLife . 13887 . 008Figure 5 . AGR2 is secreted in the extracellular medium and interacts with extracellular matrix ( ECM ) . ( A ) Proteins in the conditioned medium were subjected to immunoblotting with anti-AGR2 antibody . Transferrin ( Trf ) concentrations are shown as the loading control . One representative experiment ( n = 3 ) is shown . ( B ) Up-regulation of AGR2 protein concentrations in extracellular and intracellular lysates of HBEC infected with either empty vector ( HBEC-EV ) or AGR2 ( HBEC-AGR2 ) , as analyzed by western blot . Trf concentrations are shown as the loading control for extracellular lysates and GAPDH concentrations are shown as loading control for intracellular lysates . The transmembrane protein Calnexin ( CANX ) and the KDEL containing protein BiP are shown as control ER-resident proteins in extracellular and intracellular lysates . One representative experiment ( n = 3 ) is shown . ( C ) Levels of extracellular and intracellular AGR2 protein detected by immunoblotting in H23 organoids as compared to the levels of extracellular and intracellular AGR2 protein in HBEC organoids expressing ( HBEC-AGR2 ) or not ( HBEC-EV ) AGR2 . One representative experiment ( n = 3 ) is shown . ( D ) The stacked bars show the relative percentage of intracellular vs extracellular AGR2 . ( E , F ) Quantification of the relative signal intensity of AGR2 and two other ER-resident proteins ( CANX and Bip ) compared to GAPDH in intracellular lysates from AGR2-depleted and control H23 organoids ( E ) or from AGR2 overexpressing and control HBEC organoids ( F ) . ( G ) H23 cells were treated with CHX ( cycloheximide ) for the indicated periods ( hours ) before cell lysates and extracellular medium were prepared as described in Materials and methods . AGR2 was immunoprecipitated in intracellular and extracellular lysates with a monoclonal AGR2-specific antibody and then detected by immunoblot analysis . ( H ) The densitometric quantification of the remaining protein expressed as relative to the starting amount is shown in the diagram . ( I ) Structure model of the KTEL motif ( K172-T173-E174-L175 ) of AGR2 , PDB ID 2LNS . The four residues ( KTEL ) of the ER retention motif are shown in the zoom ( bottom panel ) . ( J ) Schematic representation of the AGR2 point or deletion mutants ( red: thioredoxin-like domain; orange: ER-retention motif ) . ( K ) Impact of the thioredoxin-like domain and of the ER-retention motif on AGR2 secretion . Quantification of the relative amount of secreted AGR2 for all the mutants . AGR2 secretion was normalized in each experiment and is compared to control non-transfected HEK-293T cells ( NT ) . Data are mean ± SEM . p<0 . 05 . ( L–M ) Cell-derived matrices were generated from HBEC-EV and HBEC-AGR2 cells after 8 days of culture . Proteins were solubilized , separated by SDS-PAGE and stained with Coomassie Blue ( L ) or immunoblotted for Laminin or AGR2 ( M ) . ( N ) Cell-derived matrices from HBEC-EV and HBEC-AGR2 cells were processed for immunofluorescence , stained with anti-laminin ( red ) and anti-AGR2 antibodies ( green ) and images were collected by microscopy . Scale bars: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 00810 . 7554/eLife . 13887 . 009Figure 5—figure supplement 1 . AGR2 expression does not alter the secretory pathway and AGR2 secreted is not O-GlcNacylated . ( A ) Proteins in the intracellular lysate in H23 organoids silenced ( Sh-AGR2 ) or not ( Sh-ctl ) for AGR2 were subjected to immunoblotting with anti-AGR2 antibody . GAPDH is shown as loading control . The transmembrane protein Calnexin ( CANX ) and the KDEL containing protein BiP are shown as control ER-resident proteins in intracellular lysates . One representative experiment ( n = 3 ) is shown . ( B ) Proteins in the intracellular lysate in HBEC infected with either empty vector ( EV ) or AGR2 , as analyzed by immunoblot with anti-AGR2 antibody . GAPDH concentrations are shown as loading control . The transmembrane protein Calnexin ( CANX ) and the KDEL containing protein BiP are shown as control ER-resident proteins in intracellular lysates . One representative experiment ( n = 3 ) is shown . ( C ) Extracellular AGR2 from medium organoid lysates of HBEC organoids infected with either empty vector ( EV ) or AGR2 was immunoprecipitated with anti-AGR2 antibody and immunoblotted with anti-AGR2 or anti-O-GlcNAc antibodies . Intracellular organoid lysates were also probed with anti-AGR2 and anti-O-GlcNAc antibodies . One representative experiment ( n = 3 ) is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 009 To further characterize the extent of AGR2 secretion , a ratiometric comparison between intracellular and extracellular AGR2 proteins was performed ( Figure 5C–D ) . AGR2 protein was immunoprecipitated from both intracellular and extracellular organoid extracts and analyzed by Western blotting ( Figure 5C ) . In adenocarcinoma organoids ( H23 ) , we found ~20% of eAGR2 and ~80% of iAGR2 ( Figure 5D ) , while in non-tumor organoids ( HBEC-EV ) , AGR2 protein was exclusively intracellular ( Figure 5D ) . In contrast , AGR2 overexpression in non-tumor organoids ( HBEC-AGR2 ) forced the secretion of eAGR2 ( ~30% in the extracellular medium ) ( Figure 5D ) . Next , to test if the distribution of AGR2 could modify other components of the secretory pathway , a ratiometric comparison of both BiP and CANX was performed in tumor organoids ( Sh-ctl ) as compared to tumor organoids silenced for AGR2 ( Sh-AGR2 ) ( Figure 5E and Figure 5—figure supplement 1A ) . Results showed that of AGR2 depletion had no impact on BiP and CANX expression . Similarly , we found no difference in the expression of BiP and CANX between non-tumor organoids overexpressing AGR2 ( HBEC-AGR2 ) or not ( HBEC-EV ) ( Figure 5F and Figure 5-—figure supplement 1B ) indicating that the secretory pathway was not altered upon changes in AGR2 expression . To further correlate the amount of AGR2 produced by tumor organoids to the capacity of cancer cells to secrete eAGR2 , we transiently blocked protein synthesis using a cycloheximide ( CHX ) pulse-chase approach and evaluated the amounts of iAGR2 and eAGR2 in tumor organoids . After 8 hr of CHX treatment , eAGR2 was practically undetectable ( Figure 5G–H ) , whereas ~60% of iAGR2 was still present . These results demonstrate that CHX treatment significantly reduces AGR2 protein synthesis which impacts on AGR2 protein secretion . Indeed , extracellular AGR2 rapidly disappeared from the culture medium through yet unknown mechanisms and became undetectable within 8 hr of CHX treatment , due to the attenuated iAGR2 synthesis and the subsequent reduction of the secreted pool . Although we have not yet identified the mechanisms by which eAGR2 disappears from the medium , several hypothesis could be postulated concerning its extracellular fate . As such the secreted protein could be i ) trapped in the ECM and become unavailable as a free protein in the medium; ii ) aggregated and therefore remain in the unsoluble fraction before biochemical analysis; iii ) degraded by extracellular proteases released by the cell or even iv ) internalized by the cells and degraded through an endolysosomal pathway . Thus , the precise mechanisms by which eAGR2 disappears remain to be further investigated . Recently , AGR2 was demonstrated to be O-glycosylated upon secretion ( Clarke et al . , 2015 ) . Protein O-GlcNAcylation was examined in non-tumor organoids overexpressing AGR2 ( HBEC-AGR2 ) or not ( HBEC-EV ) following eAGR2 immunoprecipitation ( Figure 5—figure supplement 1C ) . In intracellular organoid extracts , similar O-GlcNAc levels were observed between HBEC-EV and HEBC-AGR2 organoids ( Figure 5—figure supplement 1C ) and , as expected , the presence of AGR2 was only detected in HEBC-AGR2 organoids ( Figure 5—figure supplement 1C ) . In the extracellular medium from HEBC-AGR2 organoids , eAGR2 was detected but it was not O-glycosylated ( Figure 5-—figure supplement 1C ) . In our experimental system , we found that in contrast to BiP , AGR2 was secreted in the extracellular medium ( Figure 5A–B ) . This questioned the accessibility of the ER-retention motif present on AGR2 , namely the KTEL sequence . Although AGR2 KTEL was described to bind to KDELR1 and to a lesser extent to KDELR2 and 3 ( Alanen et al . , 2007 ) , the secretion of AGR2 might indicate a mechanism specific to this protein . Structural analysis of AGR2 showed that the KTEL motif stemmed out of the core protein and had no interaction with the rest of the protein . The protein backbone is displayed as 'tube' model with the atoms in the KTEL sequence shown in stick format ( Figure 5I ) . Residue interactions within the KTEL motif are seen more clearly: the NH3+ group of K172 forms a very clear salt bridge to the C-terminal carboxylic group of L175 , which yields a stable 'hook' shape ( Figure 5I , bottom panel ) . Thus , it appears that the KTEL motif really stands out of the core protein and is accessible to bind to its receptors . To explore the functional significance of the KTEL motif for the AGR2 secretion , HEK-293T cells which do not secrete AGR2 , were used ( Figure 5J–K ) . HEK-293T cells were transfected with different AGR2 mutant constructs ( Figure 5J ) in which the KTEL motif was mutated to either KDEL , K172D , K172A or a STOP was inserted before the KTEL ( ΔKTEL ) and the secretion of AGR2 protein was assessed by the presence of AGR2 protein in the extracellular medium . The eAGR2 levels in cells expressing the AGR2 mutants , in the extracellular milieu , were at least equal to that observed in cells expressing wild-type AGR2 . Therefore , none of our different KTEL motif mutants displayed a specific intracellular localization and all were secreted . Collectively , our results demonstrate that eAGR2 protein is secreted in the extracellular medium , independently of its KTEL motif and most likely as a soluble protein ( eAGR2 ) . AGR2 secretion results from high iAGR2 expression levels and is not a consequence of cell death or 'leak' of ER proteins into the extracellular medium . Several secreted PDIs were shown to interact with ECM proteins ( Dihazi et al . , 2013; Ilani et al . , 2013 ) . Therefore , we hypothesized , in a similar manner , that eAGR2 could interact with the ECM . To investigate the interaction between ECM and eAGR2 , HBEC cell-derived ECM was purified from HBEC cells overexpressing AGR2 ( HBEC-AGR2 ) or not ( HBEC-EV ) . Since the synthesis and organization of ECM increases with time ( Rashid et al . , 2012 ) , ECM was isolated after 8 days of culture ( Figure 5L ) . Western blotting revealed the presence of the ECM component laminin in both HBEC-EV and HBEC-AGR2 cell-derived matrices ( Figure 5M ) , however AGR2 was only detected in the ECM derived from HBEC-AGR2 cell ( Figure 5M ) . To further confirm that eAGR2 interacts with the ECM , immunofluorescence analysis of ECMs from HBEC-EV and HBEC-AGR2 cells was performed . ECM from both cell types stained positive for laminin ( Figure 5N ) . As expected , immunofluorescence staining for AGR2 revealed the presence of eAGR2 in the ECM of HBEC-AGR2 cells ( Figure 5N , lower panels ) . These results demonstrate the presence of eAGR2 in the ECM . Given that eAGR2 is secreted by tumor organoids ( Figure 5A ) and that AGR2 has been reported in the serum of cancer patients with an average level of eAGR2 ranging from 0 . 5 to 20 ng/ml ( Chen et al . , 2010 ) , we assessed whether eAGR2 might exhibit specific extracellular functions . To this end , we used different adenocarcinoma cell lines depleted for AGR2 ( A549-Sh-AGR2 , H23-Sh-AGR2 and H1838-Sh-AGR2 ) . Remarkably , the addition of recombinant human AGR2 ( 40 ng/ml ) ( Figure 6—figure supplement 1A–F ) to the medium of AGR2-depleted tumor cell lines reversed the cell growth inhibition induced by AGR2 depletion ( Figure 6A ) . Moreover , the addition of eAGR2 to the ECM of AGR2-depleted organoids restored the formation of tumor organoids ( Figure 5B–C ) . 10 . 7554/eLife . 13887 . 010Figure 6 . Extracellular AGR2 boosts the organoid-initiating frequency . ( A ) Quantification of the numbers of cells per wells following the addition of AGR2 in the extracellular medium ( +eAGR2 ) . Data are represented as mean ± SEM of at least 3 independent experiments . p<0 . 05 . ( B ) Representative brightfield images of tumor organoids grown in 3D from cells harboring Sh-AGR2 in presence ( +eAGR2 ) or not ( -eAGR2 ) of AGR2 in the ECM ( n = 3 ) . ( C ) The bar graphs show the mean of organoids per well ( mean ± SEM . ) , after 10 days in presence ( +eAGR2 ) or not ( -eAGR2 ) of AGR2 in the ECM , ( n = 3 ) . The P values are relative to untreated cells . p≤0 . 01 , p≤0 . 001 and *p≤0 . 0001 . ( D ) The bar graphs show the number of organoids per well in absence ( -eAGR2 ) of AGR2 as compared to cells in presence either of AGR2-AXXA mutant ( +eAGR2-AXXA ) or AGR2 wt ( +eAGR2 ) . ( E ) Schematic representation of AGR2 showing the distribution of AGR2’s principal domains and its mutants AXXA and SXXS . ( F ) The bar graphs show the number of organoids per well in absence ( -eAGR2 ) of AGR2 as compared to cells in presence either of AGR2-AXXA mutant ( +eAGR2-AXXA ) or AGR2 wt ( +eAGR2 ) . ( G ) The bar graphs show the number of organoids per well in the absence ( -eAGR2 ) of AGR2 as compared to cells in presence either of AGR2-SXXS mutant ( +eAGR2-SXXS ) or AGR2 wt ( + eAGR2 ) . ( H ) Quantitation of cell death in H23-Sh-AGR2 depleted cells ( Sh-AGR2 ) in absence ( -DTT ) or in the presence ( +DTT ) of DTT . Results are representative of three independent experiments . The cell death rate was determined by the percentage of caspase-3 positive cells . n . s: not significant . ( I ) Growth of cells in H23-Sh-AGR2 depleted cells ( Sh-AGR2 ) in absence ( -DTT ) or in presence of ( +DTT ) DTT ( three independent experiments ) . Data are mean ± SEM . p<0 . 05 . ( J ) The bar graphs show the number of H23-Sh-AGR2 depleted organoids per well in absence ( - eAGR2 ) of AGR2 as compared to cells in presence either AGR2 wt ( +eAGR2 ) or DTT ( +DTT ) . Data are mean ± SEM . p<0 . 05 and p≤0 . 001 . ( K ) Representative images of the organoids formed by HBEC in the presence ( +eAGR2 ) or absence ( -eAGR2 ) of AGR2 in the ECM . The bar graph shows the mean of organoids per well ( mean ± SEM . , n = 3 ) . The p values are relative to untreated cells . p≤0 . 001 . ( L ) Representative images of the organoids formed by HBEC in the presence of conditioned medium from HBEC-vector cells ( EV_CM ) or in the presence of conditioned medium from HBEC-AGR2 cells ( AGR2_CM ) in the ECM . The bar graph shows the mean of organoids per well ( mean ± SEM . , n = 3 ) . The p values are relative to untreated cells . *p≤0 . 001 . ( M ) Representative images of the organoids formed by HBEC overexpressing AGR2-AXXA ( HBEC-AGR2-AXXA ) or not ( HBEC-EV ) . The bar graph shows the mean of organoids per well ( mean ± SEM . , n = 3 ) . p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 01010 . 7554/eLife . 13887 . 011Figure 6—figure supplement 1 . AGR2 and AGR2 thioredoxin mutant ( AGR2-AXXA ) purification and activity on organoids formation . ( A ) Coomassie stained SDS-PAGE gel showing GST induction with IPTG ( left panels ) and purified GST control proteins ( right panel ) . Asterisk ( * ) indicate GST product . ( B ) Coomassie stained SDS-PAGE gel showing GST-AGR2 induction with IPTG ( left panels ) and purified AGR2-GST proteins ( right panel ) . Asterisk ( ) indicate AGR2 fused to GST product . ( C ) Coomassie stained SDS-PAGE gel showing GST-AGR2 cleaved with thrombin ( left panels ) and purified AGR2 was confirmed using an anti-AGR2 or anti-GST antibody . ( D ) The concentration of the purified AGR2 was obtained from a Bovine serum albumin ( BSA ) standard curve . ( E ) Representative images of the organoids formed following different dose of AGR2 ( + ) as compared to untreated cells ( - ) . ( F ) The bar graph shows the mean of organoids per well ( mean ± SEM ) after 10 days of culture from three independent experiments . ( G ) Coomassie stained SDS-PAGE gel showing AGR2-AXXA-GST induction with IPTG ( left panels ) and AGR2-AXXA-GST cleaved with thrombin ( right panel ) . Asterisks ( ) indicate AGR2-AXXA-GST product . ( H ) Coomassie stained SDS-PAGE gel showing AGR2-AXXA-GST cleaved with thrombin ( left panels ) and purified AGR2-AXXA was confirmed using an anti-AGR2 or anti-GST antibody . ( I ) The concentration of the purified AGR2-AXXA was obtained using a Bovine serum albumin ( BSA ) standard curve . ( J ) The bar graphs show the number of organoids per well in absence ( -eAGR2 ) of eAGR2 as compared to organoids in presence either of AGR2-AXXA mutant ( +eAGR2-AXXA ) or AGR2 wt ( +eAGR2 ) in H1838 Sh-AGR2 depleted cells . Data are mean ± SEM . p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 01110 . 7554/eLife . 13887 . 012Figure 6—figure supplement 2 . Structure model showing the CXXS ( 81–84 ) sequence . ( A ) The CXXS residues ( ball-and-stick model in bottom left and top right ) in the AGR2 crystal structure to relate to the position of these vs the KTEL motif . ( B ) Zoom in of CXXS motif along with residues 78–80 ( stick model ) . S84 forms hydrogen bonded interaction with the carboxylic acid of D79 side chain and backbone carbonyl of L78 . ( C ) Y124 ( stick model ) from an adjacent loop interacts with H83 , which in turn forms a hydrogen bond to C81; providing a good interaction network . ( D ) AXXA mutant . In silico mutation of C81 and S84 into Alanine illustrates the loss of interactions as compared to that seen in ( B ) between S84 and L78/D79 , and in ( C ) between Y127-H83-C81 . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 012 We next tested whether the extracellular role of eAGR2 requires the presence of the KTEL motif ( Figure 6D ) . Indeed , the carboxyl-terminal KTEL domain is required for intracellular AGR2 enzymatic function ( Gupta et al . , 2012 ) . Remarkably , the addition of the human recombinant AGR2-ΔKTEL ( 40 ng/ml ) to the extracellular medium of AGR2-depleted organoids was still able to restore the formation of tumor organoids ( Figure 6D ) . Therefore , the KTEL motif is not necessary for eAGR2 extracellular activity . Next , we investigated whether eAGR2 could play a direct extracellular role through its thioredoxin-like domain ( CXXS motif ) ( Figure 6E ) . To this end , we purified recombinant human AGR2-AXXA , the inactive form of AGR2 thioredoxin-like domain ( Figure 6—figure supplement 1G–J ) . AGR2-AXXA mutant , in the ECM of AGR2-depleted organoids , did not restore the formation of tumor organoids ( Figure 6F ) . However , the AGR2-AXXA mutant might be improperly folded as displayed in Figure 6—figure supplement 2 . The sequence C81-S84 is a loop connecting a β-sheet strand with an α-helix . The mutation to A could affect the structure , since we thereby loose the key interactions stabilizing the 'folding'/closure of that loop ( Figure 6—figure supplement 2 ) . S84 forms strong hydrogen bonds to both the carboxylic acid of D79 and the backbone carbonyl of L78 ( Figure 6—figure supplement 2 ) . Thus , we used another inactive form of AGR2 thioredoxin-like domain , the AGR2-SXXS mutant . In contrast to the AGR2-AXXA mutant , AGR2-SXXS mutant restored the formation of AGR2-depleted tumor organoids ( Figure 6G ) . These results show that the single cysteine residue in the AGR2 thioredoxin-like domain is not essential for the formation of tumor organoids . To further assess the redox function of eAGR2 in such process , we treated tumor organoids depleted in AGR2 ( H23-Sh-AGR2 ) with dithiothreitol ( DTT ) at a concentration of 5µM and no cell death was detected ( Figure 6H ) . The addition of DTT to the medium of AGR2-depleted H23 cells neither reversed the cell growth inhibition induced by AGR2 depletion ( Figure 6I ) nor restored the formation of tumor organoids ( Figure 6J ) . Thus , these results demonstrate that treating cells with reducing agent cannot mimic the effects of eAGR2 . We next investigated the effects of eAGR2 in non-tumorigenic HBEC ( Figure 6K ) . As expected , eAGR2 enhanced organoid growth ten-fold ( Figure 6K ) . To determine whether eAGR2 acts through an autocrine/paracrine mechanism , conditioned media from HBEC cells stably infected with empty vector ( HBEC-EV ) or overexpressing AGR2 ( HBEC-AGR2 ) were added to HBEC organoids ( Figure 6L ) . Conditioned media from control ( HBEC-EV ) did not significantly stimulate the formation of HBEC organoids . In contrast , conditioned media from HBEC-AGR2 stimulates the formation of non-tumorigenic organoids eight-fold ( Figure 6L ) . This demonstrated that eAGR2 secreted by organoids and the bacterially expressed human recombinant AGR2 are both active and have the same role on the organoid growth ( Figure 6K–L ) . Next , to confirm that eAGR2 achieved its functions through an autocrine/paracrine manner , we explored the functional effect of the AGR2-ΔKTEL protein , which is secreted and not retained by the KDEL receptors of the ER ( Gupta et al . , 2012 ) , on non-tumorigenic organoids ( Figure 6M ) . HBECs were infected with either empty vector ( HBEC-EV ) ( Figure 6M ) or AGR2-ΔKTEL containing vector ( HBEC-AGR2-ΔKTEL ) ( Figure 6M ) . AGR2-ΔKTEL overexpression increased the formation of organoids ( Figure 6M ) , thereby confirming that the expression of AGR2 dictates organoid-initiating frequency through an autocrine/paracrine manner . These results demonstrate that eAGR2 extracellular functions were independent of its thioredoxin-like domain and occurred through an autocrine/paracrine manner on organoid-initiating frequency , a phenotype of transformed tumor cells . Confocal microscopy analysis of non-tumorigenic HBEC organoids was then used to further evaluate the impact of eAGR2 on organoid architecture . We showed that eAGR2-induced re-distribution of polarization markers using confocal microscopy in HBEC organoids ( Figure 7A–D ) . As shown in Figure 7A and B ( bottom and right panels ) , the apical marker GM130 was randomly distributed throughout the HBEC organoids in presence of eAGR2 in the ECM , in contrast to its restriction to the apical localization in HBEC organoids in the absence of eAGR2 ( Figure 7A and B ( top and left panels ) . Fluorescence intensity cross-section profile ( Figure 7B ) revealed that the intensity of GM130 ( green ) had two side peaks that referred to the control HBEC organoid shell ( - eAGR2 ) . In presence of eAGR2 , GM130 distributed randomly within the HBEC organoid ( Figure 7B ) . Similarly , control HBEC organoids exhibited a lumen lined by the tight junction marker ZO-1 ( Figure 7C and D; top and left panels ) , whereas ZO-1 was no longer restricted to the lumen compartment upon eAGR2 exposure ( Figure 7C and D; bottom and right panels ) . Thus , eAGR2 in the ECM , resulted in a disruption of apicobasal polarity , of non-tumorigenic organoids . Therefore , we hypothesize that eAGR2 determines the orientation of epithelial polarity and thereby lumen formation . Hence , we next investigated the development of lumen in HBEC organoids , in the absence ( - eAGR2 ) or in the presence ( + eAGR2 ) in the ECM of eAGR2 ( Figure 7E–G ) using apical F-actin ( Figure 7E , F ) . The presence of eAGR2 , at the time of plating , gave rise to organoids unable to form a lumen ( + eAGR2 ) ( Figure 7E–G ) thereby showing that eAGR2 interferes with the formation of hollow organoid . Collectively , these data provide further evidence that eAGR2 , in the ECM , induced loss of cell polarity and lumen formation . 10 . 7554/eLife . 13887 . 013Figure 7 . Extracellular AGR2 disrupts apico-basal polarity and lumen formation . ( A , C , E ) HBEC cultured organoids in the presence ( + eAGR2 ) or absence ( -eAGR2 ) of AGR2 in the ECM , stained for GM130 ( A ) , ZO-1 ( B ) and F-Actin ( C ) . Scale bars , 50 μm . Note the repositioning or fragmentation of Golgi following the presence of eAGR2 . ( B , D , F ) Fluorescence intensity profile along the white line across the organoid in the presence ( + eAGR2 , right panel ) or absence ( -eAGR2 , left panel ) of AGR2 in the ECM , stained with DAPI and GM130 ( B ) , ZO-1 ( D ) and F-Actin ( F ) . ( G ) Quantification of organoids with lumens; n>100 for each condition , ( n=3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 013 Confocal microscopy analysis of non-tumorigenic HBEC organoids was then used to further evaluate the impact of eAGR2 on cell-cell adhesion . We investigate the possibility that eAGR2 disrupts tissue organization by affecting cell-cell adhesion . The HBEC organoids were stained with the following cell-cell adhesion markers E-cadherin ( Figure 8A ) , β-catenin ( Figure 8B ) and laminin-V ( Figure 8C ) . E-cadherin , β-catenin and Laminin-V relocalized from the basal side of the organoids to cell cytoplasm ( Figure 8A–B ) . Thus , eAGR2 in the ECM , resulted in a disruption of cell-cell contact , of non-tumorigenic organoids . 10 . 7554/eLife . 13887 . 014Figure 8 . Extracellular AGR2 disrupts cell-cell adhesion . Confocal cross-sections of HBEC organoids in the presence ( +eAGR2 ) or absence ( -eAGR2 ) of AGR2 in the ECM stained with E-Cadherin ( A ) , β-catenin ( B ) , or Laminin-V ( C ) , and DAPI ( blue ) for nucleus . Scale bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 014 Since AGR2 has been correlated as potentially involved in EMT ( Mizuuchi et al . , 2015; Ma et al . , 2015 ) and that EMT is characterized by a reduction of cell-cell adhesion and loss of apico-basolateral polarity ( Thiery , 2002 ) , we examined whether eAGR2 might regulate EMT in non-tumorigenic HBEC organoids . To this end , HBEC organoids growth with eAGR2 for 10 days or not , RNA was harvested and analysed using RT-qPCR array . The expression of 84 EMT genes was quantified and presented as a Volcano plot ( Figure 9—figure supplement 1 ) . The addition of eAGR2 to the organoids ECM significantly modified EMT transcripts such as Wnt5b ( ~thirteen-fold ) , MMP3 ( ~eleven-fold ) , MAP1B ( ~ten-fold ) , MMP9 ( ~eight-fold ) , ZEB1 ( ~three-fold ) , and VIM ( ~two-fold ) ( Figure 9A ) . Changes were also observed in genes related to ECM and cell adhesion , cell growth and proliferation , differentiation and development , migration and mobility , and cytoskeleton ( Figure 9B ) . These results were also confirmed at the protein level with the induction of Vimentin ( VIM ) , MMP9 and N-Cadherin ( CDH2 ) expression upon treatment with eAGR2 as well as decreased E-Cadherin ( CDH1 ) expression ( Figure 9C ) . The impact of eAGR2 on EMT induction is also reinforced by the relocalization of E-Cadherin and β-catenin from the basal side of the organoids to the cytoplasm ( Figure 8A and B ) . 10 . 7554/eLife . 13887 . 015Figure 9 . Extracellular AGR2 promotes EMT and invasive structures . ( A ) Transcript expression of the EMT transcriptional effectors from the array validated by qRT-PCR . Quantitative PCR levels were normalized to B2M expression . p≤0 . 05 . **p≤0 . 01 . ( B ) A focused EMT qRT-PCR array was utilized to assess EMT-regulated genes modulated by eAGR2 in HBEC cells . Samples for array data were derived from three independent replicates of the biological experiments . ( C ) Down-regulation of E-Cadherin and up-regulation of Vimentin , N-Cadherin and MMP9 protein concentrations in organoids in presence of AGR2 in the ECM ( + eAGR2 ) compared to organoids in absence of AGR2 in the ECM ( - eAGR2 ) , as analyzed by western blot . Calnexin ( CANX ) concentration is shown as loading control . One representative experiment ( n = 3 ) is shown . ( D ) Invasion phenotype induced by eAGR2 on matrigel/collagen mixture . Organoids are fixed and stained with anti-Laminin 5 antibody . Arrows show loss of laminin at the basement membrane on invasive structures . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 01510 . 7554/eLife . 13887 . 016Figure 9—figure supplement 1 . A volcano plot representation of differentially expressed gene in comparison of cultures growing in absence or presence of eAGR2 . The log Fold Change of the gene expression levels observed in the two conditions is reported on the X axis , while the –log 10 of the p-value of the statistical test of differential expression ( t-test ) is shown on the Y axis . The horizontal black dotted line represents the threshold of statistical significance of 0 . 05 for the p-value ( 1 . 3 for the -log10 p-value ) . Genes having Log Fold Change below that threshold are considered not statistically significant ( gray dots ) . Genes with p-value equal or greater than the threshold are considered biologically down-regulated ( green dots ) , up-regulated ( red dots ) or not biologically significant ( orange dots ) if their Fold Change is respectively less than or equal to 0 . 5 , greater than or equal to +2 and between 0 . 5 and 2 ( Log Fold Change <= -1 , >= +1 and between -1 and +1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13887 . 016 Matrix MetalloProteinases ( MMP ) have been previously reported to play a critical role in stimulating invasion ( Kessenbrock et al . , 2010 ) . Thus , the induction of MMP3 and 9 by eAGR2 let us to hypothesize that eAGR2 promotes invasion . Since invasion of mesenchymal cells is enhanced by stiff ECM ( Paszek and Weaver , 2004 ) , we then assessed the impact of eAGR2 on the 3D-invasive potential in stiff ECM . These experiments showed that eAGR2 disrupted laminin layers to invasive structures ( Figure 9D ) . Hence , we correlate a cell invasion phenotype with the presence of eAGR2 in the ECM of non-tumorigenic organoids . The present study reveals an important role for eAGR2 in mediating loss of cellular polarity and morphologic transformation of epithelial cells . To our knowledge , this is the first time eAGR2 is characterized in the epithelial tissue morphogenesis and tumorigenesis , and this information will increase our understanding of the biological and physiological properties of AGR2 in this field . Studies in the last decade demonstrated that despite the integrity of its KTEL signal ( ER retention motif ) ( Gupta et al . , 2012 ) , the ER-resident AGR2 is secreted in various cancers ( Chevet et al . , 2013 ) . AGR2 is retained in the ER through a mechanism involving its KTEL motif and the three KDEL receptors ( Gupta et al . , 2012; Alanen et al . , 2007 ) . To further support this aspect , we showed that AGR2 KTEL motif is accessible to bind to the KDEL receptors . Furthermore , we demonstrated that eAGR2 secretion is independent of the KTEL sequence and , is not a consequence of cell death or “leaky” release of ER proteins into the extracellular medium . Moreover , the secretory pathway was not dramatically altered upon exposure to eAGR2 . Hence , our data show that the production of AGR2 is linked to the amount of secreted eAGR2 . Recent studies have suggested that different affinities for the KDEL receptors may impact on subcellular distribution , the latter being dynamic and dependent on conditions such as cell density or growth media ( Alanen et al . , 2011 ) . This could explain how high levels of iAGR2 expression in cancer cells combined with higher cell density and/or altered growth media could impact on the amount and rate of eAGR2 secretion . We observed that eAGR2 , which is undetectable in the ECM of non-tumor organoid , is abundantly secreted by tumor organoids and is present in the ECM . Hence , these data suggest an important role for eAGR2 in controlling ECM functionality , and most likely in the remodeling of tumor microenvironment . A similar role was described for PDIA3 in remodeling of the ECM during renal fibrosis ( Dihazi et al . , 2013 ) . Moreover , high levels of eAGR2 could directly contribute to the tumorigenecity of tumor cells . Indeed , AGR2 depletion in cancer cells significantly decreased tumor cell proliferation , suggesting that high levels AGR2 in tumor cells is important for sustaining tumor cell growth . This is consistent with the observation that conditioned media from cells silenced for AGR2 have a reduced ability to stimulate proliferation of pancreatic cancer cells ( Arumugam et al . , 2008 ) . Furthermore , our data show that restoration of eAGR2 in the microenvironment of AGR2-knockdown tumor cells reversed AGR2 depletion-induced cell growth inhibition and thus indicate that the reduced cell proliferation induced by AGR2 depletion is most likely due to the reduction of eAGR2 in the tumor microenvironment . To further document the biological function of eAGR2 , we have used different human adenocarcinoma organoids genetically depleted for AGR2 and showed the decreased ability of cancer cells to form tumor organoids . This is independent of iAGR2 since the reduced tumor organoids formation in AGR2-depleted cells can be restored by addition of exogenous eAGR2 . It has been reported that the carboxyl-terminal KTEL domain is specifically required for intracellular AGR2 enzymatic function ( Gupta et al . , 2012 ) . Our data reveal that for the eAGR2 extracellular activity , the KTEL motif is not required . Remarkably , , tumor organoid formation was restored when AGR2-depleted tumor organoids were supplemented with AGR2-SXXS ( thioredoxin inactive ) , but not in the presence of reducing agent This result suggests a protein-protein interaction between eAGR2 and ECM proteins and/or proteins involved in the control of cell adhesion and cell-matrix interaction in an AGR2 thioredoxin-like domain independent fashion . The interaction of AGR2 with several proteins involved in such processes has been described in the literature . Indeed , AGR2 was found to interact with DAG1 ( Fletcher et al . , 2003 ) , LYPD3 ( Alanen et al . , 2007 ) , C4 . 4A ( Arumugam et al . , 2015 ) , however the exact mechanisms by which AGR2 interacts with those proteins remains unclear and must be further investigated . Nevertheless , our data suggest that this occurs in a thioredoxin-like domain independent fashion . Therefore , our data demonstrate that eAGR2 drives tumor organoids formation , not by changing the redox nature of microenvironment but through its presence within the ECM . However , the precise mechanisms remain to be further investigated . Our experiments provide the first evidence that eAGR2 has a crucial biological function in the tumorigenesis of cancer cells . Therefore , we hypothesized that eAGR2 could be a pro-oncogenic protein secreted in the ECM . To confirm that eAGR2 is indeed , an ECM microenvironmental pro-oncogenic protein , we investigated the effects of overexpressing AGR2 in human bronchial epithelial non-tumor cells . We found that overexpressing AGR2 enhanced cell proliferation and increased the formation of organoids . Furthermore , we demonstrated that the presence of eAGR2 in the ECM was sufficient to impact on cell’s autocrine/paracrine signaling to in turn enhance organoid-initiating frequency . Therefore , we investigated the pro-oncogenic mechanisms of eAGR2 in non-tumor organoids to mimic the conditions under which eAGR2 acts in vivo . In normal conditions , HBECs form polarized and spherical organoids with hollow lumens , however in the presence of eAGR2 in the ECM , these identical cells form large , non-polarized , undifferentiated organoids without lumen , therefore mimicking tumor organoids . Disruption of cell polarity and tissue organization is reported to be an important event for the initiation and progression of tumorigenesis . Hence , eAGR2 alone is sufficient to confer a tumorigenic phenotype in non-tumor organoids . We next examined the impact of eAGR2 on cell–cell adhesion and the development of invasive structures . We demonstrated that eAGR2: i ) promotes a cell invasion phenotype characterized by the disruption of basal laminin and , ii ) disrupts cell-cell contact characterized by the loss of E-cadherin , β-catenin and Laminin-V at the cell membrane . Adherens junctions represent a powerful invasion suppressor complex in normal epithelial cells ( Pećina-Slaus , 2003 ) . Disruption of polarized tissue structure and induction of cellular invasion are accompanied by extensive remodeling of the cellular microenvironment ( e . g . basement membrane degradation , disruption of cell-ECM interactions ) . In such context MMPs play critical roles as they stimulate tumorigenesis , cancer cell invasion and metastasis ( Stallings-Mann et al . , 2012; Arumugam et al . , 2008 ) . This is consistent with induction of MMP3 and MMP9 observed in eAGR2-treated non-tumor organoids . MMP9 may function to activate local growth factors , stimulate angiogenesis in vivo , and degrade the ECM during cell invasion , whereas MMP3 is critical for the cleavage of E-cadherin and is involved in promoting EMT and invasion in epithelial cells ( Xian et al . , 2005; Fingleton et al . , 2001; Lochter et al . , 1997 ) . This is a mechanism through which eAGR2 may act to promote cell invasion . Moreover , eAGR2 triggers the induction of Vimentin and N-Cadherin . Expression of these mesenchymal markers , together with the loss of E-cadherin and β-catenin from cell–cell contacts , as well as the invasive behavior of the organoids suggests that the non-tumorigenic epithelial cells were converted from an epithelial to a mesenchymal phenotype in presence of eAGR2 in their ECM . We also observed that eAGR2 induced the expression of other EMT-related genes . EMT is a hallmark of tumor progression associated with the acquisition of invasive and metastatic features ( Hanahan and Weinberg , 2011 ) . Although it has been only correlated that AGR2 may stimulate EMT , it is intriguing to find that eAGR2 in ECM is sufficient by itself , to induce EMT in non-tumor epithelial cells . Together , our results indicate that eAGR2 is important for epithelial morphogenesis disruption and metastatic dissemination . Thus , these results provide new insight into the possible mechanisms through which AGR2 may function as a microenvironmental pro-oncogenic protein . In conclusion , the findings from the present study demonstrate that eAGR2 is a microenvironmental regulator of epithelial tissue architecture , which has a role in the preneoplastic phenotype and thus , contributes to tumorigenicity in epithelial cells . Indeed , eAGR2 present in the ECM of non-tumorigenic organoids is sufficient , by itself , to disrupt cell polarization and organoid formation , thus leading to a disorganized phenotype with invasive properties . In epithelial cells , dynamic remodeling of the ECM is essential for the development and normal tissue homeostasis . However , loss of polarity and uncontrolled ECM remodeling leads to cancer . Together , our results highlight that eAGR2 represents a novel and unexpected microenvironmental signaling in the epithelial morphogenesis and tumorigenesis which might be explored for development of novel therapeutic strategies for epithelial cancer . Human normal bronchial epithelial cells ( HBEC ) and lung cancer cells have been previously described ( Fessart et al . , 2013 ) . HBEC ( Human Bronchial Epithelial Cells ) cells were purchased from Lonza . A549 , H1838 , H23 and HEK-293 ( CRL-1573 ) cells were purchased from American Type Culture Collection . All cell lines used for this study were tested free of mycoplasma . All cancer cells were cultured in DMEM supplemented with 10% FBS . ShRNA against AGR2 ( MSCV-LTRmiR30-PIG retroviral vector [Arumugam et al . , 2008] ) and the corresponding scramble control ( Sh-ctl ) were a gift from Pr . Lowe AW ( Stanford University , California , USA ) . The MSCV-LTRmiR30-PIG retroviral plasmid was used to produce retroviral particles in the Phoenix Ampho HEK293T cell line . After transduction into A549 , H1838 or H23 cells , selection was performed for 3 days with puromycin ( 2 μg/ml for each cell line ) . AGR2 concentrations were analyzed quantitatively by western blot and qualitatively by immunofluorescence . For AGR2 overexpression experiments , HBECs were infected with an empty vector pLenti6 . 3/V5 or a pLenti6 . 3/V5 vector expressing the coding sequence of human AGR2 . Selection was performed for 3 days with blasticidin ( 5 μg/ml ) . For AGR2 and AGR2-AXXA recombinant , AGR2 from pLenti6 . 3/V5 vector was sub-cloned into pET-60-DEST vector . Each construct was verified by DNA sequence analysis . AGR2 mutant cDNAs ( C81S , K172D , K172A , K172Stop , T173D named respectively SXXS , DTEL , ATEL , DeltaKTEL and KDEL ) ( Figure 5J ) were synthesized and cloned in pGEX-2TK for GST-AGR2 production and pcDNA4 . 1 for mammalian expression . We used commercial matrix ( Matrigel ) for the organotypic assays . The sources of the primary antibodies used in these studies were as follows: β-catenin ( BD Biosciences ) , cleaved-caspase 3 ( Asp175 ) ( Cell Signaling ) , Calnexin ( Stressgen ) , GM130 ( BD Transduction Laboratories ) , laminin V ( Millipore ) , mouse AGR2 ( Abnova , clone 1C3 ) , rabbit polyclonal AGR2 antibody ( Abcam ) , BiP antibody ( Abcam ) , GAPDH antibody ( Millipore ) , Transferrin antibody ( Cell signalling ) , and O-linked-N-acetylglucosamine antibody ( Abcam ) . The secondary antibodies used were as follows: Alexa-Fluor-488- and Alexa-Fluor-546-conjugated anti-mouse and rabbit IgGs ( Molecular Probes/Invitrogen ) . Reagents used in the study were diaminophenylindole ( DAPI ) ( Sigma-Aldrich , St Louis , MO ) , paraformaldehyde ( Sigma-Aldrich ) , DTT ( Sigma-Aldrich ) and Cycloheximide ( Sigma-Aldrich ) . Anchorage-independent growth was evaluated as described ( Lleonart et al . , 2010 ) . Briefly , 1 × 105 cells were plated in DMEM with 3 . 3% FBS containing 0 . 35% soft agar in 6-cm plates over a layer of solidified DMEM containing 0 . 7% soft agar . Medium was added twice a week to maintain humidity . After 4 weeks , colonies were stained with Crystal Violet ( 0 . 05% ) for 10 min and counted . A549-Sh-ctl ( 1 × 106 ) , A549-Sh-AGR2 ( 1 × 106 ) or A549-Sh-ctl cells ( 1 × 106 ) and A549-Sh-AGR2 ( 1 × 106 ) were injected into the tail of male RAG2 -/- gamma mice . Mice were followed weekly , and when tumors in any of the groups ( mice injected with Sh-ctl or Sh-AGR2 cells ) reached around 1 cm3 , determined by palpation ( at 2 weeks ) , all mice were killed , and their lungs were dissected and processed . A total of 44 mice ( 12 for each cell lines , A549 and H1838 ) were used for each condition ( Sh-ctl and Sh-AGR2 ) . All animal procedures met the European Community Directive guidelines ( Agreement B33-522-2/ N°DIR 1322 ) and were approved by the ethical committee . Samples of human lung cancer tissues were obtained from the Haut-Levêque University Hospital ( Bordeaux , France ) and reviewed by expert pathologist in the field ( H . Begueret ) . The number of human lung samples analyzed and their clinicopathological characteristics are described in Supplementary file 1A . These procedures were approved by the Institutional Review Board at Haut-Levêque University Hospital ( NFS96900 Certification ) . Immediately after the surgical resection , tumor specimens were fixed for 24 hr in 10% buffered formaldehyde , dehydrated and routinely embedded in paraffin . Immunohistochemical analyses were performed using 4 µm sections . All staining procedures were performed in an automated immunostainer ( Bond-III , Leica Biosystems Newcastle Ltd , Newcastle-Upon-Tyne , U . K ) using standard reagents provided by the manufacturer . Briefly , after Bond Epitope Retrieval solution 2 ( Leica Biosystems Newcastle Ltd , Newcastle-Upon-Tyne , U . K ) antigen retrieval for 20 min , de-paraffinized sections were incubated with the anti-AGR2 ( Abnova , clone 1C3 ) human monoclonal antibody at a 1:100 dilution for 15 min at room temperature . For visualization , the Bond Polymer Refine Detection kit ( Leica Biosystems Newcastle Ltd , Newcastle-Upon-Tyne , U . K ) was used according to the manufacturer’s instructions . Each immunohistochemical run contained an internal positive control ( bronchiolar epithelium or type II pneumocytes ) and an antibody negative control ( buffer , no primary antibody ) . Sections were visualized with a Nikon-Eclipse501 microscope , and images were acquired using NIS-Elements F 3 . 00 System ( Nikon Digital Sight ) . Immunohistochemical staining was scored by intensity using the following criteria: 0 , no staining; 1 , mild staining and 2 , dark staining . Quantification was done using ImageJ software and analyzing 20 images ( 20X ) per tumor . Tumors of at least six mice were used for quantification . The number of tumor foci was calculated from whole lung sections with a cut-off of >0 . 32 mm for tumor nodule diameter . Cells cultured on cover slips were fixed using 4% PFA in PBS for 15 min and then permeabilized with 0 . 25% Triton X-100 . After blocking with 1% BSA and 1% FBS in PBS , cells were incubated with the primary antibody for 1 hr at RT . Cells were then washed and incubated with secondary antibodies ( Life Technologies ) for 1 hr at RT . DAPI staining was used to visualize nucleus . For three-dimensional culture ( 3D ) organoids , cells were grown in laminin-rich basement membrane growth factor reduced Matrigel ( BDBiosciences ) ( Matrigel ) as we previously described ( Fessart et al . , 2013 ) . Confocal analysis was performed using Nikon confocal imaging system . Images were generated and converted to Tiff format . Cells were lysed in buffer A ( 0 . 05 M Tris-HCl , 0 . 15 M NaCl , 1% Triton X-100 , pH 7 . 2 ) containing protease and phosphatase inhibitors . After centrifugation , proteins in the supernatant were quantified , boiled with Laemmli buffer , resolved by SDS-PAGE and transferred to a nitrocellulose membrane . WB was performed as described previously ( Alanen et al . , 2007 ) . For conditioned medium ( CM ) , cells were grown until confluence , washed and grown in cell culture medium without FBS for 72 hr . CM was collected and concentrated using 30 kDa Centricon filters ( Millipore ) . Protein amounts were assessed by western blot . Mouse AGR2 monoclonal antibody ( Abnova , clone 1C3 ) and rabbit polyclonal Calnexin were used as primary antibodies as described previously ( Alanen et al . , 2007 ) and peroxidase-conjugated anti-rabbit and anti-mouse Ig ( Amersham ) as secondary antibodies . HBEC cells were seeded in 3D matrigel in the presence of conditioned medium from either HBEC-vector or HBEC-AGR2 cells grown in 2D . Conditioned medium was collected , filtered and centrifuged at 2800 g for 5 min . The medium was changed every forty-eight hours and the cells were incubated for an additional 48 hr . HBEC cells ( placed in 3D matrigel ) were treated with the conditioned media collected from HBEC-vector cells or HBEC-AGR2 , diluted 1:3 with fresh medium . After 48 hr , the medium was changed to fresh diluted 1:3 with the conditioned medium . Ten days later , cell growth in 3D was performed as described previously . HEK293T cells were cultured in 6-well plates for 24 hr , washed in OptiMEM ( Thermo scientific ) and incubated with 500 μL OptiMEM for transfection with wild-type or AGR2 mutants using Fugene 6 ( Promega ) according to the manufacturer’s recommendations . After 24 hr , supernatants were collected , centrifuged at 4500 g for 5 min . Transfected cells were then lysed using 30 mM Tris-HCl pH7 . 5 , 150 mM NaCl and 1% TX100 . Supernatants and cell lysates were spotted on nitrocellulose membranes using the Bio-dot mannifold ( Bio-Rad France ) . AGR2 was detected as described above . Dot intensity was quantified using the ImageJ software ( imagej . nih . gov/ij ) and results were presented as normalized secreted AGR2 . Subconfluent H23 cells were treated with cycloheximide ( 50 μg/ml ) treatment and lysis with lysis buffer ( 50 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , 2 mM EDTA , 50 mM NaF , 0 . 5% NP-40 , 1 mM Na3VO4 , 2 μg/ml aprotinin , 1 μg/ml PMSF , and 1 μg/ml leupeptin ) . An AGR2-specific mouse monoclonal antibody was used for immunoprecipitation , followed by immunoblot analysis using AGR2 antibody . To generate cell-derived matrices , HBECs were plated in 6-well plates and cultured for 8 days , changing the growth medium every two days . To purify ECM , cells were removed using the published protocol ( Rashid et al . , 2012 ) . Briefly , growth medium was aspirated and cells were washed with PBS . Cell-derived matrices were denuded of cells by dissolving the cell layer first with a solution of 0 . 5% ( v/v ) Triton X-100 in PBS and then with 25 mmol/L NH4OH in PBS , for 3 min each , followed by four washes with PBS . This ECM from HBECs in culture remained intact , firmly attached to the entire area of the tissue-culture dish and free of nuclear or cellular debris . Cell-derived matrices were recovered in Laemmli buffer by scraping , resolved by SDS-PAGE and transferred to a nitrocellulose membrane for Western blotting . For immunofluorescence , cells were plated onto coverslips and grown for 8 days before preparing the cell-derived matrices as described above . ECM was fixed with 4% paraformaldehyde and proceeds to immunofluorescence as described previously . All results were evaluated using GraphPad Prism statistical software package . Different statistical tests were used according to the type of data analyzed ( Student's t test , log-rank tests ) , as is indicated in figure legends . p≤0 . 05 was considered statistically significant . HBECs in presence ( + eAGR2 ) or in absence ( - eAGR2 ) of AGR2 were cultured on 3D Matrigel for 8 days and their total RNA was extracted using the RNeasy Mini RNA Isolation Kit ( Qiagen ) . One microgram of total RNA was reverse-transcribed to cDNA by the RT2 First Strand Kit ( Qiagen ) for each plate . Real-time PCR for RT2 Profiler PCR Array was carried out in a 25 μL solution containing cDNA , 2 × RT2 SYBR Green Mastermix , cDNA synthesis reaction and RNase-free water on an EMT Signaling Pathway RT2 Profiler PCR Array Plate ( Qiagen ) in a One-Step Plus Real-time PCR system ( Applied Biosystems ) . Real-time RT-PCR profile consists of 10 min of initial activation at 95°C followed by 40 cycles of 15-s denaturation at 95°C , and 1-min annealing and extension at 60°C and an association stage including 15-s at 95°C , 1-min at 60°C and 15-s at 95°C . The relative gene expression data was analyzed by RT2 Profiler PCR Array data analysis v3 . 5 ( Qiagen ) . Each array per plate was performed in triplicate , and contained 89 genes including 5 housekeeping genes and 84 genes involved in EMT signaling pathways . The details of these genes are listed in Supplementary file 1B . Recombinant GST fusion AGR2 proteins were expressed in Escherichia coli ( DH5α ) and purified as previously described ( Alanen et al . , 2007; Delom and Chevet , 2006 ) . Following purification , recombinant proteins were analyzed by SDS-PAGE and quantified by spectrophotometry .	Cancer cells multiply abnormally fast and therefore produce protein molecules faster than normal cells . To avoid becoming stressed by this overproduction , cancer cells make use of proteins that fold the new proteins inside the cell . One of these protein folders is called anterior gradient-2 ( or AGR2 for short ) and is produced at high levels in so-called epithelial cancers , such as breast and lung cancer . Previous research has shown that AGR2 inside cancer cells can help them grow and survive and AGR2 can also be found outside cells , such as in the blood or the urine of cancer patients . Therefore some researchers have suggested that measuring the levels of AGR2 in bodily fluids may be a useful marker for detecting cancers . Fessart et al . hypothesized that – apart from becoming a promising diagnostic tool – the AGR2 protein itself , specifically when found outside cells , might make cancer cells more aggressive . Fessart et al . used a range of techniques to test this hypothesis . For example , healthy lung cells and lung cancer cells were grown into miniature replicas of lung organs in the laboratory , and in a key experiment , AGR2 was added to the lung organoids grown from the healthy cells . The addition of AGR2 protein was enough to change the non-tumor organoids into tumor organoids and boosted their growth about ten-fold . Further experiments then revealed that AGR2 also makes cells more invasive and capable of moving , both important features of aggressive cancer cells . Overall , Fessart et al . have proven that AGR2 is a signalling molecule found outside cancer cells that makes them more aggressive . In future , more research addressing how AGR2 achieves this may lead to new therapeutic strategies against some forms of cancer .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "cancer", "biology" ]	2016	Secretion of protein disulphide isomerase AGR2 confers tumorigenic properties
Neurofibromatosis type 1 ( NF1 ) is an autosomal dominant disorder whose neurodevelopmental symptoms include impaired executive function , attention , and spatial learning and could be due to perturbed mesolimbic dopaminergic circuitry . However , these circuits have never been directly assayed in vivo . We employed the genetically encoded optical dopamine sensor dLight1 to monitor dopaminergic neurotransmission in the ventral striatum of NF1 mice during motivated behavior . Additionally , we developed novel systemic AAV vectors to facilitate morphological reconstruction of dopaminergic populations in cleared tissue . We found that NF1 mice exhibit reduced spontaneous dopaminergic neurotransmission that was associated with excitation/inhibition imbalance in the ventral tegmental area and abnormal neuronal morphology . NF1 mice also had more robust dopaminergic and behavioral responses to salient visual stimuli , which were independent of learning , and rescued by optogenetic inhibition of non-dopaminergic neurons in the VTA . Overall , these studies provide a first in vivo characterization of dopaminergic circuit function in the context of NF1 and reveal novel pathophysiological mechanisms . Neurofibromatosis type 1 ( NF1 ) is an autosomal dominant disorder of neural crest-derived tissues that affects approximately 1 in 3500 individuals worldwide and is caused by loss of one functional copy of the NF1 gene on chromosome 17 ( Wallace et al . , 1990 ) . Neurofibromin , the protein product of NF1 , inhibits Ras-dependent cellular growth and proliferation ( Basu et al . , 1992 ) and enhances cAMP signaling pathways ( Tong et al . , 2002 ) . The clinical features of NF1 include pigmentary lesions , neoplasia ( e . g . cutaneous and plexiform neurofibromas , optic gliomas , malignant peripheral nerve sheath tumors ) , cognitive and learning disabilities , peripheral neuropathy , musculoskeletal abnormalities , and gross and fine motor delays ( Cimino and Gutmann , 2018; Gutmann et al . , 2012 ) . Cognitive dysfunction is a significant source of lifetime morbidity , as up to 70% of affected individuals experience impaired executive functioning , speech and language delays , attention deficits , hyperactivity , and/or impulsivity ( Hyman et al . , 2005 ) . Furthermore , approximately one third of patients with NF1 meet DSM-V criteria for attention deficit hyperactivity disorder ( ADHD ) ( Hyman et al . , 2005; Miguel et al . , 2015 ) . Despite the societal burden of NF1-associated cognitive sequelae , their etiology has not been fully elucidated . Although homozygous genetic disruption of the Nf1 gene is embryonic lethal in mice ( Silva et al . , 1997 ) , cognitive deficits in NF1 have been successfully modeled in several transgenic and conditional knockout mouse lines ( Silva et al . , 1997; Zhu et al . , 2001; Hegedus et al . , 2007; Cui et al . , 2008; Brown et al . , 2010a; Anastasaki et al . , 2015; Omrani et al . , 2015; Li et al . , 2016; Xie et al . , 2016 ) . Heterozygous knockout mice ( Nf1+/- ) exhibit impaired spatial learning ( Costa et al . , 2001; Silva et al . , 1997 ) , which is Ras/ERK-dependent ( Costa et al . , 2002 ) , rescued by the Ras inhibitor lovastatin ( Li et al . , 2005 ) , and may be due to increased inhibitory GABA tone ( Costa et al . , 2002 ) . Additionally , the neurofibromin C-terminus is a positive regulator of G-protein-stimulated adenylyl cyclase activity ( Hannan et al . , 2006; Tong et al . , 2002 ) , and cAMP deficiency in NF1 knockout models causes altered in vitro neuronal morphology and growth , visual learning deficits , and changes in cortical architecture in mice ( Brown et al . , 2012; Brown et al . , 2010b; Hegedus et al . , 2007; Wolman et al . , 2014 ) . Attenuated dopaminergic neurotransmission in mesolimbic and nigrostriatal circuits are putative mechanisms underlying attentional , learning , and motivational deficits observed in NF1 model mice ( Diggs-Andrews and Gutmann , 2013 ) . Mesolimbic reward circuits involve the convergence of dopaminergic projections from the midbrain ventral tegmental area ( VTA ) with glutamatergic inputs from cortical and subcortical regions on medium spiny neurons in the nucleus accumbens ( NAc ) . These circuits facilitate the translation of relevant internal and external stimuli into motivated behaviors ( Wise , 2005 ) and have been implicated in the pathophysiology of ADHD and other disorders of impulse control ( Li et al . , 2006; Purper-Ouakil et al . , 2011 ) . In the optic glioma mouse model of NF1 ( OPG , a conditional Nf1 knockout in astrocytes on an Nf1+/- background ) , reduced striatal dopamine is associated with motor , exploratory , spatial learning , and attentional abnormalities ( Brown et al . , 2010a; Diggs-Andrews et al . , 2013; Anastasaki et al . , 2015 ) , which are ameliorated by treatment with the catecholamine re-uptake inhibitor methylphenidate or the dopamine precursor L-DOPA ( Brown et al . , 2010a ) . Despite these efforts , dopaminergic neurotransmission has never been investigated in NF1 models in vivo . In order to address this gap in the understanding of NF1 , we utilized the new , ultra-fast , genetically encoded dopamine sensor dLight1 ( Patriarchi et al . , 2018 ) to monitor dopamine dynamics in the lateral nucleus accumbens ( LNAc ) during motivated behavior in 129T2/SvEmsJ::C57Bl/6NTac F1 hybrid Nf1 wildtype ( Nf1+/+ ) and heterozygous knockout ( Nf1+/- ) mice . This hybrid background produces more robust behavioral phenotypes than those on a pure C57Bl/6 background ( Cui et al . , 2008; Li et al . , 2005; Shilyansky et al . , 2010 ) . Novel dopaminergic phenotypes were further parsed with patch clamp electrophysiology and optogenetics . Because previous morphological analysis has largely been restricted to neuronal cultures ( Brown et al . , 2010a; Anastasaki et al . , 2015 ) , we comprehensively characterized dopaminergic neuron structure in situ in Nf1+/+ and Nf1+/- mice using tissue clearing , tracing methods , and the novel systemic AAV-based tool Th-VAST ( catecholaminergic neuron-targeted vector-assisted spectral tracing ) . These efforts revealed distinct dopaminergic phenotypes , identified putative mechanisms governing their expression , and explored how Nf1 haploinsufficiency moderates the motivational salience of relevant environment stimuli . In order to investigate dopamine dynamics in freely behaving Nf1+/+ and Nf1+/- mice , we utilized the genetically encoded , fluorescent dopamine sensor dLight1 . 2 ( Patriarchi et al . , 2018 ) , which allows for sub-micromolar detection of extracellular dopamine concentrations with sub-second resolution and negligible sensitivity to other monoamines , GABA , and glutamate ( Corre et al . , 2018; Patriarchi et al . , 2018 ) . Fluorescent dopamine signals in the LNAc were monitored with fiber photometry ( Gunaydin et al . , 2014 ) ; this terminal field region was chosen because its afferent ventral tegmental dopaminergic inputs exhibit a high diversity of responses to both rewarding and aversive stimuli and stimulus-predictive cues ( de Jong et al . , 2019; Lammel et al . , 2011 ) . To facilitate optical dopamine measurements , an adeno-associated viral vector ( AAV9-hSyn-dLight1 . 2 ) was stereotaxically injected into the LNAc to express dLight1 . 2 in neurons , followed by implantation of a 400 µm optical fiber ( Figure 1A ) for sensor excitation and emitted photon collection via a custom photometry system ( Cho et al . , 2017 ) ( Figure 1B ) . After surgical recovery , we measured baseline differences in spontaneous dopaminergic neurotransmission by monitoring dLight1 . 2 signals ( Figure 1C , Figure 1—figure supplement 1 ) in the LNAc during 5-min epochs in which mice sat in a dark , sound-attenuating chamber . Peak analysis was performed to identify local trace prominences ( Figure 1D ) and revealed that the dopamine transient event rate was reduced in Nf1+/- mice compared to Nf1+/+ littermates ( Figure 1E ) . Baseline ( median ) fluorescence , peak amplitude , and full width at half maximal intensity ( FWHM ) was equivalent between genotypes . Because reduced LNAc dopamine content and afferent terminal TH expression have been observed in OPG mice ( Brown et al . , 2010a; Diggs-Andrews et al . , 2013 ) , we measured monoamine and monoamine metabolite levels in the NAc using high-performance liquid chromatography . We failed to detect differences in dopamine ( DA ) , serotonin ( 5-HT ) , norepinephrine ( NE ) , or their metabolites between genotypes ( Figure 1—figure supplement 2 ) . Additionally , there was no difference in dopaminergic terminal tyrosine hydroxylase expression across striatal sub-compartments ( Figure 1—figure supplement 2 ) . These findings suggest that basal differences in dLight1 . 2 event rate are not due to changes in dopaminergic terminal density or dopamine synthetic capacity . In order to further parse differences in spontaneous dopaminergic transient activity , we performed whole-cell patch clamp electrophysiological recordings in acute midbrain slices that contained the lateral ventral tegmental area ( Figure 1F ) , which is the main source of dopaminergic projections to the LNAc ( Lammel et al . , 2011 ) . Because the dependence of Nf1+/- phenotypes on genetic background precludes crossing with cell-type-specific reporter or Cre recombinase lines , we used a blood-brain barrier penetrant , systemic adeno-associated viral vector ( AAV-PHP . eB ) ( Chan et al . , 2017 ) containing a green fluorescent protein ( GFP ) transgene under control of the rat tyrosine hydroxylase promoter ( Oh et al . , 2009 ) ( AAV-PHP . eB-Th-GFP; 1 × 1011 viral genomes/mouse r . o . ; Figure 1F , right ) to label dopaminergic neurons . This allowed for visual identification during patch clamp experiments . GFP-positive cells were considered to be dopaminergic if their action potential duration was >1 ms , a previously validated threshold to distinguish dopaminergic from GABAergic neurons in the VTA ( Chieng et al . , 2011 ) . We found that putative dopaminergic neurons in Nf1+/- midbrain slices exhibited lower spontaneous whole-cell firing rates ( Figure 1G ) and required more rheobase current to elicit a spike when compared to Nf1+/+ neurons ( Figure 1H ) . This finding supports the hypothesis that phenotypic differences in baseline dLight1 . 2 event metrics are activity-dependent . Action potential threshold , duration , amplitude , and after hyperpolarization magnitude did not differ between genotypes ( Figure 1I–J , Table 1 ) . During whole-cell recordings , we also observed that Nf1+/- putative dopaminergic neurons exhibit increased input resistance ( Rm ) and decreased membrane capacitance ( Cm ) compared to Nf1+/+ littermates ( Figure 2A ) without a change in other membrane properties ( Table 2 ) . This finding was robust across experiments ( Table 2 ) . Because increased Rm could be indicative of reduced soma volume ( Torres-Torrelo et al . , 2014 ) , we manually traced over two thousand TH-positive dopaminergic somata in the VTA per genotype ( Figure 2B ) . We found that cross-sectional area , major axis length , and minor axis length were reduced in Nf1+/- mice ( Figure 2C–D , Figure 2—figure supplement 1 ) . Proportionality was maintained , however , as the soma aspect ratio was equivalent between genotypes ( Figure 2—figure supplement 1 ) . TH immunofluorescence and total neuron counts in the VTA did not differ between Nf1+/- and Nf1+/+ dopaminergic neurons ( Figure 2D ) . No phenotypic differences were observed in the adjacent substantia nigra pars compacta ( Figure 2—figure supplement 1 ) . These findings indicate that relative differences in soma size were VTA-specific and could have contributed to changes in passive membrane properties . Dendritic complexity also contributes to cell input resistance ( Bekkers and Hausser , 2007; Šišková et al . , 2014 ) , so we modified the two-component , systemic AAV-based method VAST ( Vector-Assisted Spectral Tracing ) ( Chan et al . , 2017 ) to create Th-VAST . This tool facilitates anatomical reconstruction of dendritic arbors by providing recombinase-independent , sparse , multicolor labeling of catecholaminergic neurons . VAST achieves hue diversity via stochastic expression of three tetracycline response element ( TRE ) -regulated fluorescent proteins ( XFPs; mRuby2 , mNeonGreen , and mTurquoise2 ) following systemic delivery with AAV-PHP . eB . Sparseness is subsequently tuned by titration of a co-delivered , tet-off transactivator ( tTA ) inducer vector ( Chan et al . , 2017 ) . In Th-VAST , tTA expression is targeted to catecholaminergic neurons via use of the Th promoter , and retro-orbital delivery of the XFP cocktail ( AAV-PHP . eB-TRE-XFP; 1 × 1012 vg/mouse total ) and the inducer vector ( AAV-PHP . eB-Th-tTA; 1 × 1011 vg/mouse ) produced dense multicolor labeling of Th neurons in the VTA and SNc ( Figure 2E–F , Figure 2—figure supplement 2 ) . Compared to 1 × 1012 vg/mouse AAV-PHP . eB-Th-GFP ( Chan et al . , 2017 ) , the specificity of Th-VAST vectors was lower in the VTA ( 58 . 7% vs 81% ) and SNc ( 74 . 2% vs 81% ) ( Figure 2—figure supplement 2 ) despite good XFP restriction to these areas . This likely occurred because induction of XFP expression requires very low levels of tTA , and a sub-population of VTA projection neurons have hybrid Th-GABAergic phenotypes ( Root et al . , 2014; Stuber et al . , 2015 ) . As such , spectral tracing was only performed when Th-VAST-labeled neurons were unequivocally tyrosine hydroxylase-positive . Using a lower inducer vector dose ( 6 × 109 vg/mouse ) to provide sparse labeling ( Figure 2F , right ) , we repeated Th-VAST in Nf1+/- and Nf1+/+ mice . Following two weeks of expression , we prepared and optically cleared ( using RIMS ) ( Yang et al . , 2014 ) 300 µm horizontal VTA sections that had been immunostained for TH to confirm post hoc that Th-VAST-labeled neurons were dopaminergic . After tracing in Imaris ( Figure 2G ) , Sholl analysis was performed to quantify dendritic branching by detecting neurite intersections with concentric 5 μm shells originating from the soma . No difference in dendritic complexity or total neurite length was observed between genotypes ( Figure 2H ) , which suggests that , although Nf1+/- dopaminergic neurons have smaller somata than Nf1+/+ neurons , they have similar neurite morphology . The observation that Nf1+/- dopaminergic neurons have reduced cross-sectional areas but higher rheobase requirement was unexpected , given that smaller neurons tend to be more excitable ( Torres-Torrelo et al . , 2014 ) . In order to parse these differences , we first assayed dopaminergic Ih currents , which contribute to the stability of spontaneous firing rates ( Neuhoff et al . , 2002; Seutin et al . , 2001 ) and are attenuated in hippocampal interneurons in NF1 model mice ( Omrani et al . , 2015 ) . Ih was determined by quantifying the sag current produced by a series of hyperpolarizing voltage steps from −60 mV to −130 mV in voltage clamp ( Figure 3A ) . We found that Nf1+/- dopaminergic neurons had smaller Ih current amplitudes ( Figure 3A , Figure 3—figure supplement 1 ) without a change in voltage dependence ( Figure 3B; determined by tail current analysis ) relative to Nf1+/+ littermates . Differences in Ih current amplitudes were not significant when normalized to the cell capacitance to account for cell size ( Figure 3C , Figure 3—figure supplement 1 ) , and maximum Ih current amplitude was significantly correlated with Cm across all animals and within genotypes ( Figure 3D ) . Since reduced cAMP production , which has been associated with Nf1+/- neuronal phenotypes in vitro ( Brown et al . , 2012 ) , could attenuate the Ih current , we repeated Ih measurements in Nf1+/- slices in the presence of the adenylyl cyclase activator forskolin . Addition of 20 µM forskolin to the bath solution did not significantly affect Ih magnitude or voltage dependence in Nf1+/- putative dopaminergic neurons ( Figure 3—figure supplement 1 ) . Thus , changes in Ih magnitude are likely cAMP-independent , reflective of smaller cell size , and unlikely to be the etiologic cause of reduced cell excitability . Excitation/inhibition imbalance due to increased GABAergic tone is a hypothesized mechanism governing NF1-associated cognitive deficits ( Diggs-Andrews and Gutmann , 2013 ) , and GABAA receptor agonists increase rheobase ( Rojas et al . , 2011 ) , so we investigated excitation/inhibition balance in putative dopaminergic neurons by measuring spontaneous inhibitory ( sIPSC ) and excitatory post-synaptic currents ( sEPSC ) in voltage clamp ( Figure 3E ) . We found that Nf1+/- putative dopaminergic neurons displayed increased sIPSC frequency but not sIPSC amplitude , sEPSC frequency , or sEPSC amplitude compared to Nf1+/+ neurons ( Figure 3F–G ) . Addition of 100 µM picrotoxin ( a highly selective , non-competitive GABAA receptor antagonist ) to the bath solution rescued spontaneous firing rates in Nf1+/- putative dopaminergic neurons ( Figure 3H ) without affecting passive membrane properties ( Table 2 ) . Picrotoxin also reduced rheobase to levels significantly lower than control- and picrotoxin-treated Nf1+/+ cells ( Figure 3I ) , which would be expected at baseline due to differences in soma volume . Given these findings , we next sought to determine if pharmacological inhibition of VTA GABAergic neurons was sufficient to rescue dLight1 . 2 transient rates in Nf1+/-mice . μ-opioid receptor ( MOR ) agonists , such as morphine or DAMGO , robustly increase dopaminergic neuron firing and NAc dopamine release via pre-synaptic inhibition of GABAergic neurotransmission in the VTA ( Badiani et al . , 2011; Di Chiara and Imperato , 1988; Johnson and North , 1992 ) . This model is supported by recent efforts by Lüscher and colleagues that combined NAc dLight1 monitoring and optogenetic manipulation of VTA sub-populations to examine cell-type-specific substrates of heroin ( diacetylmorphine ) reinforcement ( Corre et al . , 2018 ) . We found that pre-treating Nf1+/- mice with the mu opioid receptor agonist morphine sulfate ( 5 mg/kg , s . c . ) raised spontaneous dLight1 . 2 transient event rate but not event magnitude or FWHM relative to saline ( Figure 3—figure supplement 2 ) . This elevated event rate following MOR agonist exposure ( 0 . 54 ± 0 . 03 Hz ) was not statistically different ( unpaired t-test; t26 = 1 . 08 , p=0 . 29 ) from the spontaneous event rate in Nf1+/+ mice ( 0 . 50 ± 0 . 02 Hz ) . Thus , excitation/inhibition imbalance is a mechanism gating Nf1+/- dopaminergic excitability ex vivo , and attenuation of VTA GABAergic neurotransmission normalizes spontaneous LNAc dopaminergic neurotransmission in vivo . After measuring dLight1 . 2 signals at baseline and parsing these differences ex vivo , we next probed dopaminergic responses to salient stimuli . Because dopaminergic circuits respond strongly to rewards and reward-predictive cues ( Schultz et al . , 2015 ) , we monitored LNAc dopamine signals in water-restricted mice during consumption of 5% sucrose . In both Nf1+/- and Nf1+/+ mice , we observed robust LNAc dopamine transients time-locked to reward consumption ( Figure 4A ) that were not significantly different between genotypes ( Figure 4B ) . No difference in the number of rewards consumed during the 30 min session was observed between groups ( Figure 4C ) . We next measured the dLight1 . 2 response to social interaction , which is a positive reinforcer in mice ( Martin and Iceberg , 2015 ) . We observed large transients at the onset of interaction with a novel , sex-matched , juvenile conspecific that was independent of genotype ( Figure 4—figure supplement 1 ) . Nf1+/- and Nf1+/+ littermates also failed to display differences in preference for a novel mouse in a social preference task ( Figure 4—figure supplement 1 ) . These findings suggest that LNAc dopamine and behavioral responses to unconditioned rewards are preserved in the context of Nf1 haploinsufficiency . Dopaminergic populations have been widely studied for their role in reward learning ( Keiflin and Janak , 2015; Schultz et al . , 2015; Wise , 2004 ) , so we optically monitored dopaminergic neurotransmission during a Pavlovian conditioning assay in water-restricted Nf1+/+ and Nf1+/- mice . In this task , a 5% sucrose reward ( the unconditioned stimulus or US ) was delivered seven seconds after the beginning of a ten-second reward-predictive cue or conditioned stimulus ( the CS; a 5 kHz tone with house light illumination ) during ten , twenty-trial sessions . As each mouse learned the cue-reward association , the number of licks during CS presentation increased across sessions ( Figure 4D ) . No differences in the number of licks during the CS , the number of anticipatory licks , or the learning rate ( the slope of the linear fit of CS licks across trials ) were observed between Nf1+/+ and Nf1+/- mice ( Figure 4E , Figure 4—figure supplement 2 ) . In both genotypes , dLight1 . 2 peaks were observed in response to both CS presentation and US consumption ( Figure 4F ) , and in later trials , to the sound of the sucrose delivery pump ( Figure 4—figure supplement 2 ) . Similar to previous studies ( Patriarchi et al . , 2018 ) , dLight1 . 2 responses to the CS and US ( Figure 4G ) were enhanced and diminished , respectively , in mice that successfully learned the cue-reward association ( i . e . the learning rate was correlated with the rate of change of each feature peak across trials; ( Figure 4H , Figure 4—figure supplement 2 ) . Unexpected omission of the US following CS presentation after learning equivalently diminished the dopamine response to reward seeking in both Nf1+/- and Nf1+/+ mice ( Figure 4I ) , which is consistent with the role of dopamine in reward prediction error detection ( Schultz et al . , 2015 ) . Nf1+/- mice displayed enhanced dLight1 . 2 responses to CS presentation on the first day of testing ( Figure 4J , inset ) compared to Nf1+/+ littermates that was largest during the first trial ( Figure 4—figure supplement 2 ) . This phenotypic difference attenuated over subsequent days and re-emerged as the CS response became correlated with performance ( Figure 4J , Figure 4—figure supplement 2 ) . The magnitude of the CS response on the first day of testing did not predict task performance in later sessions ( Session one vs Session 10; Nf1+/+: R2 = 0 . 02 , p=0 . 69; Nf1+/-: R2 = 0 . 0006 , p=0 . 94 ) . Across sessions , a significant main effect of genotype on US magnitude was observed , although US magnitude was only greater in Nf1+/- mice during session 3 ( Figure 4—figure supplement 2 ) . No differences in response to the sucrose delivery pump , a purely auditory CS , were observed between genotypes across sessions ( Figure 4—figure supplement 2 ) . These findings indicate that , although CS responses are larger in Nf1+/- mice , the ability to form cue-reward associations is equivalent between genotypes and coincides with adaptive changes in dopaminergic neurotransmission with learning . Mesolimbic dopaminergic neurons are a heterogeneous population that exhibit diverse response profiles during exposure to aversive stimuli ( Ilango et al . , 2012; de Jong et al . , 2019; Lammel et al . , 2014 ) , so we next recorded dopamine dynamics in a subset of mice undergoing cued fear conditioning . In this 15-trial assay , a 10 s audiovisual CS ( house light and 3 kHz tone ) predicted a 1 s foot shock ( US ) , and the development of freezing during CS presentation was used as a proxy for learning . A previous study ( Silva et al . , 1997 ) failed to detect differences in cued fear conditioning between Nf1+/+ and Nf1+/- mice , so we employed smaller shock ( 0 . 4 mA vs . 0 . 75 mA ) and tone intensities ( 60 dB vs . 85 dB ) , shorter CS ( 10 s vs . 30 s ) and US durations ( 1 vs . 2 s ) , and a repeated trial structure ( 15 trials vs one trial ) in order to avoid a ceiling affect . Over the course of fifteen CS-US pairings , both Nf1+/+ and Nf1+/- mice exhibited trial-by-trial increases in freezing during the first seven trials that subsequently plateaued ( Figure 5A ) . In both genotypes , the expression of freezing in later trials ( average freezing time , trials 8–15 ) was correlated with the acquisition rate ( slope of the linear fit , trials 1–7; Figure 5B ) and the latency to freeze ( trials 8–15; Figure 5—figure supplement 1 ) . Compared to Nf1+/+ littermates , both the acquisition rate and average freezing time was lower in Nf1+/- mice , which could be accounted for by an increased latency to freeze ( Figure 5C ) . Qualitative review of behavioral video recordings revealed that presentation of the CS resulted in several seconds of locomotor stimulation in Nf1+/- mice that delayed freezing ( Video 1 ) , whereas Nf1+/+ mice froze with short latency at CS onset once learning had occurred ( Video 2 ) . During LNAc dLight1 . 2 monitoring , a dopamine transient was observed at the onset of CS presentation that was greatest in trial 1 , attenuated across trials , and was larger in Nf1+/- mice ( Figure 5D–E ) . In Nf1+/- but not Nf1+/+ mice , the magnitude of the CS peak during trial one was negatively correlated with acquisition rate and freezing duration and positively correlated with latency to freeze ( Figure 5F–G , Supplementary file 1 ) . During US ( shock ) delivery , we observed an initial positive dopamine transient followed by a 1 to 2 s negative anti-peak and a subsequent , broader post-US rebound that returned to baseline several seconds later ( Figure 5D , H ) . This waveform mirrors patterns of activity observed during extracellular recordings in the VTA ( Brischoux et al . , 2009 ) and GCaMP monitoring of dopaminergic axons in the LNAc ( de Jong et al . , 2019 ) . During US exposure , Nf1+/- mice had significantly larger initial US peak responses and a smaller integrated post-US rebound compared to Nf1+/+ littermates ( area under the curve; Figure 5H , Figure 5—figure supplement 1 ) . US anti-peak and post-US rebound peak responses were equivalent between genotypes across trials ( Figure 5H , Figure 5—figure supplement 1 ) . In Nf1+/+ mice , the magnitude of the US anti-peak during trial one was negatively correlated with freezing and positively correlated with latency to freeze ( Figure 5I–J , Supplementary file 2 ) . Additionally , the magnitude of the integrated post-US rebound was negatively correlated with the latency to freeze in Nf1+/- mice ( Figure 5I–J , Supplementary file 2 ) . These findings demonstrate that Nf1+/- mice exhibit altered patterns of dopaminergic neurotransmission in response to both shock-predictive cues and shock delivery , which correlated with behavioral responses during the task . Larger dopaminergic responses to the onset of the CS in the first trial was associated with longer latencies to freeze and shorter freezing durations in Nf1+/- mice , while more negative dopamine responses to shock delivery during trial one was predictive of shorter latencies to freeze and longer freezing durations in Nf1+/+ mice . In order to investigate the etiology of the dopaminergic response to the CS , we measured dLight1 . 2 responses to either a 10 s overhead light or a 3 kHz tone ( inter-trial interval: 75–90 s ) randomly presented during a 20-trial session . Both genotypes exhibited robust dopamine transients at the onset of the overhead light stimulus that returned to baseline within 1–2 s ( Figure 6A ) and decremented across trials ( Figure 6—figure supplement 1 ) . In both Nf1+/+ mice and Nf1+/- mice , 60 dB , 3 kHz auditory tone evoked dopamine transients at stimulus onset ( Figure 6B ) that were comparatively smaller than light responses ( Figure 6C ) and non-trial-dependent ( Figure 6—figure supplement 1 ) . Nf1+/- mice exhibited larger responses to light but not tone onset ( Figure 6C ) , which was confirmed when traces were analyzed as ΔF/F rather than z-score ( Figure 6—figure supplement 1 ) . These findings raise the possibility that phenotypic differences in dopaminergic CS responses are driven by reactions to an overhead visual stimulus . In order to investigate if the overhead light affected performance during cued fear conditioning , we performed the assay using a tone-only CS ( 3 kHz tone ) . During the last five trials , tone-only trials were interleaved with trials in which the house light was added to the CS ( trials 11 , 13 , 15 ) . We found that the development of cued freezing in Nf1+/- mice was equivalent to Nf1+/+ littermates across the first ten trials ( Figure 6D ) . Addition of the overhead light stimulus to the CS was sufficient to perturb the expression of freezing in Nf1+/- but not Nf1+/+ mice by increasing the latency to freeze ( Figure 6E , Video 3 ) . Thus , deficits in cued fear conditioning in Nf1+/- mice are reversible , visual stimulus-dependent , and independent of learning . Dopaminergic neurons , in addition to their role in processing rewarding or aversive outcomes , respond to salient alerting signals to modulate attentional orientation and promote appropriate motivated responses ( Bromberg-Martin et al . , 2010; Schultz and Romo , 1990 ) . Because enhanced dopaminergic responses to an overhead light may reflect increased motivational salience of an alerting visual stimulus ( Thompson et al . , 2010 ) , we performed a looming stimulus assay in Nf1+/+ and Nf1+/- mice . During this test , subjects are exposed to an expanding overhead disc designed to mimic predator approach that rapidly promotes escape and/or freezing behaviors in rodents ( Yilmaz and Meister , 2013 ) . In response to the onset of the looming stimulus , both genotypes exhibited similar sub-second reaction times ( Figure 6—figure supplement 2 ) . Compared to Nf1+/+ littermates , Nf1+/- mice were more likely to escape to the shelter at stimulus onset and exhibited shorter latencies to the first freezing episode ( Figure 6F ) . In both groups , freezing latency was not dependent on freezing location ( Figure 6—figure supplement 2 ) . No differences in the length of the first freezing episode or total freezing during the first minute after looming were observed between genotypes , although freezing was dependent on escape location: mice that escaped to the shelter exhibited shorter freezing durations than those in the open field ( Figure 6—figure supplement 2 ) . Recently , it has been shown that flight-to-shelter responses to looming stimuli are mediated by VTA GABAergic neurons and driven by excitatory projections from the ventral superior colliculus ( vSC ) ( Zhou et al . , 2019 ) . This projection innervates both dopaminergic and GABAergic neurons in the VTA ( Prévost-Solié et al . , 2019; Zhou et al . , 2019 ) and additionally regulates orientation during social interaction ( Prévost-Solié et al . , 2019 ) . In order to determine if the vSC can induce dopaminergic neurotransmission in a manner similar to overhead light exposure , we stereotaxically injected an AAV vector ( AAV5-hSyn-ChR2 ( H134R ) -eYFP ) into the vSC to express the light-gated ion channel channelrhodopsin-2 ( ChR2 ) in neurons , followed by implantation of an optical fiber for 473 nm light delivery in C57Bl/6N mice . A photometry fiber was also implanted in the ipsilateral LNAc for simultaneous dLight1 . 2 monitoring during vSC photostimulation . Similar to overhead light exposure , two seconds of vSC optogenetic stimulation ( 5 mW , 20 Hz , 5 ms pulse width ) produced time-locked , short latency dopamine release in the LNAc at stimulus onset that subsequently decayed to baseline ( Figure 6G , Figure 6—figure supplement 3 ) . In early trials , the signal decayed to sub-baseline , negative values . Discontinuation of the laser was associated with a robust dopaminergic rebound . During testing , we also observed that vSC ChR2 activation evoked vigorous escape behavior in the absence of a threatening stimulus ( Video 4 ) , as has been observed following photostimulation of vSC-to-VTA projections ( Zhou et al . , 2019 ) . These findings indicate that the SC , a visually responsive region involved in innate responses to salient visual stimuli ( Ito and Feldheim , 2018 ) , can induce rapid dopamine release in the LNAc that is qualitatively similar to overhead light exposure . Optogenetic inhibition of VTA GABAergic neurons or vSC projections in the VTA is sufficient to suppress looming stimulus responses in mice ( Zhou et al . , 2019 ) . Given that Nf1 heterozygosity is associated with excitation-inhibition imbalance in the VTA , we hypothesized that suppression of inhibitory neurotransmission during looming stimulus presentation would normalize behavioral responses in Nf1+/- mice . To test this hypothesis , we used an intersectional strategy to target GABAergic neurons in the VTA for optogenetic silencing with the light-gated chloride pump halorhodopsin ( NpHR ) during behavior . We co-injected an AAV vector to express Cre recombinase under control of the Th promoter in dopaminergic neurons ( AAV9-Th-PI-Cre-SV40 ) with either a Cre-Off NpHR ( AAV-DJ-Ef1α-DO-eNpHR3 . 0-eYFP , Nf1+/- mice ) or mCherry ( AAV-DJ-Ef1α-DO-mCherry , Nf1+/+ mice ) vector bilaterally into the VTA ( Saunders et al . , 2012 ) , followed by implantation of 300 μm optical fibers for light delivery . Because the mCherry or NpHR transgenes were double-floxed with an open reading frame ( DO ) , they would be inactivated in the presence of Cre but express normally in non-dopaminergic neurons ( Figure 6H ) . Given that ~90% of non-dopaminergic neurons are GABAergic in the VTA ( Morales and Root , 2014; Pignatelli and Bonci , 2015 ) , we reasoned that our approach would provide efficient optogenetic control of inhibitory neurotransmission in this region . Post hoc analysis revealed that only 14 . 7 ± 0 . 01% of NpHR-positive cells in the VTA expressed tyrosine hydroxylase ( Figure 6H , Figure 6—figure supplement 3 ) , indicating successful targeting of VTAnon-Th neurons in Nf1+/- and Nf1+/+ mice . In the absence of optogenetic inhibition , optical patch cable-tethered VTATh-Off-NpHR-eYFP Nf1+/- mice were more likely to escape to the available shelter in response to looming stimulus presentation and exhibited shorter latency to the first freezing episode when compared to VTATh-Off-mCherry Nf1+/+ mice ( Figure 6I ) . Delivery of 532 nm light ( 5 mW , 30 Hz , 20 ms pulse width ) to the VTA during looming stimulus presentation had no effect on VTATh-Off-mCherry Nf1+/+ mice but decreased the percentage of VTATh-Off-NpHR-eYFP Nf1+/- mice that escaped to the shelter to a level ( 33% ) that was similar to Nf1+/+ mice ( 41% ) and VTATh-Off-mCherry Nf1+/+ mice ( 42% ) ( Figure 6F–I; Video 5 ) . This shift in escape location coincided with an increase in the latency to the first freezing episode in VTATh-Off-NpHR-eYFP Nf1+/- mice that was not significantly different from VTATh-Off-mCherry Nf1+/+ mice ( Figure 6I ) . Thus , optogenetic inhibition of VTAnon-Th neurons normalizes looming stimulus responses in Nf1+/- mice and supports a role for VTA GABAergic neurons in the expression of visual stimulus response phenotypes caused by Nf1 haploinsufficiency . Developmental perturbations in mesencephalic dopaminergic circuits have been hypothesized to contribute to neurocognitive symptoms in NF1 , yet their activity has never been directly investigated in vivo . Here we leveraged the novel dopamine sensor dLight1 . 2 to assay mesoaccumbal dopamine dynamics in awake , behaving NF1 model mice and observed that the frequency of spontaneous fluorescent dopaminergic transients was lower in Nf1+/- mice . Using patch clamp electrophysiology , we showed that Nf1+/- dopaminergic neurons are less excitable and have lower spontaneous firing rates ex vivo due to increased GABAergic tone . Pharmacological or optogenetic inhibition of VTA GABAergic neurons rescued spontaneous dopaminergic and behavioral phenotypes , respectively . Given these findings , it is likely that increased tonic inhibition of Nf1+/- dopaminergic neurons reduces basal activity , whereas smaller soma volume acts as a compensatory mechanism to increase excitability during disinhibition . This would serve to facilitate bursting in response to a strong stimulus and maintain dLight1 . 2 transient amplitude , although future studies monitoring dopaminergic neuron activity with single cell resolution will be required to fully parse this hypothesis . Excitation/inhibition imbalance is present in the amygdala ( Molosh et al . , 2014; Repunte-Canonigo et al . , 2015 ) , striatum ( Shilyansky et al . , 2010 ) , and medial prefrontal cortex ( mPFC ) ( Gonçalves et al . , 2017; Shilyansky et al . , 2010 ) of Nf1+/- mice and likely contributes to deficits in working memory , contextual fear conditioning , and social memory ( Cui et al . , 2008; Molosh et al . , 2014; Shilyansky et al . , 2010 ) . Subthreshold doses of picrotoxin ( 0 . 01 mg/kg ) ( Cui et al . , 2008 ) or L-DOPA ( Wozniak et al . , 2013 ) improve cognitive performance of NF1 model mice , emphasizing the therapeutic potential of interventions that modulate mesolimbic dopaminergic circuit function . Although Nf1+/- dopaminergic neurons exhibit decreased soma size , we failed to detect changes in VTA or NAc TH immunofluorescence , as well as NAc monoamine content . This is in contrast with the optic glioma ( OPG ) mouse model of NF1 that has reduced tyrosine hydroxylase expression in the VTA ( Brown et al . , 2012; Diggs-Andrews et al . , 2013 ) and lower TH , dopamine , and phosphorylated DARPP-32 ( dopamine and cAMP-regulated phosphoprotein-32 ) in terminal fields ( Brown et al . , 2010a; Diggs-Andrews et al . , 2013; Anastasaki et al . , 2015 ) . Neurite outgrowth and growth cone areas are decreased in cultured OPG dopaminergic neurons ( Diggs-Andrews et al . , 2013 ) , yet we did not observe changes in neurite morphology in Th-VAST-labeled Nf1+/- neurons relative to Nf1+/+ littermates . Incongruence between mouse models may be due to relative differences in neurofibromin expression . Nf1+/- mice on a pure C57Bl/6 background have higher tissue neurofibromin levels compared to OPG mice , and neurofibromin dose-dependently regulates TH and cellular dopamine production in cultured mouse neurons and neural progenitor cells differentiated from patient-derived induced pluripotent stem cells ( Anastasaki et al . , 2015 ) . While the etiology of reduced soma volumes is unknown , cAMP deficiency influences the morphology of Nf1+/- and OPG neurons in vitro ( Brown et al . , 2012; Diggs-Andrews et al . , 2013 ) . Perturbations in this pathway thus represents a mechanism of interest in future efforts to characterize mesencephalic development in the context of NF1 . We also observed that Nf1+/- mice exhibit more robust dopaminergic responses to an overhead visual light stimulus , which was correlated with delayed freezing in a cued fear conditioning task . These findings are not indicative of an inability to form cue-outcome associations , since Nf1+/- mice performed similarly to Nf1+/+ littermates when conditioning was carried out in the absence of the light stimulus . NAc dopaminergic neurotransmission is necessary for the acquisition and expression of learned fear responses ( Fadok et al . , 2010; Fadok et al . , 2009 ) , so enhanced CS responses would be expected to promote not attenuate conditioned fear . More likely , exaggerated dopaminergic responses to light presentation reflect increased stimulus salience , as overhead light stimuli are aversive in mice ( Thompson et al . , 2010 ) , unexpected salient visual cues promote dopaminergic firing ( Bromberg-Martin et al . , 2010 ) , and Nf1+/- mice had more robust behavioral responses to a looming stimulus . Work by Schultz and colleagues suggests that alerting dopamine signals , unlike reward responses , enable orientation and stimulus investigation ( Schultz , 2010 ) . As such , increased visual cue salience in Nf1+/- mice may have increased freezing latency during fear conditioning by provoking an investigative locomotor response in the absence of an obvious escape location . Because the looming disc is designed to simulate a predatory environmental cue , it has a greater negative valance; in this case , increased stimulus salience would promote flight-to-shelter responses and reduce the latency to the first freezing episode . Thus , bidirectional effects of Nf1 heterozygosity on freezing latency could be consistent with a single phenotypic process and depend on visual stimulus intensity and the ability to escape . Adult and adolescent human subjects with NF1 exhibit visual processing deficits ( Hyman et al . , 2005 ) , although visual cortical areas have sparse , if any , direct connections to VTA dopaminergic neurons ( Watabe-Uchida et al . , 2012 ) . Short latency dopaminergic responses to visual stimuli are likely driven by afferents from the superior colliculus ( SC ) ( Redgrave et al . , 2010 ) , which receives direct input from retinal ganglion cells ( Dhande and Huberman , 2014 ) , responds to looming stimuli ( Zhao et al . , 2014 ) , and evokes firing of both VTA GABAergic and dopaminergic neurons in vivo while enhancing flight-to-shelter responses ( Prévost-Solié et al . , 2019; Zhou et al . , 2019 ) . Because the dopaminergic response to optogenetic vSC stimulation occurred at stimulus onset , quickly attenuated , and was followed by a post-stimulation rebound , it is likely that excitatory vSC-to-VTA projections excite and subsequently suppress dopaminergic outflow via feed-forward inhibition , producing rebound disinhibition at stimulus offset . While the initial dopaminergic peak may serve as a salience signal , flight-to-shelter responses appear to be mediated by VTA GABA neurons ( Zhou et al . , 2019 ) . In this case , increased vSC-to-VTA excitatory drive in Nf1+/- mice would be predicted to enhance LNAc dopamine release at stimulus onset , while subsequently driving signal termination and escape responses via GABAergic neurons . Thus , the role of tectal inputs to the VTA in moderating visual stimulus sensitivity in NF1 mouse models represents a promising focus for future efforts to parse disease symptomatology and may provide new insights into visual processing in patient populations . Experimental subjects were 8–12 week old 129T2/SvEmsJ::C57Bl/6NTac Nf1+/+ and Nf1+/- male and female mice that were generated via the F1 cross of 129T2/SvEmsJ male mice ( the Jackson Laboratory Stock No: 002065 ) and C57Bl/6NTac Nf1+/- female mice ( generous gift of Dr . Alcino Silva , UCLA ) . Animals were group housed ( 3–4 per group ) throughout the duration of the experiment in a vivarium on a 12 hr light/dark cycle ( lights off at 0600 hr , lights on at 1800 hr ) with ad libitum access to food and water . Fluid restricted animals were singly housed , and their water access was limited to 1 . 5 mL/day . These mice were weighed daily and were returned to ad libitum water access if their weight decline was >10% of their pre-restriction weight . Animal husbandry and experimental procedures involving animal subjects were conducted in compliance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and approved by the Institutional Animal Care and Use Committee ( IACUC ) and by the Office of Laboratory Animal Resources at California Institute of Technology under IACUC protocol 1730 . Mice were only excluded from behavioral studies if they could not complete the entire experiment due to health concerns , if there was no dynamic photometry signal 3 weeks after surgery , or if the location of the photometry fiber tip was histologically determined to be outside the LNAc ( 1 mouse ) . All behavioral experiments were performed in at least two cohorts to minimize batch effects . All experiments were performed and analyzed blinded to genotype using automated , batched data analysis scripts/software wherever possible to eliminate experimenter bias . Littermate controls were used throughout the study . Whole-cell patch-clamp recordings were performed as previously described ( Cho et al . , 2017 ) in acute brain slices . Acute 250 µm-thick horizontal slices that contained the lateral VTA were prepared on a vibratome ( VT-1200 , Leica Biosystems ) from 8 to 12 week Nf1+/+ and Nf1+/- mice that had been transcardially perfused with an ice-cold NMDG cutting solution ( Ting et al . , 2014 ) saturated with 95% O2/5% CO2 . Slices were recovered at 32°C in NMDG cutting solution for ten minutes prior to transfer to HEPES recovery artificial cerebrospinal fluid ( ACSF ) ( Ting et al . , 2014 ) for an additional 30 min of recovery . During recording , slices were continuously perfused ( 2 . 0–3 . 0 mL/min ) with 32°C , 95% O2/5% CO2 -saturated recording ACSF that contained ( mM ) : 125 NaCl , 2 . 5 KCl , 1 . 2 NaH2PO4 , 1 . 2 MgCl2 , 2 . 4 CaCl2 , 26 NaHCO3 , and 11 glucose . The medial terminal nucleus of the accessory optic track ( MT ) was used as a visual landmark to delineate the most lateral region of the VTA . Whole-cell patch clamp recordings were obtained using 3–6 MΩ patch pipettes fabricated from borosilicate capillary glass tubing ( World Precision Instruments ) and backfilled with a potassium gluconate internal solution that contained ( mM ) : 135 K gluconate , 5 KCl , 5 EGTA , 0 . 5 CaCl2 , 10 HEPES , 2 Mg-ATP , and 0 . 1 GTP . A high-chloride internal solution was used to measure IPSCs and contained ( mM ) : 128 KCl , 20 NaCl , 1 MgCl2 1 EGTA , 0 . 3 CaCl2 , 10 HEPES , 2 Mg-ATP , and 0 . 3 GTP . Signals were amplified and digitized using a MultiClamp 700B amplifier ( Molecular Devices , LLC ) and Digidata 1440 analog-to-digital converter ( Molecular Devices , LLC ) . Series resistance ( Rs ) was monitored throughout recording , and data were discarded if the uncompensated Rs exceeded 30 MΩ or the holding current at −70 mV was more negative than −200 pA . Rheobase currents were determined via the injection of a 500 pA ramp current over 500 ms from −60 mV in current clamp mode . Ih currents were measured during seven 2 s , −10 mV hyperpolarizing voltage steps from −60 mV in voltage clamp in the presence of 20 µM bicuculline methiodide and 3 mM kynurenic acid + /- 20 µM forskolin . Spontaneous firing was measured over twenty seconds of gap free recording in I = 0 mode . Spontaneous IPSCs and EPSCs were measured at −70 mV in voltage clamp . IPSCs were recorded in the presence of 3 mM kynurenic acid . Electrophysiological data were sampled at 10 kHz and filtered at 2 kHz with Clampex 10 . 4 and analyzed in Clampfit 10 . 7 ( Molecular Devices , LLC ) . Stereotaxic viral vector injections were performed in mice anesthetized with isoflurane ( 1–3% in 95% O2/5% CO2 provided via nose cone at 1 L/min ) as previously described ( Cho et al . , 2017 ) . Following anesthesia , preparation and sterilization of the scalp , and exposure of the skull surface , a craniotomy hole was drilled over the LNAc ( antero-posterior: 1 . 2 mm , medio-lateral: 1 . 6 mm relative to Bregma ) . 800 nL of the AAV9-hSyn-dLight1 . 2 vector ( titer:~4 × 1012 viral genomes/mL , produced at the UC Davis Vision Center Vector Design and Packaging Core facility; Addgene # 111068 ) was delivered into the LNAc ( antero-posterior: 1 . 2 mm , medio-lateral: 1 . 6 mm , dorso-ventral: −4 . 2 mm relative to Bregma ) using a blunt 33-gauge microinjection needle within a 10 μL microsyringe ( NanoFil , World Precision Instruments ) , a WPI microsyringe pump ( UMP3 , World Precision Instruments ) , and pump controller ( Micro4 , World Precision Instruments ) over 10 min . Following viral injection , a 5 mm long , 400 μm outer diameter mono fiber-optic cannula ( MFC_400/430–0 . 48_5 mm_ZF1 . 25_FLT , Doric Lenses Inc ) was lowered to the same stereotaxic coordinates and affixed to the skull surface with C and B Metabond ( Parkel Inc ) and dental cement . For optogenetic stimulation of the SC during dLight1 . 2 recordings , mice received a second stereotaxic injection of AAV5-hSyn-ChR2 ( H134R ) -eYFP ( UNC Vector Core ) in the SC ( antero-posterior: −4 . 0 mm , medio-lateral: 0 . 5 mm , dorso-ventral: −1 . 5 mm relative to Bregma ) , followed by implantation of a 3 mm long , 300 μm mono fiber-optic cannula ( MFC_300/330–0 . 48_3 mm_ZF1 . 25_FLT , Doric Lenses Inc; antero-posterior: −4 . 0 mm , medio-lateral: 0 . 5 mm , dorso-ventral: −1 . 3 mm relative to Bregma ) . For optogenetic inhibition of VTAnon-Th neurons , 500 nL of a 1:4 mixture of AAV9-Th-PI-Cre-SV40 ( gift of James M . Wilson , Addgene viral prep # 107788-AAV9 ) and AAV-DJ-Ef1α-DO-eNpHR3 . 0-eYFP-WPRE-pA ( gift of Bernardo Sabatini , Addgene # 37087 ) or AAV-DJ-Ef1α-DO-mCherry-WPRE-pA ( gift of Bernardo Sabatini , Addgene # 37119 ) was injected bilaterally into the VTA ( antero-posterior: −3 . 3 mm , medio-lateral: ± 0 . 5 mm , dorso-ventral: −4 . 2 mm relative to Bregma ) , followed by implantation of 5 mm long , 300 μm mono fiber-optic cannulae ( MFC_300/330–0 . 48_5 mm_ZF1 . 25_FLT; antero-posterior: −3 . 3 mm , medio-lateral: ± 1 . 84 mm , dorso-ventral: −3 . 59 mm relative to Bregma ) at angle of twenty degrees . Mice were given 1 mg/kg buprenorphine SR and 5 mg/kg ketoprofen s . c . intraoperatively and received 30 mg/kg ibuprofen p . o . in their home cage water for five days post-operatively for pain . Mice were allowed a minimum of 14 days for surgical recovery prior to participation in behavioral studies . In order to create Th-VAST , a PCR fragment containing the 2 . 5 kb rat tyrosine hydroxylase promoter ( Oh et al . , 2009 ) was subcloned into pAAV-ihSyn1-tTA-WPRE ( Addgene #99120 ) , replacing the ihSyn promoter through AflII and MluI restriction digest to create pAAV-Th-tTA-WPRE . pAAV-TRE-mRuby-WPRE ( Addgene # 99114 ) , pAAV-TRE-mNeonGreen-WPRE , pAAV-TRE-mTurquoise-WPRE ( Addgene # 99113 ) , pAAV-Th-GFP-WPRE ( Addgene # 99128 ) , pAAV-Ef1α-DO-mCherry-WPRE-pA ( Addgene # 37119 ) , and pAAV-Ef1α-DO-NpHR3 . 0-eYFP-WPRE-pA ( Addgene # 37087 ) constructs were used as previously described ( Chan et al . , 2017; Saunders et al . , 2012 ) . Virus production was performed using a published protocol ( Challis et al . , 2019 ) . In brief , HEK293T cells were triple transfected using polyethylenimine ( PEI ) to deliver viral pUCmini-iCAP-PHP . eB ( Addgene #103005 ) or pAAV-DJ-Rep-Cap ( VPK-420-DK , Cell Biolabs , Inc ) , pHelper , and transgene plasmids . Viral particles were harvested from the media and cell pellet and purified over 15% , 25% , 40% and 60% iodixanol ( OptiPrep , STEMCELL Technologies , Inc ) step gradients . Viruses were concentrated using Amicon Ultra centrifugal filters ( Millipore Sigma ) , formulated in sterile phosphate buffered saline , and titered with qPCR by measuring the number of DNase I–resistant viral genomes relative to a linearized genome plasmid as a standard . Following viral production and titering , systemic AAV vectors were administered via injection into the retro-orbital sinus during anesthesia with isoflurane ( 1–3% in 95% O2/5% CO2 provided via nose cone at 1 L/min ) , followed by administration of 1–2 drops of 0 . 5% proparacaine to the corneal surface ( Challis et al . , 2019 ) . Note: we have observed some toxicity with doses of AAV-PHP . eB-Th-tTA≥1×1011 vg/mouse when allowed to express >3 weeks; therefore , we encourage users to perform dosing and time course studies when beginning experiments with Th-VAST . Fiber photometry was used to monitor fluorescent dopamine signals using a custom system as previously described ( Cho et al . , 2017; Patriarchi et al . , 2018 ) , which allowed for dLight1 . 2 excitation and emission light to be delivered and collected via the same implanted optical fiber . Our system employed a 490 nm LED ( M490F1 , Thorlabs , Inc; filtered with FF02-472/30-25 , Semrock ) for fluorophore excitation and a 405 nm LED for isosbestic excitation ( M405F1 , Thorlabs Inc; filtered with FF01-400/40-25 , Semrock ) , which were modulated at 211 Hz and 531 Hz , respectively , controlled by a real-time processor ( RX8-2 , Tucker David Technologies ) , and delivered to the implanted optical fiber via a 0 . 48 NA , 400 µm diameter mono fiber optic patch cable ( MFP_400/430/LWMJ-0 . 48_2 m_FC-ZF1 . 25 , Doric Lenses Inc ) . The emission signal from isosbestic excitation , which has previously been shown to be calcium independent for GCaMP sensors ( Kim et al . , 2016; Lerner et al . , 2015 ) , was used as a reference signal to account for motion artifacts and photo-bleaching . Emitted light was collected via the patch cable , collimated , filtered ( MF525-39 , Thorlabs ) , and detected by a femto-Watt photoreceiver ( Model 2151 , Newport Co . ) after passing through a focusing lens ( 62–561 , Edmunds Optics ) . Photoreceiver signals were demodulated into dLight1 . 2 and control ( isosbestic ) signals , digitized ( sampling rate: 382 Hz ) , and low-pass filtered at 25 Hz using a second-order Butterworth filter with zero-phase distortion . A least-squares linear fit was applied for the 405 nm signal to be aligned with the 490 nm signal . Then , the fitted 405 nm signal was subtracted from 490 nm channel , and then divided by the fitted 405 nm signal to calculate ΔF/F values . The code to perform this function is available at: https://github . com/GradinaruLab/dLight1/blob/master/FP_Session_Processing2 . m . During behavioral experiments , the ΔF/F time-series trace was normalized using a robust z-score ( signal-signal medianmedian absolute deviation ) to account for data variability across animals and sessions . When fiber photometry was performed during behavioral testing , dLight1 . 2 signals were synchronized to the beginning of the behavioral session by delivery of TTL pulses ( via a TTL pulse generator; OTPG_4 , Doric Lenses Inc ) to the photometry system . Baseline dLight1 . 2 Measurements: Mice were tethered to the photometry patch cable , placed in a clean home cage within a sound attenuating box , and allowed to habituate for 5 min ( Lafayette Instrument Company ) . Spontaneous dLight1 . 2 signals were subsequently recorded for 5 min . This procedure was repeated at least three times per mouse . During pre-treatment experiments , morphine sulfate ( 5 . 0 mg/kg s . c . ) or saline ( s . c . ) was administered twenty minutes prior to the onset of each session . Median fluorescence was determined from the processed ΔF/F time-series trace using the median ( ) function in Matlab; peak analysis was performed on the normalized , z-scored trace using a two z-score threshold to minimize contamination by fluorescent noise . Dopamine transients were detected using the findpeaks ( ) function in Matlab , and outputs were averaged within each subject across trials . Reward Consumption and Pavlovian Conditioning: During the twenty-minute sucrose consumption assay , mice were given access to ten 50 μL sucrose ( 5% w/v ) rewards delivered every 60 s ( 0 . 5 mL total ) via a lick port in a mouse modular test chamber ( Model 80015NS , Lafayette Instrument Company ) placed within a sound-attenuating box and controlled by ABET II software ( Lafayette Instrument Company ) . Sucrose consumption was characterized by measuring the timing and number of licks at the lick spout , which was measured with an optical lickometer in the lick port . Lick bouts were defined as licking events that exceeded five licks/second and were at least 3 s removed from a previous lick bout . During the Pavlovian conditioning assay , mice were conditioned to associate the presence of a 10-s conditioned stimulus ( CS; illumination of the house-light and delivery of a 60 dB , 3 kHz tone ) with delivery of the unconditioned stimulus ( US; 50 μL 5% sucrose w/v; 1000 μL total fluid delivery/session ) 7 s after the onset of the CS . Each conditioning session consisted of twenty trials with an inter-trial interval randomly drawn from a uniform distribution between 75 and 105 s . On day 11 , the sucrose US was omitted to determine dLight1 . 2 responses to reward omission ( Day 11 , Trial 1 ) . The number of licks during CS presentation were determined for each trial using ABET II . Because US consumption often occurred during the inter-trial interval , the US timing was determined by identifying the first lick bout after reward presentation during each trial . Cued Fear Conditioning: During the cued fear conditioning task , a ten-second CS ( 60 dB , 3 kHz tone and house light illumination ) immediately preceded a 1 s , 0 . 4-mA foot shock delivered via the grid floor of the conditioning chamber ( Lafayette Instrument Company ) . The conditioning procedure was completed during 15 consecutive trials with an inter-trial interval randomly drawn from uniform distribution between 75 and 105 s . Mice were videotaped during the assay under dim red light conditions so that the amount of time freezing ( defined as the absence of any body movement ) and the latency to freeze during each trial could be quantified post hoc by two blinded reviewers . Audiovisual Stimulus Exposure: Mice were placed in a clean home cage in the sound attenuating chamber placed underneath the speaker and house light from the modular conditioning chamber . During each trial , mice were randomly presented with either a ten-second 60 dB , 3 kHz tone or house light illumination , which had equal probability of selection . Each session consisted of twenty trials , and the inter-trial interval was randomly drawn from a uniform distribution between 75 and 105 s . Social Interaction and Social Preference Assay: Test mice were placed in a clean cage and allowed to freely interact with a juvenile , sex-matched , novel , conspecific probe mouse . Social interactions were videotaped , and the onset of recording was synchronized with the photometry signal via delivery of TTL pulses . The onset of each social interaction was defined as the initiation of physical contact between mice and was terminated when mice physically disengaged . The social preference assay was performed in a 50 cm x 50 cm square , white acrylic arena that contained two identical , wire mesh-enclosed social interaction vestibules placed in the center of opposite walls of the arena . During each thirty-minute testing session , a juvenile , sex-matched , novel , conspecific probe mouse was placed in one of the vestibules , and the position of the test mouse was tracked with an overhead camera using EthoVision XT 10 ( The Noldus Company ) . The location of the probe mouse was alternated between trials . Looming Stimulus Assay: The looming stimulus assay was performed as previously described ( Yilmaz and Meister , 2013 ) . Mice were acclimated in an 87 cm x 47 . 5 cm x 30 cm ( h ) infrared-transmitting black acrylic arena in the presence of a nest/shelter for at least five minutes . The overhead looming stimulus was presented when the animal was in the center of the arena . The looming stimulus covered 5 degrees of the animal's visual field initially , expanded up to 50 degrees , and was presented five times separated by 1 s pauses on a gray background . For optogenetic manipulations , the looming stimulus was presented as described above in an arena that was optimized for behavioral testing in tethered mice that measured 50 cm x 35 cm x 30 cm ( h ) . In order to accommodate optical patch cables , the arena featured a rectangular opening adjacent to the overhead screen and a triangular nest/shelter with a 2 cm slit at the apex . During testing , laser pulses ( 532 nm , 5 mW , 20 Hz , 5 ms pulse width ) were delivered via the implanted optical fibers , which were coupled to optical patch cables ( MFP_300/330/LWMJ0 . 48_1 m_FC-ZF1 . 25 , Doric Lenses Inc ) connected to a 532 nm laser ( Changchun New Industries [CNI] Model with PSU-H-LED ) . Stimulation began 1 . 5 s prior to looming onset and continued throughout the stimulus presentation . Mice were videotaped throughout so that time to react to the disc , escape location , latency to the first freezing episode , duration of the first freezing episode , and amount of time spent freezing during the first minute after exposure could be quantified by a blinded reviewer . Optogenetic Stimulation of the Superior Colliculus: Mono fiber-optic patch cables were connected to implanted optical fibers for fiber photometry in the LNAc and optogenetic stimulation of the ipsilateral vSC . The vSC patch cable was connected to a 473 nm laser ( Changchun New Industries [CNI] Model with PSU-H-LED ) and controlled by a Doric Lenses OTPG_4 pulse generator that was triggered by the ABETII software . Mice were allowed to habituate in a clean home cage within the Lafayette sound-attenuating chamber for 5 min prior to the start of the experiment . Following habituation , dLight1 . 2 signals were recorded during twenty trials in which the vSC was stimulated with 2 s of 20 Hz , 5 ms , 473 nm laser pulses . The inter-trial interval was randomly drawn from a uniform distribution between 75 and 105 s . Free-floating brain sections were blocked for 1 hr in 10% normal donkey serum ( Millipore-Sigma , S30-M ) in phosphate buffered saline ( PBS ) with 0 . 1% Triton X-100 at room temperature , then incubated with primary antibody diluted in the blocking buffer overnight at 4°C . Sections were washed three times for 15 min in PBS . Secondary antibodies were diluted in blocking buffer , and brain sections were incubated in the secondary antibody solution for 2 hr at room temperature . Sections received three 15 min washes in PBS prior to mounting on glass slides with ProLong Diamond anti-fade mounting medium ( ThermoFisher Scientific , P36965 ) or RIMS ( refractive index matching solution; RI = 1 . 46 ) ( Yang et al . , 2014 ) . VAST sections were optically cleared overnight in RIMS prior to mounting . The following antibodies/dilutions were used: rabbit anti-tyrosine hydroxylase ( EMD Millipore , AB152 , 1:1000; for VAST ) , monocloncal mouse anti-tyrosine hydroxylase ( ImmunoStar , 22941 , 1:1000; for TH quantification , somata tracing , and cell counting ) , polyclonal chicken anti-GFP ( Aves , GFP-1020 , 1:1000 ) , Alexa Fluor 488-conjugated donkey anti-chicken IgY F ( ab’ ) two fragment ( Jackson ImmunoResearch , 703-546-155 , 1:1000 ) , Alexa Fluor 647-conjugated donkey anti-rabbit IgG Fab fragment ( Jackson ImmunoResearch , 711-606-152 , 1:1000 ) , and Alexa Fluor 647-conjugated donkey anti-mouse IgG Fab fragment ( Jackson ImmunoResearch , 711-607-003 , 1:1000 ) . Histological images were obtained using either a Keyence BZ-X fluorescence microscope or Zeiss 880 confocal microscope . Images were analyzed using BZ-X Analyzer software ( Keyence Corporation ) , ImageJ , and/or Imaris ( Bitplane ) . VTA and SNc cell counting was performed on 100 mm , coronal histological sections from Bregma −3 . 2 to −3 . 6 μm ( AP ) and averaged across sections within each mouse . Statistical analysis was performed using Matlab ( MathWorks , Inc ) and GraphPad Prism 7 ( GraphPad Software , Inc ) . All statistical tests performed on data presented in the manuscript are stated in the figure captions and provided in detail in Source data 1 . For each experiment , statistical tests were chosen based on the structure of the experiment and data set . No outliers were removed during statistical analysis . Sample sizes estimates were based on published behavioral and electrophysiological literature that utilized the 129T2/SvEmsJ::C57Bl/6NTac Nf1+/- mouse model; this was within a range commonly employed by researchers in our field using similar techniques and that which was determined via the sampsizepwr ( ) function in Matlab . Parametric tests were used throughout the manuscript; for sIPSC and ESPC data , non-parametric Mann Whitney U tests were also reported because sIPSC frequency was non-normally distributed for both genotypes ( as determined by the D’Agostino and Pearson normality test ) . In this case , the results of parametric and non-parametric hypothesis tests were congruent ( see Figure 3 caption ) . When analysis of variance ( ANOVA; one-way , two-way , and/or repeated measures ) was performed , multiple comparisons were corrected using the Bonferroni correction . Multiple t-tests were corrected with the two-stage linear step-up procedure of Benjamini , Krieger and Yekutieli with a false discovery rate of 5% . Viral vector plasmids used in this study are available on Addgene at http://www . addgene . org/Viviana_Gradinaru/ . Codes used for fiber photometry signal extraction and analysis are available at https://github . com/GradinaruLab/dLight1 . Source data is available at www . doi . org/10 . 7303/syn18904024 .	About one in 3 , 500 people have a genetic disorder called neurofibromatosis type 1 , often shortened to NF1 , making it one of the most common inherited diseases . People with NF1 may have benign and cancerous tumors throughout the body , learning disabilities , developmental delays , curvature of the spine and bone abnormalities . Children with NF1 often experience difficulties with attention , hyperactivity , speech and language delays and impulsivity . They may also have autism spectrum disorder , or display symptoms associated with this condition . Studies in mice with a genetic mutation that mimics NF1 suggest that abnormal development in cells in the middle of the brain may cause the cognitive symptoms . These midbrain neurons produce a chemical called dopamine and send it throughout the brain . Dopamine is essential for concentration and it is involved in how the brain processes pleasurable experiences . Now , Robinson et al . show that , at rest , the NF1 model mice release dopamine less often than typical mice . This happens because , when there are no stimuli to respond to , neighboring cells slow down the activity of dopamine-producing neurons in NF1 model mice . In the experiments , both NF1 model mice and typical mice were taught to associate environmental cues with rewards or punishments . Robinson et al . then measured the release of dopamine in the mice using a sensor called dLight1 , which produces different intensities of fluorescent light depending on the amount of dopamine present . This revealed that the NF1 model mice produced more dopamine in response to visual cues and had enhanced behavioral responses to these stimuli . For example , when a looming disc that mimics predators approached them from above , the NF1 model mice tried to hide in an exaggerated way compared to the typical mice . Previously , it had been shown that this type of behavior is due to the activity of the dopamine-producing neurons' neighboring cells , which Robinson et al . found is greater in NF1 model mice . Next , Robinson et al . stopped neighboring cells from interfering with the dopamine-producing neurons in NF1 model mice . This restored dopamine release to normal levels at rest , and stopped the mice from overreacting to the looming disc . The experiments help explain how the NF1 model mice process visual information . Further study of the role dopamine plays in cognitive symptoms in people with NF1 may help scientists develop treatments for the condition .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2019	Optical dopamine monitoring with dLight1 reveals mesolimbic phenotypes in a mouse model of neurofibromatosis type 1
G-quadruplexes ( G4 ) are polymorphic four-stranded structures formed by certain G-rich nucleic acids in vitro , but the sequence and structural features dictating their formation and function in vivo remains uncertain . Here we report a structure-function analysis of the complex hCEB1 G4-forming sequence . We isolated four G4 conformations in vitro , all of which bear unusual structural features: Form 1 bears a V-shaped loop and a snapback guanine; Form 2 contains a terminal G-triad; Form 3 bears a zero-nucleotide loop; and Form 4 is a zero-nucleotide loop monomer or an interlocked dimer . In vivo , Form 1 and Form 2 differently account for 2/3rd of the genomic instability of hCEB1 in two G4-stabilizing conditions . Form 3 and an unidentified form contribute to the remaining instability , while Form 4 has no detectable effect . This work underscores the structural polymorphisms originated from a single highly G-rich sequence and demonstrates the existence of non-canonical G4s in cells , thus broadening the definition of G4-forming sequences . G-rich nucleic acids can form G-quadruplexes ( G4 ) , a stable four-stranded structure formed by stacking of guanine tetrads ( G-quartets ) in the presence of coordinating cations such as K+ ( Figure 1A ) ( Davis , 2004; Neidle , 2009; Patel et al . , 2007; Sen and Gilbert , 1988 ) . This core tetrad organization is the signature of a G4 , around which a variety of conformations blossom depending on the primary sequence and physico-chemical conditions ( Chen and Yang , 2012 ) . Furthermore , competitive structural polymorphisms can result from a single nucleic acid sequence , when multiple contiguous G-tracts are available ( Phan , 2010 ) . These complexities challenge our ability to predict G4 formation from any particular sequence , let alone predict a particular structure . 10 . 7554/eLife . 26884 . 003Figure 1 . G-quadruplexes and G4-dependent minisatellite instability in S . cerevisiae . ( A ) Schematic representation of the overall G4 structure , its features and the underlying canonical G4 motif . N can be any nucleotide . G4-ligands such as Phen-DC3 bind by stacking on an outermost G-quartet . ( B ) Site of CEB1 integration in the yeast genome , near the replication origin ARS305 . CEB1 is oriented so that the G-rich strand is template for the leading strand replication machinery emanating from ARS305 . Model for G4-dependent minisatellite instability during leading-strand replication ( Lopes et al . , 2011 ) . ( C ) Example of CEB1 instability in untreated or Phen-DC3-treated WT cells and in a pif1Δ mutant . The main band above the 947 bp marker is the parental size CEB1 . The Southern blots were published previously in ( Piazza et al . , 2015 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 003 Based on pioneering biophysical knowledge , a G4 consensus motif of the form G3-5N1-7 G3-5N1-7 G3-5N1-7G3-5 ( where N can be any nucleotide ) was adopted ( Huppert and Balasubramanian , 2005; Todd et al . , 2005 ) . It imposed constraints on the G-tract number ( 4 ) and length ( 3 to 5 nt ) as well as on the length of each connecting loop ( 1 to 7 nt ) ( Figure 1A ) . These parameters established a reasonable compromise balancing false-positive ( containing sequences with several loops of >4 nt [Guédin et al . , 2010; Rachwal et al . , 2007] ) and false-negative motifs such as G4s containing only two G-quartets ( Macaya et al . , 1993; Chinnapen and Sen , 2004 ) or a single long loop together with two other short loops ( Guédin et al . , 2010 ) . This consensus was extensively used to mine genomic sequences , and estimated ~376 , 000 potential G4-forming motifs in the human genome ( Huppert and Balasubramanian , 2005; Todd et al . , 2005 ) . However , recent structural studies unveiled additional ‘non-canonical’ G4 , bearing bulges ( De Nicola et al . , 2016; Mukundan and Phan , 2013 ) , strand interruptions with snapback guanines ( Adrian et al . , 2014 ) , and incomplete tetrads ( G-triad ) ( Heddi et al . , 2016; Li et al . , 2015 ) . They result from sequences lacking four G-triplets , and thus escape the consensus . Recently , a high throughput in vitro polymerase stop assay performed on purified human genomic DNA in the presence of K+ or G4-stabilizing ligand Pyridostatin identified 716 , 310 G4-forming sites; 451 , 646 sites did not match the consensus ( Chambers et al . , 2015 ) , indicating that the false-negative rate of the initial consensus is massive . Accordingly , a new G4 prediction algorithm ( G4Hunter ) emphasizing G-richness and skewness over well-defined G-tracts and arbitrary loop lengths has been developed and its predictability ( 95% ) established upon biophysical characterization of hundreds of sequences over an extensive range of thermal stabilities ( Bedrat et al . , 2016 ) . This algorithm conservatively heightened the figure for putative G4 sequences in the human genome to ~700 , 000 , in agreement with the G4-seq assay ( Chambers et al . , 2015 ) . This re-evaluation has implications for inference of cis-acting functions of G4 and their association with other genomic and epigenomic features . Hence , biological evidence for the relevance of these non-canonical G4s is paramount . Compelling evidence for the role of G4s in various biological processes have accumulated ( for reviews see: [León-Ortiz et al . , 2014; Maizels and Gray , 2013; Rhodes and Lipps , 2015; Tarsounas and Tijsterman , 2013; Vasquez and Wang , 2013] ) . Yet , an uncertainty remains between the ability of a predicted sequence to form a G4 in vitro and exert a G4-dependent biological function in vivo . Relevant to the present study , we recently showed that , among a set of validated G4-forming variant sequences of the human minisatellite CEB25 , only the G4s with short loops preferentially containing pyrimidine were capable of inducing genomic instability in the eukaryotic model organism S . cerevisiae ( Piazza et al . , 2015 ) . These results demonstrated that only a subset of G4-forming sequences actually formed and/or exerted a biological effect; in this case , the ability to interfere with leading strand DNA replication ( Lopes et al . , 2011 ) . While the unstable CEB25-G4 motif variant bearing short loops matched the G4 consensus ( Piazza et al . , 2015 ) , we also previously reported that the human minisatellite CEB1 was similarly unstable despite the lack of a consensus G4 motif ( Lopes et al . , 2011; Piazza et al . , 2010 , 2012; Ribeyre et al . , 2009 ) . Our first biophysical study suggested that the CEB1 motif was forming a mixture of several G4 conformations in solution that could not be individually resolved ( Ribeyre et al . , 2009 ) . The structural analyses of an isolated conformation revealed a rather unique snapback scaffold with single-nucleotide loops ( Form 1 , see below and [Adrian et al . , 2014] ) , which resulted from a non-consensus G4 motif involving a G-doublet . To demonstrate the in vivo relevance of this unusual form and further resolve the variety of G4s that CEB1 likely forms , we now report our comprehensive biophysical , structural and biological structure-function analysis of the wild-type CEB1 motif and 27 mutated variants , assayed for their effects on genomic instability . This study demonstrates the existence of at least three types of non-canonical G4s in vivo and the threat they pose to genomic stability . Our experimental system assays the instability of G4-prone tandem repeats ( expressed as contraction or expansion of the number of motifs ) in two G4-stabilizing conditions: upon deletion of the G4-unwinding helicase Pif1 ( Paeschke et al . , 2013; Ribeyre et al . , 2009 ) or in cells treated with the G4-stabilizing ligand Phen-DC3 ( De Cian et al . , 2007; Monchaud et al . , 2008 ) , which inhibits G4 unwinding by Pif1 in vitro and in vivo ( Piazza et al . , 2010 ) . We summarize in Figure 1B the mechanism of G4-dependent rearrangement formation during leading strand replication . The CEB1 motif ( 39 nt ) is 77% GC-rich with a GC-skew of 77% . It comprises seven G-tracts: one G-sextet ( G9-14 ) , three G-triplets ( G2-4 , G16-18 and G20-22 ) and three G-doublets ( G24-25 , G31-32 , G34-35 ) ( Figure 2A ) . We previously showed that the CEB1 instability depends on its ability to form G4 ( s ) in vivo by simultaneously mutating the G-sextet and each G triplet ( CEB1-Gmut in Table 1 ) ( Ribeyre et al . , 2009 ) . Here , to precisely elucidate the sequences required to form G4 ( s ) and trigger CEB1 instability , we synthesized 26 new minisatellites of similar length bearing single or multiple mutations in each motif of the array ( Table 1 ) . All constructs were inserted in the vicinity of the ARS305 origin of replication , in the orientation where the G-rich strand is the template for leading strand synthesis ( Figure 1B ) ( Lopes et al . , 2011 ) . The rearrangement frequencies were measured upon mitotic growth of untreated and Phen-DC3-treated wild-type yeast cells ( WT ) as well as in pif1Δ cells as previously described ( Lopes et al . , 2011; Piazza et al . , 2010 ) ( example given Figure 1C , Materials and methods ) . The sequences , rearrangement frequencies and statistical comparisons are reported in Table 1 . First , we measured the rearrangement frequencies of the control CEB1-WT-20 and CEB1-WT-25 alleles ( 20 and 25 motifs , respectively ) . These synthetic alleles were stable in WT cells ( 0% and 2 . 5% instability , respectively ) and significantly unstable both in Phen-DC3-treated WT cells ( 10 . 9% and 12% instability , respectively ) and in the pif1Δ mutant ( 25 . 6% and 59 . 1% instability , respectively ) ( Table 1 and Figure 2B ) . The rearrangement frequencies for CEB1-WT-25 were reported previously ( Figure 1C ) ( Piazza et al . , 2015 ) . 10 . 7554/eLife . 26884 . 004Figure 2 . Determination of the G-tract requirements for CEB1 instability in S . cerevisiae . ( A ) CEB1 motif with guanine tracts colored in cyan and numbered . ( B–C ) Southern blot of CEB1-WT-20 and CEB1 variant alleles mutated for their G-tracts in WT cells treated with Phen-DC3 or in pif1Δ cells . ( B ) G-triplet and ( C ) G-sextet mutants . In most instances , several independent colonies were pooled and extracted . The number of colonies analyzed per lane and the total rearrangement frequency is indicated for each blot . The fragment sizes ( bp ) of the molecular ladders run in the first lane of each blot are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 00410 . 7554/eLife . 26884 . 005Figure 2—figure supplement 1 . Effect of loop length and sequence on CEB1 instability . Substitution of the A19 loop located between two essential G-triplets by a Thymine causes an increase in CEB1 rearrangement frequencies both in Phen-DC3-treated WT cells and in pif1Δ cells ( ANT1909 and ANT1923 , respectively; p-values vs . CEB1-WT-20 < 0 . 05 in both cases ) . Increasing the loop length in CEB1-A19TT ( strains ANT1910 and ANT1924 ) and CEB1-A19TTT ( ANT1911 and ANT1925 ) caused a gradual decrease up to non-significant instability levels for CEB1-A19TTT . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 00510 . 7554/eLife . 26884 . 006Table 1 . Genomic instabilities of CEB1 variants and the associated G4 thermal stability . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 006AlleleMotifsSequenceWT cellsWT cells + Phen-DC3pif1D cellsFormsG4 TmUV ( °C ) CEB1-1 . 8 ( Lopes et al . , 2011 ) 420 . 5% ( 384 ) 11 . 2% ( 992 ) 56 . 3% ( 119 ) WT-25 ( Piazza et al . , 2015 ) 25TGGGCTGAGGGGGGAGGGAGGGTGGCCTGCGGAGGTCCC2 . 5% ( 159 ) 12 . 0% ( 192 ) 59 . 1% ( 66 ) all71 . 16 ( 1 . 78 ) WT-20200 ( 96 ) 10 . 9% ( 192 ) 25 . 6% ( 43 ) WT-10100 ( 96 ) 4 . 2% ( 176 ) NDSingle G-tracts mutationsGmut ( Ribeyre et al . , 2009 ) 20TGCGCTGAGCGCGGAGTGAGAGTGGCCTGCGGAGGTCCC1 . 0% ( 96 ) 1 . 0% ( 192 ) ND-NA42TGCGCTGAGCGCGGAGTGAGAGTGGCCTGCGGAGGTCCCNDND0 ( 384 ) G3T22TGTGCTGAGGGGGGAGGGAGGGTGGCCTGCGGAGGTCCC1 . 0% ( 96 ) 8 . 9% ( 192 ) 20 . 0% ( 95 ) 1 , 2 , 359 . 74 ( 0 . 74 ) G ( 9 , 10 ) T24TGGGCTGATTGGGGAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 92 ) 8 . 3% ( 192 ) ND1 , 265 . 8 ( 0 . 95 ) G ( 9-11 ) T24TGGGCTGATTTGGGAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 96 ) 11 . 2% ( 188 ) 34 . 4% ( 64 ) †248 . 81 ( 2 . 65 ) G ( 9 , 10 , 14 ) T21TGGGCTGATTGGGTAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 192 ) 0 ( 192 ) †0 . 5% ( 192 ) † ( 2 ) 44 . 74 ( 0 . 31 ) G ( 10-13 ) T25TGGGCTGAGTTTTGAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 176 ) 0 ( 192 ) †0 ( 192 ) †-46 . 24 ( 1 . 08 ) G11T25TGGGCTGAGGTGGGAGGGAGGGTGGCCTGCGGAGGTCCC0 . 5% ( 192 ) 7 . 8% ( 192 ) 44 . 4% ( 36 ) 1 , 270 . 18 ( 2 . 00 ) G ( 11-12 ) T24TGGGCTGAGGTTGGAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 96 ) 0 . 5% ( 192 ) ND-NDG ( 11 , 13 ) T26TGGGCTGAGGTGTGAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 192 ) 0 ( 192 ) †0 . 5% ( 192 ) †-44 . 02 ( 0 . 78 ) G12T24TGGGCTGAGGGTGGAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 92 ) 19 . 7% ( 192 ) †10 . 2% ( 88 ) †-41 . 11 ( 0 . 81 ) G13T24TGGGCTGAGGGGTGAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 96 ) 1 . 2% ( 172 ) †2 . 1% ( 48 ) †-42 . 92 ( 0 . 98 ) G14T20TGGGCTGAGGGGGTAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 96 ) 3 . 1% ( 192 ) †13 . 3% ( 45 ) † ( 1 , 2 ) 63 . 44 ( 0 . 92 ) G17T24TGGGCTGAGGGGGGAGTGAGGGTGGCCTGCGGAGGTCCC0 ( 128 ) 0 ( 80 ) †0 ( 192 ) †-44 . 97 ( 0 . 52 ) G21T19TGGGCTGAGGGGGGAGGGAGTGTGGCCTGCGGAGGTCCC0 . 5% ( 180 ) 0 ( 192 ) †1 . 0% ( 180 ) †446 . 18 ( 1 . 14 ) G ( 24-25 ) T26TGGGCTGAGGGGGGAGGGAGGGTTTCCTGCGGAGGTCCC0 ( 192 ) 3 . 6% ( 192 ) †9 . 4% ( 96 ) †3 , 446 . 68 ( 1 . 21 ) G25T26TGGGCTGAGGGGGGAGGGAGGGTGTCCTGCGGAGGTCCC0 ( 192 ) 4 . 2% ( 192 ) †16 . 1% ( 87 ) †3 , 448 . 91 ( 0 . 71 ) G ( 31-32 ) T24TGGGCTGAGGGGGGAGGGAGGGTGGCCTGCTTAGGTCCC1 . 0% ( 96 ) 13 . 5% ( 192 ) 43 . 2% ( 37 ) All69 . 03 ( 2 . 68 ) G ( 34-35 ) T21TGGGCTGAGGGGGGAGGGAGGGTGGCCTGCGGATTTCCC5 . 0% ( 96 ) 14 . 4% ( 180 ) 17 . 6% ( 85 ) All69 . 16 ( 0 . 22 ) Double G-tracts mutationsG ( 3 , 9-11 ) T24TGTGCTGATTTGGGAGGGAGGGTGGCCTGCGGAGGTCCC0 ( 92 ) 6 . 3% ( 192 ) 8 . 3% ( 48 ) †242 . 20 ( 0 . 33 ) G ( 3 , 24-25 ) T22TGTGCTGAGGGGGGAGGGAGGGTTTCCTGCGGAGGTCCC0 ( 96 ) 3 . 7% ( 188 ) †1 . 1% ( 93 ) †335 . 38 ( 0 . 29 ) G ( 9-10 , 24-25 ) T20TGGGCTGATTGGGGAGGGAGGGTTTCCTGCGGAGGTCCC0 ( 96 ) 1 . 5% ( 192 ) †ND-43 . 92 ( 0 . 65 ) G ( 9-11 , 25 ) T24TGGGCTGATTTGGGAGGGAGGGTGTCCTGCGGAGGTCCC0 ( 96 ) 0 . 5% ( 192 ) †0 ( 48 ) †-42 . 32 ( 0 . 37 ) G ( 12 , 25 ) T24TGGGCTGAGGGTGGAGGGAGGGTGTCCTGCGGAGGTCCC0 ( 96 ) 17 . 7% ( 192 ) 11 . 5% ( 96 ) †-NDG ( 14 , 24-25 ) T20TGGGCTGAGGGGGTAGGGAGGGTTTCCTGCGGAGGTCCC0 ( 96 ) 0 ( 188 ) †ND-42 . 10 ( 0 . 49 ) p-values vs . untreated WT < 0 . 05 . † p-value vs . CEB1-WT < 0 . 05 . Forms in parenthesis denote a loop modification . Numbers in parenthesis indicate total number of colonies tested . In the Tm column , the number in parenthesis indicates the standard deviation of the triplicate Tm measurement . ND: Not determined; NA: Not applicable To address the involvement of individual G-tracts in CEB1 instability , we synthesized mutated minisatellites 19–26 motifs-long and compared their instability to the CEB1-WT allele bearing the closest number of motifs ( Table 1 ) . First , we assessed the involvement of the G2-4 , G16-18 , and G20-22 G-triplets by substituting the central guanine by a thymine . Clearly , the CEB1-G3T variant exhibited a significant instability upon Phen-DC3 treatment and PIF1 deletion ( 8 . 9% and 20 . 0% ) ( Table 1 and Figure 2B ) . Compared to the CEB1-WT of a similar size , G3T instability in each assay was not significantly different . In contrast , the CEB1-G17T and CEB1-G21T alleles remained stable in both assays ( ≤1 . 1% rearrangements ) , significantly different from CEB1-WT ( Table 1 and Figure 2B ) . Thus , the G16-18 and G20-22 triplets are critical for CEB1 instability while G2-4 is dispensable ( see below ) . To assay the role of loop length , we mutated the A19 loop ( located between two essential G-triplets ) into the more deleterious T ( Piazza et al . , 2015 ) . As expected , this single nucleotide substitution significantly increased the genomic instability of CEB1 in the Phen-DC3-treated WT cells ( from 10 . 9% to 29 . 7% ) and in the pif1Δ strain ( from 25 . 6% to 50 . 0% ) ( Figure 2—figure supplement 1 ) . Furthermore , extending the size of this loop in CEB1-A19TT and CEB1-A19TTT gradually reduced the minisatellite instability to low or background levels ( Figure 2—figure supplement 1 ) . Hence , we confirmed with the complex CEB1 sequence that G4s bearing short ( ≤3 nt ) pyrimidine-containing loops are more prone to trigger genomic instability . Next , we addressed the contribution of the G9-14 sextet by performing various combinations of G-to-T substitutions . These mutations , which interrupt G-tracts and simultaneously increase loop length ( s ) , resulted in varying degrees of stabilization . Most dramatically , the G ( 9-10 , 14 ) T , G ( 10-13 ) T , G ( 11 , 13 ) T and G13T constructs clearly abolished CEB1 instability in both conditions ( Table 1 and Figure 2C ) . The CEB1-G ( 11-12 ) T allele , only assayed in Phen-DC3-treated WT cells , also remained stable ( Table 1 ) . In contrast , the G ( 9-10 ) T , G ( 9-11 ) T and G11T mutations had no effect in both conditions ( Table 1 and Figure 2C ) . The CEB1-G14T allele exhibited an intermediate instability , significantly lower than CEB1-WT in both conditions , yet not abolished ( Table 1 and Figure 2C ) . Finally , G12T was unusual since it exhibited a significantly higher level of instability than CEB1-WT in Phen-DC3-treated WT cells ( 19 . 6 vs . 12 . 0% , p-value=0 . 05 ) and a lower instability in the absence of Pif1 ( 10 . 2 vs . 59 . 1% , p-value<0 . 01 ) ( Table 1 and Figure 2C ) . These results show that not all the Gs in the G-sextet contribute to CEB1 instability . With the exception of the G12T substitution , which is discussed below , all the mutations that disrupt the G12-14 triplet either strongly reduced or abolished CEB1 instability . In contrast , mutations of G9-11 triplets had either no or modest effects on CEB1 instability . Hence , critical for CEB1 instability is the presence of a G-triplet in the G-sextet , immediately contiguous with the two other essential G16-18 and G20-22 triplets . Shifting this G12-14 triplet by a single nucleotide away from G20-22 ( in CEB1-G ( 9-10 , 14 ) T and CEB1-G14T ) significantly reduced or abolished CEB1 instability , consistent with the stabilizing effect of increased loop length . In conclusion , our mutational analysis identified three G-triplets ( G12-14 , G16-18 , and G20-22 ) separated by single nucleotide loops as necessary for sustaining most of the CEB1 instability . Since the distant fourth G2-4 triplet is dispensable for CEB1 instability , the sequence causing the bulk of CEB1 instability escapes the G3N1-7G3N1-7G3N1-7G3 consensus . We further sought to determine the origin of the remaining guanines involved in CEB1 instability . To address the involvement of the additional G tracts , we generated mutant alleles of the G24-25 , G31-32 and G34-35 doublets , keeping the rest of the CEB1 motif intact . Strikingly , the CEB1-G ( 24-25 ) T and CEB1-G25T alleles exhibited a significant ~3 fold decrease of instability compared to CEB1-WT upon Phen-DC3 treatment ( 3 . 6% and 4 . 2% compared to 12% ) and PIF1 deletion ( 9 . 4% and 16 . 1% compared to 59 . 1% ) ( Table 1 and Figure 3A ) . In contrast , for both conditions , the CEB1-G ( 31-32 ) T and CEB1-G ( 34-35 ) T mutants were as unstable as the similarly sized CEB1-WT arrays ( Table 1 and Figure 3B ) . These results demonstrate the importance of the G24-25 doublet in CEB1 instability , contributing to ~2/3rd of the CEB1-WT instability . The unexpected role of this G-doublet is consistent with the additional genetic , biophysical and structural data reported below . 10 . 7554/eLife . 26884 . 007Figure 3 . Role of the CEB1 G24-25-doublet in CEB1 instability . ( A ) Mutation of the G24-25 doublet in CEB1-G25T or CEB1-G ( 24-25 ) T reduced by 2/3rd the minisatellite instability in both pif1Δ ( ORT7172 and ORT7173 ) and Phen-DC3-treated WT cells ( ORT7164 and ORT7165 ) . ( B ) Mutation of the G31-32 and G34-35 doublets did not affect CEB1 instability in pif1Δ cells ( ANT1965 and ANT1967 ) and Phen-DC3-treated WT cells ( ANT1950 and ANT1952 ) . ( C ) Mutation of the G-sextet in CEB1-G ( 9-11 ) T-G25T abolished the G24-25-independent instability in both WT Phen-DC3-treated ( ANT1973 ) and pif1Δ ( ANT2604 ) cells . ( D ) Mutation of the dispensable G2-4 triplet in CEB1-G ( 3 , 24–25 ) T did not decrease the G24-25-doublet-independent instability in Phen-DC3-treated WT cells ( ANT2620 ) but abolished it in a pif1Δ strain ( ANT2627 ) . ( E ) Mapping of the overlapping sets of G-tracts required for full CEB1 instability in the WT Phen-DC3-treated or pif1Δ cells . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 007 We then investigated the G-tracts requirements for the remainder of G24-25-independent instability by generating double-tract mutants . Specifically , we combined the G ( 24-25 ) T mutations with the G-sextet G ( 9-10 ) T , G ( 9-11 ) T or G14T mutations which by themselves had no detectable or only partial effects . In all cases , the remaining G24-25-independent instability was consistently abolished ( ≤1 . 5% ) in the Phen-DC3-treated cells ( Table 1 and Figure 3C ) . Accordingly , in the absence of Pif1 , the instability was also abolished in the G ( 9-11 , 25 ) T construct ( <0 . 5% ) , significantly less than the 16 . 1% observed in the single G25T mutant ( Table 1 ) . These results indicate that the remaining G24-25-independent CEB1 instability requires the G-sextet in both conditions . To further assay a potential interaction with the distant G2-4 triplet which by itself had no detectable effect , we also constructed the double-tract CEB1-G ( 3 , 24–25 ) T allele . In Phen-DC3-treated WT cells , it behaved like CEB1-G ( 24-25 ) T ( 3 . 7% vs . 3 . 6% ) ( Table 1 and Figure 3D ) . Differently , in Pif1-deficient cells this additional G3T mutation ablated the remaining G24-25-independent instability ( 1 . 1% vs . 9 . 4% or 16 . 1% in G ( 24-25 ) T and G25T , respectively , p-values<0 . 05 ) . These data indicate that in the absence of Pif1 , but not in in the WT+Phen-DC3 context , the G24-25-independent instability uniquely relies on the distant G2-4 triplet . Altogether , these results suggest that multiple G4s contribute to CEB1 instability , and these G4s are not necessarily the same in the presence of the Phen-DC3 ligand or in the absence of the Pif1 helicase ( see Discussion ) . In summary , our mutational analyses identified three partially overlapping sets of G-tracts required for full instability , two of which do not match the G4 consensus ( Figure 3E ) . One ( G3AG3AG3TG2 ) accounts for ~2/3rd of the instability in both G4-stabilizing conditions: it uniquely involves the G24-25-doublet . Another ( G6AG3AG3 ) accounts for the remaining ~1/3rd of the instability in the Phen-DC3 context only: it uniquely involves the entire G9-14-sextet and lacks a third non-null loop . The last one ( G3N4G6AG3AG3 ) accounts for the remaining ~1/3rd of the instability in the pif1Δ context only: it uniquely involves the distant G2-4 triplet separated from the rest of the G-tracts by 4 nts . Mapping of the G-tracts required for CEB1 genomic instability suggested the participation of at least three non-canonical G4s . This prompted us to conduct the biophysical characterization of the multiple G4 conformations by NMR spectroscopy . Imino proton spectra of the CEB1 motif ( 39-nt , Figure 4A , B ) or its G-rich segment ( 25-nt , Figure 4—figure supplement 1 ) evidently indicated the formation of multiple G4s , but no well-defined conformation could be identified . Consequently , we truncated the sequence to represent 4 combinations of G tracts , the structural characteristics of which were assessed using NMR and circular dichroism ( CD ) spectroscopy . Each of the sequence exhibited a clear NMR spectra corresponding to intramolecular G4s , named Forms 1 to 4 ( Figure 4B ) . Moreover , the G-to-T mutation of the first G-tract ( in 25-nt-[G3T] and 39-nt-shift-[G3T] ) ) or those on the last two G-tracts ( in 25-nt-G ( 21 , 24–25 ) T ) reduced the conformational multiplicity or potential aggregation of the 25-nt sequence and generated an imino proton spectrum resembling that of Form 1 or Form 4 , respectively ( Figure 4—figure supplement 1 ) . CD spectra of these forms exhibited a positive peak at 260 nm and a negative peak at 240 nm , characteristic of a parallel-stranded G4 ( Figure 4C and Figure 4—figure supplement 2 ) ( Wieland and Hartig , 2007 ) . A schematic representation of the four G4 structures isolated from the CEB1 motif is shown in Figure 4D . Interestingly , all these conformations exhibited non-canonical structural features . Form 1 was previously shown to be a parallel , single-nucleotide loop G4 bearing an interrupted strand and a snapback guanine with a V-shaped loop . Form 1 can exist as a monomer and dimerize at high concentration ( Adrian et al . , 2014 ) . Form 2 can be a G4 resembling Form 1 but lacks the G9 or G10 snapback guanine , leaving a vacant guanine in the lowermost quartet . Such G-triad-bearing G4s have recently been reported by us and others ( Heddi et al . , 2016; Li et al . , 2015 ) . Form 3 that incorporates only a G-sextet and two G-triplets separated by a single residue can fold into a monomeric parallel G4 with an unusual zero-nt loop and two 1-nt propeller loops . Finally , Form 4 can fold into a parallel G4 containing a 4-nt , a zero-nt , and a 1-nt loops . Depending on the salt and DNA strand concentration , Form 4 is in equilibrium between an interlocked dimeric G4 and a monomeric G4 ( Appendix 1 , Figure 4—figure supplements 3 and 4 ) . Altogether , the biophysical analyses of the highly contiguous CEB1 G-rich sub-motifs demonstrate that the CEB1 motif is able to fold into several non-canonical G4s in vitro . 10 . 7554/eLife . 26884 . 008Figure 4 . The CEB1 motif adopts multiple non-canonical G4 conformations . ( A ) Oligonucleotide sequences used to isolate individual G4 conformation in CEB1 . Guanine tracts are colored cyan and numbered within a repeat unit according to their positions . ( B ) NMR spectra and ( C ) CD spectra of the mixture of G4 resulted from the CEB1-WT motif and of isolated G4 folds Form 1 , 2 , 3 and 4 . The Form 4 spectra showed in dotted line corresponds to the dimeric form observed at high K+ and DNA strand concentration . The resonance frequency of the nuclei is expressed in part per million ( ppM ) . ( D ) Isolated G4 folding topologies result from the CEB1 sequence . The snapback guanine in Form 1 can be either G10 or G11 , with an associated V-shaped loop containing one or zero G , respectively ( Adrian et al . , 2014 ) . Form 2 resembles Form 1 but lacks the snapback guanine , and consequently exhibits a terminal G-triad . Form 3 contains a 0-nt loop between the G9-11 and the G12-14 strands . The monomeric Form 4 exhibits a 4-nt long loop between the G2-4 and G9-11 strands , and a 0-nt loop between the G9-11 and G12-14 strands . For the dimeric Form 4 see Figure 4—figure supplement 4J . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 00810 . 7554/eLife . 26884 . 009Figure 4—figure supplement 1 . Multiple G4s resulted from different G-rich fragments of CEB1 minisatellite . Conformational multiplicities were eliminated by residue specific mutations targeted at particular G-tracts . 7-nt truncation from 5’-end of 25-nt [G3T] resulted in Form 1 . 6-nt truncation from 3’-end of 25-nt [G ( 21 , 24 , 25 ) T] resulted in Form 4 . Imino proton spectra of Form 1 and Form 4 are shown in gray dotted line below 25-nt [G3T] / Form 3 [G3T] and 25-nt [G ( 21 , 24 , 25 ) T] , respectively . Guanine tracts are colored cyan and numbered according to their positions in 25-nt sequence . G-to-T mutations are highlighted in bold . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 00910 . 7554/eLife . 26884 . 010Figure 4—figure supplement 2 . CD spectra of the mutant CEB1 G4 motifs used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 01010 . 7554/eLife . 26884 . 011Figure 4—figure supplement 3 . NMR structural characterization of CEB1 Form 4 in 100 mM K+ solution . ( A ) The G-rich sequence . ( B ) TDS and ( C ) CD spectra . ( D ) Imino proton assignments from 15N-filtered spectra of samples , 2% 15N-enriched at indicated positions . ( D’ ) Imino proton spectrum following 15 min exposure in D2O solvent at 25°C . Disappearing imino peaks are marked with asterisks . ( E ) H8 proton assignments by site-specific 2H labeling at the indicated positions . The reference spectra ( ref . ) of imino and aromatic protons are shown at the top of the corresponding assignment spectra . ( F ) Through-bond correlations between guanine imino and H8 protons via 13C5 at natural abundance , using long-range J-couplings shown in the inset . ( G ) The H8/H6-H1’ sequential connectivity on NOESY spectrum ( mixing time , 300 ms ) . ( H ) The imino-H8 cyclic connectivities on NOESY spectrum ( mixing time , 300 ms ) . The tetrad arrangements ( I ) were identified from framed cross-peaks with associated label of the residue number of imino proton in the first position and that of H8 proton in the second position . Intra-residue H8/6-H1’ cross-peaks are labeled with residue numbers . Missing cross peaks are marked with asterisks . ( J ) Schematic representation of interlocked dimeric Form 4 . Anti guanines are colored cyan . The backbones of the core and loops are in black and red , respectively . Labels from different strands are in respective dark/light colors . The coexistence of monomeric and dimeric folds of Form 4 was shown using NMR and gel shift assay . ( K ) NMR spectra of Form 4 under different parameters that is , cation concentration , buffer condition and sample preparation . Imino peaks from unidentified minor form are marked with asterisks . ( L ) Mobility shift of Form 4 ( highlighted in green boxes ) in the presence of different concentrations of counter ions visualized by non-denaturing PAGE . Mobility shifts of Form 1 are highlighted in red boxes . ( M ) CD spectrum of monomeric Form 4 recorded in 5 mM K+ solution at ~4 μM strand concentration . ( N ) Schematic representation of NMR-derived interlocked dimeric and model monomeric Form 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 01110 . 7554/eLife . 26884 . 012Figure 4—figure supplement 4 . Imino spectral transition of Form 4 through K+ titration at ~0 . 1 mM strand concentration . Full transition to dimeric G4 was reached following sample annealing at >200 mM K+ . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 012 The mapping of the G-tracts required for each isolated G4 in vitro ( Figure 4A ) and the measure of the CEB1 variant instability in vivo ( Figures 2 and 3 ) allow us to infer which G4 structures underlie the instability of the CEB1 minisatellite . While mutations such as CEB1-G ( 10-13 ) T and CEB1-G ( 11 , 13 ) T disfavor all identified G4 forms , in agreement with mutagenesis data showing no instability , mutations such as in CEB1-G11T , CEB1-G ( 3 , 9–11 ) T , CEB1-G ( 3 , 24–25 ) T and CEB1-G21T would result in isolation of Form 1 + 2 , Form 2 , Form 3 and Form 4 , respectively , from the other potentially competing G4s in the full motif ( Figure 5A ) . Clearly , Form 4 is not involved in CEB1 instability as the CEB1-G21T mutant remains perfectly stable in all conditions . Form 1 + 2 accounted for 2/3rd of the instability in both contexts ( Figures 3D , 5A and B ) . Further isolation of Form 2 from Form 1 ( in CEB1-G ( 3 , 9–11 ) T and -G ( 9-11 ) T ) showed that Form 2 is sufficient in the presence of Phen-DC3 while Form 1 preferentially causes instability in the absence of Pif1 ( Figure 5A , B ) . The 0-nt containing loop Form 3 contributed to the remaining third of instability only in the Phen-DC3 context ( Figures 3D , 5A and B ) . Unfortunately , we could not isolate the putative structure ( s ) responsible for the remnant instability specific to the pif1Δ context ( Figure 5B ) . A possible explanation is that the usage of the different G-triplets and the G-sextet can fold into a mixture of several poorly stable G4s due to the incorporation of various loop lengths , up to a total of 9-nts . 10 . 7554/eLife . 26884 . 013Figure 5 . Contribution of G4 conformations to CEB1 instability . ( A ) Example of instabilities obtained in Phen-DC3-treated WT cells and in pif1Δ cells for CEB1 variants isolating single G4 conformations , normalized to the instability obtained for CEB1-WT . Since the sequence requirements for Form 2 is embedded in the sequence requirements for Form 1 , no mutation could isolate Form 1 . Its contribution can be deduced by comparing with alleles isolating Form 2 . Thermal stability and maximum loop length for each isolated form are indicated . ( B ) Summary of the CEB1 variants instability in Phen-DC3-treated WT cells and in pif1Δ cells as a function of the forms they can adopt . Each point corresponds to a different allele . The mean is shown in red . Red dots indicate possible formation of an unidentified form involving G2-4 . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 013 Our previous structure-function analysis of the CEB25 minisatellite showed that short pyrimidine loop-bearing G4s were most prone to induce genomic instability , in correlation with the structure thermal stability ( Piazza et al . , 2015 ) . Consistent with this preferential folding bias , it is remarkable that all the forms inducing CEB1 instability ( Form 1 , 2 , and 3 ) bear single/zero nucleotide loops ( Figure 4A ) , and that increasing a loop by a single nucleotide has such a profound effect on the array instability ( Figure 2—figure supplement 1 ) . To investigate the relationship between the in vivo instability of the CEB1 mutants and their in vitro thermal stability , we measured the melting temperature ( Tm ) of most CEB1 variant sequences by UV-spectroscopy in heating/cooling experiments ( Materials and methods ) . The results reported in Table 1 illustrate large differences in Tm , ranging from 71°C ( CEB1-WT ) to 35°C ( CEB1- ( G3 , 24–25 ) T ) . According to the G4 forms , the G11T mutant ( Form 1 + 2 ) has a Tm similar to CEB1-WT while all the other forms , namely Form 2 alone ( G ( 3 , 9–11 ) T ) , Form 3 ( G ( 3 , G24-25 ) T ) and Form 4 ( G21T ) exhibit low Tms ( 46°C or less ) ( Figure 5A ) . This indicates that the most prominent G4 involved in the in vivo instability has the highest Tmin vitro . More extensively , Figure 6A illustrates the relationship between the Tm of all the variants constructed in the present study and their level of instability in the WT+Phen-DC3 and pif1Δ contexts . Although a clear correlation can be established in both contexts , few notable outliers were observed , especially upon Phen-DC3 treatment ( Figure 6A and Figure 6—figure supplement 1 ) . Among the outliers , we noted that Form 4 did not induced instability despite a Tm similar to Form 2 and higher than Form 3 ( Figures 5A and 6A ) . This is consistent with our previous findings that a single loop ≥4 nt was sufficient to stabilize the array independently of the Tm ( Piazza et al . , 2015 ) . We also noted that the alleles bearing isolated Forms 2 and 3 exhibited significant instabilities in the presence of Phen-DC3 but little to none in the absence of Pif1 ( Table 1 and Figure 5 ) . To address the hypothesis that this difference results from an enhanced stabilization of these forms by Phen-DC3 , we compared the melting temperature of the isolated structures in the presence or absence of Phen-DC3 by CD spectroscopy in heating/cooling experiments ( Figure 6B ) . Indeed , we found that Phen-DC3 stabilized all forms , but to varying degrees . Namely the ΔTm for Forms 1 , 2 and 4 were similarly increased by 13 . 3°C to 17 . 7°C but Form 3 exhibited a large ΔTm of +45 . 6°C ( Figure 6B ) . The consequence of this differential effect is that Form 1 and 3 now reach a Tm of >80°C while Form 2 and 4 have a Tm hovering at ~60°C . Altogether , these results indicate that the high level of G4-induced CEB1 instability correlated with the high thermal stability of the corresponding G4 in vitro . Importantly , it revealed that Phen-DC3 differentially increases the Tm of the various G4s , allowing explaining certain quantitative discrepancies observed in the WT+Phen-DC3 and pif1Δ contexts ( see Discussion ) . 10 . 7554/eLife . 26884 . 014Figure 6 . Correlation of CEB1 instability and G4 thermal stability . ( A ) CEB1 variant instabilities correlate with the thermal stabilities of their associated G4s as determined by UV-melting , in both WT cells treated with Phen-DC3 ( left ) and in pif1Δ cells ( right ) . ( B ) Phen-DC3 differently stabilizes the isolated G4s resulting from CEB1 , as determined by CD-melting . Arrows indicate the ΔTm . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 01410 . 7554/eLife . 26884 . 015Figure 6—figure supplement 1 . Correlation between CEB1 variant instabilities upon Phen-DC3 treatment and PIF1 deletion . Genomic instabilities measured in pif1Δ strains and in Phen-DC3-treated WT strains are significantly correlated . Notable deviations are highlighted . Instabilities are represented as a percentage of CEB1-WT instability . DOI: http://dx . doi . org/10 . 7554/eLife . 26884 . 015 The various CEB1 G-tracts involved in CEB1 instability in vivo could be involved in three different G4 conformations determined in vitro: Forms 1 to 3 . These conformations exhibited non-canonical structural features , such as a V-shaped loop and snapback guanine for Form 1 , G-triad for Form 2 , and zero-nt loop for Form 3 . Hence , our structure-function analysis provides evidence for the existence of non-canonical G4s in vivo and their involvement in inducing high levels of genomic instability , as observed for the canonical CEB25-G4 motif composed of 4 G-triplets and loops of 1 nt ( Piazza et al . , 2015 ) . Among the numerous structural studies of G4s , a 0-nt loop , such as in Form 3 , were previously reported for the VEGF aptamer , although encompassing only 2 quartets ( Marušič et al . , 2013 ) ; snapback guanines occupying the vacant slot in outermost quartets such as in Form 1 were observed in the G4 of the c-MYC , c-KIT and PDGFRβ promoters ( Phan et al . , 2007 , Phan et al . , 2005 ) and the associated V-shaped loop was reported for the CHL1 intronic G4 ( Kuryavyi and Patel , 2010 ) ; Finally , G-triads such as in Form 2 have been reported for synthetic model sequences and derivatives of the human MYOG G4 ( Heddi et al . , 2016; Li et al . , 2015 ) . Interestingly , in addition to nearby guanines , the vacant slot of a triad can be filled up by freely diffusing guanine derivatives ( Park et al . , 2016 ) . As previously proposed ( Heddi et al . , 2016; Li et al . , 2015 ) , G-triad-containing G4s bear the unique property of being ( deoxy ) riboswitches sensitive to the abundance of guanine derivatives , which act as endogenous stabilizing ligands . This adds to the list of sensors and switch functions for G4s in vivo , as proposed for temperature ( Wieland and Hartig , 2007 ) and salt concentration ( Subramanian et al . , 2011 ) . A noteworthy observation regards the additivity for full CEB1 instability of the contributing Forms 2 and 3 in the presence of Phen-DC3 , and of Form 1 with an unidentified form in the absence of Pif1 ( Figure 5B ) , suggesting the lack of interference or cooperation between forms in the array . This absence of competition can be explained if each motif has an overall low propensity to fold into a G4 at the time required to interfere with replication ( Lopes et al . , 2011 ) , implying that G4 formation is a limiting step in CEB1 instability . By systematically determining the thermal stability of motif variants , and by varying lengths and sequences of single purine loops involved in all forms ( A19 ) , we confirmed with the complex CEB1 sequence that loop sequence and G-tract proximity are determinants of G4-induced array genomic instability , in correlation with G4 thermal stability . This correlation was more robust in the pif1Δ than in the Phen-DC3-treated context ( Figure 6A ) , due to notable deviations affecting Forms 2 and 3 upon Phen-DC3 treatment . These differences can be explained by the disproportionate stabilization of these forms by Phen-DC3 compared to other forms as previously reported for several other canonical G4s ( De Cian et al . , 2007 ) such as the CEB25-G4 and variants ( +9 to +14°C ) ( Piazza et al . , 2015 ) . Phen-DC3 is a universal G4 binder due to its recognition of an exposed G-quartet ( Chung et al . , 2014 ) . Consequently , its disproportionately high stabilization of certain non-canonical G4 likely results from the overcoming of the outmost quartet lability ( due to the 0-nt loop in Form 3 and the G-triad in Form 2 ) thanks to π-stacking interactions rather than from a differential G4 recognition . Thus , change in relative thermal stability induced by Phen-DC3 explains the discrepancies observed between the Phen-DC3 and the pif1Δ contexts for isolated forms , whereby the most stable forms induce instability in both contexts . These differences were not revealed with the canonical CEB25 motif and variants because of their monomorphic structures , all evenly stabilized by Phen-DC3 ( Piazza et al . , 2015 ) . Hence , according to the G4 motif , G4-ligands have the potential to exacerbate effects of otherwise more labile non-canonical G4s ( Chambers et al . , 2015 ) . In our case , the discordant behavior of CEB1-G12T and CEB1-G ( 12 , 25 ) T , which induces more instability that CEB1-WT in the Phen-DC3-treated context yet almost abolishes the instability in the Pif1-deficient context , suggests that the G12T mutation causes the formation of an uncharacterized new form that might be strongly stabilized by Phen-DC3 . As outlined in the Introduction , a new generation of G4 prediction algorithms that takes into account non-canonical G4s re-evaluated to ~700 , 000 the number of potential G4 sequences in the human genome ( Bedrat et al . , 2016 ) . The evidence for the existence of non-canonical G4s in cells revealed here gives relevance to this increased figure . However , how many of these potential G4 sequences actually form or exert a given biological function in cells remains uncertain . Indeed , a ChIP-Seq experiment using a G4-specific antibody identified only ~10 , 000 regions in human genomes , specifically enriched at nucleosome-depleted regions ( Hänsel-Hertsch et al . , 2016 ) . Similarly , our studies point at only a subset of the potential G4 sequences as being ‘at risk’ for genomic instability . The innocuousness of other sequences could either be because G4 with longer loops did not form in vivo or because they failed to interfere with leading-strand replication ( Piazza et al . , 2015 ) . This complex in vivo situation , affected both by the local context such as nucleosome occupancy , the presence of G4-stabilizing ligands , and likely dependent of the biological processes under scrutiny , argues against a ‘one-fits-all’ G4 prediction algorithm for genome data mining . With regards to genomic instability , our study identifies a subset of the most compact and stable canonical and non-canonical G4s that bears the biophysical properties required to form and hinder leading strand replication . Synthetic complete ( SC ) and Yeast-Peptone-Dextrose ( YPD ) media have been prepared according to standard protocols ( Treco and Lundblad , 2001 ) . Liquid SC media containing Phen-DC3 at 10 μM have been prepared as previously described ( Piazza et al . , 2010 ) . Relevant genotypes of the haploid Saccharomyces cerevisiae strains used in this study are listed in Supplementary file 1A . They are derived from SY2209 ( W303 RAD5+ background ) ( Fachinetti et al . , 2010 ) by Lithium-Acetate transformation ( Lopes et al . , 2011 ) . The CEB1-WT-25 ( Ribeyre et al . , 2009 ) , CEB1-G ( 9 , 10 , 14 ) T , CEB1-G ( 10-13 ) T , CEB1-G11T , CEB1-G ( 11 , 13 ) T , CEB1-G17T , CEB1-G21T , CEB1-G25T , and CEB1-G ( 25-24 ) T minisatellites have been synthesized and Sanger sequenced using a custom-made PCR-based method described previously ( Ribeyre et al . , 2009 ) . Minisatellites of similar size ( 20–26 motifs ) have been retained . The CEB1-G ( 3 , 9–11 ) T , CEB1-G ( 9-11 ) T , CEB1-G ( 9-11 ) T , G25T , CEB1-G ( 31-32 ) T , and CEB1-G ( 34-35 ) T alleles of 24 motifs have been synthesized and Sanger sequenced by GeneCust . The CEB1-WT-10 and CEB1-WT-20 ( of 10 and 20 motifs , respectively ) , CEB1-G3T , CEB1-G ( 3 , 24–25 ) T , CEB1-G ( 9-10 ) T , CEB1-G ( 9-10 , 24-25 ) T , CEB1-G12T , CEB1-G13T , CEB1-G14T and CEB1-G ( 14 , 24–25 ) T alleles ( 24 motifs ) and CEB1-A19T , CEB1-A19TT , and CEB1-A19TTT alleles ( 20 motifs ) have been synthesized and Sanger sequenced by GenScript . All minisatellites have been inserted at the same location and in the same orientation in the intergenic region between YCL048w and YCL049c ( chrIII:41801–41840 , yielding a small deletion of 39 bp ) in the vicinity of ARS305 as described previously ( Lopes et al . , 2011 ) . Briefly , transformation of a marker-less minisatellite fragment containing the appropriate flanking sequences replaced the URA3-hphMX cassette present at this location in the parental strain , allowing for the selection of the transformants ( 5FOA-resistant and Hygromycin-sensitive ) . The G-rich strand of CEB1 is on the Crick strand ( e . g . template for the leading strand replication machinery of forks emanating from ARS305 , Figure A in reference [Lopes et al . , 2011] ) . CEB1-G3T contracted to 23 motifs upon insertion in the yeast genome , CEB1-G ( 3 , 24–25 ) T contracted to 22 motifs , CEB1-G14T , CEB1-G ( 9-10 , 24-25 ) T and CEB1-G ( 14 , 24–25 ) T all contracted to 20 motifs , CEB1-G ( 34-35 ) T contracted to 21 motifs , and CEB1-A19T contracted to 19 motifs . Minisatellite instability during vegetative growth has been measured in WT , WT Phen-DC3- and pif1Δ cells as previously described ( Ribeyre et al . , 2009 ) ( Lopes et al . , 2011 ) . Briefly , untreated WT cells and pif1Δ cells from a fresh patch of cells are diluted at a concentration of 2 × 105 cells/mL in 5 mL of YPD , grown at 30°C with shacking for eight generations , spread as single colonies on YPD plates , and incubated at 30°C . The instability measurement in these cells thus corresponds to the rearrangement frequency after 35 generations . Between 48 and 192 colonies from these patches were analyzed ( see below for sample size determination ) in a single experiment . In rare instances in which an early clonal ‘jackpot’ event was present in the starting colony , we analyzed an independent patch . To measure minisatellite instability upon Phen-DC3 treatment , cells from a fresh patch were grown for 8 generations at 30°C in liquid SC containing 10 μM Phen-DC3 ( Lopes et al . , 2011 ) . Isolated colonies or pools of colonies are analyzed by Southern blot upon digestion with EcoRI that cut at each side of the minisatellite , leaving a total of 18 nt of flanking sequence . The membranes are hybridized with the Phage lambda DNA ( HindIII/EcoRI digested ladder , Promega ) and the appropriate CEB1-WT or variant probes . The signals are detected with a Typhoon Phosphorimager and quantified using ImageQuant 5 . 2 ( Molecular Dynamics ) . The elimination of secondary rearrangements ( that occurred early in the colony after plating ) and of potential early clonal events in the culture has been performed as described in ( Lopes et al . , 2011 ) . We used G*Power to compute sample size , with a α cutoff set at 0 . 05 . Given the usual range of CEB1 instability observed in untreated and Phen-DC3 treated WT cells ( based on previous studies ) , and to be able to detect instabilities of at least 5% with an α cutoff of 0 . 05 and a β power of 0 . 9 , we set the sample size at 192 colonies . This sample size also allows detecting a 2-fold decrease of instability compared to the reference CEB1-WT allele in the Phen-DC3 context . Regarding the CEB1 instability in PIF1-deleted cells , we knew from previous study that the range of instabilities was much wider , reaching very high levels ( up to 60% for CEB1-WT ) . These high instabilities prevented colony pooling for Southern blot analysis . The level of instability was hinted at upon verification of the transformants by Southern blot . Hence , when the instability was expected in the high range , we chose to sample 48 colonies , a reasonable compromise which allows detecting instability levels of 15% compared to WT cells , and instabilities 2-fold lower than in CEB1-WT with a β power of 0 . 9 . For predicted intermediate instabilities we analyzed 96 colonies , or otherwise 192 colonies as in the WT context . Unlabeled and site-specific labeled DNA oligonucleotides ( Supplementary file 1B ) were chemically synthesized on an ABI 394 DNA/RNA synthesizer . Samples were purified and dialyzed successively against 25 mM potassium chloride solution and water . Unless otherwise stated , DNA oligonucleotides were dissolved in solution containing 70 mM potassium chloride and 20 mM potassium phosphate ( pH 7 . 0 ) . DNA concentration was expressed in strand molarity using a nearest-neighbor approximation for the absorption coefficients of the unfolded species ( Cantor et al . , 1970 ) . The molecular size of the structures formed by DNA oligonucleotides was visualized by non-denaturing polyacrylamide gel electrophoresis ( PAGE ) ( Guédin et al . , 2008 ) . DNA samples were incubated in a 20 mM potassium phosphate buffer ( pH 7 . 0 ) before loading on 20% polyacrylamide gels supplemented with variable concentration of potassium chloride and run at 26°C; 40% sucrose was added before loading . Circular dichroism ( CD ) spectra were recorded on a JASCO-810 spectropolarimeter using 1 cm path length quartz cuvettes with a reaction volume of 600 µL . The DNA oligonucleotides ( ~5 µM ) were prepared in a 20 mM potassium phosphate buffer ( pH 7 . 0 ) containing 70 mM potassium chloride . For each experiment , an average of three scans was taken , the spectrum of the buffer was subtracted , and the data were zero-corrected at 320 nm . The thermal difference spectra ( TDS ) were obtained by taking the difference between the absorbance spectra of unfolded and folded oligonucleotides that were respectively recorded much above and below its melting temperature . TDS provide specific signatures of different DNA structural conformations ( Mergny et al . , 2005 ) . Spectra were recorded between 220 and 320 nm on a JASCO V-650 UV/Vis spectrophotometer using 1 cm pathlength quartz cuvettes . The DNA oligonucleotides ( ~5 µM ) were prepared in a 20 mM potassium phosphate buffer ( pH 7 . 0 ) containing 70 mM potassium chloride . For each experiment , an average of three scans weve taken , and the data were zero-corrected at 320 nm . The thermal denaturing of the CEB1-WT and its mutants was performed on JASCO UV/VIS V-650 spectrophotometer or on a CD by monitoring the UV absorption ( at 290 nm wavelength ) or the CD ellipticity ( at 260 nm wavelength ) . Prior to melting experiments , DNA samples ( 5–10 μM ) were annealed in a buffer containing 10 mM potassium chloride and 10 mM potassium phosphate ( pH 7 ) . All melting experiments were performed using the protocols described in ( Piazza et al . , 2015 ) . Melting experiments with Phen-DC3 were conducted at DNA:Phen-DC3 ratio of 1:1 . NMR experiments were performed on 600 MHz and 700MHz Bruker spectrometers at 25°C , unless otherwise specified . The strand concentration of the NMR samples was typically 0 . 2–1 . 5 mM in near-physiological conditions ( 100 mM K+ solution at pH 7 ) . Resonances for guanine residues were assigned unambiguously by using site-specific low-enrichment 15N labeling ( Phan and Patel , 2002 ) , site-specific 2H labeling ( Huang et al . , 1997 ) , and through-bond correlations at natural abundance ( Phan , 2000 ) . Spectral assignments were completed by NOESY , TOCSY , {13C-1H}-HMBC and {13C-1H}-HSQC as previously described ( Phan et al . , 2001 ) . Inter-proton distances were deduced from NOESY experiments at various mixing times . All spectral analyses were performed using the FELIX ( Felix NMR , Inc . ) program . Sample size determination for instability measurement are described in the ‘Measurement of minisatellite instability’ section . The rearrangement frequencies have been compared using a two-tailed Fisher exact test using R x64 3 . 2 . 0 ( Team , 2008 ) . A non-parametric Spearman correlation test has been used to compare thermal stability of the G4 variants and the associated CEB1 allele instabilities . In all cases , the α-cutoff for significance has been set to 0 . 05 .	Molecules of DNA encode the information needed to build cells and keep them alive . DNA is made of two strands that contain several different chemical groups known as bases arranged in different orders , like letters and words in a phrase . Generally , two DNA strands wrap around each other to make a three dimensional structure known as a double helix . However , in certain circumstances , some sequences of DNA bases can adopt alternative structures . For example , DNA sequences that contain lots of a base known as guanine may sometimes form structures called G-quadruplexes in which sets of four guanines come together . G-quadruplexes are involved in many processes in cells including regulating the activity of genes , but they can also interfere with the process that replicates the DNA at each generation . This causes the cell’s genetic information to be modified , which can damage the cell and can promote cancer . However , it is difficult to predict which DNA sequences are susceptible to form G-quadruplexes and what consequence their folding might have on the biological processes happening in cells . Recent computational and biophysical studies have shown that G-quadruplexes can form a larger variety of structures than previously known . Piazza et al . studied how some of these new “non-canonical” structures form in yeast cells and how they may interfere with DNA copying . The experiments show that a single guanine-rich DNA sequence can form several types of non-canonical G-quadruplex structures in yeast cells . This includes structures that do not have complete sets of guanines at their center or are missing loops that connect the bases to one another . Further experiments demonstrate that the threat posed by these G-quadruplexes is linked to the length of their connecting loops and how well their three-dimensional structures withstand heat . The findings of Piazza et al . identify a set of DNA sequences that are capable of forming G-quadruplexes that harm the cell . The next challenge will be to develop specific molecules that can stabilize the structures of G-quadruplexes . In the future , this avenue of research may aid the development of new treatments for cancer that target specific DNA structures .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "structural", "biology", "and", "molecular", "biophysics" ]	2017	Non-Canonical G-quadruplexes cause the hCEB1 minisatellite instability in Saccharomyces cerevisiae
Previous studies of laboratory strains of budding yeast had shown that when gene copy number is altered experimentally , RNA levels generally scale accordingly . This is true when the copy number of individual genes or entire chromosomes is altered . In a recent study , Hose et al . ( 2015 ) reported that this tight correlation between gene copy number and RNA levels is not observed in recently isolated wild Saccharomyces cerevisiae variants . To understand the origins of this proposed difference in gene expression regulation between natural variants and laboratory strains of S . cerevisiae , we evaluated the karyotype and gene expression studies performed by Hose et al . on wild S . cerevisiae strains . In contrast to the results of Hose et al . , our reexamination of their data revealed a tight correlation between gene copy number and gene expression . We conclude that widespread dosage compensation occurs neither in laboratory strains nor in natural variants of S . cerevisiae . Losses or gains of whole chromosomes , a condition known as aneuploidy , have a profound impact on cell physiology . Gene expression studies in budding yeast , fission yeast , mammalian cells , and plants revealed that this is due to the fact that changes in gene copy number result in changes in gene expression ( Chikashige et al . , 2007; Huettel et al . , 2008; Pavelka et al . , 2010; Sheltzer et al . , 2012; Stingele et al . , 2012; Torres et al . , 2010; 2007 ) . For example , in haploid budding yeast strains harboring single chromosome gains , RNA levels of more than 90% of genes located on the extra chromosome reflect the increased gene copy number ( Dephoure et al . , 2014; Torres et al . , 2007 ) . Only few genes , such as histone and some ribosomal genes defy this trend ( Dabeva and Warner , 1987; Gunjan and Verreault , 2003; Libuda and Winston , 2006; Moran et al . , 1990; Sutton et al . , 2001; Vilardell and Warner , 1997 ) . Given that aneuploidy has such a profound impact on the cell’s transcriptome and proteome it is not surprising that aneuploidy affects virtually all aspects of cell physiology , generally having a negative impact on fitness ( Hassold and Jacobs , 1984; Hodgkin , 2005; Huettel et al . , 2008; Lindsley et al . , 1972; Niwa et al . , 2006; Stingele et al . , 2012; Torres et al . , 2007; Williams et al . , 2008 ) . Aneuploidy not only affects gene expression through changes in gene copy number , the condition also causes transcriptional responses . For example , when chromosome gains or losses lead to a decrease in growth rate , a stereotypic slow-growth transcriptional response known as the environmental stress response ( ESR ) ensues ( Gasch et al . , 2000 ) . The ESR is characterized by the down-regulation of growth-promoting genes and the up-regulation of stress response genes and has been reported to occur in response to aneuploidy in many organisms including laboratory yeast strains ( Sheltzer et al . , 2012 ) . Changes in gene copy number not only can lead to transcriptional responses but also can elicit dosage compensation , a gene regulatory mechanism that specifically compensates for alterations in gene copy number at the gene expression level . Dosage compensation is best understood in the context of sex chromosome-encoded genes ( reviewed in Straub and Becker , 2007 ) . For example in mammals , an RNA-mediated mechanism silences expression of one copy of the X chromosome in females thereby equalizing X chromosome-encoded gene expression between males and females ( Lee and Bartolomei , 2013 ) . In Caenorhabditis elegans , gene expression of the two X chromosomes is reduced by half in the hermaphrodite to match the expression of the single X chromosome in males ( Meyer , 2010 ) . Dosage compensation can also affect specific loci . The perhaps best known example is the histone locus in budding yeast ( Osley and Hereford , 1981 ) . When an extra copy of the HTA1 gene ( histone H2A ) is introduced into budding yeast , mRNA turnover increases resulting in normal HTA1 transcript levels ( Moran et al . , 1990; Osley and Hereford , 1981 ) . It is important to note that dosage compensation and transcriptional responses to aneuploidy can have the same effect on a gene . Both can cause the down-regulation of a gene , but the mechanisms are distinct . Transcriptional responses to aneuploidy are elicited by an aneuploid genome affecting a biological pathway and are not restricted to the aneuploid chromosomes but impact expression of genes located throughout the genome . In contrast , dosage compensation specifically alters the expression of a gene whose copy number has been varied and its effects are thus restricted to the aneuploid chromosome . Experimental evolution studies suggest that selective pressures cause changes in karyotype such as chromosome gains or losses ( Dunham et al . , 2002; Gresham et al . , 2008 ) . However , such aneuploidies are usually transient evolutionary intermediates that , given time , are replaced with more optimal solutions ( Yona et al . , 2012 ) . A key question that arises from these studies is how prevalent whole chromosome gains and losses are in wild yeast strains and how aneuploidies affect cell physiology . Hose et al . ( 2015 ) addressed these questions . They isolated 47 wild yeast strains to identify 12 ( 26% ) that harbored whole chromosome gains and/or losses . The detailed analysis of six of these strains led them to the conclusion that aneuploidies are prevalent , stable and well-tolerated in wild yeast strains . Based on gene expression analyses , they further concluded that tolerance to aneuploidy is caused by dosage compensation mechanisms that buffer gene amplifications thereby protecting cells against the adverse effects of aneuploidy . They reported that gene-dosage compensation functions at >30% of amplified genes . To understand why dosage compensation mechanisms are rare in laboratory strains of budding yeast , but highly prevalent in wild isolates , we reevaluated the karyotype and gene expression studies performed by Hose et al . ( 2015 ) . This reexamination revealed that gene copy number and expression are tightly correlated in wild S . cerevisiae strains . We conclude that dosage compensation is a rare occurrence in both , laboratory and natural variants of S . cerevisiae . Hose et al . ( 2015 ) isolated 47 wild yeast variants and determined their karyotypes by inferring the copy number from genome sequencing data using depth of coverage . This analysis showed that 12 of these 47 strains harbor whole chromosome aneuploidies . DNA and RNA sequencing data for 6 of these 12 aneuploid strains were deposited in the NCBI Sequence Read Archive ( SRA ) under accession SRP047341 and NCBI Gene Omnibus under accession GSE61532 referenced in Hose et al . ( 2015 ) . Three of these strains harbored one or two single chromosome gains in a diploid background . Strain K9 is a diploid strain carrying an extra copy of chromosomes IX and X ( Figure 1A , G ) , YPS1009 is diploid with an extra copy of chromosome XII ( Figure 1B , G ) , and diploid strain NCYC110 carries two extra copies of chromosome VIII ( Figure 1C , G ) . In addition , Hose et al . ( 2015 ) analyzed three strains with high levels of aneuploidy . These strains were strains YJM428 , Y2189 and K1 ( Figure 1D–G ) . 10 . 7554/eLife . 10996 . 003Figure 1 . DNA and RNA copy number of six wild S . cerevisiae strains . ( A ) DNA and RNA copy number analysis of strain K9 compared to K10 . Log2 ratios of aneuploid vs . euploid DNA in the order of the chromosomal location of their encoding genes are shown on the top . DNA copy number of chromosomes IX and X are shown in red . The graph below shows the average DNA copy number per chromosome . The graph below shows RNA copy number averaged per chromosome relative to K10 ( n = 1 ) . ( B ) DNA and RNA copy number analysis of strain YPS1009 compared to YPS163 . Data are represented as in ( A ) . Error bars represent the SD of the chromosome means from three biological replicates . Medians are identical to the means . ( C ) DNA and RNA copy number analysis of strain NCYC110 compared to NCYC3290 . Data are represented as in ( A ) . Error bars represent the SD of the chromosome means from three biological replicates . Medians are identical to the means . ( D ) DNA and RNA copy number analysis of strain YJM428 compared to YJM308 . Log2 ratios of aneuploid vs . euploid DNA in the order of the chromosomal location of their encoding genes are shown on the top . DNA copy number of chromosomes XII and XVI are shown in red . Arrows indicate an amplification of part of chromosome III ( red ) and a loss of part of chromosome XV ( green ) . The graph below shows the average DNA copy number per chromosome relative to strain YJM308 . The graph below shows RNA copy number averaged per chromosome . Error bars represent the SD of the chromosome means from two biological replicates . Medians are identical to the means . Asterisk indicate significant deviations from the expected value as determined by a one sample t-test ( p < 0 . 01 ) . ( E ) DNA and RNA copy number analysis of strain Y2189 compared to Y2209 . Data are represented as in ( D ) . Error bars represent the SD of the chromosome means from two biological replicates . Medians are identical to the means . Asterisk indicate significant deviations from the expected value as determined by a one sample t-test ( p < 0 . 01 ) . Note that chromosome IV shows increased RNA copy number relative to DNA copy number . ( F ) DNA and RNA copy number analysis of strain K1 compared to K10 . Data are represented as in ( D ) . Asterisk indicate significant deviations from the expected value as determined by a one sample t-test ( p < 0 . 01 ) . Note that chromosomes I and VI exhibit an increased copy number at the DNA level but not at the RNA level . ( G ) Gene expression of six aneuploid strains ordered by chromosome . Experiments ( columns ) of two biological replicates for strains YJM428 and Y2189 , three biological replicates for strains YPS1009 and NCYC110 , and one experiment for strains K1 and K9 are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 003 We examined the karyotypes and gene expression of these strains and found the aneuploid strains K9 , YPS1009 and NCYC110 with low levels of aneuploidy to harbor relatively stable karyotypes ( Figure 1A–C ) . As discussed in more detail below , RNA levels also generally correlated well with DNA levels , with aneuploid chromosomes overall showing a corresponding increase in gene expression ( Figure 1A–C ) . It is , however , noteworthy that strain K9 which harbors extra copies of chromosome IX and X in the Hose et al . ( 2015 ) study was previously reported to be trisomic for chromosome IX only , indicating that this strain exhibits some karyotypic instability ( Kvitek et al . , 2008 ) . In contrast to the relatively stable strains K9 , YPS1009 , and NCYC110 , a different picture emerged from our analysis of strains YJM428 , Y2189 , and K1 that harbor complex karyotypes . Based on the presence of non-integer DNA copy number states , we conclude that the described aneuploidies are only present in subpopulations of cells ( Figure 1D–F ) . The comparison between RNA and DNA levels further revealed significant inconsistencies between the two data sets indicating that some strains had changed their karyotypes between the two analyses ( e . g . DNA and RNA copy numbers are very different in strains Y2189 and K1; Figure 1E , F ) . This discrepancy is problematic as Hose et al . ( 2015 ) used the standard deviations ( SD ) of the DNA measurements to establish cutoffs in their RNA data set to identify dosage compensated genes ( discussed in detail below ) . We also analyzed the karyotypes of the other six aneuploid variants UC5 , WE372 , T73 , Y3 , Y6 , and CBS7960 that were not characterized in detail by Hose and coworkers ( both Figure 1 and Supplementary file 1 in Hose et al . , 2015; log2 ratios of normalized DNA copy numbers were provided by A . Gasch ) . We found that strains T73 , which is tetrasomic for chromosome VIII ( analyzed below; Figure 4A ) , and WE372 , which is trisomic for chromosome I to harbor stable karyotypes ( Figure 2A ) . However , DNA copy numbers in strains UC5 , Y3 , Y6 , and CBS7960 exhibited non-integer DNA copy number states indicating that the strains are heterogeneous ( Figure 2B–E ) . 10 . 7554/eLife . 10996 . 004Figure 2 . Karyotypes of aneuploid wild S . cerevisiae strains Y3 , Y6 UC5 , CBS7960 , and WE372 and euploid control strains . ( A–E ) Relative DNA copy of WE372 ( A ) , Y3 ( B ) , Y6 ( C ) , UC5 ( D ) , and CBS7960 ( E ) compared to S288C . Log2 ( aneuploid vs . euploid DNA ) per gene relative ( top ) are shown in the order of the chromosomal location of their encoding genes . DNA copy numbers of amplified chromosomes are shown in red . Bar graphs ( bottom ) represent the DNA copy numbers averaged per chromosome . Asterisks indicate significant deviations from expected integral value using one sample t test ( p < 0 . 01 ) . ( F–G ) Relative DNA copy of K10 ( F ) , YJM308 ( G ) , and Y2189 ( H ) compared to S288C . Log2 ratio ( aneuploid vs . euploid DNA ) per gene are shown in the order of the chromosomal location of their encoding genes . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 00410 . 7554/eLife . 10996 . 005Figure 3 . RNA levels correlate with DNA copy number in wild and laboratory strains of S . cerevisiae . ( A ) Histogram of the log2 ratios of the RNA copy number of genes located on euploid chromosomes ( left panel , strains YPS1009 , NCYC110 , and K9 ) , genes present on trisomic chromosomes ( 3n , middle panel , YPS1009 , and K9 ) , and genes present on tetrasomic chromosomes ( right panel , NCYC110 ) , relative to euploid controls are shown . Bin size for all histograms is log2 ratio of 0 . 2 , medians are identical to means . Fits to a normal distribution ( black line ) , means and goodness of fit ( R2 ) and skewness are shown for each distribution . ( B ) The average log2 ( aneuploid vs . euploid RNA ) of triplicated genes plotted against average log2 ( aneuploid vs . euploid DNA ) in strains YPS1009 and K9 . Histogram of the log2 ratios of the DNA copy number is shown in red ( mean log2 ratio = 0 . 57 , SD = 0 . 14 , R2 = 1 . 0 , skewness = 0 . 00 ) . Histogram of the log2 ratios of the RNA copy number of is shown in blue ( median = mean = 0 . 55 , skewness = 0 . 02 ) . Fits to a normal distribution are shown ( black line ) . Numbers of genes that show RNA copy numbers lower or higher than 1 or 2 SD from the mean are shown ( separated by dotted lines ) . ( C ) Histogram of the log2 ratios of the RNA copy number of genes located on euploid chromosomes ( left panel ) , and genes present on duplicated chromosomes ( right panel ) in two disomic laboratory strains ( disome V and XVI ) relative to the euploid W303 control are shown . Bin size for all histograms is log2 ratio of 0 . 2 , medians are identical to means . Fits to a normal distribution are shown ( black line ) . Means , goodness of fit ( R2 ) and skewness are shown for each distribution . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 00510 . 7554/eLife . 10996 . 006Table 1 . DNA and RNA copy number of six wild S . cerevisiae strains . The columns describe the following parameters: Column 1: Strain name . Column 2: Identity of chromosomes amplified in each strain . Euploid represents the combined data of all euploid chromosomes in a given strain . Column 3: Reported chromosome copy number . Column 4: Number of genes quantified by RNA-seq . Column 5: Mean of the normalized log2 ratios ( aneuploid vs . euploid RNA ) . Column 6: Standard deviation ( SD ) of the normalized log2 ratios ( aneuploid vs . euploid RNA ) . Column 7: Mean of the normalized log2 ratios ( aneuploid vs . euploid DNA ) . Column 8: Standard deviation ( SD ) of the normalized log2 ratios ( aneuploid vs . euploid DNA ) . Column 9: Number of genes whose values are below two SD from the mean . Column 10: Number of genes whose values are above two SD from the mean . Column 11: Cutoff used by Hose et al . ( 2015 ) to identified genes that are dosage compensated . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 006123 45 6 7 8 9 10 11 STRAIN Chr Copy number Genes RNA Mean RNA SD DNA Mean DNA SD Number of genes RNA <2SD Number of genes RNA >2SD Cutoffs by Hose et alYJM428 -1 XII 3 525 0 . 52 0 . 63 15 XVI 4 485 0 . 95 0 . 66 9 17 Euploid 2 5087 −0 . 01 0 . 72 116 169 YJM428-2 XII 3 533 0 . 54 0 . 70 0 . 60 0 . 22 10 18 N/A XVI4 490 0 . 92 0 . 63 0 . 96 0 . 23 11 18 N/A Euploid 2 5160 −0 . 01 0 . 72 0 . 00 0 . 28 75 183 Aneuploid genes 9 ( 1% ) 14 ( 1% ) Euploid genes 36 ( 1 % ) 77 ( 1% ) Y2189-1 I 4 88 0 . 77 0 . 89 3 4 III 3 170 0 . 60 0 . 88 5 3 IX 3 216 0 . 42 0 . 91 5 9 XI 3 325 0 . 37 0 . 89 3 8 Euploid 5209 0 . 05 0 . 76 104 204 Y2189-2 I 4 89 0 . 63 1 . 01 1 . 05 1 . 04 4 3 0 . 21 III 3 167 0 . 53 0 . 90 0 . 53 0 . 55 5 6 0 . 24 IX 3 214 0 . 37 0 . 96 0 . 45 0 . 48 5 9 N/A XI 3 324 0 . 46 0 . 65 0 . 47 0 . 25 3 10 0 . 13 Euploid 2 5231 0 . 06 0 . 77 0 . 00 0 . 43 142 165 Aneuploid genes 9 ( 1% ) 15 ( 2% ) Euploid genes 50 ( 1% ) 124 ( 2% ) YPS1009-1 XII 3 511 0 . 53 0 . 62 13 27 Euploid 2 5482 0 . 00 0 . 57 132 136 YPS1009-2 XII 3 520 0 . 49 0 . 73 16 20 Euploid 2 5531 0 . 00 0 . 60 145 119 YPS1009-3 XII 3 521 0 . 56 0 . 66 0 . 62 0 . 24 11 31 0 . 10 Euploid 2 5532 0 . 00 0 . 56 0 . 00 0 . 31 130 180 Aneuploid genes 5 ( 1% ) 5 ( 1% ) Euploid genes 46 ( 1% ) 27 ( 0% ) NCYC110-1 VIII 4 288 0 . 97 0 . 61 3 14 Euploid 2 5806 0 . 00 0 . 62 69 274 NCYC110-2 VIII 4 294 0 . 93 0 . 59 4 14 Euploid 2 5919 0 . 00 0 . 61 60 247 NCYC110-3 VIII 4 292 0 . 98 0 . 58 0 . 98 0 . 16 4 14 0 . 10 Euploid 2 5890 0 . 00 0 . 57 0 . 00 0 . 12 61 254 Aneuploid genes 0 ( 0% ) 5 ( 2% ) Euploid genes 7 ( 0% ) 102 ( 2% ) K1 III 4 168 0 . 63 0 . 73 0 . 98 0 . 24 3 4 0 . 45 Euploid 2 5914 0 . 00 0 . 85 0 . 00 0 . 43 153 153 Aneuploid genes 3 ( 2% ) 4 ( 2% ) Euploid genes 153 ( 3% ) 153 ( 3% ) K9 IX 3 223 0 . 51 0 . 59 0 . 55 0 . 13 10 4 0 . 24 X 3 366 0 . 55 0 . 53 0 . 55 0 . 13 9 5 0 . 18 Euploid 2 5500 0 . 00 0 . 54 0 . 00 0 . 17 185 172 Aneuploid genes 19 ( 3% ) 9 ( 2% ) Euploid genes 185 ( 3% ) 172 ( 3% ) 10 . 7554/eLife . 10996 . 007Figure 4 . DNA and RNA copy number of euploid and aneuploid isogenic wild S . cerevisiae strains . ( A ) Plots for strains YPS163-chrVIII-2n , T73-chrVIII-4n , and YJM428-chrXVI-4n , represent the log2 ratio of their relative DNA copy number compared to their isogenic and euploid counterparts . DNA copy numbers are shown in the order of the chromosomal location of their encoding genes ( left ) . DNA copy numbers of amplified chromosomes are shown in red . Bar graphs on the right represent the RNA copy numbers averaged per chromosome for aneuploid strains relative to euploid reference strains . The average RNA copies of non-amplified chromosomes are shown in black . Amplified chromosomes , as predicted by the karyotype , are shown in blue . ( B ) Gene expression of three aneuploid strains ordered by chromosome position . Experiments ( columns ) of two biological replicates are shown . ( C ) Histogram of the log2 ratios of the DNA ( top ) and RNA ( bottom ) copy number of genes located on euploid chromosomes ( left ) and genes located on duplicated chromosomes ( right ) relative to euploid controls are shown . Bin size for all histograms is log2 ratio of 0 . 2 , medians are identical to means and all distributions show a skewness of 0 . 01 . Fits to a normal distribution are shown ( black line ) and so are means and goodness of fit ( R2 ) for each distribution . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 007 Analysis of the karyotypes of the other 35 wild strains ( both Figure 1 and Supplementary file 1 in Hose et al . ( 2015 ) ) revealed that more than half of the strains harbored karyotype profiles consistent with heterogeneity . Importantly , strains K10 , YJM308 , and Y2209 utilized as the euploid reference in the gene expression analysis of the aneuploid wild strains YJM428 , Y2189 , K1 , and K9 ( Figure 2 in Hose et al . , 2015 ) appeared to harbor heterogeneous karyotypes ( Figure 2F–H ) . In particular , strain YJM308 harbors an amplification of chromosome XV and has lost part of chromosome III ( Figure 2G ) . We conclude that only 10 . 6% ( 5 out of 47 ) of the strains analyzed by Hose et al . ( 2015 ) harbor relatively stable aneuploidies that are confined to 1 – 2 chromosomes . As all strains studied by Hose et al . ( 2015 ) were derived from single colonies , our finding of significant karyotype heterogeneity indicates that a large fraction of wild yeast strains grown under standard laboratory conditions are unstable . The observed instability and heterogeneity of many wild S . cerevisiae strains makes it likely that the aneuploidies in these wild isolates are a consequence of culturing the natural variants under laboratory conditions to which they may be ill-adapted to , instead of these strains being naturally aneuploid . Caution is therefore warranted when analyzing growth rates , gene expression patterns and phenotypes of such wild yeast strains under laboratory growth conditions . In our previous studies , we found RNA and DNA levels to be well-correlated in haploid laboratory W303 strains carrying additional chromosomes ( Dephoure et al . , 2014; Torres et al . , 2007 ) . Hose et al . ( 2015 ) reported that this coordination between DNA and RNA levels was not evident in wild budding yeast isolates . Their conclusion was based on three analyses . In the first analysis , they characterized six wild aneuploid isolates; in the second , they studied three euploid-aneuploid strain pairs; and in the third analysis , they investigated two sets of strains each comprised of a series of strains with increasing aneuploidies of one particular chromosome . To begin to understand the mechanisms that could have led to the loss of dosage compensation mechanisms in laboratory strains , we reanalyzed the data generated by Hose et al . ( 2015 ) using the methods we previously employed to examine the effects of aneuploidy on gene expression in laboratory strains . Hose et al . ( 2015 ) compared mRNA levels with DNA copy number of amplified genes across six aneuploid wild yeast strains called K9 , YPS1009 , NCYC110 , YJM428 , Y2189 , and K1 and concluded that 38% ( 838 of 2 , 204 ) of amplified genes showed lower expression than predicted by their gene copy number ( light blue points in Figure 4A in Hose et al . , 2015 ) . We reevaluated their findings . Because of karyotype heterogeneity in strains YJM428 , Y2189 , and K1 ( Figure 1D–F ) , we did not reanalyze these strains except to determine the false discovery rate discussed in detail below . Strains YPS1009 and K9 are trisomic for chromosomes XII and IX+X , respectively , while NCYC110 harbors a tetrasomy of chromosome VIII . Our analysis revealed that the expression of all genes on the aneuploid chromosomes increased proportionally with gene copy number ( Figure 1A–C , 3A , B ) . As predicted by a null model with no compensation , we found that the log2 ratios of expression values of genes encoded by the triplicated chromosomes of these strains to fit a normal distribution with a mean value very close to the predicted log2 ratio of 0 . 58 ( mean log2 ratio = 0 . 55 , R2 = 0 . 99 , Figure 3A , middle panel , Figure 3B ) for the trisomic strains and a log2 ratio of 1 ( mean log2 ratio = 0 . 95 , R2 = 0 . 97 , Figure 3A , right panel ) for the tetrasomic chromosome . No skewness in the distributions - more compensating or exacerbating - was noted as would be expected if a large fraction of the genes encoded on the aneuploid chromosome were dosage compensated ( skewness = 0 . 02 ( 3n ) and 0 . 07 ( 4n ) ; Figure 3A ) . The distribution of log2 ratios of expression values of genes encoded by euploid chromosomes also fit a normal distribution with the predicted log2 ratio of 0 ( mean log2 ratio = 0 . 00 , R2 = 0 . 99 , Figure 3A left panel ) . These data are very much in line with what is observed in aneuploid laboratory strains . RNA quantification of two disomic W303 strains ( disomes V and XVI ) showed that the log2 ratios of expression values of genes encoded by the duplicated chromosomes fit a normal distribution with a mean value very close to the predicted log2 ratio of 1 ( mean log2 ratio = 1 . 03 , R2 = 0 . 98 , Figure 3C ) . To determine how many genes were potentially subject to dosage compensation , we used 2 SD from the means of the log2 ratios of each amplified chromosome and found that between 0% ( 0 gene in NCYC110 ) and 3% ( 19 genes in K9 ) of amplified genes showed values lower than expected ( Table 1 ) . Importantly , a similar number of genes was found to exhibit higher than expected expression ( between 1% in YPS1009 and 2% in NCYC110 , Table 1 ) . Using the same cutoff on the euploid chromosomes , we found between 0 . 1% ( 7 genes in NCYC110 ) and 3% ( 153 genes in K1 ) genes with values lower than expected . The nature of the distributions of gene expression patterns ( normal distribution with expected means ) and these analyses are inconsistent with high levels of dosage compensation occurring in wild yeast strains . Instead , they indicate that gene expression proportionally increases with copy number without signs of dosage compensation in wild yeast strains . The fact that the euploid chromosomes encode the same proportion of up and downregulated genes as the aneuploid chromosomes further indicates that any effects on gene expression seen in these strains are likely to be the consequence of measurement noise or a transcriptional response elicited by the aneuploid state rather than dosage compensation . To further characterize dosage compensation in wild variants Hose et al . ( 2015 ) generated a panel of isogenic euploid and aneuploid strain pairs . They isolated a disomic strain for chromosome VIII ( YPS163-chr VIII-2n ) of the euploid strain YPS163 , and euploid versions of strain T73 , which is tetrasomic for chromosome VIII ( T73-chrVIII-4n ) and of strain YJM428 , which is tetrasomic for chromosome XVI ( YJM428-chrXVI-4n ) . They then determined DNA copy number state and gene expression levels in these strains and concluded that between 11 and 36% of genes were expressed at lower than expected levels , that is , they were dosage compensated . We compared the average chromosome copy number in the three aneuploid strains with the average RNA copy number in these strains and found that RNA levels proportionally increased with DNA copy number ( Figure 4A , B ) . The aneuploidies in the three strains represent duplications . We were , therefore , able to combine the duplicated values of the DNA and RNA copy of all the three strains . The 941 duplicated genes showed a mean log2 ratio of 1 . 02 ( SD = 0 . 29 , R2 = 0 . 99 ) for DNA copy number and a nearly identical mean log2 ratio ( mean = 0 . 97; SD = 0 . 36 , R2 = 0 . 99 ) for RNA copy number ( Figure 4C ) . Furthermore , the distribution of expression values fit a normal distribution and was indistinguishable from the distribution of the gene expression values of genes encoded by the euploid chromosomes . The standard deviations of the RNA distributions were similar for euploid and aneuploid chromosomes ( Figure 4C bottom graphs ) and each distribution showed skewness of 0 . 01 and 0 . 02 , respectively . These observations indicate that the variance of the euploid genes is the same as that of the aneuploid genes . If dosage compensation were to occur , variance and skewness , would be different between genes encoded by euploid and aneuploid chromosomes . Lastly , using 2 SD as cutoff to find potential dosage compensated genes , we identified a small number of outliers . Importantly , the number of up and downregulated outliers was similar ( Figure 5 ) . We conclude that RNA levels correlate well with DNA copy number in aneuploid strains YPS163 , T73 , and YJM428 . 10 . 7554/eLife . 10996 . 008Figure 5 . Comparison of DNA and RNA copy number distributions of strains YPS163 , T73 , and YJM428 . The average log2 ( aneuploid vs . euploid RNA ) of 941 genes located on duplicated chromosomes plotted against the average log2 ( aneuploid vs . euploid DNA ) in strains YPS163 , T73 , and YJM428 . Histogram of the log2 ratios of the DNA copy number is shown in red . Histogram of the log2 ratios of the RNA copy number is shown in blue . Fits to a normal distribution are shown ( black line ) . The number of genes that show RNA copy numbers lower or higher than 2 SD from the mean are shown ( separated by dotted lines ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 008 The third set of strains that Hose et al . ( 2015 ) analyzed was comprised of two series of yeast strains that carry increasing numbers of a specific chromosome . Starting with strain YPS1009 , which carries three copies of chromosome XII ( YPS1090-chrXII-3n ) , Hose et al . ( 2015 ) derived a euploid strain ( YPS1009-chrXII-2n ) and a strain that is tetrasomic for chromosome XII ( YPS1009-chr XII-4n; Figure 6A , B ) . Using strain NCYC110 , which carries four copies of chromosome VIII ( NCYC110-chrVIII-4n ) , they isolated a strain trisomic for chromosome VIII ( NCYC110-chrVIII-3n ) and a diploid strain ( NCYC110-chrVIII-2n; Figure 6A , B ) . They then determined DNA copy number state and gene expression levels in these strain series and concluded that 11% of genes encoded on chromosome VIII and 29% of genes encoded on chromosome XII were dosage compensated . 10 . 7554/eLife . 10996 . 009Figure 6 . RNA copy number proportionally increases with DNA copy number in aneuploid series of wild S . cerevisiae strains . ( A ) Plots for strain series YPS1009-XII-2n , YPS1009-XII-3n , YPS1009-XII-4n and strain series NCYC110-chrVIII-2n , NCYC110-chrVIII-3n , NCYC110-chrVIII-4n represent the DNA copy number compared to their euploid counterparts . DNA copy numbers are shown in the order of the chromosomal location of their encoding genes . DNA copy numbers of amplified chromosomes are shown in red . Bar graphs below represent the RNA copy numbers averaged per chromosome for aneuploid strains relative to euploid reference strains . The average RNA copies of non-amplified chromosomes are shown in black . Amplified chromosomes , as predicted by the karyotype , are shown in blue . ( B ) Gene expression of strain series YPS1009-XII-2n , YPS1009-XII-3n , YPS1009-XII-4n , and strain series NCYC110-chrVIII-2n , NCYC110-chrVIII-3n , NCYC110-chrVIII ordered by chromosome position . Experiments ( columns ) of three biological replicates are shown . ( C ) Histogram of the log2 ratios of the DNA copy number of genes located on euploid chromosomes ( top left ) and genes located on trisomic chromosomes ( top right ) in strains YPS1009-chrXII-3n and NCYC110-chrVIII-3n relative to euploid controls are shown . Fits to a normal distribution are shown ( black line ) . Histogram of the log2 ratios of the RNA copy number of genes located on euploid chromosomes ( bottom left ) and genes present on trisomic chromosomes ( bottom right ) in strains YPS1009-chrXII-3n and NCYC110-chrVIII-3n relative to euploid controls are shown . Fits to a normal distribution are shown ( black line ) . ( D ) Histogram of the log2 ratios of the DNA copy number of genes located on euploid chromosomes ( top left ) and genes located on tetrasomic chromosomes ( top right ) in strains YPS1009-chrXII-4n and NCYC110-chrVIII-4n relative to euploid controls are shown . Fits to a normal distribution are shown ( black line ) . Histogram of the log2 ratios of the RNA copy number of genes located on euploid chromosomes ( bottom left ) and genes located on tetrasomic chromosomes ( bottom right ) in strains YPS1009-chrXII-4n and NCYC110-chrVIII-4n relative to euploid controls are shown . Fits to a normal distribution are shown ( black line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 009 We found the gene expression distribution of genes located on euploid and aneuploid chromosomes to fit normal distributions without any skewness ( Figure 6C , D ) . The two trisomic strains YPS1009-chrXII-3n and NCYC110-chrVIII-3n together harbored 776 triplicated genes . Their averaged log2 ratio of DNA copy number was 0 . 57 ( SD = 0 . 15 , R2 = 0 . 99 ) and 0 . 60 ( SD = 0 . 53 , R2 = 0 . 97 ) for RNA copy number ( Figure 6C ) . A similar coordination between DNA and RNA copy number was observed in the tetrasomic strains . The mean log2 ratio of DNA copy number of genes located on the tetrasomic chromosome was 0 . 99 ( SD = 0 . 16 , R2 = 0 . 99 ) , the mean mRNA expression of genes located on the tetrasomic chromosome was 0 . 99 ( SD = 0 . 62 , R2 = 0 . 96; Figure 6D ) . Importantly , the distributions of DNA and mRNA copy number were similar for genes located on euploid , trisomic and tetrasomic chromosomes , with similar SDs and no evidence of skewness ( skewness varied between 0 . 00 and 0 . 03 ) . In summary , we were not been able to detect dosage compensation in the strains described in Hose et al . ( 2015 ) . RNA levels of genes encoded by the aneuploid chromosomes are normally distributed with the expected or close to expected mean . No difference was observed between the number of down-regulated genes located on aneuploid and euploid chromosomes . Furthermore , no skewness was observed for any of the distributions . Figure 7 shows how distributions exhibit negative values of skewness when dosage compensation occurs . Importantly , in a previous study , we were able to detect attenuation in the expression of certain genes in aneuploid yeast strains using the method employed here . In Dephoure et al . ( 2014 ) , we examined the proteomes of haploid disomic laboratory yeast strains and found that production of ribosomal proteins encoded on disomic chromosomes is significantly attenuated causing the distributions to exhibit negative skewness ( Dephoure et al . , 2014 ) . 10 . 7554/eLife . 10996 . 010Figure 7 . Theoretical distribution of RNA copy number of dosage compensated duplicated genes . The theoretical distribution of RNA copy number of duplicated genes when no dosage compensation takes place is shown in blue . The theoretical distribution of RNA copy number of duplicated genes when 30% of the genes are dosage compensated is shown in red . The fit to a normal distribution shows negative skewness values ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 010 Why did Hose et al . ( 2015 ) arrive at such different conclusions than we did ? To address this question , it is important to understand how Hose et al . ( 2015 ) analyzed and interpreted their data . We identified two problems in their data analysis . The first regards data normalization . The ratios are off by a factor of log2 = 0 . 1–0 . 2 ( normalized data utilized in Hose et al . ( 2015 ) were kindly provided by A . Gasch ) . Most normalization protocols do not take into account that aneuploid strains harbor a different number of gene copies compared to euploid strains . When this is not manually corrected , data are shifted by a factor that depends on the degree of aneuploidy and results in incorrect values as shown in Figure 8A . The degree by which the data used for analysis by Hose et al . ( 2015 ) deviate from the correctly normalized expression values is of the same magnitude as some of the cutoffs used to define dosage compensated genes ( detailed next ) . 10 . 7554/eLife . 10996 . 011Figure 8 . Evaluation of the analysis tools employed by Hose et al . ( 2015 ) . ( A ) RNA copy numbers averaged per chromosome of normalized RNA-seq data obtained by Hose et al . ( 2015 ) . Data provided by Hose et al . ( 2015 ) . ( B ) Standard deviations of RNA-seq data are greater than those of DNA-seq data . Histograms of DNA-seq RPKM and RNA-seq RPKM for strain K10 are shown . ( C ) Linear regression fits of RNA versus DNA copy number are shown for several genes identified as class 3a dosage compensated genes by Hose et al . ( 2015 ) . Eight genes from chromosome XII and two genes from chromosome VIII are shown . Average log2 ratio of aneuploid vs . euploid RNA is shown . Error bars represent SD from three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 011 The second problem with the data analysis concerns cutoffs used to define dosage compensated genes . To establish cutoffs for designating whether a gene is dosage compensated or not Hose et al . ( 2015 ) used the SD of the DNA measurements , which ranged between 0 . 1 and 0 . 45 ( Table 1 column 11 , data kindly provided by A . Gasch ) as cutoffs for the RNA measurements ( Figure 4 in Hose et al . , 2015 ) . Genes whose expression deviated by the DNA SD value from the expected RNA expression level were considered dosage compensated . This is not the correct cutoff tool because the DNA copy number measurements are less variable than mRNA measurements . As seen in Figure 8B , transcript levels can vary by several orders of magnitude depending on the expression levels of a particular gene . Therefore , the distributions of relative RNA changes will show bigger SDs than gene copy number distributions . Indeed , the RNA measurements conducted by Hose et al . ( 2015 ) show SDs between 0 . 53 and 1 . 01 ( Table 1 , column 6 ) . Employing the SD derived from the DNA measurements , which are fairly lower ( Table 1 , column 8 ) , will therefore not identify genes that are dosage compensated in a statistically significant manner ( see false discovery rate discussion below ) . This is of particular importance as genes identified in Figure 4 of Hose et al . ( 2015 ) as dosage compensated were included in a group of 245 dosage compensated genes used to establish GO term enrichments among dosage compensated genes . To determine how Hose et al . ( 2015 ) identified 838 of 2204 genes as dosage compensated we re-evaluated their analysis . Figure 4A in Hose et al . ( 2015 ) displays an unusual behavior . The null model shown by the diagonal of equal RNA and DNA in this figure did not bisect the blue ( compensated ) and magenta ( exacerbated ) points . Instead , the vast majority of points below this line were considered compensated while the vast majority of points above this line were considered not exacerbated . This suggests that there could be a high number of false positives amongst the 838 genes determined to be dosage compensated . To address this possibility , we used two methods to determine the false discovery rate . First , we scrambled the data by randomly permuting the RNA/DNA ratio between genes . We did this independently for each replicate . This preserves the RNA/DNA ratios but unlinks the values from their replicate measurements and genes . Then , we used the same effective significance cutoffs used by Hose et al . ( 2015 ) to determine the number of dosage compensated genes ( see Materials and methods ) . As this is a randomized dataset , genes identified by this method are noise and can be used to determine the number of genes the analysis method would find just by chance . Based on 10 , 000 randomizations , we determined that on average , 779 genes would have passed the threshold method used by Hose et al . ( 2015 ) by chance . This yields a false discovery rate ( FDR ) of 92 . 9% . This high false discovery rate was also seen at much lower cutoffs . The FDR was between 93 and 100% at cutoffs from 0 . 1 STD to 2 STDs . Second , we calculated the average SD for each RNA sequencing measurement . As the DNA measurements for each strain were not reported independently , we calculated the average chromosome-wide DNA error from all the sequencing data that were deposited and used the lowest of these as an estimate for all analyses . We combined these errors together ( square root of summed squares of the two composite noises ) to give a measurement noise distribution for the experiment . We then randomly sampled from a normal distribution where the SD for this normal distribution was randomly sampled from the measurement noise distribution . Using this method , we found that on average , 754 genes would have passed the effective threshold used by Hose et al . ( 2015 ) . This corresponds to a false discovery rate of 89%; this value is likely a small underestimation of FDR given our method for estimation of DNA error . We conclude that both methods that we applied to determine false discovery rate strongly suggest that only a handful , at most ~70 genes or <3% , are actually dosage compensated . These results are completely in line with previous findings from laboratory strains ( Springer et al . , 2010; Dephoure et al . , 2014; Torres et al . , 2007 ) . In a second approach to identify dosage compensated genes , Hose et al . ( 2015 ) defined genes to be dosage compensated when the RNA levels did not increase with DNA copy number in their YPS1009 ( 2N , 2N+1 chromosome XII , 2N+2 chromosomes XII ) and NCYC110 ( 2N , 2N+1 chromosome VIII , 2N+2 chromosomes VIII ) ploidy series . For this , they developed a mixture of linear regression ( MLR ) model to classify genes based on the slope and intercept of the RNA-gene copy relationships . When RNA levels did not increase proportionately as DNA copy increased as evidenced by slopes lower than 1 in the MLR model , a gene was classified as dosage compensated and categorizes as Class 3a in Table 1 in Hose et al . ( 2015 ) . Thirty genes on chromosome VIII and 142 genes on chromosome XII were identified as dosage compensated through this method . This method of identifying dosage compensated genes is problematic in several ways . First , because there are only three data points per analysis , a single deviating data point can have a significant impact on the slope . For example , a gene with values of log2 ratio = 0 . 3 , 0 . 6 and 0 . 8 representing , two , three , and four copies , respectively , will perfectly fit a straight line with the slope of 0 . 5 and hence would be classified as dosage compensated according to the criteria in Hose et al . ( 2015 ) even though none of the three data points significantly deviates from the mean value given a SD of 0 . 3 or higher ( Figure 8C , Table 2 ) . In fact , the majority ( 103 of 172 ) of class 3a genes ( Table 1 and Supplemental File 3 in Hose et al . ( 2015 ) ) fit the MLR model with slopes of 0 . 5 or higher indicating that their gene expression increases with copy number . 10 . 7554/eLife . 10996 . 012Table 2 . RNA copy number of aneuploid chromosomes in strain series NCYC110 and YPS1009 . Analysis of genes encoded by chromosome VIII in strains NCYC110-2n , NCYC110-3n , NCYC110-4n ( top ) and encoded by chromosome XII in strains YPS1009-2n , YPS1009-3n , YPS1009-4n . One SD was used as a cutoff to identified genes with lower than expected RNA levels in each biological replicate . The “All 3 replicates” line represents genes whose RNA levels are reproducibly lower than expected in 3 RNA-seq experiments . Line “Both 3n and 4n” represent the number of genes whose RNA levels are lower than expected in trisomic and tetrasomic strains . DOI: http://dx . doi . org/10 . 7554/eLife . 10996 . 012NCYC110 ChrVIII . 2n-1 ChrVIII . 2n-2 ChrVIII . 2n-3 ChrVIII . 3n-1 ChrVIII . 3n-2 ChrVIII . 3n-3 ChrVIII . 4n-1 ChrVIII . 4n-2 ChrVIII . 4n-3 Mean 0 . 04 -0 . 02 0 . 02 0 . 54 0 . 51 0 . 61 0 . 97 0 . 91 0 . 99 Number of genes 282 285 283 284 286 282 285 286 284 SD 0 . 53 0 . 56 0 . 54 0 . 51 0 . 50 0 . 54 0 . 57 0 . 55 0 . 56 Mean - 1SD 23 18 20 29 25 22 24 33 29 All 3 replicates 3 12 Both 3n and 4n 1 2 YPS1009 Chr XII-2n-1 Chr XII-2n-2 Chr XII-2n-3 Chr XII-3n-1 Chr XII-3n-2 Chr XII-3n-3 Chr XII-4n-1 Chr XII-4n-2 Chr XII-4n-3 Mean 0 . 01 0 . 04 0 . 06 0 . 57 0 . 57 0 . 60 0 . 93 0 . 96 1 . 00 Number of genes 495 500 496 498 499 499 499 499 500 SD 0 . 41 0 . 52 0 . 46 0 . 46 0 . 47 0 . 51 0 . 65 0 . 66 0 . 66 Mean - 1SD 42 56 36 47 50 46 46 52 45 All 3 replicates 8 15 17 Both 3n and 4n 37 Because of these considerations , we reanalyzed the dosage compensation in chromosomes VIII and XII by two methods . In the first , we calculated the mean and standard deviations for each of the biological replicates in the NCYC110 and YPS1009 strain series and found that only two genes on aneuploid chromosome VIII and seven genes on aneuploid chromosome XII show log2 ratios 1 SD lower than the mean in three biological replicates and were reproducibly lower when present in 3 or 4 copies . Not a single gene passed the cutoff of 2 SD below the mean . We conclude that for the majority of genes only one of the two data points supports the conclusion that a gene is expressed at lower than the expected value , calling into question that the genes identified by this approach are indeed dosage compensated . In a second approach , we defined the false discovery rate ( not determined by Hose et al . , 2015 ) to determine whether the genes identified as dosage compensated were statistically distinguishable from noise . Using the same subset of genes that Hose et al . ( 2015 ) examined , we calculated a slope based on the nine RNA measurements and matching DNA measurements ( three replicates of three strains ) for both YPS1009 and NCYC110 . Using the genes identified as dosage compensated by Hose et al . ( 2015 ) , we determined the effective cutoff of their MLR method ( see Materials and methods ) . We then randomly permuted the positions of the RNA and DNA data and recalculated the slopes for each gene . From this analysis we determined the false discovery rate was within error of 100% . We conclude that there is no significant dosage compensation in these aneuploid series . Our analyses indicate that a large fraction of wild S . cerevisiae strains are unstable and heterogeneous when grown under laboratory conditions . This result suggests that at least some wild S . cerevisiae strains may not be naturally aneuploid but could become aneuploid due to an adaptive response to laboratory growth conditions . Reevaluation of the DNA and RNA copy number data generated by Hose et al . ( 2015 ) further indicates that dosage compensation is rare in both wild and laboratory strain of S . cerevisiae . Both types of strains lack mechanisms that allow them to attenuate gene expression in response to gene copy number alterations . We conclude that wild variants of S . cerevisiae do not have mechanisms in place that protect them from changes in gene copy number . Their regulation of gene expression is thus the same as that of laboratory strains of budding yeast . We consider any chromosome whose copy number was significantly different from an integral value to be heterogeneous . To determine which chromosomes were significantly different than the nearest integer value , we used a one sample t test using the copy number of each gene on the chromosome as the input which compares a distribution of values to an expected value and then corrected for multiple hypothesis testing . In strain YJM428 , the expected value for chromosome III is 2 and the expected value for chromosome XV is 2 . In strain Y2189 , the expected value for chromosome I is 4 , and for chromosomes IX and X is 3 . In strain K1 , the expected value for chromosomes I and VI is 2 . In strains Y3 and Y6 , the expected value for chromosome VIII is 3 . In strain UC5 , the expected value for chromosome I is 2 . In strain CBS7960 , the expected value for chromosomes I and III is 1 . In strain WE372 , the expected value for chromosome I is 3 . To avoid any discrepancies in data processing , Hose and coworkers kindly provided all the relative log2 ratios of the relative DNA copy number for all 47 wild strains and for the different panel of isogenic strains . In addition , Hose and coworkers kindly provided all gene expression data utilized in their manuscript . In addition , genome sequences for 16 distinct karyotypes ( eight aneuploid and eight euploids ) could be obtained from the NIH Sequence Read Archive ( SRA ) under accession SRP047341 . Gene expression data could also be obtained from NIH GEO under accession GSE61532 . Log2 ratios provided by Hose et al . ( 2015 ) were normalized by centering the euploid chromosome ratios to 0 . This was accomplished by calculating the mean of the log2 ratios of non-duplicated genes and subtracting this factor from all data points . Chromosome copy numbers were calculated by taking the average copy number of all genes within each chromosome . For diploids copy number equals 22^ ( log2 ratio euploid vs . aneuploid ) , for haploids copy number equals 12^ ( log2 ratio euploid vs . aneuploid ) . Gene expression data of each aneuploid strain were compared to their reference genome as described in Hose et al . ( 2015 ) . Log2 ratios of aneuploid/euploid genes were normalized to the euploid chromosomes as described above for DNA . RNA copy numbers per chromosome were calculated by averaging gene copy number of the genes within each chromosome . To analyze the distributions of euploid or amplified genes , DNA and RNA log2 ratios of aneuploid/euploid we first calculated the distribution of the log2 ratios binned by a value of 0 . 2 . The frequency distributions were plotted and the data were fit to normal distribution utilizing PRISM software . Means , medians , SD , skewness and R2 of the fits are reported in each figure . Gene expression data for disomes V and XVI ( Figure 3C ) were previously published in Dephoure et al . ( 2014 ) . Gene expression profiles were visualized with Treeview .	DNA inside cells is packaged into structures called chromosomes . Different species can have different numbers of chromosomes , but when any cell divides it must allocate the right number of chromosomes to each new cell . If this process goes wrong , cells end up with too many or too few chromosomes . The presence of extra copies of the genes on the additional chromosomes can cause the levels of the proteins encoded by those genes to rise abnormally , which can in turn lead to cell damage and disease . Proteins are produced using the information in genes via a two-step process . First , the gene’s DNA is copied to create molecules of RNA , and these molecules are then translated into proteins . In many organisms , the presence of extra chromosomes in a cell is matched by a corresponding increase in the RNA molecules encoded by the extra genes . Some organisms , however , counteract this effect through a process called dosage compensation . This process inactivates single genes or whole chromosomes by various means , and ensures that normal levels of RNA are produced , even in the presence of extra genes . In 2015 , researchers from the University of Wisconsin-Madison reported that dosage compensation occurs in wild strains of budding yeast and effectively protects the yeast cells against the harmful effects of having extra chromosomes . However , these findings conflicted with earlier studies of laboratory strains of this yeast , which had reported that RNA levels increased along with gene number . Torres , Springer and Amon have re-analysed the data published in 2015 , and now challenge the findings of the previous study involving the wild yeast strains . The new re-analysis instead showed that , like in laboratory yeast strains , gene number still correlates closely with RNA levels in the wild yeast . This led Torres , Springer and Amon to conclude that , in contrast with the previous report , there is currently no evidence that dosage compensation occurs in wild strains of yeast . So why do the results of these two studies disagree ? Torres , Springer and Amon identified several issues concerning the original analysis made by the researchers from the University of Wisconsin-Madison . For example , some of the strains included in the 2015 study were unstable and were naturally losing the additional chromosomes that they’d acquired . Also , the thresholds set in the analysis to identify dosage compensated genes do not appear to have been stringent enough . Together , the new findings indicate that dosage compensation is a rare event in both wild and laboratory strains of yeast .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "cell", "biology" ]	2016	No current evidence for widespread dosage compensation in S. cerevisiae
Actin-related protein 2/3 ( Arp2/3 ) complex activation by nucleation promoting factors ( NPFs ) such as WASP , plays an important role in many actin-mediated cellular processes . In yeast , Arp2/3-mediated actin filament assembly drives endocytic membrane invagination and vesicle scission . Here we used genetics and quantitative live-cell imaging to probe the mechanisms that concentrate NPFs at endocytic sites , and to investigate how NPFs regulate actin assembly onset . Our results demonstrate that SH3 ( Src homology 3 ) domain-PRM ( proline-rich motif ) interactions involving multivalent linker proteins play central roles in concentrating NPFs at endocytic sites . Quantitative imaging suggested that productive actin assembly initiation is tightly coupled to accumulation of threshold levels of WASP and WIP , but not to recruitment kinetics or release of autoinhibition . These studies provide evidence that WASP and WIP play central roles in establishment of a robust multivalent SH3 domain-PRM network in vivo , giving actin assembly onset at endocytic sites a switch-like behavior . Nucleation-promoting factors ( NPFs ) activate the actin-related protein 2/3 ( Arp2/3 ) complex to assemble actin filaments , which are important in many cellular processes , such as morphogenesis , cell motility and endocytosis ( Badour et al . , 2003; Campellone and Welch , 2010; Rottner et al . , 2010 ) . Moreover , NPF-mediated actin nucleation can be co-opted by pathogens to mediate infection and spread ( Welch and Way , 2013 ) . How NPF activity is regulated in a spatiotemporal manner in cells is an important open question . In addition to VCA motifs ( composed of a Verprolin homology domain , Central hydrophobic region , and Acidic region ) that directly stimulate Arp2/3 complex-mediated actin nucleation , NPFs often contain additional functional elements that mediate interactions with other proteins ( Padrick and Rosen , 2010; Takenawa and Suetsugu , 2007 ) . For example , the EVH1 domain ( Ena/VASP homology ) of N-WASP ( Neural Wiskott-Aldrich Syndrome protein ) interacts with WIP ( WASP-interacting protein ) ( Donnelly et al . , 2013; Rivera et al . , 2004; Tehrani et al . , 2007 ) . In addition , most WASP family proteins , which are the best-studied NPFs , also contain a large proline-rich domain ( PRD ) that contains multiple Src homology 3 ( SH3 ) -binding proline-rich motifs ( PRMs ) . Previous studies demonstrated that dozens of SH3-domain containing ligands can bind to WASP PRMs , and that some enhance WASP NPF activity ( Donnelly et al . , 2013; Padrick and Rosen , 2010; Rivera et al . , 2004; Tehrani et al . , 2007 ) . Multivalent PRM-SH3 domain interactions between N-WASP and its ligands enable proteins to form higher order complexes in vitro on artificial membranes and undergo a phase separation , producing micrometer-size clusters ( Banjade and Rosen , 2014; Li et al . , 2012 ) . Interestingly , such clusters robustly trigger Arp2/3-mediated actin assembly ( Banjade and Rosen , 2014 ) . However , how N-WASP and WIP become highly concentrated locally to trigger actin assembly in cells is not well understood . In budding yeast , all known NPFs and over 50 other proteins are organized into plasma membrane-associated patches , which are clathrin-mediated endocytosis ( CME ) sites ( Kaksonen et al . , 2005 ) . Extensive efforts have been made to determine the functions of these NPFs and to identify their interacting partners both in vitro and in vivo ( Boettner et al . , 2011; Goode et al . , 2015 ) . In addition , a very detailed pathway has been elucidated in which each endocytic protein is recruited to endocytic sites in a predictable order and with predictable timing ( Figure 1A ) ( Lu et al . , 2016 ) . The two most important NPFs , yeast WASP ( Las17 ) and type I myosin ( Myo3 and Myo5 ) , which form a complex with yeast WIP ( Vrp1 ) and additional binding partners , can be grouped into a WASP-Myosin module ( Kaksonen et al . , 2005 ) . The WASP-Myosin module stimulates actin filament assembly and provides motor activity , facilitating membrane invagination and vesicle scission ( Galletta et al . , 2008; Lewellyn et al . , 2015; Sirotkin et al . , 2005; Soulard et al . , 2002; Sun et al . , 2006 ) . Individual components of the WASP-Myosin module are recruited to endocytic sites in a regular , timely order , in which Las17 appears first , followed by Vrp1 , and finally Myo3/5 , coinciding with the onset of robust actin filament assembly ( Sun et al . , 2006 ) . 10 . 7554/eLife . 29140 . 003Figure 1 . Two Sla1 SH3 domains and a Pan1 PRD domain share a crucial role for cell growth . ( A ) Spatial-temporal recruitment of endocytic proteins . Endocytic proteins are grouped into several modules ( Lu et al . , 2016 ) as indicated . Pan1 and End3 appear after the mid coat module proteins but slightly before Sla1 and Las17 appear ( Sun et al . , 2015 ) . * Note that proteins of the WASP-Myosin module arrive at endocytic sites with different timing . Las17 arrives with a similar timing to Sla1 , while the remaining components of the WASP-Myosin module arrive later ( Sun et al . , 2006 ) . ( B ) Synthetic genetic interaction between sla1W41AW108A and pan1∆PRD . Cell growth of indicated yeast strains was compared by spotting serial dilutions of liquid cultures on plates at 25°C , or 30°C or 37°C . ( C-E ) Analysis of sla1W41A-W108A-GFP dynamics . ( F-H ) Analysis of pan1∆PRD-GFP dynamics . C and F , Maximum fluorescence intensity of GFP-tagged patch proteins at endocytic sites ( also see Figure 1—figure supplement 1D and E ) . ( D and G ) Lifetime ( mean ± SD ) of GFP-tagged proteins . E and H , Radial kymograph representations ( for explanation , please see Figure 1—figure supplement 2 ) of GFP-tagged proteins . The scale bars are 20 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 00310 . 7554/eLife . 29140 . 004Figure 1—figure supplement 1 . Characterization of sla1W41AW108A and pan1∆PRD mutant cells . ( A ) Domain structures of Pan1 , End3 and Sla1 . EH , Eps15 homology; CC , coiled-coil domain; PRD , proline rich domain; SH3 , Src homology 3; SHD1 , Sla1 homology domain 1; SHD2 , Sla1 homology domain 2; CBM , clathrin-binding motif; SR repeats , LxxQxTG repeats . ( B ) Examining cell growth of indicated yeast strains by spotting serial dilutions of liquid cultures on plate at 30°C . ( C ) Immunoblot analysis of whole-cell extracts from indicated yeast cells . PGK ( phosphoglycerate kinase ) serves as a loading control . ( D ) Single frame from a movie in which SLA1-GFP PAN1-mCherry cells and sla1W41AW108A-GFP cells were simultaneously imaged in the GFP channel . PAN1-mCherry was used to identify wild-type cells . ( E ) Single frame from a movie in which PAN1-GFP SLA1-mCherry cells and pan1∆PRD-GFP cells were simultaneously imaged in the GFP channel . SLA1-mCherry was used to identify wild-type cells . The scale bars on cell pictures are 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 00410 . 7554/eLife . 29140 . 005Figure 1—figure supplement 2 . Flowchart for scheme used to generate radial kymograph of fluorescently labeled-proteins in a movie . Radial kymograph is suitable for presenting how a single endocytic patch moves away from plasma membrane overtime . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 005 Recent studies revealed a condition in which WASP-Myosin module driven-actin assembly is uncoupled from cortical endocytic sites in live cells ( Bradford et al . , 2015; Sun et al . , 2015 ) . Multivalent linker proteins Pan1 and End3 appear to associate with each other constitutively ( Boeke et al . , 2014; Sun et al . , 2015 ) , and together with Sla1 ( another multivalent linker protein that is recruited to endocytic sites by Pan1 and End3 ) , likely provide equivalent functions to mammalian intersectins ( ITSNs ) ( Goode et al . , 2015 ) . When Pan1 and End3 are eliminated from cells by an auxin-based degron method , endocytic sites still assemble at the cell cortex , but WASP-Myosin proteins associate with actin comet tails in the cytoplasm instead of with the endocytic sites ( Sun et al . , 2015 ) . These findings indicate that the Pan1-End3-Sla1 complex plays essential roles in coupling actin assembly to endocytic sites . However , the Pan1-End3-Sla1 complex interacts with numerous additional proteins throughout endocytosis . An important but challenging task is to identify the specific domain ( s ) and interactions required for linking actin assembly to endocytic sites . In this study , to mechanistically understand how actin polymerization is coupled to endocytic sites , we identified the key interaction ( s ) required for concentrating the actin assembly machinery at endocytic sites and we further addressed how these interactions trigger a sudden burst of actin assembly . These results not only deepen our understanding of clathrin-mediated endocytosis , but also provide general mechanistic insights into spatiotemporal regulation of actin assembly in vivo . To explore the mechanism by which actin assembly is coupled to endocytic sites at a molecular level , we sought to identify the functional domains within the endocytic linker proteins that are responsible for the process . Previous studies revealed a synthetic lethal interaction between sla1∆ and pan1∆PRD , in which the Pan1 C-terminal PRD ( proline-rich domain ) is truncated ( Barker et al . , 2007 ) ( Figure 1—figure supplement 1A ) . However , the mechanistic basis for this lethality had not been explored . The Pan1 PRD has been shown to interact with yeast type 1 myosin ( Myo3 and Myo5 ) SH3 domain and enhance Myo3/5-Vrp1 NPF activity in vitro ( Barker et al . , 2007 ) . However , sla1∆ does not display a synthetic lethal interaction with myo5CA∆myo3∆ ( Figure 1—figure supplement 1B ) , in which the type 1 myosin NPF activity is abolished ( Sun et al . , 2006 ) . Thus , pan1∆PRD does not appear to cause synthetic lethality with sla1∆ by affecting Myo3/5-Vrp1 NPF activity . To gain insights into how Sla1 and Pan1 are related functionally , it was important to identify the Sla1 functions whose loss results in the synthetic lethal interaction with pan1∆PRD . Therefore , we mutated various domains of Sla1 and crossed the mutants to pan1∆PRD . Strikingly , we found that two amino acid substitutions in Sla1 are sufficient to cause a synthetic genetic interaction with pan1∆PRD . sla1W41AW108A pan1∆PRD cells display severe growth defects at 25°C and 30°C , and are inviable at 37°C ( Figure 1B ) . W41 and W108 are the conserved tryptophan residues in two SH3 ( SRC homology 3 ) domains of Sla1 ( Figure 1—figure supplement 1A ) ( Rodal et al . , 2003 ) . Point mutation of these sites abolishes the SH3 domain interactions with PRMs ( Rodal et al . , 2003 ) . However , a point mutation ( W391A ) on the third SH3 domain of Sla1 did not show a synthetic interaction with pan1∆PRD ( Figure 1B ) . Thus , our results indicate that the first two SH3 domains of Sla1 and the PRD of Pan1 function in parallel to provide a crucial role for cell growth . We next analyzed the sla1W41AW108A and pan1∆PRD mutants separately . Immunoblotting of whole cell extracts showed that sla1W41AW108A-GFP is well expressed ( Figure 1—figure supplement 1C ) . More importantly , sla1W41AW108A-GFP appeared in cortical patches that reached fluorescence intensity levels similar to wild-type Sla1-GFP patches ( Figure 1C and Figure 1—figure supplement 1D ) . However , the sla1W41AW108A-GFP patch lifetimes were substantially longer and more variable compared to Sla1-GFP patches ( 78 . 8 ± 31 . 2 vs 24 . 1 ± 4 . 9 , p<0 . 0001 ) ( Figure 1D ) . Nevertheless , similar to the wild-type cells , sla1W41AW108A-GFP patches moved inward , off the cell cortex , at the end of their lifetime ( Figure 1E ) , indicating that endocytic internalization still takes place in this mutant . Immunoblotting showed that pan1∆PRD-GFP is also expressed at levels similar to the wild-type protein ( Figure 1—figure supplement 1C ) . In addition , cortical pan1∆PRD-GFP patches reached fluorescence intensity levels similar to wild-type Pan1-GFP patches ( Figure 1F and Figure 1—figure supplement 1E ) . Finally , pan1∆PRD-GFP patch lifetimes were slightly shorter than Pan1-GFP lifetimes ( 23 . 6 ± 4 . 2 vs 28 . 5 ± 5 . 5 , p<0 . 0001 ) ( Figure 1G ) , and they were internalized at the end of their lifetime ( Figure 1H ) . The results described above establish that neither a point mutant of two Sla1 SH3 domains nor a PRD truncation mutant of Pan1 affects either protein’s expression or cortical recruitment . Thus , the severe synthetic growth defect of an sla1 W41AW108A pan1∆PRD double mutant is caused by the loss of specific functions rather than by the absence of the mutant proteins at endocytic sites . We next analyzed the sla1 W41AW108A pan1∆PRD double mutant to determine how the Sla1 SH3 domains and the Pan1 PRD function in endocytosis . We examined cortical patch behavior of GFP-tagged Sla1 ( or sla1 mutant ) and mCherry-tagged Pan1 ( or pan1 mutant ) in sla1 W41AW108A and/or pan1∆PRD mutants . Previous studies suggested that Sla1 is recruited to endocytic sites by Pan1 and End3 ( Sun et al . , 2015; Tang et al . , 2000 ) . Sla1 ( or the sla1 mutant ) appears slightly after Pan1 ( or the pan1 mutant ) , and then internalizes in wild-type cells , pan1∆PRD cells , and sla1W41AW108A cells ( Figure 2A–D ) . Consistent with the results in Figure 1D , patch lifetimes in sla1W41AW108A cells are irregular and longer than in wild-type and pan1∆PRD cells ( Figure 2D ) . However , in the double mutants , sla1W41AW108A and pan1∆PRD colocalize as stable patches at the cell cortex ( Figure 2E ) . Most cortical patches ( 98 . 1% , 216 cortical patches from 20 cells were examined ) stay nonmotile during the entire 3 min movie , suggesting that endocytic internalization does not occur ( Figure 2D and E ) . A Lucifer yellow uptake assay confirmed a severe endocytic defect ( Figure 2—figure supplement 2 ) . Moreover , in the double mutant cells , pan1∆PRD colocalizes with Sla2 ( Figure 2F ) , a signature protein for endocytic sites , and the yeast homologue of vertebrate HIP1R ( Engqvist-Goldstein et al . , 2001 ) . These results indicate that the nonmotile cortical sla1W41AW108A/pan1∆PRD patches are nonproductive endocytic sites . We next sought to determine why the sites were nonproductive . 10 . 7554/eLife . 29140 . 006Figure 2 . Endocytic internalization is defective in sla1 W41AW108A pan1∆PRD mutant cells . ( A-C ) Single frames ( left ) from movies and circumferential kymograph representations ( for the explanation , please see Figure 2—figure supplement 1 ) for GFP- and mCherry-tagged proteins . ( A ) SLA1-GFP PAN1-mCherry cells . ( B ) SLA1-GFP pan1∆PRD-mCherry cells . ( C ) sla1W41AW108A-GFP PAN1-mCherry cells . ( D ) Lifetime ( mean ±SD ) and radial kymograph representations of GFP- and mCherry-tagged proteins for indicated strains . ( E ) Single frame ( left ) from a 3 min movie and circumferential kymograph representations of sla1W41AW108A-GFP pan1∆PRD-mCherry cells . ( F ) Single frame ( left ) from 3 min movie and circumferential kymograph representation of pan1∆PRD-GFP and Sla2-TagRFP-T in sla1W41AW108A cells . The scale bars on kymographs are 20 s . The scale bars on cell pictures are 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 00610 . 7554/eLife . 29140 . 007Figure 2—figure supplement 1 . Flowchart for scheme used to generate circumferential kymograph of fluorescently labeled patch proteins on the cell cortex in a movie . Circumferential kymographs are useful for showing how fluorescence intensity of patch proteins develops at multiple endocytic sites during the duration of the movie . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 00710 . 7554/eLife . 29140 . 008Figure 2—figure supplement 2 . Lucifer yellow ( LY ) uptake is defective in an sla1W41AW108A pan1∆PRD strain . The arrows indicate vacuoles . Note that there is no LY signal in the vacuoles or cytoplasm . 270 cells were examined . Lucifer yellow uptake was blocked in 98 . 5% of the cells . The scale bars on cell pictures are 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 008 Previous in vitro studies showed that Sla1 interacts with the yeast WASP Las17 through SH3-PRM interactions ( Feliciano and Di Pietro , 2012; Rodal et al . , 2003 ) . We therefore examined where Las17 is located relative to the nonproductive endocytic sites in sla1 W41AW108A and pan1∆PRD single and double mutants . In wild-type cells , Sla1 and Las17 accumulate and then leave endocytic sites with similar kinetics ( Figure 3A ) . In sla1W41AW108A cells , Las17 still colocalizes with sla1W41AW108A patches ( Figure 3B ) . However , Las17-GFP reaches its maximum fluorescence intensity approximately 20 s after sla1W41AW108A . More importantly , Las17-GFP fluorescence intensity reaches only half of its maximum value and continues to increase when sla1W41AW108A-mCherry fluorescence intensity has already begun to decline . Thus , in sla1W41AW108A cells , cortical recruitment of Las17 is no longer synchronized with sla1W41AW108A recruitment . These results are consistent with a role for Sla1 in Las17 recruitment ( Feliciano and Di Pietro , 2012 ) . However , Las17 is still recruited to endocytic sites in sla1W41AW108A cells , indicating that additional proteins help to recruit Las17 . In pan1∆PRD cells , Las17-GFP reaches its maximum fluorescence intensity about 5 s after Sla1-mCherry ( Figure 3C ) . Thus , Las17 recruitment is also defective in pan1∆PRD cells , although to a lesser extent than in sla1W41AW108A cells . 10 . 7554/eLife . 29140 . 009Figure 3 . Las17 ( yeast WASP ) is not recruited to cortical endocytic sites in sla1W41AW108A pan1∆PRD mutant cells . ( A-C ) Single frames ( top ) from movies and circumferential kymograph representations ( middle ) of GFP- and mCherry-tagged proteins . Averaged ( mean ±SD ) fluorescent intensity profiles for GFP- and mCherry-tagged proteins from 10 individual patches ( bottom ) . ( A ) SLA1-mCherry LAS17-GFP cells . ( B ) sla1W41AW108A-mCherry LAS17-GFP cells . Note that the fluorescence intensity profile for this strain was only analyzed for the last 50 s of the patch lifetime . ( C ) SLA1-mCherry LAS17-GFP pan1∆PRD cells . ( D and E ) A single frame and maximum intensity projection of all frames from a movie ( Video 1 ) of sla1W41AW108A-GFP LAS17-TagRFP-T pan1∆PRD cells . ( E ) Enlarged views of the boxed-areas shown in D . The arrows indicate Las17 patches in cytoplasm . ( F ) Circumferential kymograph representation of sla1W41AW108A-GFP and Las17-TagRFP-T in pan1∆PRDcells . The arrow indicates that a Las17-TagRFP-T patch transiently colocalizes with a static sla1W41AW108A-GFP patch . The scale bars on kymographs are 20 s . The scale bars on cell pictures are 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 009 In sla1W41AW108A pan1∆PRD cells , cortical sla1W41AW108A patches stay non-motile at the cell cortex , while Las17 patches move dynamically along the cell cortex and throughout the cytoplasm ( Figure 3D–F and Video 1 ) . Most cortical sla1W41AW108A patches did not recruit any detectable Las17 during a 2 min movie ( 95 . 4% , 131 patches from 14 cells were examined ) ( Figure 3F ) . In rare cases ( 4 . 6% ) , Las17 patches appeared to transiently colocalize with cortical sla1W41AW108A patches ( Figure 3F ) . However , unlike in wild-type cells , neither sla1W41AW108A patches nor Las17 patches disassemble after the transient colocalization ( Figure 3F ) . A likely explanation for such transient colocalization is that Las17 patches move along the cortex and move to or near the non-motile sla1W41AW108A patches by chance . Thus , in sla1W41AW108A pan1∆PRD cells , Las17 is no longer recruited to cortical endocytic sites . 10 . 7554/eLife . 29140 . 010Video 1 . Dynamics of sla1W41AW108A-GFP and Las17-TagRFP-T in sla1W41AW108A-GFP LAS17-TagRFP-T pan1∆PRD cells . Time to acquire one image pair is 1 . 8 s . Interval between frames is 1 . 8 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 010 Since Las17 is not recruited to cortical endocytic sites in sla1W41AW108A pan1∆PRD cells , we asked how actin dynamics are affected . Fluorescently tagged-Abp1 ( Actin binding protein 1 ) and Sac6 ( yeast fimbrin ) , which both localize to cortical actin patches , were used to monitor endocytic actin assembly in the following experiments . These actin markers were used interchangeably because neither detectably altered actin function or dynamics . In wild-type , pan1∆PRD , or sla1W41AW108A cells , the cortical actin patch marker Abp1 ( Actin binding protein 1 ) appears at endocytic sites at the end of the Sla1 ( or sla1 mutant ) lifetime , and both proteins are internalized and disappear ( Figures 4A , B and C ) . Strikingly , instead of forming cortical patches in sla1 W41AW108A pan1∆PRD cells , actin formed comet tails , as seen with Abp1-RFP ( Figure 4D ) or Sac6-GFP ( Figure 4E ) . Multifocus microscopy ( MFM ) ( Abrahamsson et al . , 2013 ) clearly demonstrated that these actin comet tails slide along the cell cortex or move through the cytoplasm over time ( Figure 4E , Video 2 ) . More importantly , these actin comet tails do not appear to originate at or associate stably with the cortical static sla1W41AW108A patches in sla1 W41AW108A pan1∆PRD cells ( Figure 4D , Video 3 ) . We examined 304 cortical static sla1W41AW108A patches in 30 sla1W41AW108A pan1∆PRD cells and found that only 32 ( 10 . 5% ) of them transiently colocalized with actin comet-tails during a 3 min movie ( Figure 4D ) . However , unlike in wild-type cells , the sla1W41AW108A-GFP patches remained stationary at the cell cortex after these rare , transient colocalizations ( Figure 4D ) . Thus , actin comet tails may move to or near the nonmotile sla1W41AW108A/pan1∆PRD patches by chance when the comet tails move along the cell cortex . 10 . 7554/eLife . 29140 . 011Video 2 . Dynamics of Sac6-GFP in sla1W41AW108A pan1∆PRD cells captured by multifocus microscopy ( MFM ) . Time interval between frames is 250 ms . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 01110 . 7554/eLife . 29140 . 012Video 3 . Dynamics of sla1W41AW108A-GFP and Abp1-RFP in sla1W41AW108A-GFP ABP1-RFP pan1∆PRD cells . Time to acquire one image pair is 1 . 8 s . Interval between frames is 1 . 8 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 01210 . 7554/eLife . 29140 . 013Figure 4 . WASP-Myosin module proteins localize at the leading tip of actin comet tails in sla1W41AW108A pan1∆PRD mutant cells . ( A-C ) Single frames ( left ) from movies and circumferential kymograph presentations ( right ) of GFP- and RFP-tagged proteins . ( A ) SLA1-GFP ABP1-RFP cells . ( B ) SLA1-GFP ABP1-RFP pan1∆PRD cells . ( C ) sla1W41AW108A-GFP ABP1-RFP cells . ( D ) Single frame ( top left ) from movie ( Video 3 ) and circumferential kymograph representations ( top right ) of sla1W41AW108A-GFP and Abp1-RFP in pan1∆PRD cells . Montages of individual actin comet tails in the boxed areas in the top left image ( bottom ) . ( E ) Dynamics of Sac6-GFP in sla1W41AW108A pan1∆PRD cells observed by multifocus microscopy ( MFM ) . An example of three simultaneously acquired Z-planes that are artificially colored in green , red or blue ( bottom panel ) . Note that the actin tails in the merged image reveal different colors depending on their positions in Z-planes . Montage of one actin comet tail from a movie ( Video 2 ) acquired in multiple Z-planes simultaneously at 1 frame/250 msec ( top right ) . Note that the actin tail changes colors over time , reflecting movement through the cytoplasm . ( F ) Single frames from movies ( Videos 4 and 5 ) of cells expressing indicated GFP-tagged protein and Abp1-RFP . The arrows indicate that WASP-Myosin module proteins localize at the leading tip of actin comet tails in sla1W41AW108A pan1∆PRD mutants . The scale bars for kymographs are 20 s . The scale bars for cell pictures are 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 01310 . 7554/eLife . 29140 . 014Figure 4—figure supplement 1 . Localization of Ede1 in sla1W41AW108A pan1∆PRD mutant cells . Single frame ( A ) from movie and circumferential kymograph representations ( B ) of Ede1-GFP and Abp1-RFP in sla1W41AW108A pan1∆PRD cells . ( C ) Montage of the boxed areas in A . The scale bars for kymographs are 20 s . The scale bars for cell pictures are 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 014 As we showed above , Las17 patches and actin comet-tails are not associated with the nonproductive cortical endocytic sites in sla1W41AW108A pan1∆PRD cells . We next determined the spatial relationship between Las17 and actin comet tails in this mutant . Interestingly , in the sla1W41AW108A pan1∆PRD mutant , Las17 patches localized at the leading tips of actin comet tails and are apparently propelled by actin assembly in a manner similar to actin rocket tails on pathogens such as vaccinia virus or Shigella flexneri ( Welch and Way , 2013 ) ( Figure 4F , Video 4 ) . Similar to Las17 , several other components of WASP-Myosin module , including Vrp1 , Myo5 , and Bbc1 , were also observed at the leading tip of actin comet tails in sla1W41AW108A pan1∆PRD cells ( Figure 4F , Video 4 , and Video 5 ) . These results show that the actin comet tails share molecular characteristics with the endocytic actin machinery . Thus , the endocytic actin machinery no longer assembles at cortical endocytic sites in sla1W41AW108A pan1∆PRD cells , but at distinct sites , the nature of which is presently obscure . Consistently , the yeast Eps15 , Ede1 , which is one of the early module proteins ( Figure 1A ) and functions in endocytic site initiation and stabilization ( Kaksonen et al . , 2005 ) , is not detected at the tip of the actin comet tails ( Figure 4—figure supplement 1 ) . These results demonstrate that the WASP-Myosin module proteins can induce actin assembly and forces independent of the upstream endocytic machinery . 10 . 7554/eLife . 29140 . 015Video 4 . Dynamics of Las17-GFP or Bbc1-GFP and Abp1-RFP in sla1W41AW108A pan1∆PRD cells . Time to acquire one image pair is 1 . 8 s . Interval between frames is 1 . 8 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 01510 . 7554/eLife . 29140 . 016Video 5 . Dynamics of Vrp1-GFP or Myo5-GFP and Abp1-RFP in sla1W41AW108A pan1∆PRD cells . Time to acquire one image pair is 1 . 8 s . Interval between frames is 1 . 8 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 016 Together , our results indicate that two Sla1 SH3 domains and the Pan1 PRD together play indispensable roles in coupling the WASP-Myosin machinery to endocytic sites . Thus , we conclude that WASP-Myosin module proteins are recruited to endocytic sites mainly through SH3-PRM interactions ( Sla1 recruits PRD-containing proteins and Pan1 recruits SH3 domain-containing proteins ) . Previous in vitro results suggested that Sla1 SH3 domains inhibit Las17 NPF activity ( Feliciano and Di Pietro , 2012; Rodal et al . , 2003 ) , while the Pan1 PRD activates Myo3/5 NPF activity ( Barker et al . , 2007 ) . Next , we addressed the recruitment and NPF regulatory roles of the Sla1 SH3 domains and the Pan1 PRD in WASP-Myosin module regulation . The sla1W41AW108A pan1∆PRD mutant provides a unique opportunity to identify the key player ( s ) that recruit the actin assembly machinery to endocytic sites and to explore which parameters are important for triggering actin assembly . Thus , we developed a strategy to artificially direct selected proteins to cortical sla1W41AW108A/pan1∆PRD sites , and then tested whether cell growth and productive endocytic actin assembly were restored . Previous studies reported that End3’s C-terminus ( end3C ) interacts with the Pan1 central region with high affinity ( Kd = 27 nM ) ( Boeke et al . , 2014; Sun et al . , 2015 ) . This C-terminal region contains less than 200 amino acids ( Figure 1—figure supplement 1A ) and its primary function is to mediate cortical recruitment through interaction with the Pan1 N-terminus ( Sun et al . , 2015; Tang et al . , 2000 ) . These features make end3C an appealing candidate to recruit Las17 to pan1∆PRD sites in sla1W41AW108A pan1∆PRD double mutants . We generated a LAS17-end3C strain and tagged Las17-end3C with GFP . The sequence encoding Las17-end3C-GFP was integrated into the LAS17 chromosomal locus so the hybrid gene was expressed from LAS17’s endogenous promoter . las17WCA∆ myo5CA∆ myo3∆ cells exhibit severe growth defects due to the loss of NPF activity from both NPFs ( Sun et al . , 2006 ) . However , LAS17-end3C-GFP myo5CA∆ myo3∆ cells grow well ( Figure 5—figure supplement 1A ) . We conclude that the end3C fusion does not interfere Las17’s NPF activity . Strikingly , Las17-end3C-GFP restored sla1W41AW108A pan1∆PRD cells to normal growth at not only 25°C and 30°C , but also the non-permissive temperature of 37°C ( Figure 5A ) . Remarkably , Las17-end3C-GFP even rescued sla1∆ pan1∆PRD from lethality at all temperatures ( Figure 5A and Figure 5—figure supplement 1B ) . Las17-end3C-GFP in LAS17-end3C-GFP sla1W41AW108A pan1∆PRD cells is expressed at levels indistinguishable from Las17-GFP in wild-type cells ( Figure 5—figure supplement 1C ) . Furthermore , there is no significant difference in patch maximum fluorescence intensity between Las17-end3C-GFP in the LAS17-end3C-GFP sla1W41AW108A pan1∆PRD mutant and Las17-GFP in wild-type cells ( Figure 5—figure supplement 1D and E ) . In addition , Las17-end3C-GFP patches only appear at the cell cortex and they develop fluorescence intensity with similar kinetics to pan1∆PRD-mCherry in the sla1W41AW108A pan1∆PRD mutant ( Figure 5—figure supplement 1F ) , indicating that end3C is sufficient to recruit Las17 to cortical endocytic sites through its interaction with pan1∆PRD when the Sla1 SH3- and Pan1 PRD- mediated interactions are absent . We conclude that artificial Las17 recruitment bypasses the requirement for the Sla1 SH3 domains and Pan1 PRD for normal cell growth . 10 . 7554/eLife . 29140 . 017Figure 5 . A WASP-end3C chimeric protein or a WIP-end3C chimeric protein restores normal growth and productive endocytic actin assembly in an sla1W41AW108A pan1∆PRD mutant . The onset of actin assembly is tightly coupled to accumulation of a threshold of WASP or WIP at endocytic sites . ( A and B ) Cell growth at 25°C , 30°C or 37°C of indicated yeast strains spotted as serial dilutions of liquid cultures on plates ( also see Figure 5—figure supplement 1B ) . ( C and E ) Single frame ( left ) from movie ( Video 6 or Video 7 ) , circumferential kymograph representation ( middle ) , and radial kymograph representations ( right ) of Las17-GFP and Abp1-RFP , or Vrp1-GFP and Abp1-RFP in wild-type cells . ( D and F ) Single frame ( left ) from movie ( Video 6 and Video 7 ) , circumferential kymograph presentation ( middle ) , and radial kymograph representation ( right ) of Las17-end3C-GFP and Abp1-RFP , or Vrp1-end3C-GFP and Abp1-RFP in sla1W41AW108A pan1∆PRD cells . The numbers shown in C-F are lifetime of GFP-tagged protein ( in green ) and Abp1-RFP ( in red ) in the indicated strains ( also see Figure 5—figure supplement 3A ) . ( G ) The average fluorescence intensity of GFP-tagged patch proteins at the moment when the Abp1-RFP signal appear at endocytic sites for the indicated strains . H and I , T1vs T2 plots for indicated strains ( for the details , please see Figure 5—figure supplement 3B ) . The scale bars in kymographs are 20 s . The scale bars on cell pictures are 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 01710 . 7554/eLife . 29140 . 018Figure 5—figure supplement 1 . The end3C fusion does not interfere with Las17’s NPF activity and is sufficient to recruit Las17 to cortical endocytic sites in sla1W41AW108A pan1∆PRD mutant cells . ( A ) Examining growth of indicated yeast strains by spotting serial dilutions of liquid cultures on plate at 30°C . ( B ) Tetrad analysis of diploids made by crossing sla1∆ to LAS17-end3C-GFP pan1∆PRD . 1–8 are tetrad dissections and a–d are haploid spore colonies . ( C ) Immunoblot analysis of whole-cell extracts from LAS17-GFP or LAS17-end3C-GFP sla1W41AW108A pan1∆PRD cells . The intensities of anti-GFP-labeled bands were normalized to that of PGK ( phosphoglycerate kinase ) loading controls . The ratio of Las17-GFP to Las17-end3C-GFP is 1 . 0: 0 . 98 . ( D ) Single frames from a movie in which LAS17-GFP ABP1-RFP cells and LAS17-end3C-GFP sla1W41AW108A pan1∆PRD cells were simultaneously imaged in the GFP channel . ABP1-RFP was used to identify wild-type cells . ( E ) Maximum fluorescence intensity of GFP-tagged patch proteins at endocytic sites in the indicated strains . ( F ) Single frames ( top left ) from movie and circumferential kymograph representations ( bottom left ) of Las17-end3C-GFP and pan1∆PRD-mCherry in sla1W41AW108A cells . Averaged ( mean ±SD ) fluorescent intensity profiles ( right ) for GFP- and mCherry-tagged proteins ( n = 15 ) . The scale bars in kymographs are 20 s . The scale bar on cell picture is 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 01810 . 7554/eLife . 29140 . 019Figure 5—figure supplement 2 . Protein expression and dynamics of various end3C fused proteins . ( A ) Immunoblot analysis of whole-cell extracts from VRP1-GFP or VRP1-end3C-GFP sla1W41AW108A pan1∆PRD cells . The intensities of anti-GFP-labeled bands were normalized to that of PGK ( phosphoglycerate kinase ) loading controls . The ratio of Vrp1-GFP and Vrp1-end3C-GFP is 1 . 00:1 . 02 . ( B ) Immunoblot analysis of whole-cell extracts for indicated strains probed with an anti-GFP antibody . ( C ) Single frames from movies of cells expressing GFP-tagged and RFP-tagged protein pairs . Montages , single-channel or merged images of single patches from movies of cells expressing GFP-tagged and RFP-tagged protein pairs . All the strains show fairly normal growth and endocytic actin patch dynamics . The scale bar on cell picture is 2 µm . ( D ) Examining growth of indicated yeast strains by spotting serial dilutions of liquid cultures on plate at 30°C . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 01910 . 7554/eLife . 29140 . 020Figure 5—figure supplement 3 . Quantification of endocytic patch protein lifetime for strains shown in Figure 5C–F and flowchart for scheme used to plot graphs shown in Figure 5H and I . ( A ) Lifetimes of GFP- and RFP-tagged proteins in the indicated strains that were presented in Figure 5C–F . ( B ) Flowchart for method used to determine T1 and T2 for each individual endocytic events to plot graphs shown in Figure 5H and I . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 020 We next fused end3C to the C-terminus of additional SH3 domain- or PRD-containing endocytic proteins ( Figure 5—figure supplement 2 ) and determined whether they could also restore normal growth to sla1W41AW108A pan1∆PRD cells ( Figure 5B ) . Intriguingly , Vrp1-end3C fully restored cell growth , similar to Las17-end3C . Bzz1-end3C ( yeast functional homolog of TOCA-1 , FBP17 , CIP4 , or PACSIN ) and Rvs167-end3C ( yeast Amphiphysin ) also rescued cell growth , but to lesser extents , particularly at 37°C . In addition , Myo5-end3C rescued sla1W41AW108A pan1∆PRD cell growth , but to a lesser extent than Bzz1-end3C or Rvs167-end3C . In contrast , Lsb4-end3C ( yeast SH3YL1a ) ( Figure 5B ) and Bbc1-end3C ( data not shown ) did not rescue cell growth . Thus , we have identified five PRD- or SH3 domain- containing endocytic proteins that can rescue cell growth of the sla1W41AW108A pan1∆PRD mutant to different extents when they are fused to end3C . Since end3C fused to Las17 ( yeast N-WASP ) or Vrp1 ( yeast WIP ) restored sla1W41AW108A pan1∆PRD to normal growth ( Figure 5A and B ) , we examined endocytic actin assembly in these two strains . As in wild-type cells ( Figure 5C and E ) , Abp1-RFP labeled cortical actin patches moved inward and then disappeared in both LAS17-end3C-GFP sla1W41AW108A pan1∆PRD ( Figure 5D ) ( Video 6 ) and VRP1-end3C-GFP sla1W41AW108A pan1∆PRD cells ( Figure 5F ) ( Video 7 ) , indicating that the actin assembly is productive and that endocytic internalization is restored in the mutant strains . Consistently , cytoplasmic actin comet tails were no longer observed in these two mutants ( Figure 5D and F , Videos 6 and 7 ) . The patch lifetimes of Las17-end3C-GFP or Vrp1-end3C-GFP in sla1W41AW108A pan1∆PRD cells , respectively , were less regular and approximately two times longer than the lifetimes of Las17 or Vrp1 in wild-type cells ( Figure 5C–F and Figure 5—figure supplement 3A ) . In contrast , the actin patch ( Abp1-RFP ) lifetimes were much less affected in LAS17-end3C-GFP sla1W41AW108A pan1∆PRD or VRP1-end3C-GFP sla1W41AW108A pan1∆PRD cells compared to wild-type cells . These results demonstrate that restoring Las17 or Vrp1 cortical localization using artificial , engineered fusions , is sufficient to direct productive actin assembly to cortical endocytic sites to compensate for loss of interactions mediated by Sla1 SH3 domains and the Pan1 PRD . Thus , our data indicate that Sla1 SH3 domains and the Pan1 PRD primarily function in recruitment , rather than the NPF regulation , of the WASP-Myosin module . 10 . 7554/eLife . 29140 . 021Video 6 . , Dynamics of Las17-GFP and Abp1-RFP in wild type cells and dynamics of Las17-end3C-GFP and Abp1-RFP in sla1W41AW108A pan1∆PRD cells . Time to acquire one image pair is 1 . 1 s . Interval between frames is 1 . 1 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 02110 . 7554/eLife . 29140 . 022Video 7 . , Dynamics of Vrp1-GFP and Abp1-RFP in wild-type cells and dynamics of Vrp1-end3C-GFP and Abp1-RFP in sla1W41AW108A pan1∆PRD cells . Time to acquire one image pair is 1 . 2 s . Interval between frames is 1 . 2 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 022 Previous studies suggest that in contrast to its mammalian N-WASP homologue , Las17’s actin nucleation promoting activity is not autoinhibited in vitro ( Feliciano and Di Pietro , 2012; Rodal et al . , 2003 ) . However , actin assembly is only observed approximately 15 s after Las17 is recruited to endocytic sites in wild-type cells ( Figure 5C ) . The interaction between the first two Sla1 SH3 domains and the Las17 PRD is thought to be important for keeping Las17 inactive until actin assembly starts ( Feliciano and Di Pietro , 2012; Rodal et al . , 2003 ) . In sla1W41AW108A pan1∆PRD cells , the Las17-end3C nucleation promoting activity can no longer be inhibited by sla1W41AW108A due to the lack of functional Sla1 SH3 domains , so actin assembly might occur prematurely . However , in the sla1W41AW108A pan1∆PRD mutant , Las17-end3C-GFP persisted at endocytic sites even longer than in wild-type cells before actin assembly was initiated ( Figure 5D ) . These results indicate that Las17 NPF activity at endocytic sites does not appear to be regulated as was previously assumed . We next investigated which parameters are important for the onset of actin assembly . We asked whether the quantity of Las17 or Vrp1 at endocytic sites predicts endocytic actin assembly initiation . The fluorescence intensity of the GFP-tagged proteins at the moment when the Abp1-RFP signal appeared ( the onset of actin assembly ) at endocytic sites was determined for the strains shown in Figure 5C–F . Interestingly , regardless of differences in lifetimes , the average intensities of the GFP-tagged Las17 and Vrp1 were all in a similar range ( 70–80% of their maximum intensity ) when actin assembly was initiated ( Figure 5G and Figure 5—figure supplement 3B ) . Importantly , the end3C-fusions did not alter expression levels of the tagged proteins ( Figure 5—figure supplement 1C and Figure 5—figure supplement 2A ) . Thus , these results establish that productive actin assembly is initiated when similar numbers of Las17 and Vrp1 are recruited to endocytic sites in wild-type and LAS17-end3C-GFP sla1W41AW108A pan1∆PRD cells . Furthermore , when the time , T1 , during at which the GFP-tagged protein reaches its average intensity , was plotted against the time T2 , when actin assembly is first detected , for numerous individual endocytic events in these strains ( Figure 5H–I , Figure 5—figure supplement 3B ) , the high R-squared values of the linear trendlines drawn for these plots indicated that a threshold accumulation ( 70–80% of their maximum quantity at endocytic sites ) of Las17 or Vrp1 at endocytic sites is tightly correlated with the onset of productive actin assembly in both wild-type and in sla1W41AW108A pan1∆PRD cells . When Las17-end3C or Vrp1-end3C is recruited to endocytic sites through interaction of its end3C domain with pan1∆PRD , other proteins associated with the cytoplasmic actin comet tails observed in sla1W41AW108A pan1∆PRD cells presumably reach cortical endocytic sites by ( direct or indirect ) interactions with the end3C-fused protein . We next quantitatively addressed how Las17-end3C or Vrp1-end3C recruit their binding partners to endocytic sites and facilitate productive actin assembly in sla1W41AW108A pan1∆PRD cells . In sla1W41AW108A pan1∆PRD cells , Las17-end3C-GFP and Vrp1-mCherry , or Vrp1-end3C-GFP and Las17-TagRFP-T , respectively , developed fluorescence intensity with identical kinetics at cortical patches ( Figure 6A and Figure 6—figure supplement 1A and B ) . Thus , when cortical localization of either Las17 or Vrp1 is restored through an end3C fusion in sla1W41AW108A pan1∆PRD cells , the other protein is recruited simultaneously . This result explains why Las17-end3C and Vrp1-end3C rescue the sla1W41AW108A pan1∆PRD mutant phenotypes to the same extent ( Figure 5B ) . Moreover , our genetic studies indicated that Vrp1 or Las17 is required for Las17-end3C or Vrp1-end3C , respectively , to restore normal growth of sla1W41AW108A pan1∆PRD cells ( Figure 6B ) . On the other hand , similar to the wild-type cells ( Figure 5C and E , Figure 6C ) , the onset of actin assembly appears to coincide with the recruitment of Myo5-GFP ( type I Myosin ) in LAS17-end3C-GFP sla1W41AW108A pan1∆PRD cells ( Figure 5D and Figure 6D ) and in VRP1-end3C-GFP sla1W41AW108A pan1∆PRD cells ( Figure 5F and Figure 6D ) . Taken together , Las17 and Vrp1 ( recruited by end3C fusions ) are necessary and sufficient to recruit the WASP-Myosin module proteins to endocytic sites and restore productive endocytic actin assembly , compensating for loss of interactions mediated by multivalent endocytic linker proteins . 10 . 7554/eLife . 29140 . 023Figure 6 . Actin assembly is triggered in a switch-like manner that corresponds to a threshold level of WASP-WIP accumulation . ( A ) Circumferential kymograph presentations of GFP- and mCherry- or TagRFP-T-tagged proteins for indicated yeast strains ( also see Figure 6—figure supplement 1 ) . ( B ) Examining growth of indicated yeast strains by spotting serial dilutions of liquid cultures on plate . ( C ) Circumferential kymograph presentations of GFP- and RFP- or TagRFP-T-tagged proteins in wild-type cells . ( D ) Circumferential kymograph presentations of GFP- and TagRFP-T-tagged proteins for indicated yeast strains . ( E ) Alignment of averaged intensity measurements of GFP- and RFP-tagged proteins from the indicated yeast cells . For more details , please also see Figure 6—figure supplement 2 . The scale bars in kymographs are 20 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 02310 . 7554/eLife . 29140 . 024Figure 6—figure supplement 1 . Two-color fluorescence intensity profiles for Las17-end3C-GFP and Vrp1-mCherry , or Vrp1-end3C-GFP and Las17-TagRFP-T in sla1W41AW108A pan1∆PRD mutant cells . ( A and B ) Fluorescence intensity profiles of GFP- and mCherry- or TagRFP-T-tagged proteins over time at endocytic sites in the indicated yeast strains . The plots shown are representative of three independent endocytic events for each strain . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 02410 . 7554/eLife . 29140 . 025Figure 6—figure supplement 2 . Alignment of averaged ( mean ±SD ) intensity measurements for GFP- and RFP-tagged proteins in the indicated yeast strains ( note different time scales in A and B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 025 To further assess the onset of actin assembly kinetically , we analyzed fluorescence intensity development of GFP- and RFP-tagged proteins in LAS17-end3C-GFP ABP1-RFP sla1W41AW108A pan1∆PRD and LAS17-GFP ABP1-RFP cells ( Figure 5C and D ) . As shown in Figure 6E and Figure 6—figure supplement 2 , Abp1-RFP joins Las17-GFP or Las17-end3C-GFP patches when the GFP-tagged proteins reach 70–80% of maximum intensity in LAS17-GFP ABP1-RFP or LAS17-end3C-GFP ABP1-RFP sla1W41AW108A pan1∆PRD cells . It takes much longer for Las17-end3C-GFP to reach 70–80% of its maximum intensity in sla1W41AW108A pan1∆PRD cells than for Las17-GFP to reach those levels in wild-type cells . This delay reflects the much longer lifetime of Las17-end3C-GFP patches observed in sla1W41AW108A pan1∆PRD cells ( Figure 5C and D , Figure 5—figure supplement 3A ) . Thus , even though Las17 NPF activity cannot be inhibited by sla1W41AW108A , Las17-end3C-GFP does not appear to initiate actin nucleation until it reaches a ‘threshold’ of 70–80% of its maximum intensity . Intriguingly , once Las17-GFP or Las17-end3C-GFP reaches an apparent ‘threshold’ level , the Abp1-RFP signal increases rapidly and reaches its maximum level at a similar rate in both wild-type and LAS17-end3C-GFP sla1W41AW108A pan1∆PRD cells ( Figure 6E ) . Thus , productive endocytic actin assembly appears to be triggered in an ‘all or nothing’ manner once the quantity of Las17 ( as well as Vrp1 , based on Figure 6—figure supplement 1A ) reaches a threshold level , regardless how recruitment occurs . Previous studies indicated that Las17 and the Myo3/5 ( type I myosin ) -Vrp1 complex are the two major NPFs for endocytic actin assembly ( Sirotkin et al . , 2005; Sun et al . , 2006 ) . Vrp1 is required for Myo3/5’s recruitment and NPF activation . As we showed above , Myo3/5 arrives at endocytic sites with similar timing to the onset of actin assembly in both wild-type and sla1W41AW108A pan1∆PRD mutant cells ( Figure 5C–F , Figure 6C and D ) . Thus , similar to actin assembly , Myo3/5 recruitment also appears to occur in a switch-like manner upon Vrp1 accumulation to a threshold level . To further explore this mechanism , we next examined how type I myosin recruitment affects the onset endocytic actin assembly . To do this , we altered the mechanism for type I myosin recruitment to endocytic sites by fusing end3C to Myo5 in myo3∆ strain . MYO5-end3C las17WCA∆ myo3∆ cells grow much better than myo5CA∆ las17WCA∆ myo3∆ cells , suggesting that NPF activity is retained in Myo5-end3C fusion protein ( Figure 5—figure supplement 2D ) . As expected , Myo5-end3C-GFP was now recruited ( through an end3C-Pan1 interaction ) to endocytic sites much earlier than in wild-type cells relative to other proteins such as Las17 ( Figure 6C and Figure 7A ) , and it had a longer lifetime ( Figure 7B ) . The timing of recruitment and lifetime of Myo5-end3C-GFP are similar to those of Las17 and Vrp1 ( Figure 7A and B ) ( Sun et al . , 2006 ) . However , actin assembly was not triggered during the first 12 . 6 ± 4 . 9 s of Myo5-end3C-GFP lifetime ( Figure 7A–C , Video 8 ) , even though Myo5-end3C and Vrp1 ( as well as Las17 ) were both present at cortical patches ( Figure 7A ) . Thus , the co-existence of type I Myosin and Vrp1 at endocytic sites is not sufficient to trigger actin assembly . Strikingly , in these mutant cells in which type I myosin arrives at endocytic sites with altered timing , the onset of actin assembly still coincided with the same apparent threshold level of Vrp1 recruitment ( Figure 7D and Figure 7—figure supplement 1 ) . These results suggest that NPF activation of type I myosin by Vrp1 also occurs in an ‘all or nothing’ manner . More importantly , these data further support a model in which recruitment of a threshold level of Las17 and Vrp1 plays a decisive role in switch-like initiation of productive actin assembly in vivo . 10 . 7554/eLife . 29140 . 026Figure 7 . The onset of actin assembly coincides with WIP recruitment to a threshold level in cells expressing Myo5-end3C-GFP . ( A ) Circumferential kymograph presentations of GFP- and RFP- or TagRFP-T-tagged proteins in the indicated yeast strains . The MYO3 gene was knocked out in all strains . ( B ) Lifetimes of GFP- and RFP-tagged proteins in the indicated strains . ( C ) Time difference between the arrival of GFP- and RFP tagged proteins in the indicated strains . ( D ) Alignment of averaged intensity profiles of GFP- and mCherry or RFP-tagged proteins in the indicated yeast cells . For more details , please also see Figure 7—figure supplement 1 . The scale bars in kymographs are 20 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 02610 . 7554/eLife . 29140 . 027Figure 7—figure supplement 1 . Alignment of averaged ( mean ±SD ) intensity measurements for GFP- and mCherry or RFP-tagged proteins from the indicated yeast strains . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 02710 . 7554/eLife . 29140 . 028Video 8 . Dynamics of Myo5-GFP and Abp1-RFP in wild-type cells and dynamics of Myo5-end3C-GFP and Abp1-RFP in myo3∆ cells . Time to acquire one image pair is 1 . 0 s . Interval between frames is 1 . 0 s . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 028 The yeast WASP Las17 PRD and the type I myosin Myo3/5 SH3 domain have previously been shown to interact with Sla1’s SH3 domains and with Pan1’s PRD in vitro , respectively , to regulate Las17 and Myo3/5-Vrp1 NPF activity ( Barker et al . , 2007; Feliciano and Di Pietro , 2012; Rodal et al . , 2003 ) . However , the results from the current study strongly indicate that the crucial in vivo function of two Sla1 SH3 domains and the Pan1 PRD is instead to recruit components of the WASP-Myosin module through formation of a robust interaction network ( discussed further in the following sections ) , triggering actin assembly and orchestrating force production at endocytic sites . We propose that the Sla1 SH3 domains play a primary role in guiding WASP-Myosin module proteins to endocytic sites by interacting with Las17 ( which in turn interacts with Vrp1 ) , while the Pan1 PRD plays a supportive role ( Figure 8A ) . In mutant cells in which either the two Sla1 SH3 domains are mutated ( Figure 3B ) or the Las17 PRD is partially truncated ( Feliciano and Di Pietro , 2012 ) , coordination of cortical Las17 and Sla1 recruitment is disrupted . In these cases , the Pan1 PRD can still interact with Myo3/5 ( Barker et al . , 2007 ) , which interacts with Vrp1 and Las17 ( Evangelista et al . , 2000 ) , to organize the actin machinery at endocytic sites ( Figure 8A ) . It is also possible that the Pan1 PRD interacts with other SH3 domain-containing proteins , which interact with Las17 . When both the Sla1 SH3 domains and the Pan1 PRD are mutated , the WASP-Myosin module components fail to be recruited to endocytic sites ( Figure 8B ) . Future studies should explore whether the Pan1 PRD and/or Vrp1 PRD interact with other SH3 domain-containing endocytic proteins besides Myo3/5 ( Figure 8A ) . 10 . 7554/eLife . 29140 . 029Figure 8 . Switch-like activation of the Arp2/3 complex mediated by SH3 domain-PRM mediated interactions during yeast CME . Models for roles of SH3 domain-PRM interactions in spatiotemporal regulation of actin assembly in wild-type cells ( A ) , in sla1W41AW108A pan1∆PRD cells ( B ) , in sla1W41AW108A pan1∆PRD LAS17-end3C-GFP or sla1W41AW108A pan1∆PRD VRP1-end3C-GFP cells ( C ) . Blue represents PRD containing proteins . Yellow represents SH3 domain-containing proteins . The gray two-way arrows indicate SH3 domain-PRM interactions previously demonstrated by in vitro assays . The dotted line with a question mark indicates possible SH3 domain-PRM interactions . The purple arrows indicate non-SH3 domain-PRM interactions that have previously been identified . The green arrow indicates the interaction between end3C and pan1∆PRD . The grey half circle indicates that multiple proteins function as a complex . The switch symbol marks the onset of actin assembly . See Discussion for description . DOI: http://dx . doi . org/10 . 7554/eLife . 29140 . 029 In our model , Pan1 and End3 recruit Sla1 ( Sun et al . , 2015 ) , and the Pan1-End3-Sla1 complex guides endocytic actin assembly to endocytic sites through an SH3-PRM interaction network . The metazoan counterpart of Pan1-End3-Sla1 is likely intersectin ( ITSN ) , which has both EH domains and SH3 domains in a single protein ( Tsyba et al . , 2011 ) . ITSN was reported to interact with N-WASP and to regulate actin assembly ( Hussain et al . , 2001; McGavin et al . , 2001 ) . Furthermore , recent studies suggest that the vaccinia virus protein A36 recruits ITSN1 to the virus prior to its actin-based motility , and that ITSN1 promotes N-WASP-dependent actin polymerization ( Donnelly et al . , 2013; Humphries et al . , 2014; Snetkov et al . , 2016 ) . Thus , the Pan1-End3-Sla1 function revealed here may reflect a general role of ITSN in spatiotemporally regulating N-WASP-dependent actin polymerization in various cellular processes . Here , we developed a novel ‘end3C fusion’ method to recruit proteins to endocytic sites independent of the Sla1 SH3 domains and the Pan1 PRD . One important feature of this method is that the end3C fusion does not affect the native expression levels of the proteins being tagged . Underproduction or overproduction of Las17 or Vrp1 tends to dramatically influence their cellular functions ( Shively et al . , 2013 ) , likely complicating interpretations . Using this powerful in vivo system , we were able to demonstrate that Las17 and Vrp1 preferentially interact with each other , and together they are necessary and sufficient to recruit the remaining WASP/Myosin module components to endocytic sites in sla1W41AW108A pan1∆PRD cells . Several lines of evidence support the conclusion that once they are recruited by the multivalent Pan1-End3-Sla1 complex , Las17 and Vrp1 in turn recruit their binding partners to endocytic sites through additional multivalent PRM and SH3-domain interactions , greatly expanding the interaction network . The Las17 PRD ( proline-rich domain ) contains 20 SH3-binding PRMs ( proline-rich motifs ) ( Feliciano and Di Pietro , 2012 ) . Vrp1 is very rich in proline and also contains more than 20 PRMs ( Donnelly et al . , 1993 ) . In budding yeast , there are only 25 SH3 domain-containing proteins in total . Remarkably , 11 of them interact with Vrp1 and/or Las17 through PRM-SH3 domain interactions ( Anderson et al . , 1998; Geli et al . , 2000; Rajmohan et al . , 2009; Tarassov et al . , 2008; Tong et al . , 2002; Tonikian et al . , 2009; Verschueren et al . , 2015 ) . These proteins arrive at endocytic sites with similar timing to , or after , Las17 and Vrp1 ( Figure 8A ) ( Boettner et al . , 2011 ) . Some Vrp1- and Las17-binding proteins themselves have multiple SH3 domains , either in their primary sequences or as the result of dimerization or oligomerization . For example , Bzz1 contains two SH3 domains and the hetero-dimeric N-BAR protein ( Kishimoto et al . , 2011 ) Rvs161/167 dimerizes through N-BAR domains ( Colwill et al . , 1999 ) . In addition , in vitro studies have shown that some of these SH3 proteins are able to bind to multiple PRMs of the Las17 PRD ( Tong et al . , 2002 ) . Multiple dimeric and oligomeric SH3 proteins can interact in a complex network with the PRMs of Las17 and Vrp1 , resulting in highly cooperative binding . The partial rescue of function when end3C is fused to Bzz1 , Rvs167 , or Myo5 ( Figure 5B ) , supports the notion that these proteins promote interactions that concentrate key actin assembly factors at endocytic sites . Cooperativity in binding of multivalent proteins would greatly increase the robustness of the network and could promote actin nucleation activity ( discussed further in the following section ) , facilitating the rapid recruitment and activation of the endocytic actin machinery ( Figure 8A ) . Interestingly , the proteins mentioned here are well conserved in mammals ( Cheng et al . , 2012; Merrifield and Kaksonen , 2014 ) and other organisms ( Lam et al . , 2001; Sirotkin et al . , 2005; Xin et al . , 2013 ) , indicating that multivalent PRM-SH3 network formation centered around WASP and WIP may be a general feature in spatiotemporal control of Arp2/3 complex-mediated actin assembly . Endocytic actin nucleation mainly depends on the yeast WASP ( Las17 ) and the type 1 myosin-WIP complex ( Myo3/5-Vrp1 ) ( Galletta et al . , 2008; Lewellyn et al . , 2015; Sirotkin et al . , 2005; Sun et al . , 2006 ) . Previous in vitro data suggest that Las17 NPF activity is constitutive ( Feliciano and Di Pietro , 2012; Rodal et al . , 2003 ) . Vrp1 is required for Myo3/5 recruitment and NPF activation ( Sirotkin et al . , 2005; Sun et al . , 2006 ) . However , since Las17 and Vrp1 arrive at endocytic sites 15–20 s prior to actin assembly and Myo3/5 recruitment , the mechanism that controls the onset of Arp2/3 activation in vivo has been mysterious . Our quantitative analysis of Las17 and Vrp1 recruitment in different genetic backgrounds provides several important new insights into NPF-mediated actin nucleation regulation by Las17 and Myo3/5-Vrp1 in vivo . Las17 NPF activity does not trigger immediate actin nucleation at endocytic sites even when we genetically disable Las17 inhibition by Sla1 in live cells ( Figure 5D ) . Co-existence of Vrp1 and type I myosin at endocytic sites is not sufficient to induce actin assembly until Las17 and Vrp1 levels rise to an apparent threshold level ( Figure 7D ) . Thus , in contrast to what was previously assumed based on in vitro studies ( Feliciano and Di Pietro , 2012; Rodal et al . , 2003; Sirotkin et al . , 2005; Sun et al . , 2006 ) , the Las17 NPF and the Vrp1-dependent type I myosin NPF activities do not appear to be active when present at low concentrations . Importantly , our quantitative analysis strongly suggested that productive actin assembly initiation is tightly coupled to accumulation of a threshold concentration of Las17 and Vrp1 at endocytic sites ( Figure 5G–5I and Figure 8 ) . Furthermore , the actin assembly rate appears to be very similar irrespective of how and when Las17 and Vrp1 are recruited to endocytic sites at sufficient levels , and assembly appears to be ‘all or nothing’ ( Figure 6E ) , leading us to propose a ‘switch-like’ activation mechanism . Previous studies proposed a hierarchical model for N-WASP NPF activation: allostery and dimerization , which control accessibility and affinity of the N-WASP VCA for the Arp2/3 complex , respectively ( Higgs and Pollard , 2000; Padrick et al . , 2008; Padrick and Rosen , 2010; Rivera et al . , 2009 ) . Our results provide support for a similar mechanism in cells . Las17 does not appear to contain a G protein binding domain , and its NPF activity is not auto-inhibited . Therefore , Las17 does not depend on allosteric control to gain ‘basal’ NPF activity , explaining previous in vitro results ( Feliciano and Di Pietro , 2012; Rodal et al . , 2003 ) . However , our results suggest that the ‘basal’ Las17 NPF activity is not sufficient to trigger actin assembly unless Las17 and Vrp1 are concentrated to a threshold level in vivo . We suggest that multivalent PRM and SH3-domain interactions between Las17 , Vrp1 and their binding partners ( discussed in the previous section ) , at endocytic sites induces formation of a higher-order complex , in which two or more VCAs are brought together to enhance affinity for the Arp2/3 complex . Consistently , previous studies using whole cell exacts found that Las17 is part of a large and biochemically stable complex ( Feliciano and Di Pietro , 2012; Soulard et al . , 2002 ) . In our proposed scenario , Las17 and Vrp1 accumulate at endocytic sties to a threshold level and their PRMs provide a high local concentration of multivalent interactions through which Myo3 and Myo5 are also recruited and activated in an ‘all or nothing’ manner . Thus , Las17 NPF activation and Vrp1-dependent Myo3/5 recruitment and NPF activation are triggered simultaneously , creating a burst of actin filament assembly , upon which the Myo3/5 motor domains can exert forces , collectively generating forces required for endocytic membrane invagination and membrane scission . The relationship between Las17 and Vrp1 recruitment and the switch-like onset of actin assembly we observe in live cells is consistent with a mechanism based on in vitro studies in which multivalent SH3-domain and PRM interactions induce a phase transition centered around N-WASP to promote local actin assembly ( Banjade and Rosen , 2014; Li et al . , 2012 ) . We speculate that the cytoplasmic WASP-Myosin module puncta propelled by actin comet tails observed when endocytic site formation is uncoupled from actin assembly ( Pan1-End3 depleted cells [Sun et al . , 2015] or sla1W41AW108A pan1∆PRD cells [Figure 4F] ) is triggered by multivalent SH3-domain and PRM interactions in the cytoplasm . Consistently , removing multi-PRM-containing proteins ( Las17 or Vrp1 ) from the cytoplasm , completely suppressed cytoplasmic WASP-Myosin module puncta formation and assembly of associated actin comet tails ( Figure 5C–F ) . Overall , our results strongly support a model in which accumulation of WASP and WIP to a threshold level at endocytic sites establishes a robust , multivalent SH3 domain-PRM interaction network ( possibly involving a transient phase separation [Banjade and Rosen , 2014; Li et al . , 2012] ) , which triggers the onset of actin assembly in a switch-like manner in vivo . It is certainly possible that the other factors , such as lipids and other endocytic modules involved in endocytic site establishment or cargo loading , also influence the network and actin assembly onset . Future efforts need to assess how such factors cooperate with WASP and WIP to facilitate productive endocytic actin assembly . Yeast strains were grown in standard rich media ( YPD ) or synthetic media ( SD ) supplemented with 0 . 2% Casamino acids . The yeast strains are listed in Supplementary file 1 . GFP , mCherry , RFP , TagRFP-T and end3C tags were integrated at the C-terminus of each gene as described previously ( Lee et al . , 2013; Longtine et al . , 1998 ) . Fluorescence microscopy was performed using a Nikon Eclipse Ti microscope ( Nikon Instruments , Melville , NY ) controlled by Metamorph ( Molecular Devices , Sunnyvale , CA ) , equipped with a Plan Apo VC 100×/1 . 4 Oil OFN25 DIC N2 objective ( with Type NF immersion oil , Nikon ) , a Perfect Focus System ( Nikon ) , and a Neo sCMOS camera ( Andor Technology Ltd . , South Windsor , CT ) ( 65 nm effective pixel size ) . For live cell imaging , cells were grown to early log phase at 25˚C . The cells in synthetic media were adhered to the surface of a concanavalin A coated ( 0 . 1 µg/ml ) coverslip . All imaging was done at room temperature . For single-channel live cell imaging , images were acquired continuously at 1 frames/sec . Two-channel movies were made using the SPECTRA X Light Engine ( Lumencor , Beaverton , OR ) for excitation with a 524/628 nm dual-band bandpass filter for GFP/mCherry emission ( Brightline , Semrock , Lake Forest , IL ) . Time to acquire one image pair is 1 . 1 s , or 1 . 3 s , or 1 . 8 s depending on the signal intensity . For multifocus imaging of Sac6-GFP labeled actin comet tails , images were acquired continuously at 4 frames/sec using multifocus microscopy ( MFM ) as described previously ( Abrahamsson et al . , 2013 ) . Image J software was used for general manipulation of images and movies , for preparing kymographs , and for data analysis and quantification . For the detailed analysis procedure , please see Figure 1—figure supplement 2 and Figure 2—figure supplement 1 . For patch lifetimes , unless otherwise stated , more than 100 patches were measured for each variant . Patches that at any point in their lifetime were too close to another patch to be clearly resolved were excluded from our analysis . For the fluorescence intensity profile , at least 10 patches were measured for each variant . These sample sizes are more than what has been used in other landmark papers in this type of study ( Bradford et al . , 2015; Kaksonen et al . , 2003; Newpher et al . , 2005 ) . For each pair of variables , pooled data of analysis were compared by a two sided Mann-Whitney test using the Prism 7 graphing software . Lucifer Yellow uptake assay was also done as previously described ( Sun et al . , 2006 ) . Cells were grown to early log phase in YPD media . Approximately 1 × 107 cells were pelleted and resuspended in 90 μl of YPD and 10 μl of 40 mg/ml Lucifer yellow CH dilithium salt . Cells were incubated for 90 min at room temperature and then washed four times in ice-cold 50 mM potassium phosphate buffer , pH 7 . 4 , containing 10 mM NaN3 and 10 mM NaF . Cells were then imaged at room temperature .	Actin is one of the most abundant proteins in yeast , mammalian and other eukaryotic cells . It assembles into long chains known as filaments that the cell uses to generate forces for various purposes . For example , actin filaments are needed to pull part of the membrane surrounding the cell inwards to bring molecules from the external environment into the cell by a process called endocytosis . In yeast , a member of the WASP family of proteins promotes the assembly of actin filaments around the site where endocytosis will occur . To achieve this , WASP interacts with several other proteins including WIP and myosin , a motor protein that moves along actin filaments to generate mechanical forces . However , it was not clear how these proteins work together to trigger actin filaments to assemble at the right place and time . Sun et al . addressed this question by studying yeast cells with genetic mutations affecting one or more of these proteins . The experiments show that WASP , myosin and WIP are recruited to sites where endocytosis is about to occur through specific interactions with other proteins . For example , a region of WASP known as the proline-rich domain can bind to proteins that contain an “SH3” domain . WASP and WIP arrive first , stimulating actin to assemble in an “all and nothing” manner and attracting myosin to the actin . Further experiments indicate that WASP and WIP need to reach a threshold level before actin starts to assemble . The findings of Sun et al . suggest that WASP and WIP play key roles in establishing the network of proteins needed for actin filaments to assemble during endocytosis . These proteins are needed for many other processes in yeast and other cells , including mammalian cells . Therefore , the next steps will be to investigate whether WASP and WIP use the same mechanism to operate in other situations .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2017	Switch-like Arp2/3 activation upon WASP and WIP recruitment to an apparent threshold level by multivalent linker proteins in vivo
To survive challenging environments , animals acquired the ability to evaluate food quality in the intestine and respond to nutrient deficiencies with changes in food-response behavior , metabolism and development . However , the regulatory mechanisms underlying intestinal sensing of specific nutrients , especially micronutrients such as vitamins , and the connections to downstream physiological responses in animals remain underexplored . We have established a system to analyze the intestinal response to vitamin B2 ( VB2 ) deficiency in Caenorhabditis elegans , and demonstrated that VB2 level critically impacts food uptake and foraging behavior by regulating specific protease gene expression and intestinal protease activity . We show that this impact is mediated by TORC1 signaling through reading the FAD-dependent ATP level . Thus , our study in live animals uncovers a VB2-sensing/response pathway that regulates food-uptake , a mechanism by which a common signaling pathway translates a specific nutrient signal into physiological activities , and the importance of gut microbiota in supplying micronutrients to animals . To survive challenging environments with fluctuating nutrient resources , animals have acquired the ability to evaluate food quality , which may lead to avoidance of food lacking certain essential nutrients or containing toxic molecules , and to alterations in developmental and metabolic programs ( Bargmann , 2006; Bjordal et al . , 2014; Chi et al . , 2016; Chng et al . , 2014; Ha et al . , 2010; Iwatsuki and Torii , 2012; Kniazeva et al . , 2015; Melo and Ruvkun , 2012; Tang and Han , 2017; Watson et al . , 2014 ) . Besides neuronal sensory systems that permit rapid feeding decision , food quality is also evaluated by the intestine-initiated systems that may be more sensitive in detecting the deficiency of certain types of nutrients including micronutrients and are capable of dictating changes in cellular/developmental programs as well as food uptake and seeking behaviors . However , the metabolic and signaling events in the intestine that are involved in evaluating the availability of specific nutrients , and the mechanisms underlying the connection of these signaling activities to food uptake/foraging behaviors , as well as developmental programs , remain largely unexplored . More specifically , although the functions of TOR complexes in responding to cellular nutrient changes ( e . g . amino acids and ATP ) have been extensively studied ( Chin et al . , 2014; Dennis et al . , 2001; Efeyan et al . , 2015; Zhu et al . , 2013 ) , the functions of these sensing activities in whole animals under true physiological conditions , including responses to nutrient variations in food , remain to be investigated . Vitamin B2 ( VB2 ) is the precursor and component of flavin mononucleotide ( FMN ) and flavin adenine dinucleotide ( FAD ) that are the redox cofactors of a large number of flavoproteins involved in various metabolic pathways ( Joosten and van Berkel , 2007; Lienhart et al . , 2013; Powers , 2003 ) . Animals obtain VB2 from diet and likely also from gut microbes , although the VB2 contribution from gut microbes has not been well documented . VB2 deficiency has been associated with various human diseases and health problems ( Powers , 2003 ) . It would thus be reasonable to speculate that a food-quality monitoring system in animals can sense VB2 in food and then regulate food response behaviors , and such a monitoring system may function in the intestine . Such a potential VB2-sensing mechanism is explored in this study . When fed live E . coli strain OP50 , a standard laboratory food for C . elegans , all hatched larvae develop to adults within 3 days at 20°C . However , when worms were fed OP50 that was killed by heat treatment ( 75°C for 90 min; HK-OP50 hereafter ) , they arrested development at early larval stages ( L1-L2 ) and failed to consume the bacterial lawn ( Figure 1A ) , suggesting that the heat-killed bacteria lacked certain nutrients or molecules required for larval growth , which is consistent with published observation suggesting that E . coli contains heat-labile nutrients required for C . elegans normal growth and longevity ( Lenaerts et al . , 2008 ) . We performed two behavior assays to assess food dwelling and food choice ( Brandt and Ringstad , 2015; Fujiwara et al . , 2002; Kniazeva et al . , 2015; Melo and Ruvkun , 2012; Shtonda and Avery , 2006 ) . The results demonstrated a strong discrimination against HK-OP50 by wild-type worms ( Figure 1B and C ) . Interestingly , in the food choice assay , the live OP50 lawn was favored until it was consumed at day 4 , when most worms then moved to the HK-OP50 and consumed that lawn ( Figure 1C ) , suggesting that worms eat heat-killed bacteria if they have first obtained some benefit from live bacteria . Moreover , when live-OP50 was added to worms grown on HK-OP50 plate for 30 days , they recovered to adults with viable progeny ( Figure 1—figure supplement 1A ) , suggesting nutrient deficiency in HK-OP50 induced a protective response from the worms similar to starvation response . 10 . 7554/eLife . 26243 . 003Figure 1 . C . elegans do not eat heat-killed bacteria without interacting with live bacteria . ( A ) Microscope images and bar graph showing that worms fed heat-killed OP50 ( HK-OP50 or HK ) arrested at L1-L2 stage three days after hatching . ( B ) Schematic drawing and quantitative data of the food dwelling assay . Circles indicate the food spot for live ( green ) and HK-OP50 ( red ) bacteria , respectively . The animals were scored 24 hr after L1 worms were placed on the food spot . Data are represented as mean ±SD . ( C ) Schematic drawing , microscopic images and quantitative data of the food choice assay . Eggs were place in the center spot ( origin ) . Live ( green ) and heat-killed ( red ) bacteria were placed on opposite sides of the plate . The percentage of worms on each spot was calculated at the indicated time . Worms moved to heat-killed bacteria and consumed it after 4 days when live food was totally consumed . Data are represented as mean ±SEM . ( D ) An assay for the effect of a small amount of live , non-edible bacteria on the growth of animals fed heat-killed bacteria . The colored diagrams show the feeding conditions . Percentage of animals that grew to L3-adult stages is indicated below each diagram . The combination of HK-OP50 and small amount live SS , neither of which alone can support growth , could support food uptake and growth . The control ( 4th column ) suggests that only live SS can promote worms to consume HK-OP50 . For bar graphs , number of worms scored is indicated in each bar . All data are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 003 10 . 7554/eLife . 26243 . 004Figure 1—source data 1 . Numerical data of Figure 1A–1D and Figure 1—figure supplement 1C . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 004 10 . 7554/eLife . 26243 . 005Figure 1—figure supplement 1 . Impact of interaction with live bacteria on the ability of C . elegans to consume heat-killed bacteria . ( A ) Worms on heat-killed OP50 ( growth for 30 days ) can be recovered once they eat live OP50 . Once live-OP50 ( yellow arrow and dotted line ) was added to worms ( red arrow ) grown on HK-OP50 plate for 30 days , they recovered to adults with viable progeny . This result suggests that nutrient deficiency in HK-OP50 induced a protective response from the worms similar to starvation response . ( B ) Representative images of animals in the assay described in Figure 1D . Only under the HK-OP50+ live SS condition , worms were able to consume all heat-killed bacteria and grow . ( C ) Cartoon illustrations , microscopic images and quantitative data showing that live bacteria do not influence the usability of heat-killed bacteria through odorants . In the middle cartoon , two petri dishes are stacked with top ( opening ) facing each other so that the worms/food on the agar pads are exposed to common air space , >100 worms were scored for each testing condition . S . S . = Staphylococcus saprophyticus . . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 005 To further investigate the impact of this interaction with live bacteria , we employed Staphylococcus saprophyticus ( SS ) , a bacterium species that is inedible for C . elegans ( Figure 1D ) . When eggs were placed on HK-OP50 lawn containing a tiny amount of live SS , the hatched worms were able to consume HK-OP50 and grow , suggesting that the live bacteria provided micronutrients that rendered the heat-killed bacteria edible ( Figure 1D and Figure 1—figure supplement 1B ) . Additional tests indicated that such potential components did not reach the worms through odorants and suggested that worms likely obtain the benefit from the small amount live bacteria through ingestion ( Figure 1—figure supplement 1C; Figure 1C ) . These data led to a speculative model that the trace amount of live bacteria , ingested by the worms , stayed in the intestine and continuously secreted micronutrients into the gut tract , which is similar to microbes in mammalian gut . Deficiency of certain nutrients in HK-OP50 may be detected by a potential food quality evaluation mechanism leading to changes in food uptake and foraging behavior . Among micronutrients tested , we found that vitamin B2 ( VB2 ) supplementation partially but significantly recovered larval growth as indicated by increased gonad length at day 7 , which indicated the increased usage of the food ( Figure 2A and Figure 2—figure supplement 1A ) . These animals fed VB2 supplemented food still grew very slowly , reaching mid to late larval stages based on gonad length and vulval development compared to that of well-fed worms , but never reach adulthood by day 7 ( Figure 2—figure supplement 1B , C ) . This slow growth indicated that VB2 supplementation does not completely compensate for the undefined deficiencies of heat-killed bacteria , which suggested that multiple essential nutrients are lacking in heat-killed bacteria . However , the prominent increase in growth suggested that VB2 may play a critical role in a food-quality evaluation pathway . Our behavior assays showed that VB2 supplementation increased the preference for heat-killed bacteria ( Figure 2B and Figure 2—figure supplement 2A ) . Using HPLC-UV ( Howe et al . , 2015 ) , we found the VB2 level was drastically reduced in worms fed HK-OP50 , compared to worms fed live OP50 , and the level was mostly recovered by VB2 supplementation ( Figure 2C ) , suggesting VB2 was very low in heat-killed bacteria . To confirm this , we measured the VB2 level in bacteria and found that VB2 level is low in heat-killed OP50 ( Figure 2—figure supplement 2C ) . 10 . 7554/eLife . 26243 . 006Figure 2 . Vitamin B2 ( VB2 ) supplementation increases the usage of heat-killed bacteria and promotes intestinal protease activity . ( A ) Scatter plot of gonad length of worms fed heat-killed OP50 supplemented with indicated vitamins scored at Day 7 . > 28 worms were scored for each sample . Representative images are shown in Figure 2—figure supplement 1A . Data are represented as mean ±SD . ( B ) Food-choice assay to test the effect of VB2 supplementation on animal dwelling behavior . The worms were scored for the ratio between the two locations at 3 day . Data are represented as mean ±SD . ( C ) HPLC-UV analysis of VB2 extracted from worms fed live , HK-OP50 and HK-OP50+VB2 supplementation . The bar graph represents the area counts of VB2 UV absorption peaks ( identified by Analyst software ) , the mean ±SEM of three independent experiments . A graph for VB2 standard from HPLC-UV analysis is shown in Figure 2—figure supplement 2B . ( D ) Scatter plot showing the effect of treating HK-OP50 with indicated enzymes , on larval growth ( by measuring gonad length ) . Only protease treatment significantly promoted growth . The number of worms scored was 52 , 37 , 52 and 52 , respectively . Data are represented as mean ±SD . ( E ) Scatter plot showing the effect of a protease inhibitor cocktail ( PIC ) on the growth of larvae fed HK-OP50 supplemented with VB2 . The number of worms scored was 32 , 44 and 18 , respectively . Data are represented as mean ±SD . ( F ) Fluorescence images and bar graph showing the impact of heat-killed bacteria and VB2 supplementation on in vivo protease activity in the intestinal tract by the EnzChek protease assay ( Hama et al . , 2009 ) . L2 stage worms were incubated with quenched BODIPY TR-X casein for 3 hr before imaging . Data are represented as mean ±SEM . p-Values were calculated by T-test and p<0 . 05 was considered a significant difference . For bar graphs , number of worms scored is indicated in each bar . All data are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 006 10 . 7554/eLife . 26243 . 007Figure 2—source data 1 . Numerical data of Figure 2A–2F and Figure 2—figure supplement 2A , D and E . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 007 10 . 7554/eLife . 26243 . 008Figure 2—figure supplement 1 . Gonad length of worms under different conditions . Under heat-killed bacteria feeding conditions , animals were very unhealthy so that stage progression may not correlate with growth well . Larval growth in this paper as indicated by increase in gonad length . ( A ) Microscopic images of larvae fed HK-OP50 with various vitamin supplements . Gonad length of each worm was measured using ImageJ at Day 7 , to evaluate progression of larval growth , which reflects the consumption of the food . The statistical data are shown in the scatter plot in Figure 2A . ( B ) Gonad length ( dotted line ) of wild-type worms fed live OP50 in different developmental stages , and the vulval morphology ( green arrow ) of L4 worms . ( C ) Gonad length of wild-type worms fed HK-OP50 or HK+VB2 food . If purely based on gonad length ( dotted line ) , worms fed HK food would be at about late L1 to early L2 , worms fed HK+VB2 food would be at about late L2 to late L3 stages , comparing to well-fed worms in ( B ) . However , the vulval morphology ( yellow arrow ) of some worms fed HK+VB2 food appeared to reach L4 stage . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 00810 . 7554/eLife . 26243 . 009Figure 2—figure supplement 2 . Impact of vitamin B2 supplementation on the ability of C . elegans to consume heat-killed bacteria . ( A ) Cartoon drawing and quantitative data of the food dwelling assay showing the impact of VB2 supplementation on this behavior . Yellow and red filled circles in the cartoon indicate the spots of HK-OP50 ( HK ) +/− VB2 supplement , respectively . L1 worms were placed on the food spot and scored for the percent that stayed within the spot after 24 hr . n=number of worms scored . Data are represented as mean ±SD . ( B ) Chromatograph from HPLC-UV analysis of the VB2 standard for the analysis shown in Figure 2C . ( C ) HPLC-UV analysis of VB2 extracted from live and heat-killed OP50 . Arrow and dotted line indicate the peak of VB2 . ( D ) BODIPY-labeled protease activity in the intestinal tract of worms fed live OP50 at four different larval stages . Fluorescence intensity was measured by ImageJ software . Data are represented as mean ±SEM . ( E ) Effects of VB2 supplementation and treatment of protease inhibitor on in vivo protease activity in worms fed HK-OP50 . Data are represented as mean ±SEM . p Values were calculated by T-test . and p<0 . 05 was considered a significant difference . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 009 To understand how VB2 affects food uptake in worms , we tested the role of the digestive system by feeding worms with HK-OP50 culture pre-treated with exogenous digestive enzymes . We found that larval growth was enhanced only by protease-treated HK-OP50 ( Figure 2D ) , indicating that exogenous protein digestion improved usability of the heat-killed bacteria . We then observed that HK-OP50 supplemented with a protease inhibitor cocktail ( PIC ) eliminated the beneficial effect of VB2 supplementation ( Figure 2E ) . These results suggest that worms fed HK-OP50 have reduced protease activity , and that the benefit of VB2 supplementation is dependent on endogenous proteases . We then used a protease detection assay ( Hama et al . , 2009; Semova et al . , 2012 ) to directly examine the in vivo protease activity in the intestinal tract ( Figure 2—figure supplement 2D and E ) . We found that the protease activity was drastically reduced in worms fed HK-OP50 , compared to worms fed live OP50 , and VB2 supplementation partially recovered the protease activity ( Figure 2F ) . Taken together , these data support that worms down-regulate digestive enzymes in response to VB2 deficiency . To identify the proteases regulated by VB2 , we examined the mRNA levels of 27 intestine-enriched , protease genes ( McGhee et al . , 2007 ) . qRT-PCR analysis showed that the mRNA levels of four genes , asp-13 , asp-14 , cpr-1 and cpr-4 , were dramatically increased in worms fed HK-OP50+VB2 compared to that in worms fed HK-OP50 ( Figure 3A ) . Among these four genes , asp-13 and asp-14 were also highly expressed in worms fed on live bacteria ( Figure 3A ) . We then found that knocking down both asp-13 and asp-14 resulted in a strong suppression of the effect of VB2 supplementation ( Figure 3B ) , suggesting that these enzymes are essential for the VB2 effect . Over-expression of both asp-13 and asp-14 behind a ubiquitously expressed rpl-28 promoter ( termed asp-13/14 OE ) ( Zhang et al . , 2009 ) significantly improved the food uptake and dwelling behavior of worms fed HK-OP50 , as determined by the gonad length , food choice and seeking assays ( Figure 3C–E ) . Moreover , intestinal-specific RNAi of asp-13/14 prevents the rescue effect of VB2 supplementation on worms fed HK-food ( Figure 3—figure supplement 1A and B ) , suggesting that intestinal function of ASP-13 and ASP14 are essential for the VB2 effect . This is consistent with the previous finding that asp-13/14 is highly expressed in the intestine ( McGhee et al . , 2007 ) and the results from our in vitro protease/protease inhibitor assay ( Figure 2D ) . These data strongly support a critical role of ASP-13 and ASP-14 in the VB2-induced increase in protease activity and uptake of heat-killed bacteria . 10 . 7554/eLife . 26243 . 010Figure 3 . Proteases ASP-13 and ASP-14 and GATA factor ELT-2 play critical roles in VB2-promoted food usability and animal growth . ( A ) Results of qRT-PCR analysis showing the expression of indicated protease genes in wild-type worms fed indicated food . Data are represented as mean ±SEM . ( B ) Impact of RNAi of indicated genes on larval growth when fed HK-OP50+VB2 supplementation . Worms with both asp-13 and asp-14 knocked down displayed strongest effect . Data are represented as mean ±SEM . ( C ) Scatter plot showing the impact of over-expression of the two protease genes behind the rpl-28 promoter , measured by gonad length of larvae fed HK-OP50 at Day 7 . n = 41 , 39 and 52 , respectively . Data are represented as mean ±SD . ( D and E ) Cartoon diagram and data from food choice ( D ) and food seeking ( E ) assays showing that overexpression of asp-13 and asp-14 ( asp-13/14 OE ) eliminated the discrimination against HK-OP50 over HK-OP50+VB2 ( D ) and improved affinity of worms towards HK-OP50 ( E ) . Data from two different overexpression lines are shown . Data are represented as mean ±SEM . ( F ) Fluorescence images and bar graph showing protease activity in the intestinal tract is dramatically decreased in worms treated with elt-2 ( RNAi ) . Data are represented as mean ±SEM . ( G ) Fluorescence images ( integrated translational ELT-2::GFP reporter strain ) and Western blot ( wild-type strain ) showing that the ELT-2 expression is prominently decreased in worms fed HK-OP50 compared to that in worms fed live OP50 , and the expression is largely recovered by VB2 supplementation . Data are represented as mean ±SEM . ( H ) Scatter plot showing larval growth measured by gonad length of worms fed heat-killed OP50 is increased by an elt-2 over-expression transgene ( OE ) at Day 7 . n = 52 for each condition . Data are represented as mean ±SD . ( I ) Fluorescence images and bar graph showing that the protease activity is increased in elt-2 overexpression ( elt-2 OE ) worms fed HK-OP50 . p-Values were calculated by T-test . For bar graphs , number of worms scored is indicated in each bar . Data are represented as mean ±SEM . p-Values were calculated by T-test and p<0 . 05 was considered a significant difference . For bar graphs , number of worms scored is indicated in each bar . All data are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 010 10 . 7554/eLife . 26243 . 011Figure 3—source data 1 . Numerical data of Figures 3B–2I and Figure 3—figure supplements 1A , B , 2A , C and E . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 011 10 . 7554/eLife . 26243 . 012Figure 3—figure supplement 1 . VB2 supplementation-induced increase in protease activity and growth rescue in worms fed HK-food depend on intestinal expression of ASP-13 and ASP-14 . Intestine-specific RNAi of asp-13 and asp-14 blocked the increases in protease activity ( A ) and gonad length ( B ) caused by VB2 supplementation on worms fed HK . Data are represented as mean ±SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 01210 . 7554/eLife . 26243 . 013Figure 3—figure supplement 2 . Roles of protease ASP-13 and ASP-14 , and GATA factor ELT-2 in VB2-induced food uptake and growth of worms fed heat-killed bacteria . ( A ) Representative images and quantitative data from the protease assay on worms treated with RNAi of indicated genes in a screen for transcription factors required for intestinal protease activity . Synchronized L1s were transferred onto plates seeded with individual RNAi bacterial clones . The protease activity in the intestinal tract of L1-L2 worms of the next generation was measured . elt-2 ( RNAi ) displayed the most dramatic decrease ( also see Figure 2D ) . Fluorescence intensity was measured by ImageJ software . Data are represented as mean ±SEM . ( B ) qRT-PCR data showing that mRNA of asp-13 and asp-14 are increased in elt-2 overexpression ( elt-2 OE ) worms fed HK-OP50 . Data are represented as mean ±SD . ( C ) Results of a food choice assay showing that over expression of elt-2 ( elt-2 OE ) eliminated the discrimination of worms against HK-OP50 over HK-OP50+VB2 . Data are represented as mean ±SEM . ( D ) ChIP-qPCR data showing enrichment of ELT-2 binding to asp-13 and asp-14 promoters . Samples from a strain expressing an ELT-2::GFP fusion transgene were subject to immunoprecipitation using either control IgG or anti-GFP antibody . Data are represented as mean ±SD . This result is consistent with the data from a recent ChIP-seq analysis suggested that asp-13 and asp-14 are likely direct targets of ELT-2 ( Mann et al . , 2016 ) . ( E ) Microscopic images and bar graph showing that overexpression of asp-13 and asp-14 significantly suppressed the larval arrest phenotype of elt-2 ( RNAi ) . Arrowheads point to two representative animals to indicate the size difference under the two conditions . Data are represented as mean ±SD . Since over-expressing asp-13 and asp-14 behind the rpl-28 promoter is expected to elude regulation by ELT-2 , this result supports that down-regulating asp-13 and asp-14 by a low ELT-2 level critically contributes to the poor food usage in worms fed HK-OP50 . All data are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 013 To identify transcription factors that regulate the expression of intestinal protease genes in response to VB2 supplementation , we combined RNAi with the protease assay to screen 18 transcription factors previously shown to be enriched in the intestine ( McGhee et al . , 2007 ) ( Figure 3—figure supplement 2A ) . The protease activity was most dramatically decreased in worms treated with elt-2 ( RNAi ) ( Figure 3F and Figure 3—figure supplement 2A ) , which is consistent with the recent report that elt-2 ( RNAi ) decreases the expression of ~50% of known protease genes , including asp-13 and asp-14 ( Mann et al . , 2016 ) . Using an integrated , translational ELT-2::GFP reporter strain which is in a glo-4 mutant background to reduce auto fluorescence ( Mann et al . , 2016 ) , we found that the ELT-2::GFP signal was decreased in worms fed HK-OP50 and the reduction was mostly recovered by VB2 supplementation ( Figure 3G ) . We also found that decreased ELT-2 on heat-killed food was recovered by VB2 supplementation in wild-type worm ( Figure 3G , Western blot ) , supporting that ELT-2 expression responds to changes in nutrient status . Using an established ELT-2 over-expression [elt-2 ( OE ) ] strain ( Mann et al . , 2016 ) , we found that over-expression of ELT-2 significantly increased both larval growth measured by gonad length and protease activity in worms fed HK-OP50 , eliminated the preference for HK-OP50+VB2 over HK-OP50 in the food choice assay , and increased the expression of asp-13 and asp-14 ( Figure 3H and I; Figure 3—figure supplement 2B and C ) . Therefore , increased expression of elt-2 is both necessary and sufficient for the impact of VB2 on protease expression and food usability . Additional tests supported the hypothesis that down-regulating asp-13 and asp-14 by a low ELT-2 level critically contributes to the poor food usage in worms fed HK-OP50 ( Figure 3—figure supplement 2D and E ) . Despite extensive studies on nutrient sensing by TORC1 ( Efeyan et al . , 2015 ) , its role in food quality evaluation processes , which include not only the detection of deficiency of a specific nutrient , but also specific downstream physiological responses in live animals , remain to be explored , particularly for micronutrients such as VB2 . We first tested if TORC1 is required for VB2-induced activation of digestive enzymes in worms fed heat-killed bacteria . We found that pretreatment with RNAi targeting either daf-15/raptor or ragc-1 , encoding two key TORC1 components ( Jia et al . , 2004; Long et al . , 2002; Robida-Stubbs et al . , 2012 ) , essentially eliminated the effect of VB2 supplementation on protease activity ( Figure 4A ) , indicating that VB2 induction of protease activity is TORC1 dependent . To address that lower TORC1 activity on protease activity derive does not simply from slowing or arresting development , we tested the protease activity of several arrested larvae ( RNAi treatment ) . Figure 4—figure supplement 1 shows that protease activity is not all reduced in arrested larvae , which indicate that lower TORC1 activity on protease activity ( Figure 4A ) is unlikely to be from generally slowing or arresting development . 10 . 7554/eLife . 26243 . 014Figure 4 . TORC1 mediates the impact of VB2 on the expression of digest enzymes and food usage . ( A ) Fluorescence images and bar graph showing that VB2-induced increase in protease activity in the digestive tract depends on TORC1 function . Data are represented as mean ±SEM . ( B ) Fluorescence images and bar graph showing that autophagy activity ( measured by the intensity of LGG-1::GFP puncta in seam cells ) increases in worms fed HK-OP50 , compared to worms fed live OP50 , and the increase is eliminated by VB2 supplementation . The inset image in each panel shows magnified area indicated by the white bar . Figure 4—figure supplement 2A and B show data supporting that LGG-1 puncta is repressed by TORC1 activity , consistent with known negative regulation of autophagy by TORC1 ( Robida-Stubbs et al . , 2012 ) . Data are represented as mean ±SEM . ( C ) Fluorescence images and bar graph showing that protease activity is increased in an nprl-3 ( lf ) mutant fed HK-OP50 . nprl-3 negatively regulates TORC1 ( Zhu et al . , 2013 ) . Data are represented as mean ±SEM . ( D ) qRT-PCR analysis showing that mRNA of asp-13 and asp-14 were increased in nprl-3 ( lf ) mutant fed HK-OP50 . Data are represented as mean ±SEM . ( E ) Scatter plot showing that larval growth is increased in the nprl-3 ( lf ) mutant worms fed HK-OP50 at Day 7 . n = 52 for each condition . Data are represented as mean ±SD . ( F and G ) Result from food choice ( F ) and food-seeking ( G ) assays showing that nprl-3 ( lf ) , which hyperactivates TORC1 , eliminated the discrimination against HK-OP50 over HK-OP50+VB2 ( F ) and improved affinity of worms toward HK-OP50 ( G ) . The wild-type data are the same as that in Figure 3D and E , as the data for both pairs of figures were generated from the same set of experiments . Data are represented as mean ±SEM . p-Values were calculated by T-test and p<0 . 05 was considered a significant difference . For bar graphs , number of worms scored is indicated in each bar . All data are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 014 10 . 7554/eLife . 26243 . 015Figure 4—source data 1 . Numerical data of Figure 4A–4C and E–G and Figure 4—figure supplements 1 , 2A , B , D , F and G . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 015 10 . 7554/eLife . 26243 . 016Figure 4—figure supplement 1 . Protease activity of several arrested larvae ( RNAi treatment ) , elt-2 RNAi as positive control . Protease activity is not all reduced in arrested larvae , which indicate that lower TORC1 activity on protease activity ( Figure 4A ) is unlikely to be from generally slowing or arresting development . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 01610 . 7554/eLife . 26243 . 017Figure 4—figure supplement 2 . LGG-1::GFP as a marker for activity downstream of TORC1 and the impact of food source on apical membrane polarity and VHA-6 localization in the intestine of C . elegans . ( A ) Bar graph showing that the high level of autophagy activity ( measured by the number of LGG-1::GFP puncta in seam cells ) in worms fed HK-OP50 ( see Figure 4B ) is suppressed by an nprl-3 ( lf ) mutation that hyperactivates TORC1 ( Zhu et al . , 2013 ) . Data are represented as mean ±SEM . ( B ) RNAi knockdown of daf-15 , which reduces TORC1 activity , caused a drastic increase in autophagy activity in young wild type larvae . Data are represented as mean ±SEM . ( C ) Fluorescence images showing that apical localization of ERM-1::GFP , a marker for the integrity of apical membrane polarity ( Zhang et al . , 2011; Zhu et al . , 2015 ) , in the intestine is not altered in worms fed heat-killed bacteria . The results suggest that the apical membrane polarity of the intestine is not defective under the poor feeding condition . ( D ) Fluorescence images and quantitative data showing that localization of a VHA-6::mCherry fusion protein is disorganized at the apical membrane of the intestine of worms fed heat-killed bacteria ( HK ) . The defect is significantly suppressed by VB2 supplementation . VHA-6 is a subunit of the vacuolar type H+-ATPase ( V-ATPase ) that has been indicated to function in TORC1 activation in mammalian cell culture studies ( Zoncu et al . , 2011 ) . VHA-6 and its localization at the apical membrane have been linked to TORC1 activity in response to defects in a lipid biosynthesis pathway in C . elegans ( Zhu et al . , 2015 ) . These data are consistent with the proposed reduction of TORC1 activity in worms fed heat-killed bacteria and recovery of TORC1 activity by VB2 supplementation . ( E ) HPLC-UV analysis of VB2 extracted from wild-type worm fed live OP50 , or wild-type , nprl-3 ( lf ) and asp-13/14 OE worms fed HK-OP50 . Arrow indicates the peak of VB2 . ( F ) Reducing TORC1 activity ( daf-15 RNAi ) led to defects in food choice . Data are represented as mean ±SEM . ( G ) Fluorescence images ( integrated translational ELT-2::GFP reporter strain ) and Western blot ( wild-type ) showing that daf-15 ( RNAi ) caused a decrease in ELT-2 expression in worms fed live OP50 . Data are represented as mean ±SEM . p-Values were calculated by T-test . For bar graphs , the number of worms scored is indicated within each bar . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 017 We then measured TORC1 activity by utilizing an established autophagy marker ( LGG-1::GFP puncta ) known to be repressed by TORC1 ( Robida-Stubbs et al . , 2012 ) . We found that worms fed HK-OP50 displayed a high level of LGG-1::GFP puncta in seam cells compared to that in worms fed live OP50 , and this increase was suppressed by VB2 supplementation ( Figure 4B ) . To support that the observed change in the LGG-1::GFP puncta level under our testing conditions reflects a specific change in TORC1 activity , we showed that the elevated LGG-1::GFP puncta level in worms fed HK-OP50 was suppressed by hyperactivation of TORC1 in a nprl-3 loss-of-function ( lf ) mutant ( Figure 4—figure supplement 2A ) . NPRL-3 is a negative regulator of TORC1 in both worms and mammalian cells ( Bar-Peled et al . , 2013; Zhu et al . , 2013 ) . Conversely , inhibition of TORC1 activity by daf-15 ( RNAi ) elevated the LGG-1::GFP puncta level in well-fed , young larvae ( Figure 4—figure supplement 2B ) . In addition , apical localization of VHA-6 , which we previously showed contributes to intestinal TORC1 activation by a lipid biosynthesis pathway ( Zhu et al . , 2015 ) , was altered in worms fed HK-OP50 and this defect was suppressed by VB2 supplementation ( Figure 4—figure supplement 2C and D ) . These data indicate that TORC1 activity is decreased in worms fed heat-killed bacteria and that VB2 supplementation significantly recovers the activity . We obtained further evidence that TORC1 acts downstream of VB2 by showing that TORC1 activation is sufficient to mimic the effect of VB2 in boosting intestinal protease activity and food usage . Specifically , we found that the nprl-3 ( lf ) mutation significantly improved intestinal protease activity , the expression of asp-13 and asp-14 , larval growth measured by gonad length and dwelling behavior of worms fed HK-OP50 ( Figure 4C–G ) ; however , this rescue was not because VB2 level was recovered as VB2 level in nprl-3 ( lf ) mutants fed heat-killed food did not increase ( Figure 4—figure supplement 2E ) . In contrast , reducing TORC1 activity ( daf-15 RNAi ) led to defects in food choice ( Figure 4—figure supplement 2F ) . Moreover , daf-15 ( RNAi ) treatment decreased the expression of the translational ELT-2::GFP reporter in the reporter strain and the endogenous ELT-2 protein in the wild-type ( Figure 4—figure supplement 2G ) , which is consistent with the observation in a previous report ( Schieber and Chandel , 2014 ) . Therefore , our data strongly support that TORC1 mediates the effect of VB2 on the expression of proteases and the usage of heat-killed bacteria . Vitamin B2 , also known as riboflavin , is the precursor of flavin mononucleotide ( FMN ) and flavin adenine dinucleotide ( FAD ) ( Joosten and van Berkel , 2007; Lienhart et al . , 2013; Powers , 2003 ) ( Figure 5A ) . Using HPLC-UV analysis , we found that both FMN and FAD levels were decreased in worms fed HK-OP50 compared to worms fed live OP50 , and their levels were partially restored by VB2 supplementation ( Figure 5B ) . Therefore , lack of sufficient FAD and FMN likely contributes to the low quality of heat-killed bacteria as a food source . 10 . 7554/eLife . 26243 . 018Figure 5 . Change in FAD may mediate the effect of VB2 on food usage by worms fed heat-killed bacteria . ( A ) Cartoon diagram to illustrate known functional relationship between VB2 ( riboflavin ) , flavin mononucleotide ( FMN ) and flavin adenine dinucleotide ( FAD ) , as well as a proposed VB2 sensing pathway in the intestine , based on this study . RFT-2 , expressed in the intestine in C . elegans , is a homolog of the riboflavin transporter three that is required for VB2 uptake ( Biswas et al . , 2013; Gandhimathi et al . , 2015 ) . FLAD-1 is a FAD synthase that catalyzes the last step of FAD biosynthesis ( Liuzzi et al . , 2012 ) . NPRL-3 represses TORC1 activity ( Zhu et al . , 2013 ) . ( B ) HPLC-UV detection peak of FAD and FMN from worms fed under indicated conditions . Figure 5—figure supplement 1A shows the graph of FAD and FMN standards . ( C ) Scatter plot showing that supplementation of FMN or FAD significantly recovered the of wild-type larvae fed HK-OP50 . The number of worms scored was 28 , 29 , 35 and 31 , respectively . Data are represented as mean ±SD . ( D ) Fluorescence images and bar graph showing that FAD supplementation elevated the in vivo protease activity in worms fed HK-OP50 and the increase is eliminated by RNAi knockdown of daf-15/raptor . Data are represented as mean ±SEM . ( E and F ) Fluorescence images and bar graph showing that RNAi knock down of flad-1/FAD synthetase sharply reduced the in vivo protease activity in wild type ( E ) . The reduction was suppressed by a nprl-3 ( lf ) mutation ( F ) that hyperactivates TORC1 . Data are represented as mean ±SEM . ( G ) RNAi knockdown of flad-1 elevated autophagy activity ( LGG-1::GFP puncta ) that is known to be repressed by TORC1 . Data are represented as mean ±SEM . ( H and I ) Fluorescence images and bar graph showing that RNAi knock down of rft-2 , encoding a riboflavin transporter ( see A ) , sharply reduced the in vivo protease activity in wild type . The sharp reduction was suppressed by a nprl-3 ( lf ) mutation that hyperactivates TORC1 . Data are represented as mean ±SEM . ( J ) RNAi reduction of rft-2 elevated the number of LGG-1::GFP puncta that marks the level of autophagy . The increase in autophagy is consistent with a reduction in TORC1 activity . Data are represented as mean ±SEM . ( K ) Measurement of ATP levels in worms . ( top ) ATP level is reduced in worms fed HK-OP50 and the reduction is partially suppressed by VB2 or FAD supplementation . ( bottom ) flad-1 ( RNAi ) or apt-2 ( RNAi ) caused reduction in ATP level in wild type , and the decrease by flad-1 ( RNAi ) , but not by atp-2 ( RNAi ) , was overcome by FAD supplement . Data are represented as mean ±SD . ( L and M ) Fluorescence images and bar graph showing that RNAi knock down of atp-2/ATP synthetase , significantly reduced the intestinal protease activity in wild type ( L ) . The reduction seen was suppressed by an nprl-3 ( lf ) mutation ( M ) . Data are represented as mean ±SEM . ( N ) RNAi knock down of atp-2 elevated the autophagy activity ( LGG-1::GFP puncta ) . Data are represented as mean ±SEM . p-Values were calculated by T-test and p<0 . 05 was considered a significant difference . For bar graphs , the number of worms scored is indicated within each bar . All data are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 018 10 . 7554/eLife . 26243 . 019Figure 5—source data 1 . Numerical data of Figure 5B–5N and Figure 5—figure supplement 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 019 10 . 7554/eLife . 26243 . 020Figure 5—figure supplement 1 . Measurement of FAD and FMN standards and role of FAD supplementation on ELT-2 expression . ( A ) HPLC-UV detection peak of FAD and FMN standards , controls for Figure 5B . ( B ) Fluorescence images and bar graph showing that FAD supplementation increases the expression of an ELT-2::GFP reporter in worms fed HK-OP50 . Data are represented as mean ±SD . DOI: http://dx . doi . org/10 . 7554/eLife . 26243 . 020 To test if FAD also mediates the VB2 impact on TORC1 and protease production , we showed that FAD supplementation improved food uptake , intestinal protease activity and ELT-2 expression , as VB2 does ( Figure 5C and D; Figure 5—figure supplement 1B ) . Furthermore , the FAD benefit is also TORC1 dependent , as pretreatment with RNAi targeting daf-15/raptor essentially eliminated the effect of FAD supplementation on protease activity ( Figure 5D ) . We then further analyzed the functional relationship between FAD and TORC1 by decreasing FAD production through RNAi of a FAD synthetase gene ( flad-1 ) and a riboflavin transporter gene ( rft-2 ) ( Biswas et al . , 2013; Gandhimathi et al . , 2015; Liuzzi et al . , 2012 ) ( Figure 5A ) . Protease activity was dramatically reduced in both RNAi-treated strains and the activity was recovered in nprl-3 ( lf ) mutants ( Figure 5E , F , H and I ) . Furthermore , both flad-1 ( RNAi ) and rft-2 ( RNAi ) significantly increased autophagy ( Figure 5G and J ) . These data support that TORC1 acts downstream of FAD to promote protease activity . The similarity of the data between flad-1 and rft-2 further supports that FAD plays the key role in mediating VB2 effect on intestinal protease activity . FAD-associated proteins are involved in many metabolic events including fatty acid metabolism , the TCA cycle and electron transport chain ( Lienhart et al . , 2013 ) . Since many of these events are related to energy production , and TORC1 is known to be an ATP sensor in mammals and C . elegans ( Chin et al . , 2014; Dennis et al . , 2001; Hay and Sonenberg , 2004 ) , we tested if the ATP level serves as a signal to relay the impact of VB2/FAD on TORC1 and downstream protease activity . We found that the ATP level was indeed dramatically reduced in worms fed HK-OP50 compared to that in worms fed live OP50 , whereas VB2/FAD supplementation partially recovered the level ( Figure 5K ) . Consistently , reducing FAD production by flad-1 ( RNAi ) in the first generation of wild-type worms also reduced ATP and the reduction was eliminated by FAD supplementation ( Figure 5K ) . Therefore , the decreased ATP level is a major cellular consequence of VB2 deficiency . To provide further evidence that ATP reduction is causal to the reduction in TORC1 and protease activity , we used RNAi to partially inactivate atp-2/ATP5B that encodes an ATP synthetase known to affect ATP production in worms ( Chin et al . , 2014 ) . We also observed a modest reduction of ATP level in the first generation of atp-2 ( RNAi ) -treated animals and the reduction was not overcome by FAD supplementation ( Figure 5K ) , consistent with atp-2 acting downstream of FAD regarding ATP production . Importantly , intestinal protease activity was prominently decreased in the second generation of atp-2 ( RNAi ) -treated animals and the reduction was partially suppressed by nprl-3 ( lf ) ( Figure 5L and M ) . Furthermore , LGG-1::GFP puncta were increased in atp-2 ( RNAi ) worms fed live OP50 ( Figure 5N ) , which is consistent with a previous observation in C . elegans ( Chin et al . , 2014 ) . Together , these data support the hypothesis that the ATP level change mediates the impact of VB2 on TORC1 and intestinal protease activity . Our data do not exclude potential impacts of other metabolic products from FAD-associated enzymes in the process . We identified an intestinal food-quality evaluation mechanism by which animals detect vitamin B2 deficiency by the FAD-ATP-TORC1-ELT-2 pathway that dictates food uptake and foraging behaviors ( Figure 5A ) . Such a mechanism with conserved functions enhances survival in wild environments by discriminating against low-quality food . Unlike rapid neuronal food responses that facilitate quick food consumption decisions , the intestinal system described here requires a rather slow , multi-step process after the ingestion of food . It should be noted that our findings do not exclude the involvement of sensory neurons in the food evaluation systems described in this study . The intestinal process may be evolved specifically to detect the deficiency of specific nutrients such as vitamins that elude the neuronal sensory system that may be honed to detect the presence of major nutrients and toxins ( Bargmann , 1997; Chandrashekar et al . , 2006; Watson et al . , 2015 ) . In addition to VB2 , we have recently shown that intestinal deficiencies in certain fatty acids and pyrimidine are also sensed by signaling systems in C . elegans to alter reproductive/developmental programs ( Chi et al . , 2016; Kniazeva et al . , 2015 ) . These examples also suggest that the protective response to nutrient deficiency by the intestine-initiated mechanism commonly includes regulation of developmental/reproductive programs , in addition to food behaviors . VB2 deficiency likely affects the functions of a large number of flavoproteins involved in many essential cellular functions; the responses by TORC1 , including protease regulation and developmental arrest , may be considered the last and emergent protective actions by the animal . TORC1 is well known for its role in sensing multiple biomolecules including amino acids , lipids and ATP ( Chin et al . , 2014; Dennis et al . , 2001; Efeyan et al . , 2015; Zhu et al . , 2013 ) . However , what this ATP-TORC1-sensing activity might do in a whole organism under physiological conditions including response to environmental challenges is a fascinating and largely unexplored question . This study provides a perfect example to address this question as we present evidence for that ATP serves as an intermediate nutrient to mediate the sensing of VB2/FAD by TORC1 . Moreover , our study may be the first to connect TORC1 sensing to a specific micronutrient in a model organism . The versatility of TORC1 in responding to the deficiency of multiple nutritional/metabolic inputs suggests that the TORC1-protease pathway may be used to evaluate the levels of multiple essential nutrients in food sources , albeit that the evaluation process for other nutrients in the intestine remain to be demonstrated in vivo . This study may also indicate the importance of live bacteria in providing micronutrients , such as VB2 , to animals and may thus be highly related to human health , as we obtain vitamins from gut microbes in addition to food sources ( Albert et al . , 1980; LeBlanc et al . , 2013 ) . Since VB2 is known to be light sensitive and heat labile in aqueous solution ( Sheraz et al . , 2014 ) , the contribution of VB2 from gut microbes could be significantly undervalued . The system used in this study for analyzing VB2 may potentially be effective to investigate the roles and sensing mechanisms for other micronutrients provided by microbes . Nematode stocks were maintained on nematode growth medium ( NGM ) plates seeded with bacteria ( E . coli OP50 ) at 20°C ( http://www . wormbook . org/ ) . The following strains/alleles were obtained from the Caenorhabditis Genetics Center ( CGC ) or as indicated: N2 Bristol ( termed wild type ) , KWN117{pha-1 ( e2123 ) III; him-5 ( e1490 ) V; rnyEx60 [pELA2 ( vha-6p:::vha-6::mCherry ) + myo-3p::GFP + pCL1 ( pha-1+ ) ]} ( RRID:WB-STRAIN:KWN117 ) , DA2123{adIs2122 [lgg-1p::GFP::lgg-1 + rol-6 ( su1006 ) ]} ( RRID:WB-STRAIN:DA2123 ) , MH4672[nprl-3 ( ku540 ) ] ( Zhu et al . , 2013 ) , SD1949{glo-4 ( ok623 ) ;gaIs290 [elt-2::TY1::EGFP::3xFLAG ( 92C12 ) + unc-119 ( + ) ]} ( RRID:WB-STRAIN:SD1949 ) , SD1965{rde-1 ( ne300 ) ; unc-119 ( ed3 ) ; Ex[unc-119 ( + ) ]} ( RRID:WB-STRAIN:SD1965 ) , SD1963{unc-119 ( ed3 ) III; rde-1 ( ne300 ) V;gaEx234[elt-2p::elt-2::GFP + unc-119 ( + ) ]} ( RRID:WB-STRAIN:SD1963 ) ; VJ402 ( erm-1p-erm-1::GFP ) ( RRID:WB-STRAIN:VJ402 ) ( Göbel et al . , 2004 ) ;VP303 {rde-1 ( ne219 ) V; [nhx-2p::rde-1 + rol-6 ( su1006 ) ]} ( RRID:WB-STRAIN:VP303 ) . Ex [rpl-28p::asp-13+rpl-28p::asp-14]; this strain was constructed for this study by standard cloning and microinjection techniques . Preparation of HK-OP50: standard overnight culture of E . coli OP50 grown in LB broth was concentrated to 1/10 vol and was then heat-killed in a 75°C water bath for 90 min . About 150 µl of the HK-OP50 was spread onto each NGM plate . For vitamin supplementation , each vitamin was dissolved in H2O to generate a stock . The stock solution was spread onto NGM plates seeded with HK-OP50 . The vendor , stock concentrations and volumes used for each vitamin are listed as follows: Thiamine hydrochloride VB1 ( Sigma T4625 , 10 mg/ml , 30 µl ) , ( − ) -Riboflavin/VB2 ( Sigma R9504 0 . 3 mg/ml , 50 µl ) , Nicotinamide/VB3 ( Sigma , N3376 , 30 mg/ml , 30 µl ) , D-Pantothenic acid hemicalcium salt/VB5 ( Sigma 21210–5 G-F , 30 mg/ml , 30 µl ) , Pyridoxine hydrochloride/VB6 ( Sigma P9755 , 4 mg/ml , 30 µl ) , Folic acid/VB9 ( Sigma F-8259 , 2 mg/ml , 30 µl ) , VB12 ( Sigma V2876 , 30 ug/ml , 50 µl ) and L-Ascobric acid/VC ( Sigma A0278 , 110 mg/ml , 30 µl ) . For HK-OP50 plates with a small amount of live bacteria , liquid cultures of bacteria were grown overnight at 37°C in LB broth . E . coli OP50 and S . saprophyticus ( ATCC 15305 ) were diluted to the same OD600 and 0 . 1 µl of the live bacteria was added onto the center of a lawn of HK-OP50 on NGM plates . For treatment of HK-OP50 culture with enzymes , a standard overnight culture of HK-OP50 was pelleted and then re-suspended to the initial volume with S-medium . Enzymes were prepared as stock solutions in H2O and then added to the HK-OP50 in S-medium . After incubation at 37°C ( 210 rpm ) for 90 min , about 150 µl of this enzyme-treated HK-OP50 was added to wells of a 96-well plate , where synchronized L1 larvae were seeded into each well . The source of the enzyme , stock concentration and the volume added to 1 ml of HK-OP50/S-medium are: amylase ( Sigma A3403-500KU , 10–50 mg/ml , 80 µl ) , lipase ( Sigma L1754-5G , 26 mg/ml , 80 µl ) and protease ( Sigma P6911-100MG , 8 . 2 mg/m , 80 µl ) . For treatment of heat-killed OP50 with protease inhibitor , a protease inhibitor cocktail ( PIC , Roche-4693124001-Complete , Mini Protease Inhibitor Cocktail Tablets ) was prepared as a stock ( 1 tablet/ml ) in H2O . 200 µl of the PIC stock solution was spread onto the top of the HK-OP50 lawn on NGM plates . In typical experiments , eggs were added to plates immediately after the bacterial food was placed on the plate . To assay for larval growth , gonad length was measured at day 7 . Three to ten replicates for each condition were performed for each experiment , and the experiments were duplicated on different days . For the food dwelling assay ( Kniazeva et al . , 2015 ) , 20 µl of bacteria were spotted onto the center of 6 cm NGM plates . Synchronized L1s were seeded on the center of the lawn . After one day , animals were scored as either off the lawn or on the lawn ( including those on the border ) . In the experiments using VB2 supplementation , 15 µl of VB2 stock solution was added on the top of HK-OP50 spot . For the food choice assay ( Kniazeva et al . , 2015 ) , 20 µl of bacteria were spotted onto two opposite spots on 6 cm NGM plates . The eggs were seeded on the center of plate . The number of animals at each location was scored after the time indicated in the figures . This value was divided by the total number of worms on both locations to determine the percentage in each location . In the experiments using VB2 supplementation , 20 µl of VB2 stock solution was added on the top of HK-OP50 spot . The food-seeking assay ( Figures 3E and 4G ) followed the protocol described in Kniazeva et al . ( 2015 ) . Worms were counted 3 days after placing the eggs on the origin . Three to ten replicates for each condition were performed for each assay , and the experiments were duplicated on different days . Animals were grown on live OP50 , and eggs were collected by bleaching and then washing in M9 buffer . Synchronized L1 larvae were obtained by allowing the eggs to hatch in M9 buffer for 18 hr . Eggs or synchronized L1s were seeded to plates prepared for specific assay and incubated at 20°C for 7 days . Animals were washed off the plates , mounted on agarose pads , and examined under Nomarski optics . Gonad length was measured by the ImageJ software . In some cases , larval stage was determined based on the size of the worms . In vivo imaging of worm intestinal protease activity was performed following a published protocol ( Hama et al . , 2009 ) with minor modifications . Briefly , quenched BODIPY TR-X casein ( EnzChek Protease Assay Kit , Invitrogen ) was dissolved in 0 . 1 M sodium bicarbonate ( pH 8 . 3 ) at a concentration of 1 mg/ml and stored at −20°C . Aliquots of EnzChek were diluted to 400 μg/ml in dH2O . Reconstituted EnzChek was added into the tube with animals at a final concentration of 20 μg/ml and incubated for 3 hr at room temperature with shaking . Animals were then washed three times with M9 , mounted on agarose pads , and examined under a Nomarski optics . The ﬂuorescent intensity of whole individual larvae was calculated by the Image-J software . For RNA isolation , C . elegans at the same stage ( L1-L2 ) grown on indicated food plates were quickly collected from NGM plates and washed three times with M9 buffer , followed by the addition of 250 ul TRIzol ( Invitrogen ) . Lysates were preserved at 80°C until RNA isolation . RNA was isolated with two chloroform extractions ( 50 μl each ) , followed by isopropanol precipitation ( 125 μL ) of the aqueous phase , and a single wash of the resulting pellet with 70% ethanol ( 250 μL ) . RNA pellets were dried in a tissue culture hood and re-suspended in RNase-free water , and then puriﬁed of contaminating DNA by DNaseI treatment ( TURBO DNA-free Kit , Invitrogen ) followed by cleanup using QIAGEN RNeasy columns . cDNA for RT-PCR experiments was synthesized using 300 ng of total RNA template , SuperScript III , and oligo ( dT ) 12–18 primer , according to the manufacturer’s protocol ( Life Technologies ) . qRT-PCR reactions were performed in triplicate using the Applied Biosystems StepOnePlus Real-Time PCR system and Rotor-Gene SYBR Green PCR Kit ( QIAGEN ) and fold-change calculations were performed manually . A Student’s t test was used to evaluate statistical signiﬁcance . Primers can be found in Supplementary file 1 . SD1963{unc-119 ( ed3 ) III; rde-1 ( ne300 ) V;gaEx234[elt-2p::elt-2::GFP + unc-119 ( + ) ]} worms were grown on heat-killed OP50/NGM plates and harvested . The ChIP was carried out as described in ( Mukhopadhyay et al . , 2008 ) . Primers can be found in Supplementary file 1 . The ERM-1::GFP ( Göbel et al . , 2004 ) or VHA-6::mCherry transgenic adults were bleached and eggs were collected and seeded onto indicated plates ( HK-OP50 , HK-OP50+VB2 or live OP50 ) and grown for 3 days . The apical protein localization ( ERM-1::GFP or VHA-6::mCherry ) was carried out according to the published method ( Zhu et al . , 2015 ) . DA2123 animals carrying an integrated GFP::LGG-1 translational fusion gene ( Kang et al . , 2007 ) , were used to quantify levels of autophagy . The DA2123 adults animals were bleached and the resulting eggs were placed on the indicated food condition ( HK-OP50 , HK-OP50+VB2 or live OP50 ) . After 3 days , the animals at the equivalent stage were mounted on a 2% agar pad with a drop of M9 ( plus 25 nM NaN3 ) . Fluorescence images were captured by Nomarski microscopy . GFP-positive puncta were counted in 2–10 seam cells from approximately 50 animals in three independent experiments . ATP level from worms was carried out as described ( Sagi and Kim , 2012 ) . In brief , worms grown on HK-OP50 or HK-OP50+VB2 for 3 days were collected , washed four times with M9 , suspended in lysis buffer ( 10 mM Tris-HCl ph8 . 0 , 25 mM NaCl , 1 mM EDTA , 1 X protease inhibitor ) , boiled for 20 min and quickly frozen in −80°C . All samples were processed on the same day . A Luminescent ATP Detection Assay Kit ( Abcam ) was used to measure ATP concentrations according to the manufacturer’s instructions . ATP concentrations were normalized to absolute protein concentrations . Each assay was repeated in triplicate , and the average ATP concentration and SD were calculated . To analyze the effect of RNAi on growth and protease activity of worms fed HK-OP50 , eggs from worms grown on OP50 were collected by bleaching , washed three times in M9 buffer , and then allowed to hatch in M9 buffer for 18 hr . The synchronized L1 worms were then transferred onto plates seeded with bacteria expressing double-stranded RNA of individual genes . After the worms grew to adults on the RNAi plate , eggs were collected again by bleaching , washed three times in M9 buffer , then seeded onto assay plates with the food to be tested . To measure larval growth , worms grew for 7days before scoring . For the in vivo protease activity assay , worms grew for 3 days before scoring . To screen for transcription factors that are involved in regulating protease activity in response to changes in feeding conditions by RNAi , synchronized L1 worms were transferred onto plates seeded with bacteria expressing double-stranded RNA of individual genes . The protease activity in the intestinal tract of L1-L2 worms of the next generation was measured as described above . To assay the effect of over-expressing asp-13 and asp-14 in worms treated with elt-2 ( RNAi ) , L4 stage wild type and asp-13/asp-14 O/E worms were transferred onto plates seeded with elt-2 ( RNAi ) . Phenotypes were scored for the next generation . For intestine-specific RNAi of asp-13 and asp-14 , the strain VP303 {rde-1 ( ne219 ) V; [nhx-2p::rde-1 + rol-6 ( su1006 ) ]} ( Espelt et al . , 2005 ) was used . To assay the ATP level , synchronized L1 worms were transferred onto plates seeded with bacteria expressing double-stranded RNA of individual genes . The ATP level in L4 worms was measured as described above . Analysis of flavins from worms was carried out as described ( Liuzzi et al . , 2012 ) . In brief , worms were harvested , washed in M9 buffer , suspended in lysis buffer ( 10 mM Tris-HCl PH8 . 0 , 25 mM NaCl , 1 mM EDTA , 1 X protease inhibitor ) and lysed by sonication . The lysate was centrifuged at 13 000 g for 1 min . Vitamin B2 , FMN , and FAD content of the supernatant were analyzed by a Shimadzu Prominence modular HPLC system using a Supelcosil LC-18-THPLC column ( 25 cm ×4 . 6 mm , 5 µm particle size ) ( Sigma-Aldrich ) following the manufacturer’s instructions and detected by an inline UV/Vis detector ( SPD-10A , Shimadzu ) . The identity of each flavin peak was determined by a retention time comparison with the relative flavin standards ( Sigma-Aldrich ) . To measure ELT-2 protein level , second generation of daf-15 RNAi-treated worms ( L3 larvae arrested ) or worms fed different food were analyzed by standard Western blot methods , and probed with anti-ELT-2 monoclonal antibody 455-2A4 ( Wiesenfahrt et al . , 2016 ) ( Developmental Studies Hybridoma Bank , University of Iowa ) and anti-Actin ( Sigma-A2066 ) as a loading control . Analysis of fluorescent reporter expression and gonad phenotypes were performed under Nomarski optics on a Zeiss Axioplan2 microscope with a Zeiss AxioCam MRm CCD camera . Plate phenotypes were observed using a Leica MZ16F dissecting microscope with a Hamamatsu C4742-95 CCD camera . All statistical analyses ( except quantification of the VHA-6::mCherry fluorescent signal ) were performed using Student’s t-test and p<0 . 05 was considered a significant difference . Statistical analyses of the VHA-6:mCherry fluorescent signal was performed using the χ two test .	Animals are able to sense changes in the quality of their diet , which allows them to adapt to environments where the availability of nutrients fluctuates . For example , if a food lacks essential nutrients , or contains toxic molecules , animals can avoid it . If food is scarce , they can alter their metabolism to compensate . This assessment is done in the intestine , but exactly how it works is not fully understood , especially when it comes to vitamins . Animals require Vitamin B2 to grow and remain healthy . This vitamin is involved in the process by which cells make chemical energy , which is needed to fuel many biological processes . Vitamin B2 is found in the diet , and is also produced by bacteria living in the gut . Here , Qi et al . used a worm called Caenorhabiditis elegans to examine how the gut detects this vitamin , and what impact this has on how the worm behaves and develops . The experiments show that when the worms’ diet includes live bacteria , they developed normally from larvae into adults . However , if the worms were fed only bacteria that had been killed by cooking and exposure to light – which damages Vitamin B2 – they stopped eating , shut down the production of two key digestive enzymes and stopped growing . Supplying these worms with extra vitamin B2 , or extra digestive enzymes , stimulated the worms to start eating the cooked bacteria . Further experiments show that when Vitamin B2 is scarce , the levels of chemical energy in the worms’ cells decrease . This drop in energy is sensed by a complex of molecules called TORC1 , which triggers the changes in the worms’ metabolism and behavior . The findings of Qi et al . indicate that gut microbes can play important roles in providing micronutrients like Vitamin B2 . Many of the molecular pathways used by these worms have equivalents in humans . Therefore , further research on these pathways in worms may help us to understand how the human body responds to nutrients and how metabolic diseases may alter these pathways .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2017	A vitamin-B2-sensing mechanism that regulates gut protease activity to impact animal’s food behavior and growth
Eukaryotic genes generate multiple RNA transcript isoforms though alternative transcription , splicing , and polyadenylation . However , the relationship between human transcript diversity and protein production is complex as each isoform can be translated differently . We fractionated a polysome profile and reconstructed transcript isoforms from each fraction , which we term Transcript Isoforms in Polysomes sequencing ( TrIP-seq ) . Analysis of these data revealed regulatory features that control ribosome occupancy and translational output of each transcript isoform . We extracted a panel of 5′ and 3′ untranslated regions that control protein production from an unrelated gene in cells over a 100-fold range . Select 5′ untranslated regions exert robust translational control between cell lines , while 3′ untranslated regions can confer cell type-specific expression . These results expose the large dynamic range of transcript-isoform-specific translational control , identify isoform-specific sequences that control protein output in human cells , and demonstrate that transcript isoform diversity must be considered when relating RNA and protein levels . Eukaryotic genes can produce a staggering diversity of messenger RNA products . In humans , there is a median of five annotated transcript isoforms per gene , with more than 80 annotated isoforms in some cases ( Ensembl release 75 ) . These transcript isoforms arise from the combined action of alternative transcription initiation , splicing and cleavage , and polyadenylation , which are often cell type specific ( Barbosa-Morais et al . , 2012; Mele et al . , 2015; Merkin et al . , 2012; Wang et al . , 2008; Xiong et al . , 2015; Zheng and Black , 2013 ) . As each of these transcript isoforms may in principle harbor a unique set of translational control elements , gene-level expression , which averages out the translational potential of each individual transcript isoform , may not be an accurate measure of protein levels . Since changes in both translation and splicing are linked to numerous human disorders , it is critical to understand the relationship between the two ( Maslon et al . , 2014; Piccirillo et al . , 2014; Ruggero , 2013; Xiong et al . , 2015 ) . Eukaryotic mRNAs are decorated with diverse sequence features that can control translation , which can vary between transcript isoforms ( de Klerk and 'tHoen , 2015; Ingolia et al . , 2011; Sterne-Weiler et al . , 2013 ) . Classic examples of translational control include upstream ORFs ( uORF ) -mediated translational control of yeast GCN4 ( Hinnebusch , 2005 ) , protein binding such as the iron regulatory protein ( Gray and Hentze , 1994 ) , and the action of micro-RNAs ( Nottrott et al . , 2006; Wilczynska and Bushell , 2015 ) or DEAD-box proteins such as eIF4A and Ded1 ( Chuang et al . , 1997; Hinnebusch and Lorsch , 2012; Sen et al . , 2015 ) . Alternative 5′ leader sequences , uORFs , and select tandem 3′ untranslated region ( UTR ) isoforms have been demonstrated to influence protein production ( Brar et al . , 2012; Hinnebusch , 2005; Ingolia et al . , 2011; Mayr and Bartel , 2009; Sandberg et al . , 2008; Zhang et al . , 2012 ) . Any of these features may in principle be different between transcript isoforms , but the prevalence and dynamic range of isoform-specific translational control across the human genome is currently unknown . Previous work measuring genome-wide translation in human cells has focused largely on the relationship between gene-level mRNA abundance and protein levels , which is blind to the contribution of transcript isoforms . Ribosome profiling is not well-suited for measuring transcript isoform-specific translation , primarily due to the short ~30 bp length of ribosome-protected fragments ( Ingolia , 2014 ) . Prior attempts to characterize isoform-specific translation have measured the effects of 5′ end diversity in yeast ( Arribere and Gilbert , 2013 ) and 3′ end diversity in mouse cells ( Spies et al . , 2013 ) , or splicing differences between cytoplasmic and aggregate polysomal mRNAs ( Maslon et al . , 2014; Sterne-Weiler et al . , 2013 ) . However , sequencing just the ends of transcripts cannot distinguish between transcript isoforms of the same gene harboring degenerate termini . In addition , aggregating polysome fractions averages lowly- and highly-ribosome-associated messages . Therefore , a different strategy is required to understand how the diversity of the human transcriptome impacts translational output . Here , we adapt a classic approach of polysome profiling coupled with global gene expression analysis ( Arava et al . , 2003 ) to measure transcript-isoform specific translation using deep sequencing , which we term Transcript Isoforms in Polysomes sequencing ( TrIP-seq ) . By using high gradient resolution and sequencing depth , this approach yields polysome profiles for over 60 , 000 individual transcript isoforms representing almost 14 , 000 protein coding genes . We observe frequent intron retention on ribosome-associated transcripts , even in high-polysome fractions , identifying a population of retained but not nuclear-detained introns ( Boutz et al . , 2015 ) . Properties of 3′ untranslated regions predominate over the 5′ leader sequence as the driving force behind differential polysome association for transcript isoforms of the same gene among the transcript features tested . We show that regulatory sequences differentially included in transcript isoforms of the same gene are modular and can trigger differences in the translation of reporters spanning two orders of magnitude . These findings provide a lens through which to ascribe functional consequences to RNA-seq-generated transcriptomes . Moreover , TrIP-seq analysis uncovers regulatory elements that can be utilized to tune translation of synthetic messages robustly in cells . We determined the ribosomal association of transcript isoforms by sequencing transcripts cofractionating with different numbers of ribosomes with sufficient depth to determine isoform abundances , as was performed at the gene level in yeast ( Arava et al . , 2003 ) . We treated HEK 293T cells with cycloheximide to stall translation and fractionated the cytoplasm into ribosome-containing samples including one to eight or more ribosomes ( Figures 1A and Figure 1—figure supplement 1A; see Materials and methods for details ) . We made RNA sequencing libraries from each fraction in biological duplicate and obtained transcript-level abundances using the Cufflinks suite ( Figure 1—source data 1 and 2; [Trapnell et al . , 2010] ) . Clustering of the samples recapitulates the gradient order ( Figure 1B ) , indicating the polysome profile was accurately fractionated . Four subgroups emerge from this clustering: the 80S ( monosome ) , low polysomes ( two-four ribosomes ) , high polysomes ( five-eight+ ribosomes ) , and total cytoplasmic RNA . We tested the robustness of the clustering of samples by computing the average Jaccard distance between clusters from data subjected to three different resampling methods , which was ≥0 . 75 , suggesting stable clusters ( Figure 1—figure supplement 1G; Materials and methods ) . This suggests that in cells , transcript isoforms are predominantly poorly- or highly-ribosome associated , causing low polysomes to cluster away from high polysomes , which could be have numerous biological origins including highly abundant short ORFs . 10 . 7554/eLife . 10921 . 003Figure 1 . Transcript Isoforms in Polysomes sequencing ( TrIP-seq ) measures transcript isoform specific translation . ( A ) HEK 293T cells were treated with cycloheximide and the cytoplasmic fraction was extracted and applied to a sucrose gradient , which was further fractionated into individual polysomes that were converted into sequencing libraries . ( B ) Intersample clustering recapitulates the gradient order of polysomes , indicating the sequenced fractions are faithful to the gradient profile . ( C , D ) RT-PCR analysis and transcript-level quantification for ACTB and ATF4 ( C ) and three transcripts of EEF1B2 ( D ) demonstrating concordance of sequencing and transcript-specific RT-PCR . TPM – transcripts per million . *nonspecific amplicon . ( E ) Members of transcript classes with more than 100 reads in indicated fractions show that coding genes are represented across the polysome while noncoding genes are preferentially in low polysome fractions . ( F ) Spearman’s correlation ( RS ) between gene ( left ) or isoform ( right ) read counts from ribosome profiling or TrIP-seq . See also Figure 1—figure supplement 1 and Materials and methods for calculation of TrIP-seq polysome counts . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 00310 . 7554/eLife . 10921 . 004Figure 1—source data 1 . Gene-level abundances for all Ensembl 75 annotated human genes across all sequenced polysome fractions . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 00410 . 7554/eLife . 10921 . 005Figure 1—source data 2 . Transcript isoform abundances for all Ensembl 75 annotated human transcripts across all sequenced polysome fractions . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 00510 . 7554/eLife . 10921 . 006Figure 1—figure supplement 1 . Extended TrIP-seq validation . ( A ) A large version of a representative polysome profile showing fractions collected . ( B ) A diagram of the alternative transcript isoforms of the gene EEF1B2 . Coding regions are wide boxes , noncoding regions are narrow boxes , and introns are lines . Upstream open reading frames ( uORFs ) are indicated by gray bars , with red octagons marking stop codons . The uORF for EEF1B2-201 is in-frame with the annotated start codon . Differentially included regions are shown in yellow , arrows show the direction of transcription , and select introns have been removed for clarity as indicated by gaps . ( C ) A replicate RT-PCR gel of EEF1B2 showing the presence of EEF1B2-003 in high polysomes . ( D ) The distribution of coding isoforms and noncoding genes across polysome fractions . ( E ) Replicate correlations between the weighted polysome sum of TrIP-seq read counts at the isoform level . RS: Spearman’s correlation . ( F ) Correlation between TrIP-seq polysome counts and HEK 293T cell iBAQ protein abundance . ( G ) The Jaccard distance between clusterings of resampled data from Figure 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 00610 . 7554/eLife . 10921 . 007Figure 1—figure supplement 2 . Extended validation of TrIP-seq isoform abundances across polysome fractions using qRT-PCR . ( A ) qRT-PCR ( top ) and TrIP-seq abundances ( bottom ) for three EEF1B2 isoforms . ( B ) As in ( A ) except for the gene SRSF5 . qRT-PCR abundance is calculated as a fold change versus cytoplasmic abundance . Note that systematic normalization differences exist between the TrIP-seq and qRT-PCR abundances since the TrIP-seq abundances are from repeat and rRNA subtracted data , while the input for qRT-PCR is total RNA . NTC: no template control . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 00710 . 7554/eLife . 10921 . 008Figure 1—figure supplement 3 . Read Tracking Across the Sequencing Analysis Pipeline . "Not RMSK" refers to reads not mapping to repeatmasker sequences , and "not abundant" is reads not mapping to rRNA and the mitochondrial chromosome , etc ( Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 008 To confirm the sequencing data , we first analyzed beta-actin ( ACTB ) , which is known to be heavily translated with many ribosomes on each message , and migrates with high polysomes in a gradient ( Figure 1C; ( Sterne-Weiler et al . , 2013; Zhang et al . , 2015 ) . However , ACTB only has one transcript isoform , so we examined ATF4 , which has one dominant isoform exhibiting low-polysome association in TrIP-seq data and by RT-PCR ( Figure 1C ) , consistent with it being translationally repressed in the absence of cellular stress ( Harding et al . , 2000 ) . We then analyzed eukaryotic elongation factor 1 beta 2 ( EEF1B2 ) , which has three isoforms with the same coding sequence but alternative 5′ leaders as a representative gene that exhibits isoform-specific ribosome association ( Figure 1—figure supplement 1B ) . The isoform-specific polysome profiles and RT-PCR conducted from polysome fractions agree qualitatively , validating the accuracy of the isoform-level quantifications ( Figure 1D ) . Weaker amplification of the longer isoform ( EEF1B2-003 ) in high polysomes may be due to PCR bias towards smaller amplicons ( Walsh et al . , 1992 ) , and replicate experiments show EEF1B2-003 present in high polysomes ( Figure 1—figure supplement 1C ) . We additionally validated the TrIP-seq data using qRT-PCR from polysome fractions and find that the two measurements agree for both EEF1B2 and SRSF5 ( Figure 1—figure supplement 2 ) . Both coding mRNAs and long noncoding RNAs ( lncRNAs ) associate with ribosomes ( Ingolia et al . , 2014; van Heesch et al . , 2014 ) , so we measured their polysomal abundance . Protein-coding genes and isoforms are found across the polysome , but noncoding and lncRNA genes are predominantly in the low polysome fractions indicating that they are generally weakly ribosome associated or in other large macromolecular complexes ( Figure 1E and Figure 1—figure supplement 1D ) . To determine whether TrIP-seq is a measure of translation as opposed to cryptic association with large macromolecular complexes or stalled ribosomes ( Darnell et al . , 2011; Ishimura et al . , 2014 ) , we compared TrIP-seq data to ribosome profiling and proteomics datasets . First , we reanalyzed a ribosome profiling dataset from HEK 293T cells and compared the number of ribosome-protected fragments to a weighted sum of the TrIP-seq polysome reads ( Materials and methods; nribo ≥2 ) ( Sidrauski et al . , 2015 ) . We observe a strong correlation between ribosome profiling and TrIP-seq at the gene level ( RS = 0 . 88; Figure 1F ) . However , at the isoform level , the correlation decreases ( RS = 0 . 47 ) , which is worse than the correlation between TrIP-seq replicates ( RS = 0 . 90; Figure 1—figure supplement 1E ) , suggesting the discrepancy is not due to variability in TrIP-seq but instead is likely due to known issues quantifying transcript isoforms in ribosome profiling ( Ingolia , 2014 ) . We next computed the translation efficiency ( TE ) of ribosome profiling and TrIP-seq data by dividing by cytoplasmic RNA levels . TE measured by the two methods correlated with RS = 0 . 24 , perhaps due to variability in culture conditions or in either technique . We compared the data of Sidrauski et al to an unpublished ribosome profiling dataset in 293T cells from the Yoon-Jae Cho lab . Ribosome-protected fragments between experiments correlate with RS = 0 . 78 , while TE has a substantially lower correlation of RS = 0 . 41 , suggesting biological or technical variability disproportionately affects TE . We then compared TrIP-seq data to protein abundances from HEK 293T cells , and find that the two datasets correlate ( RS = 0 . 57 , Figure 1—figure supplement 1F , ( Geiger et al . , 2012 ) , which is better than RNA-seq ( RS = 0 . 45; not shown ) . The observations that most translational stalling in mammalian cells is transient and translation elongation rates are homogeneous ( Ingolia et al . , 2011 ) , TrIP-seq appears to underestimate rather than overestimate protein abundance ( Figure 1—figure supplement 1F ) , and TrIP-seq correlates well at the gene-level with ribosome profiling ( Figure 1F ) suggest that TrIP-seq primarily measures translating ribosome association as opposed RNAs bound to other large complexes . To extract global trends in isoform-specific translation , we hierarchically clustered transcript isoform polysome profiles and selected eight clusters that are representative of general trends in the data ( Figure 2; Materials and methods ) . The depth of sequencing ( Figure 1—figure supplement 3 ) , augmented by the fractionation strategy , enables detection of 62 , 703 transcript isoforms in the polysome profile ( Figure 2—source data 1 ) . Isoforms in the observed clusters exhibit diverse average patterns across polysomes ( Figure 2A , B ) , from clusters 1 and 2 , which contain isoforms primarily in high polysomes , to cluster 3 with isoforms in the middle , to clusters 6 and 7 where isoforms are in low polysomes . Independent clustering of the two biological replicate TrIP-seq datasets shows that the high- and low-polysome clusters ( 1 , 2 and 6 ) appear more robust when comparing between averaged and individual replicate clusterings ( Figure 2—figure supplement 1A ) . Many clusters have similar total polysome abundance but different distributions , indicating that to obtain accurate measurements of isoform-specific translatability it is crucial to fractionate the polysome profile . 10 . 7554/eLife . 10921 . 009Figure 2 . Clustering of transcript distributions yields eight major clusters with diverse behavior across the polysome profile . ( A ) The average relative abundance of all isoforms in each cluster across polysomes is shown . Error is s . d . ( B ) Hierarchical clustering of 62 , 703 transcript isoform distributions across the polysome profile and cytoplasmic fraction . Yellow: above isoform average , cyan: below isoform average . ( C ) Transcript type distribution per cluster from Ensembl-annotated biotypes . Dotted lines mark the abundance of each transcript type in all isoforms that went in to the clustering . See also Figure 1—source data 1 and 2 and Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 00910 . 7554/eLife . 10921 . 010Figure 2—source data 1 . Variance stabilized transcript isoform abundances ( see Materials and methods ) and cluster number for all Ensembl 75 annotated human transcripts across all sequenced polysome fractions . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 01010 . 7554/eLife . 10921 . 011Figure 2—figure supplement 1 . Further information on TrIP-seq clusters . ( A ) Clustered transcript distributions from each biological replicate show robust clustering for the well-translated clusters 1 , 2 and 6 . The Jaccard distance is plotted ( the intersection divided by the union ) of each replicate cluster versus the pooled replicate clusters . ( B ) GO terms significant at the Benjamini-corrected p-value of 0 . 05 are indicated; clusters 3 and 4 had no enriched terms . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 011 Surprisingly , the poorly translated cluster 7 , which has ~9 , 000 transcripts , contains a similar number of annotated retained intron or protein-coding isoforms ( Figure 2C ) . We queried the gene ontology ( GO ) terms associated with each cluster and found that the poorly translated clusters 6 and 7 are enriched for translation and splicing genes ( Figure 2—figure supplement 1B ) , implicating the alternative splicing-nonsense-mediated decay ( AS-NMD ) pathway ( Jangi et al . , 2014; Lareau et al . , 2007 ) . However , not all retained intron transcripts are subject to NMD ( Boutz et al . , 2015; Gohring et al . , 2014 ) , and we also observe retained intron transcripts in the best-translated clusters ( 1 and 2 ) , indicating at least some of the introns may be translated . In sum , clustering of isoform abundance distributions across polysomes immediately provides insight into the diverse patterns of ribosome association by individual isoforms and transcript types in cells , and reveals intron-rich clusters of transcripts associated with polysomes that escape nuclear detention . We reasoned we could extract features regulating the translation of a gene by comparing transcript isoforms of the same gene that are well- or poorly-translated . We selected 24 features to explore that are involved in translational control , such as the length of the coding sequence and untranslated regions , predicted secondary structure , and microRNA binding sites . As their average polysome profiles and composition are broadly similar ( Figure 2 ) , we merged clusters 1 and 2 to generate a larger pool of high-polysome isoforms ( Figure 3A ) , and compared these to cluster 6 , representing poorly translated isoforms in low polysomes . We then compared transcript isoforms of the same gene that are in high- and low-polysome clusters , or gene-linked isoforms , to extract transcript features that influence translation . The number of gene-linked isoforms per feature per set varies between 569 and 6491 . We then measured the effect size and calculated statistical significance for all features between gene-linked isoforms ( Figures 3B and Figure 3—figure supplement 1A; Materials and methods; [Cliff , 1993] ) . 10 . 7554/eLife . 10921 . 012Figure 3 . Effect of transcript features on polysome association . ( A ) Meta-transcript distributions for two high polysome clusters ( 1 and 2 ) and one low polysome cluster ( 6 ) . Clusters 1 and 2 were pooled for the analysis in ( B ) . ( B ) The distance between distributions for 24 different transcript features evaluated for transcripts strongly or weakly associated with polysomes . Distance is the nonparametric effect size , measured as the dimensionless quantity Cliff’s d ( see Materials and methods ) and error bars are bootstrapped 95% confidence intervals . All differences except 5′ UTR and transcript length are significant at the p = 0 . 001 level based on two-tailed Mann-Whitney U-tests ( Figure 3—figure supplement 1A ) . See Materials and methods for a description of all features and how they were tabulated . UTR – untranslated region; CDS – coding sequence . ( C ) Enrichment of either cognate ( ATG start codon ) or non-cognate ( non-ATG ) uORFs in high polysome versus low polysome clusters . Density is uORFs per 100 isoforms . See also Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 01210 . 7554/eLife . 10921 . 013Figure 3—figure supplement 1 . Further details of high versus low polysome associated transcript isoform comparisons . ( A ) Empirical cumulative distribution functions ( eCDFs ) for all features compared , with p-values calculated using the Mann-Whitney U test . High polysome transcripts are from the union of clusters 1 and 2 , while low polysome transcripts are from cluster 6 . ( B ) Meta-isoform distributions for transcripts with 5′ leaders of indicated length ranges . Error bars and shaded gray area are upper and lower quartiles . ( C ) Percent of each TrIP-seq cluster with the indicated 5′ leader length . Note the increase in 5′ leader length in the poorly translated cluster 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 013 We find that longer coding sequences and highly abundant versions of gene-linked isoforms are biased towards high polysomes , likely because shorter coding sequences cannot accommodate as many ribosomes ( Figure 3B and Figure 3—figure supplement 1A ) . However , the length of the coding sequence is not the sole determinant of polysome association , since the ribosome density ( measured by dividing the weighted sum of TrIP-seq polysome reads by the number of cytoplasmic reads ) is also higher in gene-linked isoforms found in high polysomes ( Figure 3B ) . We also find that transcripts from gene-linked isoforms with more exons also tend to be better translated , as was observed using reporter genes ( Nott et al . , 2003; 2004 ) . This is not due to overall transcript length , as there is no significant difference between transcript length in gene-linked isoforms , reflecting a positive influence of splicing on translation . Highly translated isoforms contain fewer rare codons on average , but contain more stretches of rare codons , perhaps indicating a need for translational pausing ( Figure 3B ) . The observation that exons promote translatability has been shown for select genes , and here we show that this extends across the human genome . One of the strongest effects seen on polysome association of gene-linked isoforms comes from the length and content of the 3′ UTR ( Figure 3B and Figure 3—figure supplement 1A ) . Specifically , we find that gene-linked 3′ UTRs in low polysomes are considerably longer than those on high polysomes ( mean length 1551 nt versus 982 nt ) . There are numerous regulatory elements contained within 3′ UTRs , including microRNA binding sites , AU-rich elements , and protein-binding sites ( Bartel , 2009; Szostak and Gebauer , 2013 ) . The fraction of the 3′ UTR containing AU-rich elements is increased in low polysome gene-linked isoforms , possibly due to translational repression ( Brooks and Blackshear , 2013; Moore et al . , 2014 ) . Conserved predicted binding sites for miRNAs ( Garcia et al . , 2011 ) are also more abundant in poorly translated gene-linked isoforms , but this could be due to a correlation between 3′ UTR length and the number of miRNA binding sites . We , therefore , filtered the miRNAs to those bound to AGO1 or AGO2 in HEK 293T cells ( Ender et al . , 2008 ) . In this HEK 293T-expressed miRNA set , the difference for both the number of conserved miRNA binding sites and their score increases between gene-linked isoforms , suggesting that miRNA binding may be functionally relevant in this context ( Figure 3B and Figure 3—figure supplement 1A ) . Increased AU-rich elements and miRNA binding should lead to decreased half-life for low polysome isoforms , which is borne out in half-life comparisons ( Figure 3B and Figure 3—figure supplement 1A; [ ( Tani et al . , 2012] ) . Surprisingly , gene-linked isoforms in high polysomes have increased predicted structure both at the cap and in 75-nucleotide windows in the 5′ leader . There are at least three possibilities that could explain this result . First , RNA structures can both repress or promote translation in a context dependent manner ( Xue et al . , 2015 ) , so it is possible that some isoforms are being driven to high polysomes by recruitment of transacting factors . Second , recent studies of global RNA secondary structure have consistently observed different structure in cells than in vitro ( Rouskin et al . , 2014; Spitale et al . , 2015 ) . Lastly , HEK 293T cells may translate messages with inhibitory 5′ leaders more efficiently than other cell types . In contrast , we find no significant dependence for the length of the 5′ leader region on polysome association based on this comparison ( Figures 3B and Figure 3—figure supplement 1A ) . However , isoforms containing 5′ leaders over 1000 nucleotides long are poorly ribosome-associated relative to shorter 5′ leaders ( Figure 3—figure supplement 1B ) and cluster 7 , which is associated with few ribosomes , contains longer 5′ leaders ( Figure 3—figure supplement 1C ) . It is likely that more subtle features of the 5′ leader also influence translatability , as in the examples in Figure 1D . Upstream open reading frames ( uORFs ) can positively or negatively influence translation ( Brar et al . , 2012; Calvo et al . , 2009; Ferreira et al . , 2013; Hinnebusch , 2005 ) . We , therefore , counted the number of experimentally determined uORFs ( Wan and Qian , 2014 ) in each cluster and observe surprisingly complex behavior . Cognate uORFs ( those starting with ATG ) promote or repress translation , while noncognate uORFs generally promote translation ( Arribere and Gilbert , 2013; Brar et al . , 2012 ) . Along these lines , we find that cluster 1 ( high polysomes ) is enriched for cognate and noncognate uORFs while cluster 6 ( low polysomes ) is enriched for cognate uORFs . However , cluster 2 ( high polysomes ) is enriched for only noncognate uORFs , indicative of the complex and idiosyncratic behavior of uORFs . Taken together , we find predominant influences for the 3′ UTR and the number of introns in determining polysome occupancy of gene-linked transcript isoforms in human cells , while a diversity of other features can influence translatability to a lesser extent . RNA-seq characterizes the isoform diversity of a sample , but not the functional consequences of this diversity . We reasoned that TrIP-seq data could be used to predict which isoform changes are likely to lead to translation changes in other systems . We chose a human embryonic dataset containing 124 individual cells from seven preimplantation developmental stages and reprocessed these data to directly compare to TrIP-seq data ( Materials and methods; [Yan et al . , 2013] ) . We mapped transcripts expressed during each developmental stage onto the TrIP-seq clusters , to attempt to gain insight into global translational properties in human embryos ( Figure 4A ) . In embryos , our analysis predicts a shift towards the extremes of translation , with an increase in the percent of transcript isoforms that are both highly ( e . g . clusters 1 and 2 ) and lowly ( e . g . cluster 6 ) translated isoforms in HEK 293T cells ( Figure 4A ) . Localized translation is widespread in development , and it is possible that the observed increase in poorly translated isoforms reflects a greater need for translational control ( Besse and Ephrussi , 2008; Jung et al . , 2014 ) . It is also possible that translation in early embryos is differentially regulated than in HEK 293T cells , which is now testable by applying TrIP-seq to other cell types , yielding cell type-specific translational control programs . 10 . 7554/eLife . 10921 . 014Figure 4 . Predicted isoform-specific translational control changes during human embryogenesis . ( A ) For each embryonic stage , transcripts are mapped onto all eight TrIP-seq clusters , as in Figure 2 . The percentage of transcripts mapping to each cluster per stage is then calculated , and compared to the percentage for TrIP-seq data . hESC – human embryonic stem cell; p0 , p10 – passage zero or passage ten . Error is s . d . between single cells at each embryonic stage or between TrIP-seq biological replicates . ( B ) Expression levels for two transcripts of CSDE1 across embryonic development and polysome fractions demonstrating a switch in translational status . Error is S . D . TPM – transcripts per million . ( C ) Diagram of two transcript isoforms of CSDE1 . Regions different between the two isoforms are in yellow and select shared intronic regions have been shortened for clarity . See also Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 01410 . 7554/eLife . 10921 . 015Figure 4—figure supplement 1 . Clustering of human preimplantation embryo data ( A ) Hierarchical clustering of 45 , 895 transcript isoforms across human preimplantation embryogenesis with the stages indicated below . Yellow: above isoform average , cyan: below isoform average . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 01510 . 7554/eLife . 10921 . 016Figure 4—figure supplement 2 . Additional examples of transcript isoforms that exhibit differential expression in human embryos and differential translation in TrIP-seq data . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 016 We then collected genes with isoforms that change between early and late embryonic stages . Clustering of abundances for 45 , 895 isoforms across embryogenesis yielded seven clusters with varying profiles , which we then analyzed by comparing transcripts of the same gene located in different embryonic clusters as before ( Figure 4—figure supplement 1A ) . These isoforms were filtered by those that move between low- and high-polysome clusters in the TrIP-seq data , yielding 366 isoform pairs belonging to 270 genes that are developmentally regulated and exhibit differential translation in HEK 293T cells . For example , the CSDE1 gene expresses one isoform in oocytes , early development and in human ES cells , which is poorly translated ( CSDE1-002 ) , but a second appears following fertilization that is well-translated ( CSDE1-007; Figure 4B , C ) . As global zygotic transcription begins in the two-cell stage ( Vassena et al . , 2011; Yan et al . , 2013 ) , it is possible that the alternative CSDE1 isoform is supplied by the spermatid ( Fischer et al . , 2012; Soumillon et al . , 2013 ) or that it may be precociously transcribed . We also present three other genes ( EIF4A2 , RNF170 , and TBC1D15 ) that show stage-specific expression in the embryo data and differential polysome association in HEK 293T cells ( Figure 4—figure supplement 2 ) . Therefore , widespread changes at the transcriptome level during human embryogenesis may produce concomitant changes in protein production , which can now be predicted using TrIP-seq data . Engineering the translation of a transcript without altering its coding sequence is desirable when introducing exogenous mRNA to cells or patients . We hypothesized that the regulatory features we discovered by comparing highly to lowly translated isoforms ( Figure 3B ) should be transferrable to an arbitrary reporter . Nine different 5′ leaders and eight different 3′ UTRs derived from gene-linked isoforms ( Figure 3 ) were appended onto Renilla luciferase with a synthetic poly-A60 tail , which were individually in vitro transcribed , capped , and 2′-O-methylated ( Figure 5—figure supplement 1A and Materials and methods ) . We elected to transcribe RNA both because plasmid-encoded transcripts can be heterogeneous and to mimic a scenario where one is delivering RNA to affect cell activity or for therapeutic intervention ( Kormann et al . , 2011; Warren et al . , 2010 ) . All tested 5′ leader sequences modulate protein production by the luciferase reporter in concordance with the observed polysome association by TrIP-seq , when transfected into HEK 293T cells ( Figure 5A , B ) . The three 5′ leaders from EEF1B2 alter luciferase production in a stepwise manner by roughly a factor of 20 . The TrIP-seq profiles from NAE1 are the most distinct among those tested , and the two luciferase constructs differ in output by two orders of magnitude . Surprisingly , the two 5′ leaders from RICTOR differ by only nine nucleotides , yet still exhibit differential protein production , possibly due to altered local RNA secondary structure near the 5′ cap ( Figure 5—figure supplement 1B ) . Both 5′ leaders from SRSF5 contain two uORFs , but they are close ( SRSF5-002 ) or far ( SRSF5-005 ) from the start codon ( Figure 5—figure supplement 1A ) , suggesting reinitiation following uORF translation may be impacting luciferase production ( Grant et al . , 1994; Hinnebusch , 2005 ) . Not only do these data provide strong evidence that TrIP-seq data can be used to predictably tune the output of heterologous mRNAs using isoform-specific untranslated regions , it additionally validates that the polysome profiles observed are connected to translational output . 10 . 7554/eLife . 10921 . 017Figure 5 . Regulatory UTR sequences are sufficient to control translation of a heterologous reporter . HEK 293T cells were transfected with indicated RNAs and Renilla luciferase units are plotted . ( A–D ) Renilla light units normalized to control UTR mRNA ( A , C ) and corresponding traces from TrIP-seq data ( B , D ) for 5′ leaders ( A , B ) or 3′ UTRs ( C , D ) exhibit differential protein production over two log-units in cells . Inset graphs show comparisons between mock-transfected cells and the lowest output mRNAs . Error bars are S . E . M . from at least four biological and three technical replicates ( twelve total; A ) or three biological and three technical replicates ( nine total; B ) . ( B , D ) Error is S . D . Rluc – Renilla luciferase units; UTR – untranslated region . ( E ) Luciferase fold change versus the average number of ribosomes on each transcript , computed by averaging the plots in B and D . See also Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 01710 . 7554/eLife . 10921 . 018Figure 5—figure supplement 1 . Diagrams of isoforms differentially represented in gigh versus low polysomes . ( A ) Isoform diagrams for the genes CCNE2 , NAB1 , NAE1 , NDC1 , RICTOR , and SRSF5 . Coding regions are wide boxes , noncoding regions are narrow boxes , introns are lines , and upstream open reading frames ( uORFs ) are indicated by gray bars . Differentially included regions are shown in yellow , arrows show the direction of transcription , and select introns have been removed for clarity as indicated by gaps . ( B ) Secondary structures of two RICTOR 5′ leader sequences predicted by RNAfold . RNA bases are colored by base pairing probability . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 018 We additionally selected eight 3′ UTRs from four sets of gene-linked isoforms to test their ability to control translation , and find good agreement with the TrIP-seq polysome abundance for two out of four 3′ UTR pairs ( Figure 5C , D ) . In all cases , replacing the 3′ UTRs leads to decreased protein production compared to the short 3′ UTR in the control RNA . The paired 3′ UTRs of NAB1 and RICTOR each differ by two kilobases , and a factor of 35 and 44 in luciferase output , respectively , indicating that 3′ UTRs can strongly modulate protein production ( Mayr and Bartel , 2009; Sandberg et al . , 2008 ) . However , the two other paired 3′ UTRs , despite also differing by two kilobases each , are not distinguishable at the protein production level in cells . The two tested CCNE2 isoforms differ in both the 5′ and 3′ UTRs , so it is possible that the isoform-level translational control is occurring via the 5′ UTR , and there is a small but not significant difference between the two NDC1 3′ UTRs , suggesting regulation of this 3′ UTR may be subtle . We also find a positive relationship between the average number of ribosomes on each transcript isoform in TrIP-seq and the luciferase fold change ( Figure 5E ) . In sum , we show that elements found to control translation at the isoform level using TrIP-seq can be grafted on to heterologous coding sequences to control translation over a range of two orders of magnitude in human cells . As different cell types contain different macromolecules , they may also translate messages with the same regulatory features differently . We , therefore , tested the panel of gene-specific untranslated regions fused with luciferase ( Figure 5 ) in four additional cell lines: A549 ( lung carcinoma ) , K-562 ( chronic myelogenous leukemia ) , MCF-7 ( breast adenocarcinoma ) , and Hep G2 ( hepatocellular carcinoma ) . We additionally tested HEK 293T cells at a shorter timepoint of 2 hr , to explore the role of RNA stability in the observed luciferase output . Remarkably , the luciferase production from all tested 5′ leader reporters was qualitatively similar in all six conditions ( Figure 6A ) . However , the difference between the two RICTOR 5′ leader sequences is decreased . In contrast , the 3′ UTR reporters show considerably more variability ( Figure 6B ) . Calculation of the coefficient of variation of all reporters across all six conditions highlights the greater variability found in 3′ UTR reporters ( Figure 6—figure supplement 1A ) . We additionally compared HEK 293T cells at 2-hr and 8-hr post-transfection to ascertain differences in RNA stability . Most 5′ leader reporters are similar between these two timepoints ( Figure 6A ) , but 3′ UTR reporters show larger changes . Specifically , NAB1-001 and RICTOR-001 generate much more luciferase at the 2-hr timepoint , suggesting the long UTRs of these genes may promote RNA degradation leading to decreased protein at eighteen hours . 10 . 7554/eLife . 10921 . 019Figure 6 . Translational control by transcript 5′ leader sequences is robust across cell types . ( A , B ) Five cell lines in six conditions were transfected with Renilla luciferase fused to the UTR indicated and luciferase units normalized to control UTR mRNA . Both 5′ leaders ( A ) and 3′ UTRs ( B ) were tested . Error is S . E . M . between three technical and three biological replicates ( nine total ) per condition . See also Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 01910 . 7554/eLife . 10921 . 020Figure 6—figure supplement 1 . Changes in protein production conferred by 5′ UTRs are more robust across cell lines than those conferred by 3′ UTRs . Plotted is the coefficient of variation ( the standard deviation divided by the mean ) for the indicated UTR fused to luciferase across five cell lines and HEK 293T cells at 2 hr and 18 hr ( Figure 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 020 Thus , translational control by the 5′ leader sequence is robust to changes in macromolecular composition across the cell lines tested , perhaps because the abundance of most translation initiation factors is high ( Kulak et al . , 2014 ) , which will buffer against small changes in their expression level . However , factors interacting with 3′ UTRs , specifically miRNAs , can vary considerably between cell types , and this variability may be the cause of the diverse behavior of the 3′ UTR reporters across cell lines . Therefore , at least for the panel tested here , transcript 5′ leaders confer robust control of protein production across cell types , while transcript 3′ UTRs may be better suited to tune the production of a protein to a particular cell type . In this work , we showed that the dynamic range of transcript-isoform-specific translational control spans at least two orders of magnitude in human cells ( Figure 5 ) , indicating that it is crucial to take isoform-level effects into account when assessing translation in organisms with extensive alternative transcript processing ( Figure 7A ) . Globally , we find that alternative 3′ UTRs broadly influence translation , with longer isoforms of the same gene associated with lower protein production ( Figure 3B ) . We further show that regulatory regions are sufficient to control the translation of unrelated coding sequences , enabling predictable tuning of translational output of arbitrary genes ( Figures 5 and 6B ) . We demonstrate that translational control conferred by a panel of 5′ leaders is more robust across cell types than 3′ UTRs , suggesting predictable control of protein production . Future work focused on measuring isoform-specific translation in different cell types will yield cell type-specific regulatory sequences , which could then be used to engineer cell type-specific translation of mRNA , as in derivation of pluripotent stem cells ( Warren et al . , 2010 ) , or mRNA therapeutics ( Kormann et al . , 2011 ) . 10 . 7554/eLife . 10921 . 021Figure 7 . Transcript processing has widespread effects on protein production . ( A ) The mapping between gene and protein is complex , and RNA processing has a strong influence on the translatability of individual transcript isoforms and therefore protein production . ( B ) Regulatory elements can be discovered by comparing transcript isoforms that are differentially translated , and subsequently fused to heterologous genes to control protein output in cells . ORF – open reading frame . DOI: http://dx . doi . org/10 . 7554/eLife . 10921 . 021 Several observations indicate that TrIP-seq is a faithful measure of translation in cells . First , TrIP-seq and ribosome profiling are highly correlated at the gene level ( Figure 1F ) and both ribosome profiling and TrIP-seq correlate with mass spectrometric measurements of protein abundance ( [Ingolia et al . , 2009] , Figure 1—figure supplement 1F ) . Second , protein-coding genes are enriched in high ribosomal fractions ( Figure 1E ) , while lncRNAs are found primarily in the low ribosomal fractions ( Figure 1E and Figure 1—figure supplement 1D ) . Third , a systematic investigation of translational stalling in mouse embryonic stem cells found this phenomenon to primarily be transient and not lead to ribosome accumulation on transcripts ( Ingolia et al . , 2011 ) , although in specific circumstances it can be more widespread ( Darnell et al . , 2011; Ishimura et al . , 2014; Richter and Coller , 2015 ) . Fourth , known highly- and lowly-translated transcript isoforms are enriched in high- and low-polysome fractions in TrIP-seq data ( Figure 1C ) . Fifth , reads derived from stalled polysomes would artificially inflate the apparent translatability , however , TrIP-seq primarily underestimates rather than overestimates protein abundance ( Figure 1—figure supplement 1F ) . Sixth , cryptic 'pseudo-polysomes' induced by miRNA complexes would preferentially enhance polysome association for long 3′ UTR transcript isoforms ( Maroney et al . , 2006; Nottrott et al . , 2006; Olsen and Ambros , 1999; Thermann and Hentze , 2007 ) , which is the opposite of what is observed ( Figure 3B ) . Lastly , transcript isoform changes found to lead to differential ribosome association using TrIP-seq are sufficient to modulate the protein output from a reporter RNA in a manner predicted by the TrIP-seq data ( Figure 5 ) , indicating that transcript-specific ribosome association is a correlate of protein output . Therefore , we conclude that TrIP-seq is a measure of transcript-specific translation . What are the mechanistic foundations of the observed transcript-specific translation ? Each of the different untranslated regions tested by reporter assays operate using different mechanisms . For example , the three EEF1B2 transcript isoforms all share the same uORF , but alternative splicing changes the distance between the uORF and the start codon , which can influence the efficiency of downstream initiation ( Grant et al . , 1994; Hinnebusch , 2005 ) . Globally , we found a large difference between the 3′ UTR lengths of poorly- versus well-translated isoforms by querying the prevalence of features likely to influence translation ( Figures 3 and 5 ) . There are at least three possibilities for how 3′ UTRs could influence translation . First , miRNA-mediated regulation targets the 3′ UTR ( Mayr and Bartel , 2009; Sandberg et al . , 2008 ) . Second , numerous RNA binding proteins target the 3′ UTR through AU-rich elements or other specific binding sites and are known to influence translation , mRNA decay and even protein localization ( Berkovits and Mayr , 2015; Szostak and Gebauer , 2013; Zhao et al . , 2014 ) . Lastly , it is possible that increasing the length of the 3′ UTR influences the impact of the mRNA closed loop on translation initiation or reinitiation ( Amrani et al . , 2008; Costello et al . , 2015 ) . Targeted work on specific 3′ UTR isoform sets could test the different possibilities for 3′ UTR-mediated translational control . We additionally demonstrated isoform-specific translational control for select 5′ leaders ( Figures 1 and 5 ) as was shown genome-wide in yeast , where 5′ leaders are short enough to directly sequence ( Arribere and Gilbert , 2013 ) , and find that globally uORFs can both up- and down-regulate translation ( Figure 3 ) . However , we find a surprisingly minor global dependence of other 5′ leader features tested on translation . The mechanisms of translational control of transcript isoforms therefore follow some general trends but are likely highly idiosyncratic; these data empower investigations into the mechanisms underlying transcript-specific translational control for thousands of human genes . We predicted the consequences of transcript isoforms observed during human embryonic development on protein production ( Figure 4 ) . We show that even small changes to transcripts can dramatically affect protein production ( Figure 5 ) , so changes at the isoform level ( Figure 4 ) can considerably affect protein abundances even if the gene-level RNA expression remains similar . The application of TrIP-seq to different cell types should yield cell type-specific translation enhancer and repressor elements , contributing to the understanding of cell type-specific translational control , and augmenting our ability to precisely engineer translation in complex systems . If simpler sample preparation is desired , it is likely sufficient to sequence the cytoplasmic and monosome fractions , as well as pooled two-four ribosome and five-eight+ ribosome fractions , as these cluster together ( Figure 1B ) . Accumulation of TrIP-seq data in additional cell types may enable building a holistic model of translational control with isoform resolution in human cells . We showed that the translation of an arbitrary gene , Renilla luciferase , can be controlled in a manner predicted by TrIP-seq data over two orders of magnitude ( Figure 5 ) and in different cell types ( Figure 6 ) . These data can thus be used to select regulatory regions to control translation , without redesign of the coding sequence of the message ( Figure 7B ) . It may prove superior to use 3′ UTRs to design cell type specific translation , as the repertoire of miRNAs and RNA-binding proteins that may affect translation through the 3′ UTR vary between cell types ( Figure 6B ) . Indeed , in Caenorhabditis elegans , 3′ UTRs are sufficient to specify germline-specific expression of the attached ORF ( Merritt et al . , 2008 ) . Even without engineering of unnatural mRNAs , it may be possible to use antisense oligonucleotides to direct splicing of poorly- or well-translated isoforms to adjust protein expression in situ ( Kole et al . , 2012 ) , for example to downregulate the oncogenic NAE1 gene [Figure 5A , Xie et al . , 2014] ) , which is currently being targeted by small molecules ( Luo et al . , 2012; Wu and Yu , 2015 ) . We anticipate the ability to tune the translational output of mRNA will facilitate research and therapeutic uses of designed and endogenous mRNA molecules . Two independently passaged , biological replicate 15 cm dishes of HEK 293T cells obtained from the University of California , Berkeley cell culture facility were grown to ~70% confluency in DMEM + 10% FBS . The cell line was authenticated by DDC Medical ( Fairfield , OH ) and were verified to be free of mycoplasma contamination . Cells were actively growing when harvested . The media was aspirated and replaced by PBS + 100 μg/ml cycloheximide and incubated at 37°C for 10 min . We chose cycloheximide because it induces rapid protein synthesis arrest ( Han et al . , 2014 ) , has been successfully used in ribosome profiling of HEK 293T cells ( Sidrauski et al . , 2015 ) , and 100 μg/ml is ~100 times higher than the concentration required to inhibit protein synthesis in reticulocytes ( Godchaux et al . , 1967 ) . Each dish was then placed on ice , media aspirated , and replaced by ice cold PBS + 100 μg/ml cycloheximide . Cells were scraped , pelleted at 16 , 000× g for 30 s , and re-suspended in three pellet-volumes ice cold hypotonic lysis buffer ( 10 mM HEPES pH 7 . 9 , 1 . 5 mM MgCl2 , 10 mM KCl , 0 . 5 mM DTT , 1% Triton X-100 and 100 μg/ml cycloheximide ) ( Folco et al . , 2012 ) . After 10 min , cells were lysed on ice by ten strokes through a 26-gauge needle and nuclei were pelleted at 1 , 500× g for 5 min . Lysate from ~15 million cells ( one dish ) was layered on top of triplicate 10–50% ( w/v ) sucrose gradients ( 20 mM HEPES:KOH pH 7 . 6 , 100 mM KCl , 5 mM MgCl2 , 1 mM DTT and 100 μg/ml cycloheximide ) made using a Biocomp Instruments ( Canada ) gradient master . Gradients were centrifuged for 2 hr at 36 , 000 RPM in a SW-41 rotor , punctured , and manually peak fractionated using real-time A260 monitoring with a Brandel ( Gaithersburg , MD ) gradient fractionator and ISCO ( Lincoln , NE ) UA-6 detector . RNA was extracted from pooled technical triplicate sucrose gradient fractions by ethanol precipitation followed by acid phenol:chloroform extraction . Direct phenol:chloroform extraction was precluded by phase inversion in high sucrose fractions . RNA was then DNase treated , acid phenol:chloroform extracted , and ethanol precipitated . Cytoplasmic RNA was TRIzol extracted ( Life Technologies , Grand Island , NY ) and ethanol precipitated . Total RNA integrity was verified using a BioAnalyzer ( Agilent , Santa Clara , CA ) . Ribosomal RNA was then depleted using Ribo-Zero ( Illumina , San Diego , CA ) and biological duplicate sequencing libraries were generated using the TruSeq RNA Sample Prep v2 kit ( Illumina ) without the poly-A selection steps . An equal mass ( 100 ng ) of rRNA-depleted RNA was used as input to each individual library preparation . Libraries were verified using a BioAnalyzer and quantified using a Qubit ( Life Technologies ) prior to pooling for sequencing . Library insert sizes were typically ~150 ± ~25 bp . Pooled libraries were 75-bp paired-end sequenced on an Illumina HiSeq 2500 and runs of the same library in different lanes or flowcells were merged . Adapters were trimmed using Cutadapt v1 . 5 ( Martin , 2011 ) followed by subtractive alignments against the repeatmasker ( RMSK ) database ( retrieved from UCSC on 2/11/2015 ) and abundant sequences from the Illumina iGenomes project ( e . g . ribosomal RNA and the mitochondrial chromosome ) using Bowtie2 v2 . 2 . 4 ( Langmead and Salzberg , 2012 ) . Unaligned reads were then aligned to the Ensembl release 75 transcriptome using Tophat v2 . 0 . 13 with parameters "-r 5 --mate-std-dev 50 -g 100 --report-secondary-alignments" ( Trapnell et al . , 2009 ) . Mapping percentages for each pipeline stage are presented in Figure 1—figure supplement 3 . Transcript isoform level abundances were calculated using Cuffquant v2 . 2 . 1 ( Roberts et al . , 2011; Trapnell et al . , 2010 ) , normalized between samples with Cuffnorm v2 . 2 . 1 , and transcripts per million ( TPM ) ( Wagner et al . , 2012 ) values were calculated according to: TPMi=RPKMi 106∑g∈all genes RPKMg Biological duplicate datasets were processed independently . Quantified abundances are in Figure 1—source data 1 and Figure 1—source data 2 . TrIP-seq plots are available for all Ensembl GRCh37 isoforms at http://meru . qb3 . berkeley . edu/tripseq . Cuffnorm counts for each replicate were subjected to a variance stabilizing transformation ( VST ) using the DESeq2 R package to correct for heteroscedasticity ( Love et al . , 2014 ) ; the VST approaches log2 for large counts but compresses low counts to suppress Poisson noise . Variance stabilized counts were averaged between replicates , filtered such that the mean across all nine samples was greater than one ( roughly translating to 100 reads ) , and then mean subtracted to generate relative expression values . Inter-row distance was computed using Spearman’s rank correlation , and hierarchical clustering was performed using the fastcluster R package ( Müllner , 2013 ) with Ward’s agglomeration method . The resulting dendrogram was split at increasing heights into subtrees until clusters contained similar overall trends , generating the large clusters presented in Figure 2 . The samples were clustered similarly except with Euclidean distance and complete agglomeration . The R packages magrittr , dendextend Galili , 2015 , and ggplot2 Wickham , 2009 were used to generate figures shown and are available through CRAN . Transcript isoforms in each cluster are in Figure 2—source data 1 . The intersample clustering in Figure 1B was performed by hierarchically clustering the VST-transformed isoform-level counts as above between samples . The statistical significance of the resulting clustering was analyzed by measuring the Jaccard distance between clusterings of these data after subjecting the data to subsampling by bootstrap , jittering ( adding random noise to each point ) , or replacing random points by noise using the R package fpc . The mean Jaccard distances of 100 such subsamplings as well as the average are presented in Figure Figure 1—figure supplement 1G S1G; a Jaccard distance ≥ . 75 is considered a stable cluster by the R package fpc . Isoform-specific 5′ and 3′ untranslated regions were amplified from anchored oligo-dT primed cDNA libraries from HEK 293T cells and Gibson cloned ( Gibson et al . , 2009 ) into a vector based on pUC57 containing Renilla luciferase and a synthetic polyA60 tail ( pA60; Fukaya and Tomari , 2011 ) . Gibson cloning was performed such that untranslated regions were precisely cloned next to the ATG but contained two guanosines as the 5′-most nucleotides for T7 transcription for 5′ leaders , or the stop codon used by the isoform , and the six nucleotides CTGCAG at the 3′ end of the 3′ UTR immediately preceding the polyA60 tail for 3′ UTRs . The differences between cloned isoforms are shown in Figure 5—figure supplement 1 , which was generated using IGV ( Thorvaldsdottir et al . , 2013 ) . The entirety of all untranslated regions was verified using dideoxy sequencing . Transcription templates were generated by PCR using Phusion polymerase ( NEB , Ipswich , MA ) , size verified by gel electrophoresis , and gel purified . Transcription was performed using T7 polymerase with 1 μg template in a buffer containing 7 . 5 mM each NTP , 1 μg pyrophosphatase ( Roche , Pleasanton , CA ) , 30 mM DTT , 35 mM MgCl2 , 2 mM spermidine , 0 . 01% Triton X-100 and 30 mM Tris pH 8 . 1 for four hours at 37°C followed by DNase treatment with RQ1 RNase-free DNase ( Promega , Madison , WI ) for 30 min . Transcription products were purified by ethanol precipitation and a Zymo ( Irvine , CA ) Clean & Concentrator column , followed by simultaneous capping using Vaccinia capping enzyme ( NEB ) and 2′-O-methylation ( NEB ) . Capped products were purified using a Zymo Clean & Concentrator column , followed by size verification on glyoxylated samples ( Ambion , Foster City , CA ) using an agarose gel , and full-length 3′ UTR-containing RNAs were gel purified from an agarose gel ( Zymo ) . The concentration of RNAs containing 5′ untranslated regions was determined using A260 , which was normalized by the intensity of the full-length product on an agarose gel , and then by the molar ratio of the construct to empty pA60 . Molar-adjusted amounts of each RNA relative to 100–200 ng of pA60 were transfected into three technical triplicate wells of ~50% confluent cells in a 96-well plate using the TransIT-mRNA reagent ( Mirus , Madison , WI ) . The concentration of RNAs containing 3′ UTRs was determined using a Qubit RNA HS assay ( Life Technologies ) , normalized for the molar ratio of the construct to empty pA60 , and molar-adjusted amounts of RNA relative to 7 ng of pA60 were transfected using TransIT-mRNA . A reduced amount of the 3′ UTR RNAs was used due to decreased yield of the longest 3′ UTRs . HEK 293T cells were grown in DMEM + 10% FBS , Hep G2 cells were grown in EMEM + 10% FBS , MCF7 cells were grown in DMEM:F12 + 10% FBS , A549 cells were grown in F12-K media + 10% FBS , and K-562 cells were grown in RPMI media + 10% FBS . Cells were harvested after ~18 hr ( Figure 5 ) or 2 hr ( Figure 6 ) and Renilla luminescence was measured ( Promega ) . The Ensembl release 75 annotation set from the Illumina iGenomes project was first decomposed into 5′ leader , start codon , CDS , 3′ UTR , and whole-transcript regions . Length , GC-content , and number of exons were computed directly . Cytoplasmic expression and the median expression from the 80S through the eight+ ribosome fraction were extracted from TrIP-seq data . Transcript halflife data were derived from HeLa cell measurements ( Tani et al . , 2012 ) . The structure was computed using RNALfold from the ViennaRNA package ( Lorenz et al . , 2011 ) in a 75-nt window . Codon usage statistics were downloaded from http://www . kazusa . or . jp/codon/ on 11/20/2014 and the minimum codon frequency is the average of codon usage across a five-codon window . The fraction of AU-elements is calculated as the percentage of the 3′ UTR that is of repeating A or U nucleotides for more than 5nt in a row . TargetScan 6 . 2 scores ( Garcia et al . , 2011 ) were downloaded from http://targetscan . org and parsed for the properties indicated , and an identical comparison was performed after filtering the miRNA list by those expressed in HEK 293T cells ( Ender et al . , 2008 ) . All feature tabulation was performed using custom Python programs , which are available through GitHub at https://github . com/stephenfloor/tripseq-analysis . Isoforms belonging to the same gene present in different clusters were compiled , yielding gene-linked isoforms . For example , if the gene A has isoform 001 in cluster one and isoforms 002 and 003 in cluster two , features of isoform 001 are added to the cluster one set and features of 002 and 003 are added to the cluster two set and 001–002 as well as 001–003 would be gene-linked isoforms . Features in each set were then compared for statistical significance using the Mann-Whitney U test and visualized as empirical cumulative distribution functions ( Figure 3—figure supplement 1A ) . The effect size was then computed between distributions using Cliff’s d , which is a nonparametric , dimensionless measure of the distance between distributions ( Cliff , 1993 ) . Cliff’s d is a measure of the number of times that a point xi in one distribution is greater than all points xj in the second distribution , or d=# ( xi>xj ) −# ( xi<xj ) mn where # denotes the number of times , the two distributions are of sizes n and m , and xi and xj are items of the two sets . Cliff’s d is also related to the Mann-Whitney U statistic , by d=2Umn−1 Cliff’s d and the boostrap confidence intervals shown in Figure 3 were computed using the R package orddom , which is available through CRAN . Ribosome profiling was performed in HEK 293T cells ( Sidrauski et al . , 2015 ) . These data were downloaded from the NCBI ( GEO: GSE65778 ) and reprocessed by subtractive Bowtie 2 alignment to rRNA and RMSK sequences , and then mapped onto the human transcriptome using Tophat , as for the TrIP-seq data . Aligned reads were quantified at the gene and isoform level using Cuffquant onto the Ensembl release 75 transcriptome and normalized with Cuffnorm as above , to facilitate direct comparison to the TrIP-seq data . The number of reads derived from polysomes for TrIP-seq data was computed as the sum of read counts for fractions containing two to eight+ ribosomes multiplied by the number of ribosomes in each fraction: TrIP−seq polysome counts=∑i=17i ni+24n8 where ni is the number of reads in the ith polysome fraction averaged for both biological replicates . The factor of 24 is applied to the last fraction based on a ceiling of ~40 ribosomes per transcript which then leads to an average of 24 ribosomes in the eighth peak ( 40 + 8 / 2 = 24 ) . The number of ribosomes in the final peak is not directly measured by the sucrose gradients used Figure 1—figure supplement 1A ) . RNA sequencing datasets of human preimplantation embryos ( Yan et al . , 2013 ) were downloaded from the NCBI ( GEO: GSE36552 ) . Reads were converted to FASTQ using fastq-dump ( NCBI ) and then processed as for TrIP-seq data by adapter trimming with Cutadapt , subtractive alignment to the RMSK and abundant sequences , and transcriptome alignment with Tophat to Ensembl release 75 . Transcript abundances were calculated using Cuffquant and normalized with Cuffnorm . Data from individual cells were processed independently and averaged for each stage to generate the data shown in Figure 4 and Figure 4—figure supplement 1 . The data were reprocessed for consistency to facilitate direct abundance comparisons . Cytoplasmic lysate was fractionated using a sucrose gradient as for the RNA-seq libraries and RNA was extracted from each fraction . Libraries of cDNA were then synthesized from these fractions using random primers ( Applied Biosystems ) from an equal amount of RNA as measured using a Qubit ( Life Technologies; RNA HS assay ) . Template RNA was removed using RNase H treatment ( NEB ) and cDNAs were purified over an oligo cleanup column ( Zymo ) . PCR was then performed for 25 ( ACTB ) or 30 ( EEF1B2 , ATF4 ) cycles with gene-specific primers ( below ) using Taq Titanium ( Clontech ) using an equal amount of cDNA input into each PCR , as measured using a Qubit ( ssDNA assay ) . Reactions were then run on a 1% agarose gel and stained with SYBR-Gold ( Life Technologies ) . Gels are representative of three biological replicates . Primers used for RT-PCR are: ACTB-forward AGAGCTACGAGCTGCCTGAC ACTB-reverse AGCACTGTGTTGGCGTACAG EEF1B2-forward TTCCCGTCATCTTCGGGAGCCGT EEF1B2-reverse CTTTTCAGGTCTCCGAAACCCATGG ATF4-forward GGCTCTGCAGCGGCAACCCC ATF4-reverse CGACTGGTCGAAGGGGGACA RNA was extracted from polysome fractions as for TrIP-seq . qRT-PCR was performed using the SuperScript III Platinum SYBR Green One-Step kit ( Life Technologies ) with 1ng input RNA in a 20 ul reaction volume . CT values were converted to fold changes over the cytoplasmic abundance and plotted . Primers used for qRT-PCR are: EEF1B2-001-fwd AGCCGTGGAGCTCTCGGATA EEF1B2-001-rev AGCTCTTGTCCGCCAGGTAA EEF1B2-003-fwd TCCAAACGCAACGAAAGGTCC EEF1B2-003-rev CGCCGGACCGAAGGTTAAAG EEF1B2-201-fwd AGCCGTGGAGCGTGGG EEF1B2-201-rev TCGGCTGTATCCGAGAGCTG SRSF5-002-fwd GACCCCGTCCGGTAGGAAGTACTAGCC SRSF5-002-rev CAATATCTCTTATCCGTCCATATCCC SRSF5-005-fwd GGTGAGTGGCTCACTTTGAGGGCAAG SRSF5-005-rev CGAATCAACTGCGCTCATTAGACGC The DAVID server was used to calculate gene ontology ( GO ) terms for biological processes ( BP ) associated with the individual clusters ( Jiao et al . , 2012 ) . Transcripts associated with each cluster were input into DAVID and GO BP terms with a Benjamini-corrected p-value of <0 . 05 were tabulated . Raw sequencing reads for all samples are available through the NCBI via the GEO Accession ID GSE69352 .	To produce a protein , a gene’s DNA is first copied to make molecules of messenger RNA ( mRNA ) . The mRNAs pass through a molecular machine known as the ribosome , which translates the genetic code to make a protein . Not all of an mRNA is translated to make a protein; the “untranslated” regions play crucial roles in regulating how much of the protein is produced . In animals , plants and other eukaryotes , many mRNAs are made up of small pieces that are “spliced” together . During this process , proteins are deposited on the mRNA to mark the splice junctions , which are then cleared when the mRNA is translated . Many different mRNAs can be produced from the same gene by splicing different combinations of RNA pieces . Each of these mRNA “isoforms” can , in principle , contain a unique set of features that control its translation . Hence each mRNA isoform can be translated differently so that different amounts of the corresponding protein product are produced . However , the relationship between the variety of isoforms and the control of translation is complex and not well understood . To address these questions , Floor and Doudna measured the translation of over 60 , 000 mRNA isoforms made from almost 14 , 000 human genes . The experiments show that untranslated regions at the end of the mRNA ( known as the 3′ end ) strongly influence translation , even if the protein coding regions remain the same . Furthermore , the data showed that mRNAs with more splice junctions are translated better , implying an mRNA has some sort of memory of how many junctions it had even after the protein markers have been cleared . Next , Floor and Doudna inserted regulatory sequences from differently translated isoforms into an unrelated “reporter” gene . This dramatically changed the amount of protein produced from the reporter gene , in a manner predicted by the earlier experiments . Untranslated regions at the beginning of the mRNAs ( known as the 5′ end ) controlled the amount of protein produced from the reporter consistently across different types of cells from the body . On the other hand , the 3′ regions can tune the level of protein production in particular types of cells . Floor and Doudna’s findings demonstrate that differences between mRNA isoforms of a gene can have a big effect on the level of protein production . Changes in the types of mRNA made from a gene are often associated with human diseases , and these findings suggest one reason why . Additionally , the ability to engineer translation of an mRNA using the data is likely to aid the development of mRNA-based therapies .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "computational", "and", "systems", "biology" ]	2016	Tunable protein synthesis by transcript isoforms in human cells
We introduce an Interaction- and Trade-off-based Eco-Evolutionary Model ( ITEEM ) , in which species are competing in a well-mixed system , and their evolution in interaction trait space is subject to a life-history trade-off between replication rate and competitive ability . We demonstrate that the shape of the trade-off has a fundamental impact on eco-evolutionary dynamics , as it imposes four phases of diversity , including a sharp phase transition . Despite its minimalism , ITEEM produces a remarkable range of patterns of eco-evolutionary dynamics that are observed in experimental and natural systems . Most notably we find self-organization towards structured communities with high and sustained diversity , in which competing species form interaction cycles similar to rock-paper-scissors games . We observe an immense diversity in natural communities ( Hutchinson , 1961; Tilman , 1982; Huston , 1994 ) , but also in controlled experiments ( Maharjan et al . , 2006; Gresham et al . , 2008; Kinnersley et al . , 2009; Herron and Doebeli , 2013; Kvitek and Sherlock , 2013 ) , where many species continuously compete , diversify and adapt via eco-evolutionary dynamics ( Darwin , 1859; Cody and Diamond , 1975 ) . However , the basic theoretical models ( Volterra , 1928; Tilman , 1982 ) predict that both ecological and evolutionary dynamics tend to decrease the number of coexisting species by competitive exclusion or selection of the fittest . This apparent contradiction between observations and theory gives the stunning biodiversity in communities the air of a paradox ( Hutchinson , 1961; Sommer and Worm , 2002 ) and hence has begotten a long , ongoing debate on the mechanisms underlying emergence and stability of diversity in communities of competitive organisms ( Hutchinson , 1959; Huston , 1994; Chesson , 2000; Sommer and Worm , 2002; Doebeli and Ispolatov , 2010 ) . To identify candidate mechanisms that could resolve the problem of generation and maintenance of diversity , the basic theoretical ecological and evolutionary models have been extended by numerous features ( Chesson , 2000; Chave et al . , 2002 ) , including spatial structure ( Mitarai et al . , 2012; Villa Martín et al . , 2016; Vandermeer and Yitbarek , 2012 ) , spatial and temporal heterogeneity ( Caswell and Cohen , 1991; Fukami and Nakajima , 2011; Hanski and Mononen , 2011; Kremer and Klausmeier , 2013 ) , tailored interaction network topologies ( Melián et al . , 2009; Mougi and Kondoh , 2012; Kärenlampi , 2014; Laird and Schamp , 2015; Coyte et al . , 2015; Grilli et al . , 2017 ) , predefined niche width ( Scheffer and van Nes , 2006; Doebeli , 1996 ) , adjusted mutation-selection rate ( Johnson , 1999; Desai and Fisher , 2007 ) , and life-history trade-offs ( Rees , 1993; Bonsall et al . , 2004; de Mazancourt and Dieckmann , 2004; Gudelj et al . , 2007; Ferenci , 2016; Posfai et al . , 2017 ) . However , it is still unclear which features are essential to explain biodiversity . For instance , diversity is also observed under stable and homogeneous conditions ( Gresham et al . , 2008; Kinnersley et al . , 2009; Maharjan et al . , 2012; Herron and Doebeli , 2013; Kvitek and Sherlock , 2013 ) . So far , models of eco-evolutionary dynamics have been developed in three major categories: models in genotype space , like population genetics ( Ewens , 2012 ) and quasispecies models ( Nowak , 2006 ) ; models in phenotype space , like adaptive dynamics ( Doebeli , 2011 ) and webworld models ( Drossel et al . , 2001 ) ; and models in interaction space , like Lotka-Volterra models ( Coyte et al . , 2015; Ginzburg et al . , 1988 ) and evolving networks ( Mathiesen et al . , 2011; Allesina and Levine , 2011 ) . Each of these categories has strengths and limitations and emphasizes particular aspects . However , in nature these aspects are entangled by eco-evolutionary feedbacks that link genotype , phenotype , and interaction levels ( Post and Palkovacs , 2009; Schoener , 2011; Ferriere and Legendre , 2013; Weber et al . , 2017 ) . In a closed system of evolving organisms mutations , that is , evolutionary changes at the genetic level ( Figure 1a ) , can cause phenotypic variations if they are mapped to novel phenotypic traits in phenotype space ( Figure 1b ) ( Soyer , 2012 ) . These variations have ecological impact only if they affect biotic or abiotic interactions of species ( Figure 1c ) ; otherwise they are ecologically neutral . The resulting adaptive variations in the interaction network change the species composition through population dynamics . Finally , frequency-dependence occasionally selects strategies that adapt species to their new environment ( Schoener , 2011; Moya-Laraño et al . , 2014; Hendry , 2016; Weber et al . , 2017 ) . Thus , we have a link from interactions to eco-evolutionary dynamics , suggesting that we do not need to follow all evolutionary changes at the genetic or phenotypic level if we are interested in macro-eco-evolutionary dynamics , but only those changes that affect interactions . In this picture , evolution can be considered as an exploration of interaction space , and modeling at this level can help us to study how complex competitive interaction networks evolve and shape diversity . This neglect of genetic and phenotypic details in interaction-based models ( Ginzburg et al . , 1988; Solé , 2002; Tokita and Yasutomi , 2003; Shtilerman et al . , 2015 ) equals a coarse-graining of the eco-evolutionary system ( Figure 1 ) . This coarse-graining not only reduces complexity but it should also make the approach applicable to a broader class of biological systems . Interaction-based evolutionary models have received some attention in the past ( Ginzburg et al . , 1988; Solé , 2002 ) but then were almost forgotten , despite remarkable results . We think that these works have pointed to a possible solution of a hard problem: The complexity of evolving ecosystems is immense , and it is therefore difficult to find a representation suitable for the development of a statistical mechanics that enables qualitative and quantitative analysis ( Weber et al . , 2017 ) . Modeling at the level of interaction traits , rather than modeling of detailed descriptions of genotypes or phenotypes , coarse-grains these complex systems in a natural way so that this approach may be helpful for developing a biologically meaningful statistical mechanics . The first eco-evolutionary interaction-based model was introduced by Ginzburg et al . ( 1988 ) based on Lotka–Volterra dynamics for competitive communities . Instead of adding species characterized by random coefficients , taken out of some arbitrary species pool , they made the assumption that a new mutant should be ecologically similar to its parent , which means that phenotypic variations that are not ecologically neutral generate mutants that interact with other species similar to their parents ( Figure 1 ) . Thus , speciation events were simulated as ecologically continuous mutations in the strength of competitive interactions . This model , although conceptually progressive , was not able to produce a large stable diversity , possibly because diversity requires components not included in this model . Therefore subsequent interaction-based models supplemented it with ad hoc features to specifically increase diversity , such as special types of mutations ( Tokita and Yasutomi , 2003 ) , addition of mutual interactions ( Tokita and Yasutomi , 2003; Yoshida , 2003 ) , enforcement of partially connected interaction graphs ( Kärenlampi , 2014 ) , or imposed parent-offspring niche separation ( Shtilerman et al . , 2015 ) . While these models generated , as expected , higher diversity than the original Ginzburg model , they could not reproduce key characteristics of real systems , for example emergence of large and stable diversity , diversification to separate species and mass extinctions . Of course , the use of ad hoc features that deliberately increase diversity also cannot explain why diversity emerges . An essential component missing in the previous interaction-based models had been a constraint on strategy adoption . In real systems such constraints prevent the emergence of Darwinian Demons , that is , species that develop in the absence of any restriction and act as a sink in the network of population flow . Among all investigated features responsible for diversity , mentioned above , life-history trade-offs that regulate energy investment in different life-history strategies are fundamentally imposed by physical laws such as energy conservation or other thermodynamic constraints , and thus present in any natural system ( Stearns , 1989; Gudelj et al . , 2007; Del Giudice et al . , 2015 ) . These physical laws constrain evolutionary trajectories in trait space of evolving organisms and determine plausible evolutionary paths ( Fraebel et al . , 2017; Ng'oma et al . , 2017 ) , i . e . combinations of strategies adopted or abandoned over time . Roles of trade-offs for emergence and stabilization of diversity have been investigated in previous eco-evolutionary studies ( Posfai et al . , 2017; Rees , 1993; Bonsall et al . , 2004; de Mazancourt and Dieckmann , 2004; Ferenci , 2016; Gudelj et al . , 2007 ) and experiments ( Stearns , 1989; Kneitel and Chase , 2004; Agrawal et al . , 2010; Maharjan et al . , 2013; Ferenci , 2016 ) . It has been shown , for example , that if metabolic trade-offs are considered , even at equilibrium and in homogeneous environments , stable coexistence of species becomes possible ( Gudelj et al . , 2007; Beardmore et al . , 2011; Maharjan and Ferenci , 2016 ) . Here , we introduce a new , minimalist model , the Interaction and Trade-off-based Eco-Evolutionary Model ( ITEEM ) , with simple and intuitive eco-evolutionary dynamics at the interaction level that considers a life-history trade-off between interaction traits and replication rate , that means , better competitors replicate less ( Jakobsson and Eriksson , 2003; Bonsall et al . , 2004 ) . To our knowledge , ITEEM is the first model which joins these two elements , the interaction-space description with a life-history trade-off , that we deem crucial for an understanding of eco-evolutionary dynamics . We use ITEEM to study development of communities of organisms that diversify from one ancestor by gradual changes in their interaction traits and compete under Lotka-Volterra dynamics in well-mixed , closed system . We show that ITEEM dynamics , without any ad hoc assumption , not only generates large and complex biodiversity over long times ( Herron and Doebeli , 2013; Kvitek and Sherlock , 2013 ) but also closely resembles other observed eco-evolutionary dynamics , such as sympatric speciation ( Tilmon , 2008; Bolnick and Fitzpatrick , 2007; Herron and Doebeli , 2013 ) , emergence of two or more levels of differentiation similar to phylogenetic structures ( Barraclough et al . , 2003 ) , occasional collapses of diversity and mass extinctions ( Rankin and López-Sepulcre , 2005; Solé , 2002 ) , and emergence of cycles in interaction networks that facilitate species diversification and coexistence ( Buss and Jackson , 1979; Hibbing et al . , 2010; Maynard et al . , 2017 ) . Interestingly , the model shows a unimodal ( ‘humpback’ ) course of diversity as function of trade-off , with a critical trade-off at which biodiversity undergoes a phase transition , a behavior observed in nature ( Kassen et al . , 2000; Smith , 2007; Vallina et al . , 2014; Nathan et al . , 2016 ) . By changing the shape of trade-off and comparing the results with a no-trade-off model , we show that diversity is a natural outcome of competition if interacting species evolve under physical constraints that restrict energy allocation to different strategies . The natural emergence of diversity from a bare-bone eco-evolutionary model suggests that a unified treatment of ecology and evolution under physical constraints dissolves the apparent paradox of stable diversity . ITEEM is an individual-based model ( Black and McKane , 2012; DeAngelis and Grimm , 2014 ) with simple intuitive updating rules for population and evolutionary dynamics . A simulated system in ITEEM has Ns sites of undefined spatial arrangement ( no neighborhood ) , each providing permanently a pool of resources that is sufficient for the metabolism of one organism . The community is well-mixed , which means that the probability for an encounter is the same for all pairs of individuals , and that the probability of an individual to enter a site ( i . e . to access resources ) is the same for all individuals and sites . We start an eco-evolutionary simulation with individuals of a single strain occupying a fraction of the Ns sites , and then carry out long simulations for millions of generations . Note that in the following , to facilitate discourse , we use the term strain for a group of individuals with identical traits , whereas the term species denotes a monophyletic cluster of strains with some intraspecific diversity ( for a discussion on application of these terms in this study see Appendix 1 , Species and strains ) . Over time t , measured in generations , the number of individuals , Nind ( t ) , number of strains , Nst ( t ) , and number of species , Nsp ( t ) , change by ecological ( birth , death , competition ) and evolutionary dynamics ( mutation , extinction , diversification ) . Every generation or time step consists of Ns sequential replication trials of randomly selected individuals , followed at the end by a single death step . In the death step all individuals that have reached their lifespan at that generation will vanish . Lifespans of individuals are drawn at their births from a Poisson distribution with overall fixed mean lifespan λ . This is equivalent to an identical per capita death rate for all strains . For comparison , simulations with no attributed lifespan ( λ=∞ ) were carried out , too; in this case the only cause of death is defeat in a competitive encounter . At each replication trial , a randomly selected individual of a strain α can replicate with probability rα . Age of individuals plays no role in their reproduction and thus a newborn individual can be selected and replicate with the same probability as adult individuals . With a fixed probability μ the offspring mutates to a new strain α′ . Then , the newborn individual is assigned to a randomly selected site . If the site is empty , the new individual will occupy it . If the site is already occupied , the new individual competes with the current holder in a life-or-death struggle . In that case , the surviving individual is determined probabilistically by the ‘interaction’ Iαβ , defined for each pair of strains α , β . Iαβ is the survival probability of an α individual in a competitive encounter with a β individual , with Iαβ∈[0 , 1] and Iαβ+Iβα=1 ( Grilli et al . , 2017 ) . All interactions Iαβ form an interaction matrix I ( t ) that encodes the outcomes of all possible competitive encounters in this probabilistic sense . Row α of I defines the ‘interaction trait’ Tα= ( Iα1 , Iα2 , … , IαNst ( t ) ) of strain α , with Nst ( t ) the number of strains at time t . If strain α goes extinct , its interaction elements must be removed , i . e . the αth row and column of I are deleted . Conversely , if a mutation of α generates a new strain α′ , its trait vector is obtained by adding a small random variation to the parent trait , that is Tα′=Tα+η , where η= ( η1 , ⋯ , ηNst ( t ) ) is a vector of independent random variations , drawn from a zero-centered normal distribution of fixed width m . With this , I grows by one row and column . The new elements of the matrix are:Iα′β=Iαβ+ηβ , Iβα′=1−Iα′β , ( 1 ) Iα′α′=0 . 5 , where β=1 , ⋯ , Nst ( t ) and thus α′ inherits its interactions from α , but with a small random modification . Evolutionary variations in ITEEM generate mutants that are ecologically similar to their parents . Such variations can represent any phenotypic variation that influences interactions of strains with their community and thus changes their relative competitive abilities ( Thompson , 1998; Thorpe et al . , 2011; Bergstrom and Kerr , 2015; Thompson , 1999 ) . With Equation 1 we assume that all the interaction terms of the new mutant can change independently . To implement trade-off between competitive ability and fecundity , we introduce a relation between competitive ability C , defined as average interaction ( 2 ) C ( Tα ) =1Nst ( t ) −1∑β≠α Iαβ , and replication rα ( for fecundity ) . When Nst=1 , competitive ability of that single strain is set to zero . To study the influence of trade-off between competitive ability and replication , we systematically change its shape by varying a parameter δ ( 0≤δ<1 ) ( Figure 2 ) . For details of trade-off function and its effect on trait distribution and relative fitness see Appendix 1 , Trade-off . Trade-off functions can be concave ( δ<0 . 5 ) , linear ( δ=0 . 5 ) , or convex ( δ>0 . 5 ) . The trade-off function ties better competitive ability to lower fecundity and vice versa . The extreme case δ=0 makes r=1 and thus independent of C , which means no trade-off . We compare ITEEM results to the corresponding results of a neutral model ( Hubbell , 2001 ) , where we have formally evolving trait vectors Tα but fixed and uniform replication probabilities and interactions . Accordingly , the neutral model has no trade-off . ITEEM belongs to the well-established class of generalized Lotka-Volterra ( GLV ) models in the sense that the population-level approximation of the stochastic , individual-based ecological dynamics of ITEEM leads to the competitive Lotka-Volterra equations ( Appendix 1 , Generalized Lotka–Volterra ( GLV ) equation ) . Thus the results of the model can be interpreted in the framework of competitive GLV equations that model competition for a renewable resource pool and summarize all types of competition ( Gill , 1974; Maurer , 1984 ) in the elements of the interaction matrix I ( see above ) , i . e . these elements represent the resultant negative effect of all competitor populations on each other . Our model also allows to study speciation in terms of network dynamics . The interaction matrix I defines a complete dominance network between coexisting strains . In this network the nodes are strains ( α , β ) , and the directed edges connecting them indicate direction and strength of dominance , i . e . sign and size of Iαβ−Iβα , respectively . Thus , the elements of the weighted adjacency matrix of this network are defined as either Wαβ=Iαβ−Iβα , if α is the superior competitor in the pairwise encounter with β ( Iαβ>Iβα ) , or otherwise as Wαβ=0 . With this definition all Wαβ are in [0 , 1] . Accordingly , for the dominance network of species , we computed directed edges between any two species , i and j , by averaging over edges between all pairs of strains belonging to these species , that is Wijsp=W¯αβ for all strains α and β in the ith and jth species , respectively . The strength and direction of dominance edges indicate the effective flow of population between species . As we consider a trade-off between replication and competitive ability in the framework of GLV equations , we can distinguish between r- and α-selection ( Gill , 1974; Kurihara et al . , 1990; Masel , 2014 ) . r-selection selects for reproductive ability , which is beneficial in low density regimes , while α-selection selects for competitive ability and is effective at high density regimes under frequency-dependent selection . α-selection , first introduced by Gill ( Gill , 1974 ) , can be realized by acquisition of any kind of ability or mechanism that increases the chance of an organism to take over resources , to prevent competitors from gaining resources ( Gill , 1974 ) , or helps the organism to tolerate stress or reduction of contested resource availability ( Aarssen , 1984 ) . α-selection is different from K-selection; although both are effective at high density , the latter is limited to investments in efficient and parsimonious usage of resources ( Masel , 2014 ) . The source code of the ITEEM model is freely available at GitHub ( Farahpour , 2018; copy archived at https://github . com/elifesciences-publications/ITEEM ) . Our first question was whether ITEEM is able to generate and sustain diversity . Since we have a well-mixed system with initially only one strain , a positive answer implies sympatric diversification: the emergence of new species by evolutionary branching without geographic isolation or resource partitioning . In fact , we observe that during long-time eco-evolutionary trajectories in ITEEM new , distinct species emerge , and their coexistence establishes a sustained high diversity in the system ( Figure 3a ) . Remarkably , the emerging diversity has a clear hierarchical structure in the phylogeny tree and trait space: at the highest level we see that the phylogenetically separated strains ( Figure 3a and Appendix 1 , Species and strains ) appear as well-separated clusters in trait space ( Figure 3b ) similar to biological species . Within these clusters there are sub-clusters of individual strains ( Barraclough et al . , 2003 ) . Both levels of diversity can be quantitatively identified as levels in the distribution of branch lengths in minimum spanning trees in trait space ( Appendix 1 , SMST and distribution of species and strains in trait space ) . This hierarchical diversity is reminiscent of the phylogenetic structures in biology ( Barraclough et al . , 2003 ) . Overall , the model shows evolutionary divergence from one ancestor to several species consisting of a total of hundreds of coexisting strains ( Figure 3c ) . This evolutionary divergence in interaction space is the result of frequency-dependent selection without any further assumption on the competition function , for example a Gaussian or unimodal competition kernel ( Dieckmann and Doebeli , 1999; Doebeli and Ispolatov , 2010 ) , or predefined niche width ( Scheffer and van Nes , 2006 ) . In the course of this diverging sympatric evolution , diversity measures typically increase and , depending on trade-off parameter δ , high diversity is sustained over hundreds of thousands of generations ( Figure 3d , and Appendix 1 , Diversity over time ) . This observation holds for several complementary measures of diversity , no matter whether they are based on abundance of strains or species , or on functional diversity , i . e . quantities that measure the spread of the population in trait space ( Appendix 1 , Functional diversity ( FD ) , functional group and functional niche ) . The observed pattern of divergence contradicts the long-held view of sequential fixation in asexual populations ( Muller , 1932 ) . Instead , we see frequently concurrent speciation with emergence of two or more species in quick succession ( Figure 3a ) , in agreement with recent results from long-term bacterial and yeast cultures ( Herron and Doebeli , 2013; Maddamsetti et al . , 2015; Kvitek and Sherlock , 2013 ) . ITEEM systems self-organize toward structured communities: the interaction matrix of a diverse system obtained after many generations has a conspicuous block structure with groups of strains with similar interaction strategies ( Figure 3e ) , and these groups being well-separated from each other in trait space ( Figure 3b ) ( Sander et al . , 2015 ) . This fact can be interpreted in terms of functional organization as the interaction trait in ITEEM directly determines the functions of strains and species in the community ( Appendix 1 , Functional diversity ( FD ) , functional group and functional niche ) . This means that the block structure in Figure 3e corresponds to self-organized , well-separated functional niches ( Whittaker et al . , 1973; Rosenfeld , 2002; Taillefumier et al . , 2017 ) , each occupied by a cluster of closely related strains . This niche differentiation among species , which facilitates their coexistence , is the result of frequency-dependent selection among competing strategies . Within each functional niche the predominant dynamics , determining relative abundances of strains in the niche , is neutral . Speciation can occur when random genetic drift in a functional group generates sufficiently large differences between the strategies of strains in that group , and then selection forces imposed by biotic interactions reinforce this nascent diversification by driving strategies further apart . We observe as characteristic of the dynamics of the dominance network W ( see Model ) the appearance of strong edges as diversification increases trait distance ( or dissimilarity ) between species ( Figure 3f ) ( Anderson and Jensen , 2005 ) . Three or more directed edges in the dominance network can form cycles of strains in which each strain competes successfully against one cycle neighbor but loses against the other neighbor , a configuration corresponding to rock-paper-scissors games ( Szolnoki et al . , 2014 ) . Such intransitive dominance relations have been observed in nature ( Buss and Jackson , 1979; Sinervo and Lively , 1996; Lankau and Strauss , 2007; Bergstrom and Kerr , 2015 ) , and it has been shown that they stabilize a system driven by competitive interactions ( Allesina and Levine , 2011; Mathiesen et al . , 2011; Mitarai et al . , 2012; Laird and Schamp , 2015; Maynard et al . , 2017; Gallien et al . , 2017 ) . We find in ITEEM networks that the increase of diversity coincides with growth of mean strength of cycles ( Figure 3d , g and Appendix 1 , Intransitive dominance cycles ) . Note that these cycles emerge and self-organize in the evolving ITEEM networks without any presumption or constraint on network topology . Formation of strong cycles could also hint at a mechanistic explanation for another phenomenon that we observe in long ITEEM simulations: Occasionally diversity collapses from medium levels abruptly to very low levels , usually followed by a recovery ( Figure 3d ) . Remarkably , dynamics before these mass extinctions are clear exceptions of the generally strong correlation of diversity and average cycle strength . While the diversity immediately before mass extinctions is inconspicuous , these events are always preceded by exceptionally high average cycle strengths ( Appendix 1 , Collapses of diversity ) . Because of the rarity of mass extinctions in our simulations we currently have not sufficient data for a strong statement on this phenomenon , however , it is conceivable that the emergence of new species in a system with strong cycles likely leads to frustrations , i . e . the newcomers cannot be accommodated without inducing tensions in the network , and these tensions can destabilize the network and discharge in a collapse . The extinction of a species in a network with strong cycles will probably have a similar effect . This explanation of mass extinctions would be consistent with related works where collapses of diversity occur if maximization of competitive fitness ( here: by the newcomer species ) leads to a loss of absolute fitness ( here: break-down of the network ) ( Matsuda and Abrams , 1994; Masel , 2014 ) . This is a special case of the tragedy of the commons ( Hardin , 1968; Masel , 2014 ) that happens when competing organisms under frequency-dependent selection exploit shared resources ( Rankin and López-Sepulcre , 2005 ) , as it is the case in ITEEM . The eco-evolutionary dynamics described above depends on lifespan and trade-off between replication and competitive ability . This becomes clear if we study properties of dominance network and trait diversity . Figure 4a relates properties of the dominance network to the trade-off parameter δ , at fixed lifespan λ . Specifically , we plot two indicators of community structure against trade-off parameter δ , namely mean weight of dominance edges ⟨ W ⟩ , and mean strength of cycles ρ . Figure 4b summarizes the behavior of diversity as function of δ and lifespan λ . For this summary , we chose ten parameters that quantify different aspects of diversity , for example richness , evenness , functional diversity , and trait distribution , and then averaged over their normalized values to obtain an overall measure of diversity ( color bar in the figure ) . The full set of parameters is detailed in Appendix 1 , Diversity indexes and parameters of dynamics for different trade-offs and lifespans . The resulting phase diagram gives us an overview of the community diversity for different trade-off parameters δ and lifespans λ . The diagram shows a weak dependency of diversity on λ and a strong impact of δ , with four distinct phases ( I-IV ) from low to high δ as described in the following . Without trade-off ( δ=0 ) , strains do not have to sacrifice replication for better competitive abilities . Any resident community can be invaded by a new mutant with relatively higher C that does not have to compensate with a lower r . These mutants resemble Darwinian Demons ( Law , 1979 ) , i . e . strains or species that can maximize all aspects of fitness ( here C and r ) simultaneously and would exist under physically unconstrained evolution . Such Darwinian Demons can then be outcompeted by their own mutant offspring’s that have higher C and the same r . Thus we have sequential predominance of such strategies with constantly changing traits and improving competitiveness , but no diverse network emerges . As we increase δ from this unrealistic extreme into phase I ( 0<δ≲0 . 2 ) coexistence is facilitated . However , the small δ still favors investing in relatively higher competitive ability as a low-cost strategy to increase fitness . In this phase ⟨W⟩ and ρ ( Figure 4a ) slightly increase: biotic selection pressure exerted by inter-species interactions starts to generate diverse communities ( left inset in Figure 4b , Appendix 1 , Diversity indexes and parameters of dynamics for different trade-offs and lifespans ) . When δ increases further ( phase II ) , trade-off starts to force strains to choose between higher replication or better competitive abilities . Extremes of these quantities do not allow for viable species: sacrificing r completely for maximum C stalls population dynamics , whereas maximum r leads to inferior C . Thus strains seek middle ground values in both r and C . The nature of C as mean of interactions ( Equation 2 ) allows for many combinations of interaction traits with approximately the same mean . Thus , in a middle range of r and C , many strategies with the same overall fitness are possible , which is a condition of diversity ( Marks and Lechowicz , 2006 ) . From this multitude of strategies , sets of trait combinations emerge in which strains with different combinations keep each other in check , for example by the competitive rock-paper-scissors-like cycles between species described above . An equivalent interpretation is the emergence of diverse sets of non-overlapping compartments or functional niches in trait space ( Figure 3b , e ) . Diversity in this phase II is the highest and most stable ( middle inset in Figure 4b , Appendix 1 , Diversity indexes and parameters of dynamics for different trade-offs and lifespans ) . As δ approaches 0 . 7 , ⟨ W ⟩ and ρ plummet ( Figure 4a ) to interaction values comparable to the noise level m ( see Model ) , and a cycle strength typical for the neutral model ( horizontal light green ribbon in Figure 4a ) , respectively . The sharp drop of ⟨ W ⟩ and ρ at δ≈0 . 7 is reminiscent of a phase transition . As expected for a phase transition , the steepness increases with system size ( Appendix 1 , Size of the system ) . For δ≳0 . 7 , weights of dominance edges never grow and no structures , for example cycles , emerge . Diversity remains low and close to that of a neutral system . The sharp transition at δ≈0 . 7 which is visible in practically all diversity measures ( between phases II and III in Figure 4b , see also Appendix 1 , Diversity indexes and parameters of dynamics for different trade-offs and lifespans ) is a transition from a system dominated by biotic selection pressure to a neutral system . In high trade-off phase III , a small relative change in C produces a large relative change in r ( Appendix 1 , Strength of trade-off function ) . For instance , given a resident strain R with r and C , a closely related mutant M increases the fitness by adopting a relatively high r while paying a relatively small penalty in C ( see Appendix 1 , Strength of trade-off function for the relative impacts of the traits ) , and therefore will invade R . Thus , diversity in phase III will remain stable and low , and is characterized by a group of similar strains with no effective interaction and hence no diversification to distinct species ( right inset in Figure 4b and Appendix 1 , Diversity indexes and parameters of dynamics for different trade-offs and lifespans ) . In this high trade-off regime , lifespan comes into play: here , decreasing λ can make lives too short for replication . These hostile conditions minimize diversity and favor extinction ( phase IV ) . There is a well-known but not well understood unimodal relationship ( ‘humpback curve’ ) between biomass productivity and diversity: diversity as function of productivity has a convex shape with a maximum at middle values of productivity ( Smith , 2007; Vallina et al . , 2014 ) . This productivity-diversity relation has been reported at different scales in a wide-range of natural communities , for example phytoplankton assemblages ( Vallina et al . , 2014 ) , microbial ( Kassen et al . , 2000; Horner-Devine et al . , 2003; Smith , 2007 ) , plant ( Guo and Berry , 1998; Michalet et al . , 2006 ) , and animal communities ( Bailey et al . , 2004 ) . This behavior is reminiscent of horizontal sections through the phase diagram in Figure 4b , though here the driving parameter is not productivity but trade-off . However , we can make the following argument for a monotonic relation between productivity and trade-off shape . First we note that biomass productivity is a function of available resources ( Kassen et al . , 2000 ) : the larger the available resources , the higher the possible productivity . This allows us to argue in terms of available resources . For eco-evolutionary systems with scarce resources , species with high replication rates will have low competitive ability because for each individual of the numerous offspring there is little material or energy available to develop costly mechanisms that increase competitive ability . On the other hand , if a species under these resource-limited conditions produces competitively constructed individuals it cannot produce many of them . This argument shows a correspondence between a resource-limited condition and high δ for trade-off between replication and competitive ability . At the opposite , rich end of the resource scale , evolving species are not confronted with hard choices between replication rate and competitive ability , which is equivalent to low δ . Taken together , the trade-off axis should roughly correspond to the inverted resource axis: high δ for poor resources ( or low productivity ) and low δ for rich resources ( or high productivity ) ; a detailed analytical derivation will be presented elsewhere . The fact that ITEEM produces this frequently observed humpback curve proposes trade-off as underlying mechanism of this productivity-diversity relation . Observation of eco-evolutionary trajectories as in Figure 3 suggested the hypothesis that speciation and extinction events in ITEEM simulations do not occur at a constant rate and independently of each other , but that one speciation or extinction makes a following speciation or extinction more likely . Such a frequency-dependence occurs if emergence or extinction of one species creates the niche for emergence and invasion of another species , or causes its decline or extinction ( Herron and Doebeli , 2013 ) . Without frequency-dependence such evolutionary events should be uncorrelated . To test for frequency-dependent selection we checked whether the probability distribution of inter-event times ( time intervals between consecutive speciation or extinction events ) is compatible with a constant rate Poisson process , i . e . a purely random process , or whether such events are correlated ( Appendix 1 , Frequency-dependent selection ) . We find that for long inter-event times the decay of the distribution in ITEEM simulations is indistinguishable from that of a Poisson process . However , for shorter times there are significant deviations from a Poisson process for speciation and extinction events: at inter-event times of around 104 the probability decreases for a Poisson process but significantly increases in ITEEM simulations . Thus , the model shows frequency-dependent selection with the emergence of new species increasing the probability for generation of further species , and the loss of a species making further losses more likely . This behavior of ITEEM is similar to microbial systems where new species open new niches for further species , or the loss of species causes the loss of dependent species ( Herron and Doebeli , 2013; Maddamsetti et al . , 2015 ) . The above analysis illustrates a further application of ITEEM simulations . Eco-evolutionary trajectories from ITEEM simulations can be used to develop analytical methods for the inference of competition based on observed diversification patterns . Such methods could be instrumental for understanding the reciprocal effects of competition and diversification . Mutations are controlled in ITEEM by two parameters: mutation probability μ , and width m of trait variation . In simulations , diversity grew faster and to a higher level with increasing mutation probability ( μ=10−4 , 5×10−4 , 10−3 , 5×10−3 ) , but without changing the overall structure of the phase diagram ( Appendix 1 , Mutation probability ) . One interesting tendency is that for higher μ , the lifespan becomes more important at the interface of regions III and IV ( high trade-offs ) , leading to an expansion of region III at the expense of the hostile region IV: long lifespans in combination with high mutation probability establish low but viable diversity at large δ . The humpback curve of diversity over δ is observed for all mutation probabilities . Thus , the diversity in ITEEM is not a simple result of a mutation-selection balance but trade-off plays an important role in shaping diversity in trait space . The width of trait variation , m , influences both the speed of evolutionary dynamics and the maximum variation inside species , i . e . clusters of strains . The smaller m the slower the dynamics and the smaller the clusters . However extreme values of m can completely suppress the diverging evolution: Very small variations are wiped out by rapid ecological dynamics , and very large variations disrupt selection forces by imposing big fluctuations . The neutral model introduced in the Model section has no meaningful interaction traits , and consequently no meaningful competitive ability or trade-off with fecundity . Instead , it evolves solely by random drift in trait space . Similarly to ITEEM , the neutral model generates clumpy structures of traits ( Appendix 1 , Neutral model ) , though here the clusters are much closer and thus the functional diversity is much lower . This can be demonstrated quantitatively by the size of the minimum spanning tree of populations in trait space that are much smaller for the neutral model than for ITEEM at moderate trade-off ( Appendix 1 , Neutral model ) . The clumpy structures generated with the neutral model do not follow a stable trajectory of divergent evolution , and , hence , niche differentiation cannot be established . In a neutral model , without frequency-dependent selection and trade-off , stable structures and cycles cannot form in the community network , and consequently , diversity cannot grow effectively ( Appendix 1 , Neutral model ) . The comparison with the neutral model points to frequency-dependent selection as a promoter of diversity in ITEEM . For high trade-offs ( region III in Figure 4b ) , diversity and number of strong cycles in ITEEM are comparable to the neutral model ( Figure 4a ) . In established eco-evolutionary models , organisms are described in terms of one or a few phenotype traits . In contrast , the phenotype space of real systems is often very high-dimensional; competitive species in their evolutionary arms race are not confined to few predefined phenotypes but rather explore new dimensions in that space ( Maharjan et al . , 2006; Maharjan et al . , 2012; Zaman et al . , 2014; Doebeli and Ispolatov , 2017 ) . Coevolution systematically pushes species toward complex traits that facilitate diversification and coexistence ( Zaman et al . , 2014; Svardal et al . , 2014 ) , and evolutionary innovation frequently generates phenotypic dimensions that are completely novel in the system ( Doebeli and Ispolatov , 2017 ) . Complexity and multi-dimensionality of phenotype space have recently been the subject of several experimental and theoretical studies with different approaches that demonstrate that evolutionary dynamics and diversification in high-dimensional phenotype trait space can produce more complex patterns in comparison to evolution in low-dimensional space ( Doebeli and Ispolatov , 2010; Gilman et al . , 2012; Svardal et al . , 2014; Kraft et al . , 2015; Doebeli and Ispolatov , 2017 ) . For example , it has been shown that the conditions needed for frequency-dependent selection to generate diversity are satisfied more easily in high-dimensional phenotype spaces ( Doebeli and Ispolatov , 2010 ) . Moreover , the level at which diversity saturates in a system depends on its dimensionality , with higher dimensions allowing for more diversity ( Doebeli and Ispolatov , 2017 ) , and the probability of intransitive cycles in species competition networks grows rapidly with the number of phenotype traits . The conventional way to tackle this problem is to use models with a larger number of phenotype traits . However , this is not really a solution of the problem because this still confines evolution to the chosen fixed number of traits , and it also makes these models more complex and thus computationally less tractable . As will be discussed below , interaction-based models such as ITEEM offer a natural solution to this problem by mapping the system to an interaction trait space that can dynamically expand by the emergence of novel interaction traits as eco-evolutionary dynamics unfolds . Interaction-based eco-evolutionary models rely on the assumption that phenotypic evolution can be coarse-grained to the interaction level ( Figure 1 ) . This means that regardless of the details of phenotypic variations , we just study the resultant changes in the interaction network . In an eco-evolutionary system dominated by competition this is justified because phenotypic variations are relevant only when they change the interaction of organisms , directly or indirectly; otherwise they do not impact ecological dynamics . The interaction level is still sufficiently detailed to model macro-evolutionary dynamics that are dominated by ecological interactions . A transition from phenotype space to interaction space requires a mapping from the former to the latter , based on the rules that characterize the interaction of individuals with different phenotypic traits . As a concrete example , we might consider the competition kernel of adaptive dynamics models ( Doebeli , 2011 ) that determines the competitive pressure of two individuals with specific traits . That formalism describes well how , after mapping phenotypic traits to the interaction space , ecological outcome eventually is determined by interactions between species . In Appendix 1 , Phenotype-interaction map , some properties of this mapping are discussed . In the first interaction-based model by Ginzburg et al . ( 1988 ) , emergence of a new mutant was counted as speciation , and it was shown that simulating speciation events as ecologically continuous mutations in the strength of competitive interactions resulted in stable communities . However the Ginzburg model produced stable coexistence of only a few similar interaction traits , without branching and diversification to distinct species . As outlined in the introduction , subsequent interaction-based models tried to solve this problem by supplementing the Ginzburg model with some ad hoc features . For example , Tokita and Yasutomi ( 2003 ) mixed mutualistic and competitive interactions , and showed that only local mutations , i . e . changes in one pair-wise interaction rate , can produce stable diversity . Recently , Shtilerman et al . ( 2015 ) enforced diversification in purely competitive communities by imposing a large parent-offspring niche separation . To our knowledge , ITEEM is the first interaction-based model in which , despite its minimalism and without ad hoc features , diversity gradually emerges under frequency-dependent selection by considering physical constraints of eco-evolutionary dynamics . In all previous interaction-based models , eco-evolutionary dynamics has been divided into iterations over two successive steps: each first step of continuous population dynamics , implemented by integration of differential equations , was followed by a stochastic evolutionary process , namely speciation events and mutations , as a second step . However , in nature these two steps are not separated but intertwined in a single non-equilibrium process . Hence , the artificial separation necessitated the introduction of model components and parameters that do not correspond to biological phenomena and observables . In contrast , individual-based models like ITEEM operate with organisms as units , and efficiently simulate eco-evolutionary dynamics in a more natural and consistent way , with parameters that correspond to biological observables . Life-history trade-offs , like the trade-off between replication and competitive ability , now experimentally established as essential to living systems ( Stearns , 1989; Agrawal et al . , 2010; Masel , 2014 ) , are inescapable constraints imposed by physical limitations in natural systems . Our results with ITEEM show that trade-offs fundamentally impact eco-evolutionary dynamics , in agreement with other eco-evolutionary models with trade-off ( Huisman et al . , 2001; Bonsall et al . , 2004; de Mazancourt and Dieckmann , 2004; Beardmore et al . , 2011 ) . Remarkably , we observe with ITEEM sustained high diversity in a well-mixed homogeneous system . This is possible because moderate life-history trade-offs force evolving species to adopt different strategies or , in other words , lead to the emergence of well-separated functional niches in interaction space ( Gudelj et al . , 2007; Beardmore et al . , 2011 ) . Given the accumulating experimental and theoretical evidence , the importance of trade-off for diversity is becoming more and more clear . ITEEM provides an intuitive and generic conceptual framework with a minimum of specific assumptions or requirements . This makes the results transferable to different systems , for example biological , economical and social systems , wherever competition is the driving force of evolving communities . Put simply , ITEEM shows generally that in a bare-bone eco-evolutionary model withal standard population dynamics ( birth-death-competition ) and a basic evolutionary process ( mutation ) , diverse set of strategies will emerge and coexist if physical constraints force species to manage their resource allocation . Despite its minimalism , ITEEM reproduces in a single framework several phenomena of eco-evolutionary dynamics that previously were addressed with a range of distinct models or not at all , namely sympatric and concurrent speciation with emergence of new niches in the community , mass extinctions and recovery , large and sustained functional diversity with hierarchical organization , spontaneous emergence of intransitive interactions and cycles , and a unimodal diversity distribution as function of trade-off between replication and competition . The model allows detailed analysis of eco-evolutionary mechanisms and could guide experimental tests . The current model has important limitations . For instance , the trade-off formulation was chosen to reflect reasonable properties in a minimalist way . This should be revised or refined as more experimental data become available . Secondly , individual lifespans in this study came from a random distribution with an identical fixed mean . Hence we have no adaptation and evolutionary-based diversity in lifespan . This limits the applicability of the current model to communities of species that have similar lifespans , and that invest their main adaptation effort into growth or reproduction and competitive ability . Furthermore , our model assumes an undefined pool of steadily replenished shared resources in a well-mixed system . This was motivated by the goal of a minimalist model for competitive communities that could reveal mechanisms behind diversification and niche differentiation , without resource partitioning or geographic isolation . However , in nature , there will in general be few or several limiting resources and abiotic factors that have their own dynamics . For this scenario , which is better explained by a resource-competition model than by the GLV equation , it is possible to consider resources as additional rows and columns in the interaction matrix I and in this way to include abiotic interactions as well as biotic ones . In an interaction-based model like ITEEM the interaction terms of the mutants change gradually and independently ( Equation 1 ) . This assumption of random exploration of interaction space can be violated , for example , in simplified models with few fixed phenotypic traits . Further studies are necessary to investigate the general properties and restrictions of the map between phenotype and interaction space . In Appendix 1 , Phenotype-interaction map we briefly introduced and discussed some properties of this map .	A patch of rain forest , a coral reef , a pond , and the microbes in our guts are all examples of biological communities . More generally , a community is a group of organisms that live together at the same place and time . Many communities are composed of a large number of different species , and this diversity is maintained for long times . Although diversity is a key feature of biological communities , the mechanisms that generate and maintain diversity are not well understood . Research had hinted at links between diversity and the trade-offs that species are subject to . For instance , there is a trade-off between competitiveness and reproduction: if there are limited resources in the environment a species may either produce many offspring that are not very competitive , or fewer , more competitive offspring . Farahpour et al . have now simulated the development of communities of organisms that reproduce , compete , and die in a uniform environment . Crucially , these computational simulations introduced a trade-off between competitive ability and reproduction . The simulations show that the form of trade-off has a fundamental impact on diversity: moderate trade-offs favor diversity , whereas extreme trade-offs suppress diversity . The simulations also revealed mechanisms that underlie how diversity is generated . In particular , cyclic relationships emerge where one species dominates another but is also dominated by a third , similar to the rock-paper-scissors game . Since Farahpour et al . used a bare-bone model with only a few essential features the results could apply to a larger class of community-like systems whose evolution is driven by competition . This includes economic and social systems as well as biological communities .	[ "Abstract", "Introduction", "Model", "Results", "Discussion" ]	[ "ecology", "computational", "and", "systems", "biology" ]	2018	Trade-off shapes diversity in eco-evolutionary dynamics
Prp3 is an essential U4/U6 di-snRNP-associated protein whose functions and molecular mechanisms in pre-mRNA splicing are presently poorly understood . We show by structural and biochemical analyses that Prp3 contains a bipartite U4/U6 di-snRNA-binding region comprising an expanded ferredoxin-like fold , which recognizes a 3′-overhang of U6 snRNA , and a preceding peptide , which binds U4/U6 stem II . Phylogenetic analyses revealed that the single-stranded RNA-binding domain is exclusively found in Prp3 orthologs , thus qualifying as a spliceosome-specific RNA interaction module . The composite double-stranded/single-stranded RNA-binding region assembles cooperatively with Snu13 and Prp31 on U4/U6 di-snRNAs and inhibits Brr2-mediated U4/U6 di-snRNA unwinding in vitro . RNP-disrupting mutations in Prp3 lead to U4/U6•U5 tri-snRNP assembly and splicing defects in vivo . Our results reveal how Prp3 acts as an important bridge between U4/U6 and U5 in the tri-snRNP and comparison with a Prp24-U6 snRNA recycling complex suggests how Prp3 may be involved in U4/U6 reassembly after splicing . Pre-mRNA splicing is catalyzed by a multi-subunit RNA-protein ( RNP ) enzyme , the spliceosome , which facilitates two successive transesterification reactions ( steps 1 and 2 ) that lead to the removal of an intron and the ligation of its flanking exons . For each splicing event , a spliceosome is newly formed via the stepwise recruitment of small nuclear ( sn ) RNPs ( U1 , U2 , U4 , U5 and U6 in the case of the major spliceosome ) and numerous non-snRNP proteins to a pre-mRNA substrate ( Wahl et al . , 2009; Will and Lührmann , 2011 ) . During canonical cross-intron spliceosome assembly , U1 and U2 snRNPs bind the 5′-splice site ( 5′SS ) and the branch point region of an intron , respectively ( Mount et al . , 1983; Kramer et al . , 1984; Parker et al . , 1987; Wu and Manley , 1989; Zhuang and Weiner , 1989 ) . Formation of this A complex is followed by the recruitment of a preformed U4/U6•U5 tri-snRNP , yielding the pre-catalytic B complex ( Bindereif and Green , 1987; Cheng and Abelson , 1987; Konarska and Sharp , 1987; Deckert et al . , 2006; Fabrizio et al . , 2009 ) , which requires extensive conformational and compositional rearrangements to form a catalytically active spliceosome . Catalytic activation includes the disruption of the U1/5′SS interaction ( Konforti et al . , 1993 ) and the separation of U4 snRNA from U6 snRNA ( Konarska and Sharp , 1987; Yean and Lin , 1991 ) , which are extensively base-paired in the tri-snRNP and in the B complex , leading to the dissociation of U1 snRNP , U4 snRNA and all U4/U6-associated proteins . These rearrangements give rise to the Bact complex ( Fabrizio et al . , 2009; Bessonov et al . , 2010 ) and , upon further remodeling , to the B* complex ( Warkocki et al . , 2009 ) , in which U6 snRNA forms a central part of the spliceosome's active site . The B* complex facilitates the first step of splicing , the ensuing C complex ( Konarska et al . , 2006; Bessonov et al . , 2008; Fabrizio et al . , 2009 ) catalyzes the second step of splicing , after which the spliceosome releases its products and the remaining snRNPs and non-snRNP factors are recycled . To participate in further rounds of splicing , the U4/U6•U5 tri-snRNP must be reassembled by initial dimerization of U4 and U6 snRNPs followed by association with U5 snRNP ( Stanek and Neugebauer , 2006 ) . Association of U4 and U6 snRNPs is mediated in part by base pairing between their respective snRNAs , which form two inter-molecular helices ( stems I and II ) that are separated by a U4 5′-stem-loop ( 5′SL; Figure 1A ) . U4/U6 base pairing is mutually exclusive with the U6 snRNA conformation in the activated spliceosome . Reannealing of U4 and U6 snRNAs after splicing thus requires the Prp24 assembly chaperone in yeast ( Raghunathan and Guthrie , 1998 ) or its SART3 ortholog in human ( Bell et al . , 2002 ) , which transiently bind U6 snRNA , as well as the LSm proteins ( Achsel et al . , 1999; Rader and Guthrie , 2002; Verdone et al . , 2004 ) , which bind and remain at the 3′-end of U6 snRNA ( Beggs , 2005; Zhou et al . , 2014 ) . In addition , the U4-specific Prp3 protein is required for U4/U6 di-snRNP and U4/U6•U5 tri-snRNP formation ( Anthony et al . , 1997 ) but molecular mechanisms underlying its functions are poorly understood . 10 . 7554/eLife . 07320 . 003Figure 1 . Protein and RNA requirements for Prp3 binding to U4/U6 di-snRNA . ( A ) Schematic presentation of yeast U4/U6 di-snRNA ( yU4—gold; yU6—orange ) and domain organizations of yeast and human Prp3 . Regions corresponding to the C-terminal U4/U6-binding fragments ( CTF ) of the proteins are indicated by black lines above the schemes . ( B ) ESMA monitoring binding of hPrp3 protein variants ( as MBP or GST fusions; 25 µM ) to hU4/U6stem II+13nt ( scheme on the left ) . hPrp3 constructs are indicated above the lanes . FL—full-length; N—residues 1–442; C—residues 195–683; CTF—residues 484–683 . Bands are identified on the right; RNA—unbound RNA; RNPs—RNA–protein complexes . ( C ) EMSA titrations monitoring binding of yPrp3CTF variants and yPrp3DUF1115 ( proteins indicated at the left of the gels ) to yU4/U6stem II+13nt ( scheme on the top ) . Protein concentrations in each lane are indicated above the first gel . Bottom: quantification of the data above . The data were fit to a single exponential Hill equation ( fraction bound = A[protein]n/ ( [protein]n + Kdn ) : A , fit maximum of RNA bound; n , Hill coefficient ) ( Ryder et al . , 2008 ) . Errors indicate standard errors of the mean of at least two independent experiments . ( D ) Isothermal titration calorimetry monitoring interactions between yPrp3CTF variants or yPrp3DUF1115 ( proteins indicated above and on the left of each panel ) to yU4/U6stem II+13nt ( scheme in the first panel ) . The proteins , in particular yPrp3DUF1115 and yPrp3CTF , R322A , tended to aggregate when added in excess of available RNA binding sites , giving rise to background signals at the ends of some runs . Data points in gray in the DUF1115 analysis were omitted during the fitting . Deduced binding stoichiometries ( N ) , Kd′s , enthalpies ( ΔH ) and entropies ( ΔS ) of the interactions are listed in the lower parts of the panels . ( E ) Binding of yPrp3CTF ( 20 µM ) to the indicated fragments of yU4/U6 ( schemes above the gels; arrows indicate sequential shortening of RNA elements ) . Lanes 1–14—shortening of the yU6 3′-overhang . Lanes 15–20—shortening of stem II . ( F ) EMSA titrations monitoring binding of binding of yPrp3CTF or yPrp3DUF1115 to yU4/U6stem II+13nt bearing yU6 C69 G ( top two gels ) or yU4 C12 G ( bottom two gels ) exchanges that restore Watson–Crick base pairing ( schemes on the far left; mutant residues highlighted in green ) . Proteins are indicated at the left of the gels . Protein concentrations in each lane are indicated above the first gel . Bottom: quantification as in panel 1C . CTF-wt U4/U6—reference copied from panel 1C . Errors indicate standard errors of the mean of at least two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 00310 . 7554/eLife . 07320 . 004Figure 1—figure supplement 1 . Protein sequence comparisons . Multiple sequence alignments of Prp3 orthologs , acylphosphatases ( AcyP ) and blue light photoreceptors containing the BLUF domain . Sequences were aligned using the program MUSCLE ( Edgar , 2004 ) , manually adjusted and displayed using ALSCRIPT ( Barton , 1993 ) . Symbols below the alignments denote secondary structure elements assigned using PDBsum ( Laskowski et al . , 1997 ) based on the crystal structures of yPrp3CTF ( this study ) , Bos taurus AcyP ( PDB ID 2ACY ) ( Thunnissen et al . , 1997 ) and the R . sphaeroides BLUF domain ( PDB ID 2BYC ) ( Jung et al . , 2005 ) . Cyan triangles—yPrp3296−469 residues in contact with yU6 snRNA; red diamonds—residues in the flavin-binding pocket of the R . sphaeroides BLUF domain , yellow circles—phosphate-binding and active site residues in B . taurus AcyP . Residue numbers refer to the yPrp3 sequence . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 00410 . 7554/eLife . 07320 . 005Figure 1—figure supplement 2 . RNA sequence comparisons . ( A , B ) Multiple sequence alignments of U6 regions ( B ) forming stem II and the 3′-overhang and the U4 region ( C ) forming stem II in U4/U6 di-snRNAs . Residues in red form non-Watson–Crick pyrimidine–pyrimidine base pairs in the lower parts of stem II . The U6 sequences for H . sapiens ( gi: 669633321 ) , M . musculus ( gi: 90855862 ) , D . rerio ( gi: 159248718 ) , X . tropicalis ( gi: 284925142 ) , D . melanogaster ( gi: 8768 ) , C . elegans ( gi: 6894 ) , S . pombe ( gi: 159883886 ) and S . cerevisiae ( gi: 4512 ) as well as the U4 sequences for H . sapiens ( gi: 36 , 174 ) , M . musculus ( gi: 175433 ) , D . rerio ( gi: 111306314 ) , D . melanogaster ( gi: 45557156 ) , C . elegans ( gi: 6893 ) , S . pombe ( gi: 5084 ) and S . cerevisiae ( gi: 172644 ) were obtained from NCBI ( http://www . ncbi . nlm . nih . gov ) . The U4 sequence for X . tropicalis was extracted from the X . tropicalis 8 . 0 Genome ( http://www . xenbase . org/entry ) by using H . sapiens U6 as a query sequence . Sequences were aligned using the program MUSCLE ( Edgar , 2004 ) , manually adjusted and displayed using Geneious ( http://www . geneious . com ) . The relevant U6 ( B ) and U4 regions ( C ) were extracted according to the full-length sequence alignments and secondary structure predictions . ( C ) Schemes of U4/U6 stem II from selected organisms , highlighting ( red ) the non-Watson–Crick pyrimidine–pyrimidine base pairs in the lower parts of stem II . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 005 Human ( h ) and yeast ( y ) Prp3 form a complex with the respective Prp4 proteins ( Ayadi et al . , 1998; Gonzalez-Santos et al . , 2002 ) . hPrp3 can be crosslinked to the U6 snRNA portion of a U4/U6 di-snRNA complex comprising the U4 5′SL , an intact stem II and a U6 3′-overhang ( Nottrott et al . , 2002 ) and can pull down U4 and U6 snRNAs from nuclear extract ( Gonzalez-Santos et al . , 2002 ) , pointing towards direct Prp3-snRNA interactions . hPrp3 also interacts with the U5-specific proteins hPrp6 and hSnu66 ( Liu et al . , 2006 ) . Except for an N-terminal region , Prp3 is highly conserved from yeast to human and contains a C-terminal domain of unknown function ( DUF1115; PFAM ID PF06544; Figure 1A; Figure 1—figure supplement 1 ) . Recent homology modeling has predicted a ferredoxin-like fold for the human Prp3 DUF1115 domain ( Korneta et al . , 2012 ) . Here , we demonstrate that C-terminal regions of Prp3 , including DUF1115 , bind U4/U6 di-snRNA fragments containing stem II and a U6 3′-overhang and elucidated crystal structures of a yPrp3 C-terminal region alone and in complex with a U4/U6 di-snRNA fragment . Structure-guided mutations that led to reduced U4/U6 interaction in vitro and to reduced cell viability also reduced U4/U6•U5 tri-snRNP levels and splicing in vivo . Our results indicate how Prp3 functionally bridges U4/U6 and U5 in the tri-snRNP by Prp3-RNA interactions on one side and Prp3-protein interactions on the other . Moreover , a comparison with the structure of a Prp24-U6 snRNA complex ( Montemayor et al . , 2014 ) suggests how Prp3 may initiate the handover of U6 snRNA from the Prp24 recycling factor to U4 snRNP during U4/U6 reassembly . Previous studies have shown that human ( h ) Prp3 in the context of a hPrp3-hPrp4-hCypH complex contacts U6 snRNA in a region that forms stem II and a U6 single-stranded 3′-overhang ( Nottrott et al . , 2002 ) . To test whether hPrp3 alone is sufficient for stable RNA binding and to delineate hPrp3 elements required for complex formation , we produced full-length hPrp3 ( hPrp3FL ) and fragments lacking ca . 200–250 residues from either end ( hPrp3N—residues 1–442; hPrp3C—residues 195–683 ) as N-terminal maltose-binding protein [MBP] fusion proteins . We then tested binding of these proteins to a hU4/U6 construct containing stem II ( fused by a GAAA tetraloop ) and a 13 nucleotide [nt] U6 3′-overhang ( hU4/U6stem II+13nt ) in electrophoretic mobility shift assays ( EMSAs ) . Only the full-length protein and hPrp3C , but not hPrp3N , bound hU4/U6stem II+13nt ( Figure 1B , lanes 2–4 ) . Trypsin treatment of the hPrp3C-hU4/U6stem II+13nt complex gave rise to a protein fragment containing residues 484–683 ( hPrp3CTF ) as shown by mass spectrometric fingerprinting and N-terminal sequencing . hPrp3CTF contains the predicted DUF1115 domain ( residues 540–683 ) and a conserved , preceding peptide rich in basic amino acid residues ( Figure 1A; Figure 1—figure supplement 1 ) . Binding of recombinant hPrp3CTF ( as a glutathione S-transferase [GST] fusion ) to hU4/U6stem II+13nt was comparable to hPrp3FL or hPrp3C ( Figure 1B , lane 5 ) . These results show that the C-terminal ca . 200 amino acids of hPrp3 encompass the protein elements that mediate stable U4/U6 binding . All following experiments were performed with yeast ( y ) factors . To test whether the C-terminal U4/U6 di-snRNA-binding region is conserved in yPrp3 , we produced a protein comprising the 177 C-terminal residues of yPrp3 ( residue 296–469; yPrp3CTF ) . yPrp3CTF bound a yU4/U6 duplex containing the complete stem II and a 13-nt yU6 3′-overhang ( yU4/U6stem II+13nt ) with an apparent dissociation constant ( Kd , app ) of 0 . 92 µM as determined by EMSA ( Figure 1C , first gel and quantification ) and with a Kd of 110 nM as determined by isothermal titration calorimetry ( ITC; Figure 1D , top left ) . The lower apparent affinity estimated by EMSA is likely due to the presence of non-specific tRNA competitor in this assay . The yPrp3 DUF1115 domain ( residues 325–469; yPrp3DUF1115 ) lacking the preceding basic peptide showed ca . twofold reduced affinity for yU4/U6stem II+13nt in EMSA ( Kd , app 1 . 83 µM; Figure 1C , second gel and quantification ) and a ca . 3 . 5-fold lower affinity in ITC ( Kd 380 µM; Figure 1D , top right ) . Exchange of conserved arginine residues at positions 304 and 322 in the N-terminal basic peptide of yPrp3CTF ( Figure 1—figure supplement 1 ) , which could directly interact with the negatively charged RNA backbone , reduced affinities for yU4/U6stem II+13nt in both EMSA ( Kd , app 1 . 58 µM and 2 . 63 µM for yPrp3CTF , R304A and yPrp3CTF , R322A , respectively; Figure 1C , third and fourth gels and quantification ) and ITC ( Kd 260 nM and 400 nM for yPrp3CTF , R304A and yPrp3CTF , R322A , respectively; Figure 1D , bottom left and right ) to a similar extent as removal of the entire preceding peptide . Furthermore , the thermodynamic signatures of the interactions involving yPrp3DUF1115 , yPrp3CTF , R304A and yPrp3CTF , R322A changed considerably compared to yPrp3CTF . The DUF1115 domain as well as both arginine-to-alanine variants ( in particular yPrp3CTF , R304A ) exhibited significantly less favorable interaction enthalpies and more favorable ( or less unfavorable ) interaction entropies compared to yPrp3CTF ( Figure 1D ) . One explanation for these observations could be that the N-terminal peptide is a flexible element in yPrp3CTF , which becomes immobilized ( loss in conformational entropy ) by RNA contacts ( gain in interaction enthalpy ) upon binding of yU4/U6stem II+13nt . In any case , these observations show that both the DUF1115 domain and the preceding basic peptide contribute to the RNA binding . Next , we further probed the RNA requirements for stable binding . yPrp3CTF efficiently bound a yU4/U6 constructs bearing full-length stem II and yU6 3′-overhangs of at least eight nts ( Figure 1E , lanes 1–8 ) , while further shortening of the yU6 3′-overhang led to progressively reduced binding ( lanes 9–14 ) . Removal of the first two A-U base pairs from the 5′-end of stem II had no consequence for binding of yPrp3CTF ( Figure 1E , lanes 15–18 ) , but reduced binding was seen when seven base pairs , including a non-canonical C69U6-C12U4 pair , were removed ( Figure 1E , lane 20 ) . A non-canonical pyrimidine–pyrimidine base pair is conserved in the corresponding stem II regions of U4/U6 from other organisms ( Figure 1—figure supplement 2 ) . When we converted yU6 C69 or yU4 C12 of the non-canonical C-C pair in stem II to G’s to allow Watson-Crick base pairing at these positions , the affinity of yPrp3CTF was reduced about twofold in EMSA titrations ( Figure 1F , first and third gels and quantification ) . In contrast , the DUF1115 domain alone did not exhibit reduced affinity to the mutant RNAs compared to the wt ( Figure 1F , second and fourth gels and quantification ) . These findings suggest that yPrp3CTF recognizes portions of the U6 3′-overhang as well as parts of yU4/U6 stem II , including the non-canonical C69U6-C12U4 pair . Binding of Prp3 to U4/U6 stem II would be expected to stabilize the U4/U6 duplex , which is unwound by the Brr2 helicase during spliceosome catalytic activation . To test relative contributions of the yPrp3 DUF1115 domain and the preceding basic peptide to stabilization of yU4/U6 , we therefore assessed the influence of yPrp3CTF and yPrp3DUF1115 on Brr2-mediated U4/U6 unwinding . Addition of yPrp3CTF reduced the rate of yBrr2-mediated yU4/U6 unwinding about fivefold , while yPrp3DUF1115 showed almost no effect ( ca . 1 . 2-fold reduction; Figure 2A , B ) . 10 . 7554/eLife . 07320 . 006Figure 2 . Effects on Brr2-mediated U4/U6 unwinding and interplay with other U4/U6 proteins . ( A ) Native gels monitoring yU4/U6 di-snRNA unwinding by yBrr2 in the absence of other proteins ( top ) , in the presence of yPrp3CTF ( middle ) or yPrp3DUF1115 ( bottom ) . Asterisks indicate radioactive label on yU4 snRNA . ( B ) Quantification of the data in ( A ) . The data were fit to a single exponential equation: % duplex unwound = A{1 − exp ( −ku t ) }; A—amplitude of the reaction; ku—apparent first-order rate constant; t—time . Amplitudes and rate constants are listed . Errors indicate standard errors of the mean of four independent experiments . ( C ) Binding of yPrp3DUF1115 ( left three lanes of each panel ) or yPrp3CTF ( right three lanes of each panel ) to yU4/U6fused ( lanes 1–6; left ) , yU4/U6fused pre-bound to Snu13 ( lanes 7–12; middle ) or yU4/U6fused pre-bound to Snu13 and Prp311−462 ( lanes 13–18; right ) under otherwise identical conditions . All three panels are from the same gel and were regrouped . Schemes of yU4/U6fused alone or pre-bound by proteins are shown on the top . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 006 Together , these analyses show that Prp3 orthologs contain a conserved , C-terminal region ( Prp3CTF ) that is necessary and sufficient for stable binding to U4/U6 di-snRNAs and that upper parts of stem II and at least eight nts of the U6 3′-overhang are required for stable binding . Prp3CTF comprises the DUF1115 domain and a preceding basic peptide , both of which contribute to RNA binding . Previous assembly and structural studies had shown that the U4/U6-specific proteins Snu13 , Prp31 and Prp3 bind U4/U6 di-snRNAs in a cooperative manner ( Nottrott et al . , 2002; Liu et al . , 2007 ) . To see whether the C-terminal U4/U6-binding region of Prp3 is sufficient to sense pre-bound Snu13 and Prp31 proteins , we compared binding of yPrp3CTF and yPrp3DUF1115 to a fused yU4/U6-like RNA ( yU4/U6fused; Figure 2C ) alone or pre-bound to ySnu13 or ySnu13 and yPrp311−462 under identical conditions . Both yPrp3 variants bound the ySnu13-yPrp311−462-yU4/U6fused complex ( Figure 2C , lanes 13–18 , right ) more efficiently than the ySnu13-yU4/U6fused complex ( lanes 7–12 , middle ) , which in turn was bound better than the naked RNA ( lanes 1–6 , left ) . However , yPrp3CTF interacted more stably than yPrp3DUF1115 with the naked RNA ( lanes 4–6 vs lanes 1–3 ) and with the ySnu13-yU4/U6fused complex ( lanes 10–12 vs lanes 7–9 ) . The sensitivity of the assay was insufficient to resolve possible binding differences to the ySnu13-yPrp311−462-yU4/U6fused complex ( lanes 13–18 ) . These results show that yPrp3CTF binds cooperatively with ySnu13 and yPrp31 to yU4/U6 , in part based on the basic peptide preceding the DUF1115 domain . Sequence analyses of the C-terminal U4/U6-binding portions of Prp3 proteins did not reveal any obvious similarities to known RNA-binding domains . We therefore determined the crystal structure of yPrp3CTF at 2 . 7 Å resolution ( Table 1 ) . The protein crystallized with three monomers per asymmetric unit ( Figure 3—figure supplement 1A ) . The electron density allowed modeling of residues 335–467 , corresponding to the predicted DUF1115 domain . Apart from the C-terminal two residues , the N-terminal 39 residues ( 296–334; comprising the preceding basic peptide ) could not be traced , indicating that they indeed constitute an intrinsically disordered or flexibly attached element , as surmised based on the ITC experiments . 10 . 7554/eLife . 07320 . 007Table 1 . Crystallographic dataDOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 007Data setyPrp3296-469yPrp3325-469yPrp3296-469-yU4/U6stem II-2Data collection Wavelength ( Å ) 0 . 918400 . 918410 . 97968 Space groupC2221P65C2 Unit cell parameters a , b , c ( Å ) 87 . 7 , 161 . 2 , 105 . 256 . 1 , 56 . 1 , 86 . 8144 . 7 , 59 . 6 , 109 . 8 α , β , γ ( ° ) 90 . 0 , 90 . 0 , 90 . 090 . 0 , 90 . 0 , 120 . 090 . 0 , 118 . 5 , 90 . 0 Resolution ( Å ) 50 . 0–2 . 70 ( 2 . 80–2 . 70 ) 50 . 0–2 . 00 ( 2 . 12–2 . 00 ) 30 . 0–3 . 25 ( 3 . 33–3 . 25 ) Reflections Unique20 , 036 ( 1993 ) 10 , 459 ( 1622 ) 24 , 895 ( 1844 ) Completeness ( % ) 97 . 3 ( 99 . 0 ) 99 . 3 ( 96 . 2 ) 97 . 9 ( 98 . 5 ) Redundancy3 . 4 ( 3 . 3 ) 11 . 4 ( 11 . 3 ) 2 . 0 ( 1 . 9 ) Rmeas†0 . 066 ( 0 . 791 ) 0 . 072 ( 0 . 459 ) 0 . 017 ( 0 . 146 ) I/σ ( I ) 13 . 31 ( 1 . 11 ) 24 . 96 ( 5 . 94 ) 5 . 38 ( 0 . 90 ) CC ( 1/2 ) ‡–99 . 9 ( 95 . 5 ) 99 . 4 ( 42 . 0 ) Refinement Resolution ( Å ) 30 . 00–2 . 70 ( 2 . 77–2 . 70 ) 24 . 28-2 . 00 ( 2 . 20–2 . 00 ) 29 . 78-3 . 25 ( 3 . 38–3 . 25 ) Reflections Number18 , 932 ( 1146 ) 10 , 454 ( 2424 ) 24 , 894 ( 2631 ) Completeness ( % ) 95 . 8 ( 80 . 5 ) 99 . 3 ( 97 . 0 ) 98 . 3 ( 98 . 0 ) Test set ( % ) 5 . 25 . 05 . 0 R factors§ Rwork ( % ) 20 . 7 ( 37 . 0 ) 16 . 1 ( 16 . 6 ) 24 . 9 ( 36 . 9 ) Rfree ( % ) 26 . 2 ( 42 . 7 ) 22 . 1 ( 21 . 9 ) 29 . 9 ( 38 . 9 ) RMSD# Bond lengths ( Å ) 0 . 0100 . 0130 . 007 Bond angles ( ° ) 1 . 3161 . 2781 . 006 Ramachandran plot¶ ( % ) Favored96 . 5798 . 5690 . 87 Allowed2 . 641 . 447 . 98 Outlier0 . 790 . 001 . 14Values for the highest resolution shell in parentheses . †Rmeas = Σh[n/ ( n − 1 ) ]1/2Σi\|Ih − Ih , i\|/ΣhΣiIh , I , where Ih is the mean intensity of symmetry-equivalent reflections and n is the redundancy . ‡CC ( 1/2 ) is the percentage of correlation between intensities from random half-data sets . §R = Σhkl\|\|Fobs\| − \|Fcalc\|\|/Σhkl\|Fobs\|; Rwork − hkl ∉ T; Rfree − hkl ∈ T; T—test set . #Root-mean-square deviation from target geometries . ¶Calculated with MolProbity ( http://molprobity . biochem . duke . edu/ ) . The core of the yPrp3 DUF1115 domain resembles the homology model of the corresponding human domain ( Korneta et al . , 2012 ) ( root-mean-square deviation [RMSD] of 2 Å for 87 common Cα atoms ) . The structure exhibits a five-stranded mixed β-sheet with two α-helices ( α1 and α2 ) running parallel to the β-strands on one side of the sheet and one ( α3 ) on the other ( Figure 3A ) . The first 98 residues of the structure ( residues 335–432 ) form a α/β sandwich with a ferredoxin-like topology and an additional , long β-hairpin ( residues 403–421 , comprising strands β3a and β3b ) inserted between helix α2 and strand β4 . The ferredoxin-like fold is further expanded by an extra β-strand ( β5 ) , a helix ( α3 ) and a following loop at the C-terminus . In two of the three independent molecules , the β3a/β3b hairpins adopt very similar conformations with their long axes oriented perpendicular to the direction of the strands in the central β-sheet . Residues D405 and D418 at the bases of the hairpins coordinate an Yt3+ ion . The third molecule lacks electron density for large parts of the hairpin ( residues 407–421 ) , suggesting that crystal packing and a bound metal ion may have stabilized this normally flexible element in the first two molecules . 10 . 7554/eLife . 07320 . 008Figure 3 . Structural overviews . ( A ) Structure of a yPrp3CTF . Secondary structure elements and termini are labeled . Dashed line indicates residues contained in the construct but not visible in the electron density . ( B ) Structure of a yPrp3CTF-yU4/U6stem II+10nt complex with the protein in the same orientation as in Figure 3A . yPrp3CTF—cyan; yU4—gold; yU6—orange . Dashed line indicates residues contained in the construct but not visible in the electron density . Sections 1–3 , between which the yU6 3′-overhang changes direction on the surface of yPrp3CTF , are indicated by black lines . Residues in the yU6 3′-overhang are numbered . ( C ) Electrostatic surface potential of yPrp3CTF in complex with yU4/U6stem II+10nt . Blue—positive; red—negative . Units kT/e with k—Boltzmann'’s constant , T—absolute temperature , E—charge of an electron . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 00810 . 7554/eLife . 07320 . 009Figure 3—figure supplement 1 . Structural comparisons . ( A ) Ribbon plots of all crystallographically independent , superimposed structures of the yPrp3 DUF1115 domain . Isolated yPrp3DUF1115—white; isolated yPrp3CTF—gray colors; RNA-bound yPrp3CTF—cyan and blue . ( B ) Structural superimposition of the two crystallographically independent yPrp3CTF-yU4/U6stemII+10nt complexes based on the protein portions . yPrp3296−469—cyan ( complex 1 ) and blue ( comply 2 ) ; yU4—gold ( complex 1 ) and light gray ( complex 2 ) ; yU6—orange ( complex 1 ) and dark gray ( complex 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 00910 . 7554/eLife . 07320 . 010Figure 3—figure supplement 2 . Phylogenetic analysis . ( A ) Structure-based phylogenetic tree of ferredoxin-like fold proteins . Red and gray boxes indicate structures that recognize and do not recognize nucleic acids , respectively . The position of the Prp3 ssRNA-binding domain is indicated by a cyan box . ( B–H ) Comparison of the yPrp3 DUF1115 domain alone ( B ) and in complex with RNA ( F ) to B . taurus AcyP ( C; PDB ID 2ACY ) ( Thunnissen et al . , 1997 ) , the R . sphaeroides BLUF domain ( D; PDB ID 2BYC ) ( Jung et al . , 2005 ) , a structural model of the hPrp3 DUF1115 domain ( E ) ( Korneta et al . , 2012 ) , the C-terminal RRM of the human spliceosomal U1A protein in complex with stem-loop I of U1 snRNA ( G; PDB ID 1URN ) ( Oubridge et al . , 1994 ) , and Epstein–Barr viral DNA-binding domain in complex with DNA ( H; PDB ID 1B3T ) ( Bochkarev et al . , 1998 ) . All structures are shown with the ferredoxin-like core elements in the same orientation . The cores are shown in color , additional elements in light gray . Nucleic acids—orange . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 010 We also determined the crystal structure of yPrp3DUF1115 , lacking the preceding basic peptide , at 2 . 0 Å resolution ( Table 1 ) . While the ordered parts of yPrp3CTF and yPrp3DUF1115 are very similar ( RMSD of 0 . 48 Å for 131 common Cα atoms; Figure 3—figure supplement 1A ) , the β3a/β3b hairpin in yPrp3DUF1115 is positioned closer to the globular part of the protein , showing that its relative positioning is indeed flexible . To elucidate the molecular mechanism underlying U4/U6 di-snRNA binding by yPrp3CTF , we determined its crystal structure in complex with yU4/U6stem II+10nt at 3 . 25 Å resolution ( Table 1; Figure 3B ) . Residues 335–468 of yPrp3CTF and all nts ( G81-U90 ) of the yU6 3′-overhang were well ordered in the two complexes contained in an asymmetric unit . The electron density for the yU4/U6 stem II regions was less well defined , in particular in the part of stem II distal to the duplex-to-single strand junction in one of the complexes . The electron density was consistent with the stem II regions adopting standard A-form geometry in both complexes but did not allow us to model in detail possible local deviations , for example , around a non-canonical C69U6-C12U4 base pair . Protein–RNA interactions in the well-defined portions of the two crystallographically independent complexes were largely identical ( Figure 3—figure supplement 1B ) . In both complexes , the single-stranded ( ss ) yU6 3′-overhang arches across helix α1 and strand β5 of yPrp3CTF , running below the long β3a/β3b hairpin ( Figure 3B ) . Its binding surface on the protein is carpeted by an electropositive surface potential ( Figure 3C ) . The yU6 3′-overhang can be divided into three sections , between which its backbone changes directions on the surface of yPrp3CTF ( Figure 3B ) . Nts G81-A83 ( section 1 ) stack on the terminal U80U6-A1U4 base pair of stem II and run diagonally from the N-terminus of helix α2 to the N-terminus of helix α1; nts C84-G86 ( section 2 ) are positioned perpendicular across the N-terminal end of helix α1; nts U87-U90 ( section 3 ) run along the exposed edge of strand β5 and the C-terminus of helix α1 . As a consequence , the end of the U6 3′-overhang is directed via two ca . 90 ° kinks back towards stem II . A RNA structural similarity search using the RNA Bricks database ( Chojnowski et al . , 2014 ) failed to uncover a case of a similar RNA redirection on the surface of a single protein domain . Notably , the unusual doubly-kinked RNA conformation is stabilized by protein elements that expand the core ferredoxin fold in yPrp3 , that is , the β3a/β3b hairpin , strand β5 and helix α3 ( details below ) . All nts of the yU6 3′-overhang ( G81-U90 ) contained in the present structure , except the first G81 residue , directly contact yPrp3CTF in one or both complexes , consistent with the important role of this part of U6 for stable Prp3 binding in yeast and human . In one yPrp3CTF-yU4/U6stem II+10nt complex , the guanidinium group of R399 ( helix α2 ) forms ionic interactions to the backbone phosphate of A82 and the side chain amino group of K355 ( helix α1 ) binds to the A83 phosphate ( Figure 4A ) . In the complex lacking these interactions , the corresponding backbone regions of yU6 and the stem II duplex are slightly pulled away from the protein , presumably by crystal packing interactions . The exocyclic N6 group of A83 approaches the side chain carboxyl of E362 ( helix α1 ) and its base stacks on the side chain of F354 ( helix α1; Figure 4A ) . The extracyclic amino group ( N4 ) of C84 is hydrogen bonded to the backbone carbonyl of E407 ( β3a/β3b hairpin ) and the base is thereby held sandwiched between H409 ( β3a/β3b hairpin ) and P350 ( helix α1; Figure 4B ) . The A83 and C84 backbone phosphates approach the side chain of K351 ( helix α1; Figure 4B ) . These interactions splay out A83 and C84 , stabilizing the first kink in the yU6 3′-overhang . The guanidinium group of R353 ( helix α1 ) engages in hydrogen bonds to the Watson-Crick face of C85 ( atoms O2 and N3 ) and to the Hoogsteen face of G86 ( atoms O6 and N7; Figure 4C ) . Furthermore , N1 of G86 is contacted by the side chain carboxyl of D374 ( β1-β2 loop ) , its base engages in cation-π stacking to the guanidium group of R371 ( strand β2 ) and its 2′-hydroxyl group is hydrogen bonded to the backbone carbonyl of M442 ( strand β5; Figure 4C ) . O4 of U87 is hydrogen bonded to the side chain amide of Q457 ( helix α3 ) and its base stacks on F441 ( strand β5; Figure 4D ) . The O2 atom of U88 hydrogen bonds with the backbone NH of W440 ( strand β5 ) , positioning the base laterally on the F441 side chain . The U87-U88 backbone region encircles the side chain of K357 ( helix α1 ) , which contacts the anionic phosphate oxygens of U87 and the O3 atom of U88 ( Figure 4D ) . The interactions involving K357 , W440 , F441 and Q457 stabilize the second kink in the yU6 3′-overhang . The side chain amino group of K361 ( helix α1 ) contacts the U89 anionic phosphate oxygens in one complex ( Figure 4E ) and interacts with O2 , O2′ and O3′ of U88 as well as the phosphate and O5′ of U89 in the other complex ( Figure 4F ) . The last nt , U90 , adopts two different conformations in the two complexes; in one case , its N3 , O2 and O4 atoms are hydrogen bonded to the side chain hydroxyl of S364 ( α3-β2 loop; Figure 4E ) , while in the other case it interacts with the N1 and N6 positions of nt A82 at the beginning of the yU6 3′-overhang ( Figure 4F ) . 10 . 7554/eLife . 07320 . 011Figure 4 . Details of the yPrp3-yU4/U6 di-snRNA interaction . ( A–H ) Close-up views of yPrp3–RNA contacts . The protein and the RNA are shown as semi-transparent cartoons ( yPrp3CTF—cyan; yU6—orange; yU4—gold ) with interacting residues as sticks ( colored by element; carbon or phosphorus—as the respective molecule; nitrogen—blue; oxygen—red , sulfur—yellow ) . Black dashed lines—hydrogen bonds or salt bridges . Panels ( G ) and ( H ) show overlays of unbound ( protein—gray ) and complex ( protein—cyan; yU6 snRNA—orange ) situations . Rotation symbols indicate the views relative to Figure 3B . ( I ) Binding of yPrp3CTF ( 20 µM ) to wt and mutant versions ( indicated above the gel ) of U4/U6stem II+13nt ( scheme on the left; mutated region in green ) . ( J ) Variants of yPrp3CTF ( 20 µM; indicated above the gel ) binding to yU4/U6stem II+10nt ( scheme on the left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 011 The overall structure of the DUF1115 domain is globally unchanged compared to the structure of isolated yPrp3CTF or yPrp3DUF1115 ( RMSD of 0 . 80–0 . 86 Å for 133 common Cα atoms ) but there are local adjustments accompanying RNA binding , which may contribute to the binding specificity . Upon RNA binding , the β3a/β3b hairpin moves closer to the yPrp3CTF core domain to engage in direct interactions with the RNA and the H409 side chain in this element turns on top of the C84 base ( Figure 4G ) . Within the yPrp3CTF core domain , the side chain of F441 is rotated ca . 90°C out of its intra-molecular position in the isolated protein to stack on U87 . Instead , the U88 base moves into the original F441 position , from where it contacts the W440 backbone ( Figure 4H ) . Furthermore , the K357 side chain adopts an alternative conformation upon RNA binding to interact with the anionic phosphate oxygens of U87 and the O3′ atom of U88 ( Figure 4H ) . Although not directly contacted by the protein , the bases of G81 and A82 mediate a continuous π-stack from the terminal base pair of stem II to the yPrp3-bound portion of the yU6 3′-overhang . Consistent with this stacking being important for yPrp3 binding , replacement of these purines in yU4/U6stem II+13nt with a smaller pyrimidine ( C ) led to reduced affinity to yPrp3CTF ( Figure 4I , lanes 2 and 3 ) . Sequence-specific interactions are seen between R353 and the Watson-Crick edge of C85 as well as the Hoogsteen edge of G86 ( Figure 4C ) , suggesting that the protein reads out parts of the RNA sequence . Consistently , mutation of C85 to A or G86 to C in yU4/U6stem II+13nt significantly reduced or essentially abrogated the interaction with yPrp3CTF , respectively ( Figure 4I , lanes 4 and 5 ) . Correspondingly , a yPrp3CTF variant bearing a R353A exchange showed essentially no binding to yU4/U6stem II+10nt anymore ( Figure 4J , lane 2; this and all other yPrp3CTF variants discussed below could be produced and purified like the wild type [wt] protein , suggesting that none of the tested mutations interfered with proper protein folding ) . We also exchanged several additional yPrp3 residues that directly contact yU4/U6stem II+10nt in our structure and tested binding of the corresponding yPrp3CTF variants to this RNA ( Figure 4J ) . R399A ( lane 12 ) and F441H ( lane 14 ) essentially abrogated the interaction with the RNA , while K357A ( lane 6 ) strongly and F354A/H ( lanes 3 and 4 ) , K361A ( lane 9 ) and S364R ( lane 11 ) weakly destabilized the complex ( indicated primarily by larger fractions of unbound yU4/U6stem II+10nt in the corresponding lanes ) . Single residue exchanges K355A ( lane 5 ) , M358A/E ( lanes 7 and 8 ) , E362A ( lane 10 ) and H409E ( lane 13 ) of yPrp3CTF showed no significant or only very mild reduction in RNA binding under the chosen conditions , indicating that individually the interactions involving these residues are not essential for stable RNA complex formation . Nts G81 , A82 , C85 and G86 , which upon mutation led to reduced yPrp3CTF binding , are highly evolutionarily conserved in U6 snRNAs ( Figure 1—figure supplement 2 ) . Likewise , R353 , R399 and F441 , where substitutions strongly affected RNA binding , are conserved in Prp3 orthologs ( Figure 1—figure supplement 1 ) . We conclude that the mode of Prp3-U4/U6 interaction seen in our structure is present in all eukaryotes . The ferredoxin superfamily currently encompasses 59 subfamilies . Previous bioinformatics analyses indicated that within the superfamily the Prp3 DUF1115 domain is most homologous to acylphosphatase ( AcyP ) and BLUF domains ( Korneta et al . , 2012 ) , which do not recognize nucleic acids . However , the sequence comparison did not reveal the evolutionary relationship between the Prp3 ssRNA-binding domain and other members of the ferredoxin superfamily . We therefore calculated a phylogenetic tree based on pairwise structural comparisons with representative ferredoxin fold structures ( Figure 3—figure supplement 2A ) . The tree is in agreement with sequence analyses , showing that the Prp3 ssRNA-binding domain is most similar to AcyP and BLUF domains . For example , residues 335–467 of yPrp3CTF ( Figure 3—figure supplement 2B ) spatially align with cow AcyP ( PDB ID 2ACY; Figure 3—figure supplement 2C ) with a RMSD of 2 . 5 Å over 90 common Cα atoms ( Z-score 10 . 1 ) and show a comparable similarity to the BLUF domain of the BlrB photoreceptor from Rhodobacter sphaeroides ( PDB ID 2BYC; RMSD of 2 . 0 Å for 84 common Cα atoms; Z-score of 10 . 0; Figure 3—figure supplement 2D ) . On the other hand , the Prp3 domain is only distantly related to other ferredoxin fold proteins that bind RNAs , such as RRMs ( Figure 3—figure supplement 2F , G ) or ribosomal proteins S6 , S10 , and L10 . The closest nucleic acid-binding relative of the Prp3 ssRNA-binding domain is the IS608 transposase domain , which recognizes DNA ( Figure 3—figure supplement 2H ) . To test the importance of the observed yPrp3-yU4/U6 interactions for the function of yPrp3 in vivo , we used a haploid yeast strain , in which the chromosomal copy of the PRP3 gene was deleted and the essential protein was produced from a counter-selectable URA3-marked plasmid . We then shuffled plasmids carrying mutant prp3 genes into this strain , selected against the URA3 plasmid and monitored the effects of yPrp3 variants on cell viability , snRNP levels and pre-mRNA splicing . yPrp3 proteins bearing mutations R304A , R322A or both ( in the basic peptide preceding DUF1115 ) as well as variants bearing R399A or F441A exchanges ( yU6 3′-overhang-binding DUF1115 domain ) led to mild temperature sensitive ( ts ) growth ( Figure 5A ) . A R399A/F441A double mutant strongly exacerbated this effect ( Figure 5A ) . These observations are consistent with the mutated residues contributing to functionally important yPrp3-yU4/U6 interactions , as seen in our binding assays and yPrp3CTF -yU4/U6stem II+10nt crystal structure . To trace the origin of the growth defects , we monitored the effects of the mutations on splicing in vivo at 37°C , using endogenous U3 ( pre- ) snoRNA as a reporter . Slight accumulation of unspliced U3 pre-snoRNAs was seen in strains producing the R304A , R399A , R304A/R322A or R399A/F441A variants of yPrp3 compared to the wt ( Figure 5B ) , indicating that the observed growth defects likely root in inefficient splicing . Finally , monitoring snRNP levels by Northern blotting of cellular extracts spread out on a glycerol gradient showed that the levels of U4/U6 di-snRNP and of isolated U5 snRNP were increased at the expense of U4/U6•U5 tri-snRNP in strains producing Prp3 variants R304A/R322A and R399A/F441A ( Figure 5C , D ) . At the same time , Western blotting revealed reduced levels of yPrp3 associated with U4/U6•U5 tri-snRNP in the mutant strains ( Figure 5E ) . These results suggest that reduced tri-snRNP levels are the cause of the reduced splicing activity in the mutant strains and that efficient yPrp3 binding to U4/U6 di-snRNAs via its C-terminal region is important for U4/U6•U5 tri-snRNP stability . 10 . 7554/eLife . 07320 . 012Figure 5 . Consequences of yPrp3 variants in vivo . ( A ) Growth of yeast strains producing the indicated yPrp3 variants at various temperatures . Serial dilutions of liquid cultures were spotted on YPD agar plates and incubated at the indicated temperatures for 1–2 days . ( B ) In vivo splicing assays , monitoring levels of U3 pre-snoRNAs in yeast strains producing wt yPrp3 or the indicated yPrp3 variants . Cells were grown at 37°C for 5 hr before total RNA was extracted and U3 mature or pre-snoRNAs were detected by primer extension using a radiolabeled DNA oligonucleotide complementary to a region in U3 exon 2 . ( C , D ) Northern blotting of cellular extracts of wt ( top ) or the indicated prp3 mutant yeast strains ( middle and bottom ) . Cellular extracts were separated on 10–30 % ( v/v ) glycerol gradients . Even-numbered gradient fractions ( indicated above the blots ) were probed with radiolabeled DNA oligomers complementary to snRNA regions . Positions of various snRNPs on the gradients are indicated below the blots . ( D ) Quantification of the data shown ( C ) . ( E ) Western blots of the same gradients . Odd numbered gradient fractions ( indicated above the blots ) were probed with anti-ySnu114 ( top ) and anti-yPrp3 ( bottom ) antibodies . Positions of various snRNPs on the gradients are indicated below the blots . Dotted boxes—yPrp3 signals in the U4/U6•U5 tri-snRNP fractions . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 012 Here , we showed that Prp3 orthologs contain a C-terminal U4/U6 di-snRNA binding region that encompasses a DUF1115 domain and a preceding peptide rich in basic residues . Our crystal structure and targeted mutational analyses demonstrate that the DUF1115 domain acts as a ssRNA-binding domain that specifically recognizes the first 10 nts of the U6 3′-overhang . While our structural analysis did not reveal how the preceding peptide is involved in U4/U6 binding , several pieces of evidence indicate that it binds along U4/U6 stem II: The peptide is important for full binding of Prp3 C-terminal portions to U4/U6 ( Figure 1C , D ) but it is not a structural element of the DUF1115 domain and thus does not act via stabilizing the RNA-binding fold of DUF1115 . Single Arg-to-Ala exchanges in the peptide lead to reduced RNA binding ( Figure 1C , D ) , indicating that these residues may directly contact the RNA . RNA constructs with shortened stem II or in which a conserved non-canonical base pair was converted to a Watson-Crick pair show reduced binding to Prp3CTF ( Figure 1E , F ) , which cannot be explained by the observed DUF1115 domain contacts ( exclusively to the U6 3′-overhang ) . Finally , Prp3CTF , but not the DUF1115 domain , inhibits Brr2-mediated U4/U6 unwinding ( Figure 2A , B ) , suggesting that the peptide stabilizes contacts between U4 and U6 , presumably by binding the stem II duplex . Residues in both the DUF1115 domain and the preceding basic peptide of Prp3 as well as nts in U4/U6 , which we identified as being important for stable complex formation , are evolutionarily highly conserved , suggesting that the composite ds/ss U4/U6 di-snRNA-binding region of Prp3 is present in all eukaryotes . DUF1115 exhibits a ferredoxin-like core similar to AcyP/BLUF proteins but evolutionarily remote from other nucleic acid-binding domains in the ferredoxin superfamily . As we failed to detect a DUF1115-like domain by sequence comparisons in hundreds of other RNA-binding proteins recently identified in transcriptome-wide screens ( Baltz et al . , 2012; Castello et al . , 2012 ) , DUF1115 most likely represents a spliceosome-specific ssRNA-binding domain . Due to the pronounced reorientation that the U6 3′-overhang experiences on the surface of the Prp3 DUF1115 domain , the domain likely is an important architectural element in the U4/U6 di-snRNP and/or the U4/U6•U5 tri-snRNP . The mode of Prp3 binding to U4/U6 di-snRNAs revealed herein helps to rationalize a large body of mutational analyses on U6 snRNAs in yeast and human . Previously , the functional importance of various U6 snRNA regions was studied by site-directed mutagenesis . hU6 snRNA bearing a C62G mutation , which converts the C-U mismatch in U4/U6 stem II into a G-U wobble pair , supported only low levels of splicing ( Wolff and Bindereif , 1993 ) . Likewise , the corresponding C69G mutation in yU6 snRNA led to reduced splicing activity ( Ryan and Abelson , 2002 ) . We showed that a yU4/U6 stem II construct bearing the C69G mutation exhibits reduced binding to yPrp3 and according to our structural model this region of the U4/U6 di-snRNAs is recognized by the peptide preceding the Prp3 DUF1115 domain . Therefore , a defect in Prp3 binding may underlie the splicing defects of these U6 snRNA mutations in yeast and human . Residues of the highly conserved UGA motif located in the terminus of U6 stem II and beginning of the 3′-overhang were shown to be important for U4/U6•U5 tri-snRNP stability and splicing . Deletion of nts U74-G75-A76 in hU6 snRNA ( corresponding to residues U80-G81-A82 in yU6 ) reduced the levels of U4/U6•U5 tri-snRNP and assembled spliceosomes to less than 10% of wt and abrogated splicing activity ( Wolff and Bindereif , 1992 ) . An A76C variant of hU6 snRNA showed reduced interaction with U4 snRNA , concomitant with reduced spliceosome assembly and low splicing activity ( Wolff and Bindereif , 1995 ) . Furthermore , yU6 snRNA bearing a U80G exchange almost completely abolished formation of U4/U6 di-snRNPs and U4/U6•U5 tri-snRNPs and was defective in splicing ( Fabrizio et al . , 1989; Ryan and Abelson , 2002; Ryan et al . , 2002 ) . The G81C point mutation in yU6 snRNA strongly blocked splicing in vitro ( Madhani et al . , 1990; Ryan and Abelson , 2002 ) and caused a substantial accumulation of free U6 snRNP deficient in U4/U6 di-snRNP assembly ( Ryan and Abelson , 2002 ) . The effects of some of these mutations were suggested to root in a stabilization of intra-molecular base pairing in U6 snRNA . However , our observation that G81C and A82C exchanges in yU6 snRNA led to weak binding of yPrp3 in vitro suggests that aberrant Prp3 binding may also contribute to these phenotypes . The continuous stacking of nts in the transition region between stem II and the U6 3′-overhang seen in our structure may properly orient these elements for stable Prp3 binding . 3′-truncated yU6 snRNAs containing residues 1–94 , 1–91 or 1–88 allowed reconstitution of 35–40 % of wt splicing activity , while only 20% and 4% of the wt splicing activity were regained with yU6 snRNA molecules further shortened to residues 1–86 or 1–85 , respectively , and yU6 snRNAs containing only residues 1–81 or 1–80 fully blocked splicing ( Madhani et al . , 1990; Ryan et al . , 2002 ) . The phenotypes associated with the deletion of distal yU6 3′-overhang residues likely originate from the removal of binding sites for the Prp24 assembly chaperone and the LSm protein complex . However , the exacerbated effects when yU6 snRNA was shortened to 86 nts or less correlate very well with our finding of reduced yPrp3 binding to yU4/U6 constructs bearing sequentially shortened U6 3′-overhangs or the C85A and G86C point mutations . Reduced Prp3 binding may also contribute to the reduced hU4/U6 di-snRNP levels seen previously in the presence of hU6 snRNA bearing a C79U exchange in the 3′-overhang ( Wolff and Bindereif , 1995 ) , as we observed weak yPrp3 interaction with a yU4/U6 bearing a variant residue at the analogous position ( C85A ) . The binding of the U4/U6-specific proteins Snu13 , Prp31 and the Prp3-Prp4 ( -CypH ) complex to U4/U6 di-snRNAs is highly interdependent . Snu13 binds a K-turn motif in the U4 5′SL ( Nottrott et al . , 1999; Vidovic et al . , 2000 ) , serving as an assembly initiating protein for the subsequent incorporation of the other components ( Nottrott et al . , 2002 ) . An explanation for the ordered binding of Snu13 and Prp31 was provided by structural studies that showed how hSnu13 and the hU4 5′SL provide a composite binding platform for the hPrp31 Nop domain ( Liu et al . , 2007 , 2011 ) , while the mechanism by which hSnu13 leads to enhanced interaction of hPrp3 with U4/U6 di-snRNAs ( Nottrott et al . , 2002 ) remained enigmatic . Our results indicate that Prp3 contacts RNA elements along stem II , that is , in the vicinity of the U4/U6 3-way junction . Snu13 binds one branch of the 3-way junction ( the U4 5′SL ) , which could exert a stabilizing effect on the 3-way junction and stem II and lead to improved binding of Prp3 . This effect could be further enhanced by Prp31 , whose U4/U6 contacts extend into the lower part of stem II ( Schultz et al . , 2006 ) . It is also conceivable that the Prp3 stem II-binding peptide directly contacts Snu13 and/or Prp31 on the RNAs and that the latter two proteins even modulate how the peptide interacts with stem II . Our in vivo analyses of the effects of yPrp3 mutations that interfere with stable yU4/U6 binding in vitro , show that the yPrp3-yU4/U6 interactions we observe in our complex structure are important for U4/U6•U5 tri-snRNP stability and , likely as a consequence , for pre-mRNA splicing . The role of Prp3 in mediating tri-snRNP stability can be understood from our present data , showing that the C-terminal portions of Prp3 maintain interactions with the U4/U6 di-snRNAs , and previous yeast 2-hybrid ( Y2H ) analyses , showing that Prp3 contacts the U5 snRNP proteins Prp6 and Snu66 ( Liu et al . , 2006 ) . By combining these functions , Prp3 apparently acts as a bridge , linking the U4/U6 di-snRNP and the U5 snRNP via Prp3-RNA interactions on the U4/U6 side and Prp3-protein interactions on the U5 side ( Figure 6A ) . 10 . 7554/eLife . 07320 . 013Figure 6 . Models for splicing-associated functions of Prp3 . ( A ) Function of Prp3 as a bridge in the U4/U6•U5 tri snRNP . U4/U6 di-snRNP—orange; U5 snRNP—red; snRNAs—black sticks; U4/U6 stem II and U6 3′-overhang—gold and orange sticks . Prp3 uses its C-terminal region to bind U4/U6 stem II and the U6 3′-overhang on the U4/U6 di-snRNP side ( this work ) and interacts with proteins Prp6 or Snu66 on the U5 snRNP side ( Liu et al . , 2006 ) . ( B ) Structure of yPrp24 in complex with yU6 snRNA ( Montemayor et al . , 2014 ) . yPrp24 RRM 1–4 domains—white , light gray , dark gray and black , respectively . Regions of yU6 forming stem I , stem II and the 3′-overhang in yU4/U6 are shown in different blue colors . Cold-sensitive A62G mutation and its paired nucleotide , C85—magenta . ( C ) Model for the function of Prp3 during U4/U6 di-snRNP reassembly after splicing . ( i ) Upon release from the spliceosome , U6 snRNA is bound by Prp24 and LSm proteins . ( ii ) Recruitment of Prp3 and association with its cognate U6 regions , which are partially exposed in the Prp24-U6 complex . Emerging Prp3-U6 interactions initiate detachment of U6 snRNA from Prp24 . ( iii ) Incorporation of pre-assembled U4 snRNP may complete Prp24 displacement and assembly of U4/U6 di-snRNP . U4 incorporation may be aided by Prp3 stabilizing the emerging stem II and by the cooperative binding of Snu13 , Prp31 and Prp3 to U4/U6 di-snRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 07320 . 013 After their release from the spliceosome , U4 and U6 snRNAs have to be re-decorated with proteins and reassembled into a di-snRNP before they can participate in further rounds of splicing . This recycling requires the Prp24 protein in yeast and the related SART3 protein in human . Recently , a crystal structure of near full-length Prp24 in complex with a large portion of U6 snRNA was elucidated ( Montemayor et al . , 2014 ) . While not contained in the Prp24-U6 structure , the distal 3′-end of U6 snRNA is available in that complex for binding the LSm proteins ( Figure 6B ) , consistent with the observation of a stable Prp24-LSm-U6 snRNA complex in yeast ( Vidal et al . , 1999 ) . In contrast , U6 regions that form stem I , stem II and the U6 3′-overhang in the assembled U4/U6 di-snRNP are sequestered in an internal U6 stem loop ( ISL ) and by Prp24-U6 interactions ( Figure 6B ) . As Prp24 does not harbor an NTP-dependent RNA helicase activity , the question arises how the U6 ISL can be unwound and U6 handed over from Prp24 to U4 snRNP . Notably , regions of U6 snRNA that form stem II and the U6 3′-overhang in the U4/U6 di-snRNP ( i . e . , Prp3-binding elements ) are largely exposed on the surface of the Prp24-U6 snRNA complex ( Figure 6B ) . Furthermore , the Prp24-U6 structure represents an artificially stabilized situation , obtained after introducing a cold-sensitive A62G mutation in U6 , which impedes unwinding of the U6 ISL ( Fortner et al . , 1994 ) by sequestering C85 in a non-natural Watson–Crick base pair ( magenta in Figure 6B ) . We have shown here that C85 is a crucial contact residue of Prp3 , which also latches onto the surrounding nts in the context of U4/U6 di-snRNP via its DUF1115 domain . Thus in the wt situation , the U6 ISL on Prp24 can most likely “breathe” in the region surrounding C85 , whereby this region could become available for Prp3 binding . Consistent with this idea , chemical probing studies showed that these Prp3-binding nts are only weakly protected in the Prp24-LSm-U6 snRNA particle ( Karaduman et al . , 2006 ) . We therefore suggest that binding of Prp3 to its cognate U6 3′-overhang region traps and subsequently extends and stabilizes spontaneous local unwinding of the internal stem loop ( ISL ) on the Prp24-LSm-U6 snRNA complex ( Figure 6C ) . We envision that Prp24 displacement is completed by entry of a pre-assembled U4 snRNP ( Figure 6C ) . Prp3 could also support this step by stabilizing the emerging U4/U6 stem II via its stem II-binding peptide . Furthermore , the cooperativity we detect in binding of Snu13 , Prp31 and Prp3 to U4/U6 di-snRNA may help to complete assembly of the U4/U6 di-snRNP . Taken together , our results support an important role of Prp3 in the handover of U6 snRNA from Prp24 to U4 snRNP during U4/U6 di-snRNP reassembly after splicing . DNA encoding full-length hPrp3 or the hPrp3C fragment ( residues 195–683 ) with C-terminal His6-tags were PCR-amplified from pGADT7-hPRP3 ( Liu et al . , 2006 ) and subcloned into vector pMAL-c2t , in which the Factor Xa cleavage site following the N-terminal MBP tag of pMAL-c2x ( New England Biolabs , Ipswich , MA ) was replaced by a TEV cleavage site . The inserts of these and all other plasmids were verified by sequencing . A DNA fragment encoding hPrp3N ( residues 1–442 ) was cloned into pETM-41 ( EMBL , Heidelberg , Germany ) for production of the protein bearing an N-terminal His6-MBP fusion . Plasmids were transformed into Escherichia coli BL21 ( DE3 ) -RIL cells ( Stratagene , La Jolla , CA ) and expressed at 16–18°C using the auto-induction method ( Studier , 2005 ) . Target proteins were double affinity purified using Ni-NTA beads and amylose resin followed by gel filtration using a Superdex 200 26/60 column ( GE Healthcare , Munich , Germany ) in 20 mM Tris–HCl , pH 7 . 5 , 200 mM NaCl , 2 mM DTT . RNase A ( Qiagen , Hilden , Germany ) was included in the initial purification steps for digestion of host RNAs . DNA fragments encoding hPrp3CTF ( residues 484–683 ) or yPrp311−462 were introduced via BamHI and XhoI restriction sites into pGEX-6P ( GE Healthcare ) for production as N-terminal , PreScission-cleavable GST fusion proteins . GST-hPrp3CTF and GST-yPrp311−462 were produced at 18°C in E . coli BL21 ( DE3 ) -RIL cells using the auto-induction method and purified using glutathione-Sepharose beads ( GE Healthcare ) . The fusion proteins were eluted from beads in buffer containing 10 mM reduced L-glutathione and further purified via a Superdex 75 16/60 gel filtration column ( GE Healthcare ) in 20 mM Tris–HCl , pH 7 . 5 , 200 mM NaCl , 2 mM DTT . DNA constructs encoding yPrp3CTF ( residues 296–469 ) , yPrp3DUF1115 ( residues 325–469 ) or ySnu13 were amplified from PRP3 or SNU13 synthetic genes ( GENEART AG , Regensburg , Germany ) , respectively , and subcloned into pETM-11 ( EMBL , Heidelberg ) . Mutant versions of pETM-11-PRP3CTF were generated by site-directed mutagenesis using the QuikChange II kit ( Agilent Technologies , Santa Clara , CA ) . The plasmids were transformed into E . coli BL21 ( DE3 ) -RIL cells and expressed at 18–25°C for two days using the auto-induction method . N-terminally His6-tagged proteins were affinity purified using 1 ml or 5 ml HisTrap FF columns ( GE Healthcare ) . 1 M LiCl buffer was included in the wash step to remove unspecifically bound nucleic acids . His6-tagged proteins used in some EMSA and unwinding assays were directly applied to a Superdex 75 16/60 column in 10 mM Tris–HCl , pH 7 . 0 , 150 mM NaCl , 2 mM DTT . For EMSA titrations , ITC and crystallization , TEV protease was used to remove the N-terminal His6-tag before the size-exclusion chromatography step . Production of selenomethionine ( SeMet ) -substituted His6-yPrp3CTF in BL21 ( DE3 ) -RIL cells was carried out in M9 minimal medium supplemented with trace metals and vitamins ( van den Ent and Lowe , 2000 ) . At an OD595 of 0 . 6 , 50 mg/l of SeMet ( Fisher Scientific , Hampton , NH ) , 100 mg/l of lysine , threonine and phenylalanine and 50 mg/l of leucine , isoleucine and valine were added . After 15 min , the temperature was reduced to 20°C and 0 . 32 mM IPTG was added for induction overnight . SeMet-substituted protein was purified in the same way as the native protein . Full-length yBrr2 bearing an N-terminal His6-tag was produced using a recombinant baculovirus in insect cells as described for hBrr2 ( Santos et al . , 2012 ) . Briefly , a 800 ml infected High FiveTM cell pellet was resuspended in 50 mM HEPES , pH 8 . 0 , 600 mM NaCl , 2 mM β-mercaptoethanol , 0 . 05% NP-40 , 20% glycerol , 10 mM imidazole , supplemented with protease inhibitors ( Roche , Penzberg , Germany ) and lyzed by sonication using a Sonopuls Ultrasonic Homogenizer HD 3100 ( Bandelin ) . His6-yBrr2 was captured from the cleared lysate on a 5 ml HisTrap FF column ( GE Healthcare ) and eluted with a linear gradient from 10 to 250 mM imidazole . The eluted protein was diluted to a final concentration of 80 mM sodium chloride and loaded on a Mono Q 10/100 GL column ( GE Healthcare ) equilibrated with 50 mM Tris–HCl , pH 8 . 0 , 50 mM NaCl , 5% glycerol , 2 mM β-mercaptoethanol . His6-yBrr2 was eluted with a linear 50 to 600 mM sodium chloride gradient and further purified by gel filtration on a Superdex 200 26/60 column ( GE Healthcare ) in 40 mM Tris–HCl , pH 8 . 0 , 200 mM NaCl , 20% glycerol , 2 mM DTT . Full-length yU4 and yU6 snRNAs were synthesized by in vitro transcription using T7 RNA polymerase and PCR-generated templates . The transcripts were purified using the RNeasy Midi Kit ( Qiagen ) . RNA duplexes were prepared by combining 30 nM of [32P]-5′-end labeled U4 snRNA with a fivefold molar excess of unlabeled U6 snRNA in annealing buffer ( 40 mM Tris–HCl , pH 7 . 5 , 100 mM NaCl ) . The solution was heated to 80°C for 2 min and cooled to 25°C over 90 min . 12 . 5 mM MgCl2 were added to the solution at 70°C . The annealed duplex was separated from ss U4 and U6 snRNAs by 6% native PAGE . The duplex was eluted from the gel , phenol-chloroform extracted , ethanol precipitated and resuspended in annealing buffer . yU4/U6fused was prepared by in vitro transcription and purified via a Mono Q column , followed by gel electrophoresis on an 8% denaturing ( 7 M urea ) polyacrylamide gel , eluted and precipitated by isopropanol . All other RNAs were chemically synthesized ( Dharmacon , Lafayette , CO ) . Complementary oligonucleotides were annealed before use by resuspension in H2O , mixing , heating to 95°C for 2 min , slow cooling to room temperature followed by incubation on ice . Typically , [32P]-5′-end labeled RNAs were mixed with recombinant proteins in 20 mM Tris–HCl , pH 7 . 0 , 150 mM NaCl , 2 mM DTT , 0 . 5 µg/µl tRNA . For EMSAs involving hPrp3FL , hPrp3N and hPrp3C , samples were incubated in 10 mM Tris–HCl , pH 7 . 5 , 200 mM NaCl , 2 mM DTT , 0 . 33 µg/µl tRNA and 0 . 2 µg/µl heparin . For hPrp3CTF the same buffer with 67 ng/µl heparin was used . RNP complexes were allowed to form for 30 min at 4°C and then separated on a 5–6 % ( 60:1 ) polyacrylamide gel . For EMSA titrations of yPrp3CTF variants and yPrp3DUF1115 , proteins lacking the N-terminal His6-tag were employed . EMSA experiments were performed as described above ( pre-incubation in 20 mM HEPES , pH 6 . 8 , 150 mM NaCl , 0 . 33 µg/µl tRNA ) with the indicated concentrations of proteins . Radioactive bands were visualized by autoradiography and quantified with the Image Quant 5 . 2 software ( GE Healthcare ) . Apparent Kd values were obtained by fitting the resulting data points to a single exponential Hill equation ( fraction bound = A[protein]n/ ( [protein]n + Kdn ) ; A , fit maximum of RNA bound; n , Hill coefficient ) ( Ryder et al . , 2008 ) using GraphPad Prism ( GraphPad Software , Inc . , La Jolla , CA ) . For monitoring binding cooperativity of yPrp3CTF and yPrp3CTF with Snu13 and Prp311−462 , [32P]-5′-end labeled RNA oligonucleotides were mixed with recombinant proteins in 20 mM HEPES-NaOH , pH 7 . 5 , 200 mM NaCl , 3 mM DTT , 0 . 33 µg/µl tRNA , 67 ng/µl acetylated BSA , 13 ng/µl heparin at 4°C for 30 min and separated on a 4% ( 30:1 ) polyacrylamide gel . Radioactive bands were visualized by autoradiography using a phosphoimager ( Molecular Dynamics , Sunnyvale , CA ) . Proteins without affinity tags were used for ITC experiments . Proteins and RNA were buffer exchanged to 20 mM HEPES , pH 6 . 8 , 150 mM NaCl by dialysis and their concentrations were determined via the absorbances at 280 and 260 nm , respectively . yPrp3CTF variants ( 50 µM ) or yPrp3DUF1115 ( 200 µM ) were used as titrants , yU4/U6stem II+13nt ( fused by a GAGA tetraloop; 300 µl at 5 µM or 25 µM ) as the analyte . Measurements were conducted at 20°C on a MicroCalTM iTC200 ( GE Healthcare ) , with 16 injections of 2 . 5 µl each with 180 s intervals between injections . Titrant heats of dilution were subtracted and data were fit using MicroCal Origin 7 ( GE Healthcare ) . 1 . 75 nM yU4/U6 di-snRNAs without or with 20 µM of yPrp3 variants were pre-incubated with RNA at 30°C for 3 min in 40 mM Tris–HCl , pH 7 . 5 , 50 mM NaCl , 8% glycerol , 0 . 5 mM MgCl2 , 100 ng/µl acetylated BSA , 1 U/µl RNasin , 1 . 5 mM DTT before the addition of 50 nM yBrr2 . The reactions were initiated by adding 1 mM ATP/MgCl2 and further incubated at 30°C for 0–90 min . Aliquots were taken at the indicated time points and reactions were stopped with one volume of 40 mM Tris–HCl , pH 7 . 4 , 50 mM NaCl , 25 mM EDTA , 1% SDS , 10% glycerol , 0 . 05% xylene cyanol , 0 . 05% bromophenol blue . Samples were loaded on a 6% native PAGE and run at 10 W for 2 hr . RNA bands were visualized by autoradiography and quantified with the Image Quant 5 . 2 software . Data were fit to a single exponential equation ( fraction unwound = A{1-exp ( -ku t ) } ) ; A , amplitude of the reaction; ku , apparent first-order rate constant for unwinding; t , time using GraphPad Prism ( GraphPad Software , Inc . ) . The wt yPRP3 gene was PCR-amplified from plasmid EWB2235 , cloned into vector pRS314 and used to introduce the desired mutations by the QuikChange site-directed mutagenesis strategy ( Stratagene ) . Wt and mutant plasmids were transformed into yeast strain EWY2845 ( prp3::LEU2; his3Δ200; leu2-3112; lys2-810; trp1-1; ura3-52 [PRP3/YCp50]; kindly provided by John L . Woolford , Jr . , Carnegie Mellon University , Pittsburgh , USA ) that harbors the PRP3 gene on a counter-selectable URA3 plasmid . Transformants were selected in a medium lacking tryptophan , followed by three times streaking on 5-FOA plates at 25°C to counter-select the URA3 plasmid . Growth phenotypes of the yeast cells were assessed by spotting about 5 × 105 cells and tenfold serial dilutions on YPD agar plates and incubating at 16°C , 30°C or 37°C for 1–2 days . To analyze the effect of prp3 mutations on splicing in vivo , the yeast cells producing wt or variant forms of yPrp3 were used to inoculate YPD medium at an OD600 of 0 . 05 and the cultures were further incubated at 37°C for 5 hr . For monitoring the levels of unspliced and spliced U3 ( pre- ) snoRNAs , 8 µg of total RNA from each sample were used for primer extension as described ( Mozaffari-Jovin et al . , 2013 ) . Yeast cells expressing wt or ts variants of yPrp3 were grown in YPD medium at 30°C . Whole cell extract prepared from each strain was incubated at 37°C ( the non-permissive temperature for the Prp3 ts variants ) for 30 min , diluted with equal volume of G100 buffer ( 20 mM HEPES-KOH , pH 7 . 0 , 100 mM KCl , 0 . 2 mM EDTA ) and fractionated by ultracentrifugation on a 12 ml 10–30 % ( v/v ) glycerol gradient in G100 buffer . Subsequent to ultracentrifugation at 37 , 000 rpm for 15 hr in a Sorvall TST41 . 14 rotor , the distribution of spliceosomal U4 , U5 and U6 snRNPs across the gradient fractions was monitored by Northern blotting and quantified as described previously ( Mozaffari-Jovin et al . , 2013 ) . The association of yPrp3 with tri-snRNP was analyzed by Western blotting of gradient fractions and immunostaining using antibodies against yPrp3 and ySnu114 and the Amersham ECL detection kit ( GE Healthcare ) . Crystallization was performed using the sitting-drop vapor-diffusion method at 18°C . Protein yPrp3CTF was concentrated to 15 mg/ml in 10 mM Tris–HCl , pH 7 . 5 , 200 mM NaCl , 1 mM DTT . The best crystals were obtained by mixing 1 μl of protein solution with 0 . 2 µl of 0 . 1 M yittrium ( III ) chloride hexahydrate or praseodymium ( III ) acetate hydrate and 0 . 8 μl of reservoir solution ( 0 . 1 M succinic acid , pH 7 . 0 , 16 . 4% PEG 3350 ) . Protein yPrp3325−469 was concentrated to 15 mg/ml in 10 mM HEPES , pH 6 . 8 , 100 mM NaCl , 2 mM DTT . The best crystals were obtained using a 1:1 mixture of protein solution and reservoir buffer consisting of 0 . 1 M HEPES , pH 7 . 5 , 10% PEG 8000 . Prior to data collection , both types of crystals were cryoprotected in reservoir solution supplemented with 27% ( v/v ) ethylene glycol and flash-cooled in liquid nitrogen . yU4/U6stem II+10nt was slowly added to the SeMet-substituted yPrp3CTF in a 1:1 molar ratio and incubated at 4°C for 15 min . The mixture was then chromatographed on a Superdex 75 16/60 column in 10 mM Na HEPES , pH 6 . 8 , 1 mM DTT . Peak fractions containing the yPrp3CTF-yU4/U6stem II+10nt complex were pooled and concentrated to 7 . 5 mg/ml . Crystals were grown by mixing 1 μl of complex solution with 0 . 2 µl of 0 . 5 M sodium fluoride and 0 . 8 μl of reservoir solution ( 0 . 22 M DL-malic acid , pH 6 . 8 , 16 . 5% PEG 3350 ) . For diffraction data collection , crystals were transferred to a 1:1 mixture of Paratone-N and paraffin oil as a cryoprotectant and flash-cooled in liquid nitrogen . All X-ray diffraction data were collected at 100 K on beamline BL14 . 2 at the BESSY II storage ring ( Berlin , Germany ) . Data were processed with the HKL package ( Minor et al . , 2006 ) or XDS ( Kabsch , 2010 ) . The structure of yPrp3CTF was solved by a praseodymium ( III ) multiple anomalous dispersion experiment as described ( Puehringer et al . , 2012 ) and refined against the higher resolution data from an yttrium ( III ) -derivatized crystal . The structures of yPrp3DUF1115 and of a yPrp3CTF-yU4/U6stem II+10nt complex were solved by molecular replacement using structure coordinates of yPrp3CTF as search models with the programs MOLREP ( Vagin and Teplyakov , 2010 ) and PHASER ( McCoy , 2007 ) , respectively . Models were manually completed and adjusted with COOT ( Emsley et al . , 2010 ) and the structures were refined using REFMAC5 ( Murshudov et al . , 2011 ) and phenix . refine ( Afonine et al . , 2012 ) . To preserve A-form geometry in the stem II portion of the yPrp3CTF-yU4/U6stem II+10nt structure , an A-form duplex was used as a reference model and the A-form geometry was restrained during the refinement . Electrostatic surface potentials were calculated with APBS ( Unni et al . , 2011 ) . All structure figures were created using PyMOL ( http://www . pymol . org/ ) . For phylogenetic analyses , a representative set of protein structures containing ferredoxin-like folds was assembled using the SCOP database ( Murzin et al . , 1995 ) . Pairwise structural similarity Z-scores were calculated for all structures and the yPrp3 ssRNA-binding domain using DaliLite ( Holm and Park , 2000 ) . The Z-score reciprocals were used to build a UPGMA tree with the neighbor tool of the PHYLIP package ( Felsenstein , 1989 ) . The tree was visualized with Evolview ( Zhang et al . , 2012 ) and annotated manually . Coordinates and diffraction data have been deposited with the Protein Data Bank ( www . pdb . org ) under accession codes 4YHU ( yPrp3CTF ) , 4YHV ( yPrp3DUF1115 ) and 4YHW ( yPrp3CTF-yU4/U6stem II+10nt ) and will be released upon publication .	Proteins are built following instructions contained within the DNA of gene sequences . This genetic information is copied into short-lived molecules , called messenger RNAs ( or mRNAs ) , which move away from the DNA and are then decoded by the molecular machines that build proteins . However , mRNA sequences often have to be edited before they are used . Another molecular machine , called a spliceosome , carries out some of this editing . A spliceosome is formed from a number of smaller subunits , including three RNA-protein particles that each contain one RNA molecule ( called U1 , U2 and U5 ) , and one particle that contains two RNA molecules ( called U4 and U6 ) . These subunits must assemble around an unedited mRNA in a particular order so that the spliceosome can work correctly . Once the mRNA has been edited , and the spliceosome has performed its job , these complexes need to disassemble so that they are ready to be reassembled around a new mRNA molecule . A protein called Prp3 is known to be involved in these assembly , disassembly and reassembly steps . However , it is unclear how this protein performs these activities . Liu et al . have now used structural biology and biochemical techniques to determine the three-dimensional structure of Prp3 , and have shown that this protein has a “two-part” binding site that binds to the RNA molecules in the U4/U6 subunit of the spliceosome . Further analyses revealed that one of these features is only found in Prp3 and not in other types of RNA-binding proteins . Together with previous work , Liu et al . also reveal that Prp3 can serve as a ‘bridge’ between the U4/U6 and U5 subunits of the spliceosome , and suggest how these features allow the two subunits to group together before they are incorporated into a spliceosome . Notably , certain mutations in the gene for the Prp3 protein lead to a human eye disease called retinitis pigmentosa . In the future it will be important to investigate if the above activities are affected in the mutant variants of the Prp3 protein .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	A composite double-/single-stranded RNA-binding region in protein Prp3 supports tri-snRNP stability and splicing
Mutations , deletions , and changes in copy number of mitochondrial DNA ( mtDNA ) , are observed throughout cancers . Here , we survey mtDNA copy number variation across 22 tumor types profiled by The Cancer Genome Atlas project . We observe a tendency for some cancers , especially of the bladder , breast , and kidney , to be depleted of mtDNA , relative to matched normal tissue . Analysis of genetic context reveals an association between incidence of several somatic alterations , including IDH1 mutations in gliomas , and mtDNA content . In some but not all cancer types , mtDNA content is correlated with the expression of respiratory genes , and anti-correlated to the expression of immune response and cell-cycle genes . In tandem with immunohistochemical evidence , we find that some tumors may compensate for mtDNA depletion to sustain levels of respiratory proteins . Our results highlight the extent of mtDNA copy number variation in tumors and point to related therapeutic opportunities . Human cells contain many copies of the 16-kilobase mitochondrial genome , which encodes 13 essential components of the mitochondrial electron transport chain and ATP synthase . Alterations of mitochondrial DNA ( mtDNA ) , via inactivating genetic mutations or depletion of the number of copies of mtDNA in a cell , can impair mitochondrial respiration and contribute to pathologies as diverse as encephelopathies and neuropathies ( El-Hattab and Scaglia , 2013 ) , and the process of aging ( Balaban et al . , 2005; Finkel and Holbrook , 2000 ) . In cancer , a number of studies have examined the role of mtDNA mutations in carcinogenesis ( Wallace , 2012; Ju et al . , 2014; Larman et al . , 2012; He et al . , 2010 ) . However , the contribution of changes in the gross number of mtDNA genomes in a tumor ( i . e . the ‘mtDNA copy number’ ) to tumor development and progression has not been adequately investigated . In contrast to the fixed ( diploid ) copy number of the nuclear genome , many copies of mtDNA exist within each cell , and these levels can fluctuate . Because mitochondria undergo a constant process of fusion and fission , it is difficult to meaningfully determine the number of mtDNA molecules per mitochondrion . Instead , studies have focused on measuring mtDNA copy number per cell , with estimates for humans that vary between a few hundred and over one hundred thousand copies , depending on the tissue under examination ( Wai et al . , 2010 ) . Furthermore , because mtDNA serves as a template for the transcription of essential electron transport chain complexes , the quantity of mtDNA in a cell may serve a surrogate marker for the cell’s capacity to conduct oxidative phosphorylation if the copy number of mtDNA is rate-limiting . For instance , a recent study estimated that energy-intensive tissues such as cardiac and skeletal muscle contained between 4000 and 6000 copies of mtDNA per cell , while liver , kidney , and lung tissues averaged between 500 and 2000 copies ( D'Erchia et al . , 2015 ) . Mitochondrial dysfunction plays several distinct roles in cancer ( Schon et al . , 2012; Wallace , 2012; Larman et al . , 2012 ) . First , the normal functions of mitochondria ( e . g . respiration ) may be subverted to support the growth of the tumor . A canonical example of this is the observation that many tumors suppress mitochondrial respiration in favor of increased uptake of glucose and secretion of lactate ( ‘the Warburg effect’ ) , a phenomenon which has found clinical utility for imaging of tumors using FDG-PET ( Vander Heiden et al . , 2009 ) . Second , mitochondria are susceptible to mutations in nuclear- and mitochondrially-encoded genes , and a subset of tumors are known to be caused by mutations of the mitochondrial enzymes FH , SDH , and IDH ( King et al . , 2006 ) . Furthermore , mtDNA dysfunction affecting the electron transport chain can lead to generation of excess reactive oxygen species ( ROS ) , contributing to tumor cell metastasis ( Ishikawa et al . , 2008 ) . To date , no comprehensive analysis of mtDNA copy number changes in tumors has been completed , despite a large literature of isolated reports ( Yu , 2011 ) . Large-scale studies of mtDNA in cancer have instead focused on the analysis of mutations and heteroplasmy , largely ignoring the contribution of mtDNA copy number variation to the development and progression of tumors . Here , we use whole-genome and whole-exome sequencing data to examine changes in mtDNA copy number across a panel of cancer types profiled by The Cancer Genome Atlas ( TCGA ) consortium . Using the resulting mtDNA copy number estimates , we ask fundamental questions about mtDNA and cancer . We investigate whether evidence of the Warburg effect can be found in patterns of mtDNA accumulation or depletion . We further examine the connection between gene expression levels and mtDNA copy number , and identify a subset of mitochondrially-localized metabolic pathways exhibiting a high degree of co-expression with mtDNA levels . Finally , we ask whether gross variations of mtDNA copy number are linked to the incidence of somatic alterations ( including mutations and copy number alterations ) across cancer types . Altogether , our results shed light on the contribution of aberrant mitochondrial function , through changes in mtDNA content , to cancer . 10 . 7554/eLife . 10769 . 003Figure 1 . Summary of methods . ( A ) Reads were analyzed to determine the number aligning to each chromosome . Relative abundance of mitochondrial DNA was calculated as the ratio of mtDNA reads to nuclear DNA reads , and corrected for tumor purity and ploidy . The results of these calculations were employed in three different types of analysis . ( B ) Comparisons across samples profiled by both whole exome and whole genome sequencing provided validation of mtDNA copy number estimates . ( C ) Pairs of matched tumor/adjacent-normal samples were compared to uncover patterns of mtDNA accumulation and depletion . ( D ) Using all data available ( including tumor samples lacking matched normal samples ) , statistical associations between mtDNA copy number and ( 1 ) mutation/copy number alterations , and ( 2 ) gene expression , were calculated . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 003 To estimate the copy number of mtDNA in a tumor sample , we implemented a computationally efficient and fast approach based on comparing the number of sequencing reads aligning to ( 1 ) the mitochondrial ( MT ) genome and ( 2 ) the nuclear genome . Comparable approaches have been used to estimate somatic copy number alterations within the nuclear genome in cancer [for a review , see Zhao et al . ( 2013 ) ] . The approach assumes that regions of the genome of equal ploidy should be sequenced to comparable depth . In a normal human cell , the autosomal nuclear genome is at a fixed ( diploid ) copy number . Thus , by calculating the ratio of reads aligning to the mitochondrial and nuclear genomes , respectively , it is possible to estimate mtDNA ploidy relative to a diploid standard . This approach to assaying mtDNA copy number has been proposed and implemented by others in prior work ( Guo et al . , 2013; D'Erchia et al . , 2015; Samuels et al . , 2013 ) . To estimate mtDNA copy number , we calculated the ratio of ( 1 ) the number of sequencing reads mapping to the MT genome ( rm ) to ( 2 ) the number of reads mapping to the nuclear genome ( rn ) ( Equation 1 ) . Because tumor cells can exhibit large-scale genomic amplifications and deletions , and may be infiltrated by stromal and immune cells , we applied a ploidy/purity correction ( ‘R’ ) , described in detail in the Materials and methods . This calculation yields the relative mtDNA copy number m . Assuming two samples have been processed in identical manners , the sample with a higher value of m contains more copies of mtDNA ( Guo et al . , 2013; D'Erchia et al . , 2015 ) . In line with previous studies ( e . g . [Ju et al . , 2014] ) , we observed significant variation in mean mtDNA copy number between sequencing centers , as well as between each batch ( i . e . , each TCGA plate ID ) within a single sequencing center . We applied a batch correction to control for this effect ( see Materials and methods ) . ( 1 ) m=rmrn×R We applied this method to whole exome sequencing ( WXS ) and whole genome sequencing ( WGS ) data from 22 distinct TCGA studies ( Figure 2 , see Materials and methods for further details on data collection ) . To validate estimates of mtDNA copy number , we compared estimates from samples submitted to both WXS and WGS . Although mitochondrial reads are abundant in both WGS and WXS , the two sequencing methods capture mtDNA at different efficiencies: exome sequencing involves the targeted enrichment of exonic regions prior to sequencing and does not target mtDNA ( Samuels et al . , 2013 ) , while WGS sequences cellular DNA in an unbiased manner . If our approach to estimating mtDNA copy number is accurate , then we expect that the two sequencing platforms should offer comparable estimates of mtDNA copy number across a panel of samples , i . e . , samples with high mtDNA copy number in WGS should have similarly high mtDNA copy number in WXS . We compared mtDNA copy number estimates in 1110 samples across 8 tumor types profiled by both WXS and WGS , controlling for sequencing center and TCGA plate ID . We confirmed that across all combinations of cancer types and sequencing centers , WXS and WGS offer significantly correlated estimates of mtDNA copy number ( Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 10769 . 004Figure 2 . Summary of data . Whole-exome and whole-genome sequencing data were obtained from 22 TCGA studies . Abbreviations for each cancer type follow the standard TCGA nomenclature . The data were processed at four different sequencing centers , each of which was analyzed separately . Over 1000 samples were paired instances of tumor/adjacent-normal tissue from the same patient , which were used to quantify changes in mtDNA content across tumors . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 00410 . 7554/eLife . 10769 . 005Figure 2—figure supplement 1 . Comparison of mtDNA copy number estimates of samples profiled by both whole genome ( WGS ) and whole exome ( WXS ) sequencing . Each point is one sample , X-axis indicates copy number from whole exome sequencing , Y-axis indicates copy number from whole genome sequencing . All correlations are statistically significant ( Spearman ρ p-value , BLCA: 2×10-14; BRCA: 6×10-18; GBM: 2×10-13; HNSC: 1×10-18; KIRC: 3×10-12; LUAD: 4×10-10; THCA: 4×10-23; UCEC: 2×10-31 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 005 Do tumors have different numbers of copies of mtDNA compared to normal tissue ? We investigated whether tumor samples showed a significant change in mtDNA content , relative to matched normal tissues . To do so , for each pair of tumor/adjacent-normal samples collected from the same patient , sequenced at a single sequencing center and within the same batch ( 1090 pairs in total ) , we calculated the ratio ( 2 ) r=log2⁢ ( mTmN ) where mT and mN are the mtDNA copy number estimates in tumor and normal tissues , respectively . We then used non-parametric Wilcoxon signed rank tests to assess whether each cancer type was signficantly enriched for tumor samples with higher or lower mtDNA content than matched normal tissue . The analysis was restricted to 15 cancer types for which we had at least 10 matched tumor/normal pairs . To ensure a meaningful comparison , we only used adjacent-normal tissue ( and not blood ) for the analysis . We elected to focus on analyzing whole-exome sequencing data , for which we had the largest number of samples . A complete list of all calculations is available in Supplementary file 1 . Strikingly , seven of the fifteen tumor types analyzed showed a statistically significant ( BH-corrected Mann-Whitney p-value <0 . 05 ) decrease in mtDNA abundance in tumor samples ( Figure 3A ) . These ‘mtDNA-depleted’ tumor types included bladder ( BLCA ) , breast ( BRCA ) , esophogeal ( ESCA ) , head and neck squamous cell ( HNSC ) , kidney ( both clear-cell , KIRC , and papillary , KIRP , but not chromophobe , KICH , subtypes ) , and liver ( LIHC ) cancers . Despite a tendency towards mtDNA depletion , all tumor types contained at least one sample with higher mtDNA content than adjacent normal tissue . Nevertheless , the depletion effect was exceptionally strong in several tumor types: except for a handful of WGS samples , nearly all bladder tumors were depleted of mtDNA . Similarly , 87% of clear-cell kidney tumor samples contained less mtDNA than their normal tissue counterparts . In contrast , a single tumor type , lung adenocarcinoma ( LUAD ) , showed statistically significant mtDNA accumulation . In cases where sufficient numbers of both WGS and WXS data were available ( bladder , breast , head and neck , clear-cell kidney , thyroid , endometrial , and lung adenocarcinomas ) , we observed a consistent effect across samples processed by both platforms ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 10769 . 006Figure 3 . Many tumor types show depletion of mtDNA in tumor samples , relative to adjacent normal tissue . Normalized histograms and density plots illustrate log2 ratio of mtDNA content in tumor tissue , to mtDNA content in normal tissue . Each row is a different tumor type . Statistical significance of trends is assessed using a Wilcoxon sign rank test , and p-values are corrected using the Benjamini-Hochberg procedure . Cancer types displaying significant depletion/accumulation of mtDNA are colored in blue/red . Seven of fiteen tumor types show a significant depletion of mtDNA content ( a shift of the distribution to the left of the dashed line ) , relative to normal tissue . One tumor type , lung adenocarcinomas , shows an increase in mtDNA content , relative to normal tissue . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 00610 . 7554/eLife . 10769 . 007Figure 3—source data 1 . Correlation of mtDNA copy number with ESTIMATE ( Yoshihara et al . , 2013 ) stromal and immune scores . Third column indicates whether tumor mtDNA content or log ratio of tumor to normal mtDNA content was used . p-values are uncorrected for multiple hypothesis testing . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 00710 . 7554/eLife . 10769 . 008Figure 3—figure supplement 1 . mtDNA tumor:normal copy number ratio using whole-genome sequencing ( WGS ) data . Trends in WGS are in agreement with those in Figure 2 in whole exome sequencing ( WXS ) . All data for this figure can be found in Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 00810 . 7554/eLife . 10769 . 009Figure 3—figure supplement 2 . Correlation between tumor mtDNA copy number and ESTIMATE immune scores . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 00910 . 7554/eLife . 10769 . 010Figure 3—figure supplement 3 . Correlation between tumor mtDNA copy number and ESTIMATE stromal scores . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 01010 . 7554/eLife . 10769 . 011Figure 3—figure supplement 4 . Correlation between tumor/normal mtDNA copy number ratio and ESTIMATE immune scores . Horizontal line indicates a log ratio of zero ( i . e . , tumor and normal tissue have identical copy number ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 01110 . 7554/eLife . 10769 . 012Figure 3—figure supplement 5 . Correlation between tumor/normal mtDNA copy number ratio and ESTIMATE stromal scores . Horizontal line indicates a log ratio of zero ( i . e . , tumor and normal tissue have identical copy number ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 012 Our estimates of tumor mtDNA content are based on sequencing DNA of bulk tumor tissue , which includes stromal and immune cell infiltration . While the effect of purity on mtDNA copy number is partially accounted for by the correction factor R , it is still possible that tumor impurity may bias the calculation of copy number ( for example , if immune cells have lower mtDNA copy number than adjacent-normal tissue ) . Although our computational method is unable to deconvolve sequencing traces arising from tumor cells vs . infiltrate , we nevertheless investigated the statistical association between tumor mtDNA content and stromal/immune infiltration . We obtained estimates of stromal and immune cell infiltration based on gene expression data for eight tumor types calculated using the ESTIMATE algorithm ( Yoshihara et al . , 2013 ) . We correlated these values with estimates of mtDNA copy number , with full results reported in Figure 3—figure supplements 2–5 and Figure 3—source data 1 , with p-values reported as uncorrected for multiple hypothesis testing . We observed a recurrent but weak negative correlation between tumor mtDNA content and immune infilitration score . Further detailed study is required to trace the contribution of infiltrating cells to mtDNA content . To further relate changes in mtDNA content to clinical progression of disease , we used Cox regression to determine if tumor mtDNA copy number was a significant predictor of patient survival ( Figure 4 ) . In total , five cancer types showed statistically significant association between patient survival and mtDNA content . In three cancer types ( adrenocortical carcinoma , p-value 0 . 026; chromophobe renal cell carcinoma , p-value 0 . 053; and low-grade glioma , p-value 0 . 009 ) , high tumor mtDNA content was associated with better survival . The opposite trend , of poor survival in patients with high tumor mtDNA , was observed in clear-cell renal cell carcinoma ( p-value 0 . 023 ) and melanoma ( p-value 0 . 043 ) . The finding regarding KICH is particularly intriguing given the central role mitochondrial dysfunction has been proposed to play in the disease ( Davis et al . , 2014 ) . That mtDNA copy number correlates with better or worse survival , depending on cancer type , suggests that other confounding factors strongly tied to survival , such as the presence of somatic mutations , may influence mtDNA levels . In a later section , we will investigate this hypothesis . 10 . 7554/eLife . 10769 . 013Figure 4 . mtDNA content is significantly associated with patient survival in ( A ) adrenocortical ( ACC ) and ( B ) kidney chromophobe carcinoma ( KICH ) . For visualization purposes , patients are partitioned into two groups , based on tumor mtDNA copy number relative to the median mtDNA copy number across all tumor samples in the cancer type . Cox regression identified a significant association between high tumor mtDNA and better survival in these two tumor types ( ACC , p-value 0 . 026; KICH , p-value 0 . 053 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 013 Proteins encoded in mtDNA localize exclusively to the mitochondrial electron transport chain and ATP synthase , and fluctuations in mtDNA copy number are well-known to influence the level of transcription of these genes . It has also been observed that complete depletion of mtDNA in cell lines by exposure to ethidium bromide affects a number of additional signaling pathways ( Chandel and Schumacker , 1999 ) . Thus , we were compelled to ask if changes in mtDNA content narrowly influenced changes in the expression of oxidative phosphorylation genes , or if they were more broadly connected to the other functions of mitochondria . Our approach to this question was to search for gene sets whose transcriptional signatures were highly correlated to mtDNA copy number . To do so , we calculated the non-parametric Spearman correlation between the expression of each gene and mtDNA copy number , and then used the mean-rank gene set test implemented in limma ( Law et al . , 2014 ) to identify gene sets which were significantly enriched for highly correlated genes . The approach was applied in an unbiased manner to all Reactome gene sets in the Canonical Pathways group from the MSigDB database ( Liberzon et al . , 2011 ) . In general , each tissue exhibited specific gene sets which were strongly correlated to mtDNA copy number levels . However , when aggregating across all cancer types , mitochondrially-localized metabolic pathways showed the most frequent significant correlation with mtDNA abundance ( Figure 5 and Supplementary file 2 , Worksheet Fig5Data ) . This recurrent positive correlation between expression of mitochondrial genes and mtDNA copy number across many tumor types served as a second , independent validation that estimates of mtDNA copy number reflected in vivo mtDNA ploidy . We also calculated the correlation between mtDNA copy number and the expression of TFAM , a critical transcription and replication factor which binds to mtDNA in nucleoids , and found a significant positive correlation ( Spearman p-value <0 . 05 ) in 34% of studies ( Figure 5—source data 1 ) . 10 . 7554/eLife . 10769 . 014Figure 5 . Gene set analysis identifies pathways correlated to mtDNA content . ( A ) Correlations between all genes and mtDNA content are calculated . Then , gene sets enriched for high/low correlation coefficients are identified . ( B ) mtDNA copy number is most strongly correlated to metabolic pathways including respiratory electron transport and the TCA cycle , which are localized to the mitochondria . Enrichment score corresponds to the -log10 p-value of the statistical enrichment test , accounting for the sign of the correlation ( i . e . positive or negative correlation ) . Red blocks indicate an enrichment for positive correlation , blue blocks indicate an enrichment for negative correlation . The top ten most frequently positively correlated gene sets across all studies are depicted . Full results are available in Supplementary file 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 01410 . 7554/eLife . 10769 . 015Figure 5—source data 1 . Correlation of mtDNA copy number estimates from WXS and expression of TFAM . Only cases with at least 20 samples were evaluated . Approximately 35% of studies show statistically significant positive correlation . Log10 range indicates the number of orders of magnitude separating the maximal and minimal value of mtDNA copy number in that cohort . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 01510 . 7554/eLife . 10769 . 016Figure 5—figure supplement 1 . The top ten gene sets most frequently negatively correlated with mtDNA copy number across all studies are depicted . Enrichment score corresponds to the -log10 p-value of the statistical enrichment test , accounting for the sign of the correlation ( i . e . , positive or negative ) . Red blocks indicate an enrichment for po sitive correlation , blue blocks indicate an enrichment for negative correlation . Full results are available in Figure 5—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 016 In line with expectation , we found that the ‘TCA Cycle and Respiratory Electron Transport’ gene set was the most frequently correlated to mtDNA copy number ( 1st out of 674 gene sets ) . Among the remaining top positively correlated gene sets , many were metabolism-related , and included mitochondrial beta oxidation of fatty acids , and branched chain amino acid ( BCAA ) catabolism . BCAAs ( valine , leucine , isoleucine ) are essential amino acids whose catabolism depends on the activity of an enzyme complex , branched chain α-keto acid dehydrogenase , located in the inner mitochondrial membrane ( Hutson et al . , 1988 ) , and prior work has demonstrated that dietary supplementation of BCAA’s promotes mitochondrial biogenesis ( Valerio et al . , 2011; D'Antona et al . , 2010 ) . Furthermore , a recent study has shown that elevated plasma levels of BCAAs are found 2 to 5 years before a cohort of patients developed pancreatic ductal adenocarcinoma ( Mayers et al . , 2014 ) . A number of gene sets showed recurrent negative correlation to mtDNA copy number ( Figure 5—figure supplement 1 and Supplementary file 2 , Worksheet Fig5Data ) . Several of these gene sets , including those related to mRNA processing and the cell cycle , are associated with known non-metabolic functions of mitochondria in the cell . In particular , the replication of mitochondria and mtDNA is intimately linked to the cell cycle ( Chatre and Ricchetti , 2013 ) , and the nucleotide precursors to mtDNA are in part produced de novo , via a pathway that is only active during the S phase of the cell cycle ( Sigoillot et al . , 2003 ) . Several immune pathways , including those related to interferon signaling , are also frequently negatively correlated with mtDNA content . This is interesting in light of the role that mitochondria play in innate immunity ( West et al . , 2011; Weinberg et al . , 2015 ) . Of particular interest is a recent report by West and colleagues ( West et al . , 2015 ) , demonstrating that mtDNA stress induced by depletion of TFAM triggered the innate immune response via interferon-stimulated genes and anti-viral signaling . Of the seven tumor types shown to be depleted of mtDNA in Figure 3 , five ( BLCA , BRCA , ESCA , HNSC , KIRC ) exhibit a negative correlation between expression of immune system genes and tumor tissue ( but not necessarily normal tissue ) mtDNA content . A subset of tumor types did not show strong positive correlation between mtDNA copy number and expression of mitochondrial metabolic genes . In some cases , this was the result of an apparently dominant correlation with another pathway . Interestingly , in prostate adjacent normal tissue , the expression of mitochondrial respiratory genes was anti-correlated to mtDNA content ( see Supplementary file 2 ) . We speculate that this effect may be associated with the unique mitochondrial metabolism of prostate epithelia , which secrete large amounts of citrate generated in the mitochondria , rather than oxidizing it further and using the resulting NADH in the respiratory electron transport chain ( Costello et al . , 1997; 2004 ) . The landscape of genetic events driving tumors is diverse , and the presence and activity of these genetic lesions is now being used in design of clinical trials and development of new treatments ( Rubio-Perez et al . , 2015 ) . We sought to understand whether mtDNA abundance was associated with the incidence of particular mutations/copy number alterations ( CNAs ) in patient samples . To do so , we evaluated whether patients with a particular genetic lesion showed statistically significant increases or decreases in tumor mtDNA abundance , compared to wild-type samples . We restricted analysis to whole-exome sequencing data and which were not under embargo by the TCGA as of March 2015 . All results for the analysis are reported in Figure 6 and Supplementary files 3 and 4 . 10 . 7554/eLife . 10769 . 017Figure 6 . mtDNA content is correlated to the incidence of certain mutations and copy number alterations . Each point corresponds to a single alteration ( e . g TP53 mutation ) . Direction of arrow indicates whether alteration increases or decreases mtDNA content . X-axis in ( A ) and ( C ) indicates the fraction of samples in a cancer-type that contained the alteration ( i . e . , ≈ 20% of LGG samples were 10q deleted ) . ( A ) 73 out of 1896 copy number alterations ( CNAs ) tested were found to be significantly associated with mtDNA content ( Mann-Whitney p-value <0 . 05 ) . CNAs from the UCEC cancer type were excluded because of strong association between mtDNA copy number and the ‘copy-number high’ serous-like subtype in UCEC , shown in ( B ) . ( B ) The UCEC serous-like subtype displays a marked increase in tumor mtDNA copy number , relative to endometrial tumors of other subtypes . ( C ) Relatively few mutations ( 3 out of 3954 tested ) associate significantly with tumor mtDNA content ( Mann-Whitney p-value <0 . 05 ) . The UCEC associations are likely the result of the correlation to the serous-like subtype . ( D ) IDH1 and PTEN mutation status is linked to tumor mtDNA copy number in LGG . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 01710 . 7554/eLife . 10769 . 018Figure 6—figure supplement 1 . mtDNA copy number levels in kidney chromophobe carcinoma are elevated in samples with truncating mtDNA mutations . Samples are sorted by increasing mtDNA copy number on the X-axis , with percentile indicated on the Y-axis . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 018 The most apparent result of our analysis was the association of a large number of CNAs in endometrial carcinomas ( UCEC ) with increased mtDNA abundance . Recent work by the TCGA proposed a subtype stratification of endometrial carcinomas based on mutation and CNA frequency ( Kandoth et al . , 2013 ) . Among these subtypes is a serous-like ‘copy-number-high’ subtype with large numbers of somatic CNAs . We obtained the UCEC subtype classifications and confirmed that serous-like endometrial carcinomas exhibited substantially higher mtDNA copy number than all other subtypes ( Mann-Whitney p-value 7×10-6 , Figure 6 ) , explaining the large number of associations we observed . TP53 mutations are enriched in the serous-like subtype , and these mutations also showed statistically significant association with mtDNA abundance ( BH-corrected p-value 0 . 012 ) . After removing associations in UCEC , we were left with a small number of statistically significant mutations and CNAs associated with mtDNA abundance . Among these , the strongest signal arose from increased tumor mtDNA content in IDH1-mutant low grade gliomas ( Figure 6 , BH-corrected p-value 0 . 012 ) . Both IDH1 and IDH2 activating mutations induce production of the so-called ‘onco-metabolite’ 2-hydroxyglutarate , which competitively inhibits α-ketoglutarate-dependent histone demethylases and 5-methylcytosine hydroxylases , inducing a hypermethylation phenotype ( Turcan et al . , 2012; Xu et al . , 2011 ) . Surprisingly , IDH2 mutations showed no statistically significant change in mtDNA abundance , suggesting that the effect is specific to the cytosolic isoform IDH1 . Notably , mutations in PTEN were associated with a significant decrease in mtDNA abundance ( BH-corrected p-value 0 . 033 ) . These results echo a complementary finding by Navis and colleagues ( Navis et al . , 2013 ) , who reported that a mutant IDH1 R132H oligodendroglioma xenograft model displayed high densities of mitochondria and increased levels of mitochondrial metabolic activity . They proposed that an increase in mitochondrial mass would increase activity of mitochondrial IDH2 and compensate for loss of activity introduced by mutant IDH1 . Finally , prompted by a recent report implicating mutations in mtDNA itself with the pathology of kidney chromophobe carcinomas ( KICH ) ( Davis et al . , 2014 ) , we investigated the connection between mtDNA copy number and mtDNA mutations in KICH . Using somatic mtDNA mutation calls provided by the TCGA ( Davis et al . , 2014 ) , we examined whether mtDNA-mutated samples were likely to have more or fewer mtDNA copies than unmutated samples . We found that samples with mtDNA indels contained much higher quantities of mtDNA than unmutated samples ( Mann-Whitney U-test p-value 0 . 002 , Figure 6—figure supplement 1 ) . The same effect was not found when examining only single nucleotide variants . These results suggest that the presence of inactivating mtDNA mutations may induce increased mtDNA replication , perhaps as a response to inadequate mitochondrial energy production . So far , our findings have indicated that a number of tumor types appear to be depleted of mtDNA relative to normal tissue , and that in some ( but not all ) cases , the amount of mtDNA in a sample is correlated to the expression of respiratory genes . However , in some cancer types ( e . g . bladder ) , tumors exhibited depletion of mtDNA ( Figure 3 ) , but expression of mitochondrial genes was not correlated to mtDNA copy number ( Figure 5 ) . This discrepancy is reminiscent of prior work describing mtDNA depletion which was not accompanied by a drop in respiratory activity or mitochondrial protein expression . Instead , a compensation of respiratory activity was described in cases of mtDNA depletion caused by either genetic alterations ( Seidel-Rogol and Shadel , 2002; Barthélémy et al . , 2001; Dorado et al . , 2011 ) or reverse-transcriptase inhibitors ( Kim et al . , 2008; Miró et al . , 2004; Stankov et al . , 2007 ) . To investigate whether mtDNA depletion was associated with a concurrent decrease of mitochondrial protein expression , we examined the abundance of a mitochondrial protein using immunohistochemistry ( IHC ) ( Thermo Fisher Scientific Mitochondria Ab-2 , Clone MTC02 , see Materials and methods ) in 3 tumor/normal pairs of clear-cell renal cell carcinoma , papillary renal cell carcinoma , and high-grade muscle-invasive urothelial bladder carcinoma ( corresponding to TCGA studies KIRC , KIRP , and BLCA respectively; Figure 7 , and Figure 7—source data 1 ) . In KIRC , which was the most strongly mtDNA-depleted tumor type in Figure 3 , we found significant depletion of mitochondrial protein in all tumor samples compared to adjacent normal renal parenchyma . In KIRP , for which 69% of paired samples were depleted of mtDNA in Figure 3 , we observed a more subtle depletion of mitochondrial protein in 2/3 tumor samples , compared to adjacent normal renal parenchyma . In BLCA , we found that 2/3 BLCA tumors showed increased levels of mitochondrial protein , which contrasted with Figure 3 , where nearly all samples showed evidence of mtDNA depletion . 10 . 7554/eLife . 10769 . 019Figure 7 . Top panel depicts H&E stains , and bottom panel depicts immunohistochemistry with antibody against mitochondrial protein . In all H&E stains , red ‘T’ indicates tumor tissue , while blue ‘N’ indicates normal tissue . Orientation of tumor/normal tissue is mirrored in bottom panel . ( A ) H&E-stained section shows clear cell renal cell carcinoma ( top left , KIRC Sample 1 from Figure 7—source data 1 ) with the classical features of tumor nests with clear cytoplasm , separated by intricate , branching vascular septae , and adjacent non-neoplastic renal parenchyma ( lower right ) . ( B ) KIRC Sample 1 immunohistochemical staining with MITO Ab2 antibody reveals markedly lower mitochondrial content ( cytoplasmic , brown granular positivity ) in clear cell RCC compared to normal tubules . ( C ) H&E-stained section shows papillary renal cell carcinoma type 1 ( KIRP Sample 3 ) with tumor ( top right ) and normal tubules ( lower left ) . ( D ) KIRP Sample 3 immunohistochemical stain with MITO Ab2 antibody shows KIRP with a slightly weaker positivity compared to normal tubules . ( E ) H&E-stained section showing invasive high grade urothelial carcinoma ( lower left ) with sheets of tumor cells in the lamina propria and the overlying normal urothelium ( top right ) . ( F ) Immunohistochemical staining with MITO Ab2 antibody reveals slightly higher mitochondrial staining in urothelial carcinoma compared to normal urothelium . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 01910 . 7554/eLife . 10769 . 020Figure 7—source data 1 . Table of results of immunohistochemistry for mitochondrial protein . The H-score for KIRC was significantly less than matched normal renal parenchyma in 3/3 samples . The H-score for KIRP was marginally lower than the H-score than matched normal renal parenchyma in 2/3 samples . The H-score for BLCA was higher than the H-score for matched normal urothelial tissue in 2/3 samples . DOI: http://dx . doi . org/10 . 7554/eLife . 10769 . 020 Collectively , our results from IHC regarding mitochondrial protein expression agree with those from sequencing in 2/3 cancer types ( KIRC and KIRP ) . In a third cancer type ( BLCA ) , mtDNA depletion as quantified by sequencing is not mirrored by a synchronous down-regulation of mitochondrial protein levels . As mentioned earlier , our results from gene expression analysis ( Figure 5 ) indicate that mtDNA copy number is correlated to mitochondrial respiratory gene expression in KIRC and KIRP , but not BLCA . In fact , in BLCA , the gene sets most strongly correlated to mtDNA copy number were associated with the cell cycle and immune response . This suggests that other mechanisms compensate for the depletion of mtDNA in BLCA ( and potentially in other cancer types ) , which is further discussed in the concluding section . Taken together , these results support the notion that factors besides mtDNA copy number can determine the rate of mitochondrial transcription , and that mtDNA depletion is not sufficient evidence to conclude that mitochondrial respiration is down-regulated in a tumor . In this study , we have investigated the variation of mtDNA copy number levels across many tumor types , arriving at several intriguing observations . Across nearly half of the tumor types we studied , we found evidence for depletion of mtDNA , relative to adjacent normal tissues . Orthogonal measurements of transcription levels ( via RNA-Seq ) and mitochondrial protein levels ( via IHC ) in a subset of these samples linked this variation to downregulation of mitochondrially-localized metabolic pathways , in some but not all tumor types . Our findings of gross changes in mtDNA content in tumors echo a number of prior but isolated observations , largely based on quantitative PCR measurements and with substantially smaller sample sizes , of mtDNA copy number changes in cancers ( see [Yu , 2011] for a thorough review ) . For example , oncocytomas ( not analyzed in this work ) are well-known to be characterized by the excessive accumulation of mitochondria ( Tickoo et al . , 2000 ) . Furthermore , decreases in mtDNA copy number have been reported in breast cancer ( Mambo et al . , 2005; Fan et al . , 2009 ) , liver cancer ( Lee , 2004 ) , and clear-cell kidney cancers ( Meierhofer et al . , 2004; Nilsson et al . , 2015 ) . While the majority of our observations agree with prior work ( when comparing to [Yu , 2011] ) , some of our results are in contradiction to prior studies . The discordance between findings seems in part due to inadequate sample sizes , and incomplete or unavailable matched normal tissue . For example , in contrast to ( Mambo et al . , 2005 ) and ( Wang et al . , 2005 ) , we find no clear increase or decrease in mtDNA content in thyroid or endometrial carcinomas , respectively . However , ( Mambo et al . , 2005 ) profiled 20 paired thyroid tumors , versus 66 paired thyroid tumors in this report; and ( Wang et al . , 2005 ) utilized unpaired samples of tumor and normal endometrial tissue ( Wang et al . , 2005 ) , versus 32 paired samples here . We further showed that mtDNA ploidy alone cannot be used as a surrogate for the respiratory activity of a tumor sample . The literature contains several reports of mtDNA copy number depletion without reduction in mitochondrial transcription/respiratory activity , both in vitro and in vivo . In ( Seidel-Rogol and Shadel , 2002 ) , HeLa cells depleted of mtDNA by culture in ethidium bromide showed substantial mitochondrial transcription despite the fact that mtDNA , TFAM , and mitochondrial RNA polymerase were all at depleted levels . There , the authors suggest that an excess of TFAM and mitochondrial RNA polymerase prior to depletion may ensure that , even once depleted , transcription is sustained . Another report examined mtDNA depletion as a result of thymidine kinase 2 deficiency in mice , and observed a down-regulation of the mitochondrial transcriptional terminator MTERF3 in heart tissue . As a result , the expression of mitochondrial transcripts ( ND6 and COX1 ) increased in heart tissue , as did the ratio of the levels of these transcripts to mtDNA levels . The consequence of this transcriptional compensation was that the heart tissue was spared from respiratory deficiency ( Dorado et al . , 2011 ) . In tandem with our report , these findings emphasize a nuanced connection between mtDNA copy number and respiratory gene expression . We would argue strongly that future studies investigating changes in mtDNA in tumors should quantify mtDNA protein expression in parallel with estimating mtDNA copy number . A number of related open questions remain to be resolved , including what mechanisms determine the incidence and/or extent of compensation to mtDNA depletion , and what the consequences of mtDNA depletion may be when such compensation takes place ( e . g . upregulation of the immune response ) . While mtDNA depletion or accumulation may typify certain cancer types , we further identified that subsets of patient samples , characterized by the presence of particular somatic mutations/copy number alterations , were enriched/depleted in mtDNA . The presence of activating IDH1 mutations ( in low grade gliomas ) or a large number of copy number alterations ( in serous-like endometrial carcinomas ) is strongly correlated to high tumor mtDNA content . If these tumors ( and others with increased mtDNA content ) have an increased dependence on mitochondrial metabolism to proliferate , using mitochondrially-targeted therapies ( e . g . metformin ) may be a therapeutic opportunity . Similarly , vulnerability to mitochondrially-targeted therapies might arise from disabling passenger mutations in genes required for mtDNA copy number maintenance ( e . g . DNA polymerase gamma ) . Both hypotheses should be amenable to investigation in carefully chosen cell line models of cancer . A number of reports have now described extensive genetic heterogeneity of some tumor types ( e . g . kidney cancers [Gerlinger et al . , 2012] ) , where spatially distinct biopsies isolated from the same patients have non-overlapping somatic alterations . However , no reports have examined how mitochondrial DNA mutations and copy number vary spatially across a tumor . Variation of this kind , if it exists , might reflect functional diversity in mitochondrial metabolic activity and signaling in different regions of a tumor . Alternately , it would be of particular interest to trace the time-evolution of mtDNA content in a single patient over the course of treatment . As critical players in immunity , signaling , and metabolism , we suspect that mitochondria will inevitably play a role in the evolution of resistance to therapeutic intervention . Whole exome sequencing ( WXS ) and whole genome sequencing ( WGS ) BAM files for 22 distinct TCGA studies were obtained from the TCGA CGHub repository ( Figure 2 ) ( Wilks et al . , 2014 ) . We restricted our analyses to sequence data aligned to GRCH37 using the mitochondrial Cambridge Reference Sequence ( CRS ) . We focused only on primary tumor , adjacent normal tissue , and normal blood samples ( ‘01’ , ‘11’ , and ‘10’ in the sample type field of the TCGA barcode ) . We further restricted our analyses to samples which were not whole-genome amplified prior to sequencing ( i . e . , we only used samples containing ’D’ in the analyte field of the TCGA barcode ) , because such amplification could potentially bias the relative abundances of mitochondrial and nuclear DNA in the sample . Samtools ( Li et al . , 2009 ) was used to extract reads aligning to the mitochondrial genome meeting the following critieria: ( 1 ) passed quality-control , ( 2 ) were not marked as duplicate reads , ( 3 ) were properly paired , and ( 4 ) were aligned with Phred-scaled mapping quality ( MAPQ ) >30 . The number of such reads aligning to the mitochondrial genome was compared to the number of such reads aligning to the nuclear genome . The pipeline described above includes a number of controls to ensure that mtDNA copy number estimates are not influenced by nuclear integrations of mitochondrial sequences ( NUMTs ) ( Hazkani-Covo et al . , 2010 ) . A direct result of restricting analysis to properly paired reads is that reads whose mate mapped to a different chromosome are removed prior to copy number calculation . Furthermore , by requiring a conservative Phred-scaled minimal mapping quality of 30 ( equivalent to a 99 . 9% likelihood that reads are aligned to the correct genomic location ) , reads with homology to nuclear-encoded NUMTs are removed prior to copy number calculations . Prior work has established that more lenient mapping quality thresholds of 20 are sufficient for accurately calling mtDNA copy number ( Ding et al . , 2015 ) A complete list of all copy number estimates is available in Supplementary file 1 . Affymetrix SNP6 arrays for tumor and normal samples were acquired for 22 cancer types from the TCGA . Arrays for each individual cancer type were processed together , quantile-normalized and median polished with Affymetrix power tools using the birdseed algorithm to obtain allele-specific intensities . PennCNV ( Wang et al . , 2007 ) was used to generate log R ratio and B-allele frequencies for each tumor . ASCAT ( Van Loo et al . , 2010 ) was used to generate allele-specific copy number and estimate tumor ploidy and purity using matched arrays from tumor and normal tissue . In order to estimate mtDNA copy number in Equation 1 , we compared the number of reads aligning to the mitochondrial genome to the number of reads aligning to a genome of known ploidy . For samples of normal tissue , we assumed this known ploidy was equal to 2 . For tumor tissue which may be infilitrated by stromal/immune cells and copy-number altered , we need to correct for the ‘effective ploidy’ of the sample . We define this correction factor to be ( 3 ) RTumor=Purity×Ploidy+ ( 1-Purity ) ×22 where the purity and ploidy values are obtained from ASCAT , as described above . When a sample is composed of pure normal tissue , R=1 . Inspection of mtDNA copy number results indicated a potential association between mtDNA copy number and processing batch . This is consistent with prior reports , e . g . ( Ju et al . , 2014 ) , which described large variation in efficiency of mtDNA depletion in exome sequencing in a sequencing-center-dependent manner . We separately examined the log10 mtDNA copy number for each TCGA plate ID for ( 1 ) blood , and ( 2 ) tissue-derived ( tumor and adjacent-normal tissue ) samples . Kruskal-Wallis tests using either blood or tissue-derived mtDNA copy number indicated significant differences in median mtDNA copy number between TCGA plates in 21/22 whole exome sequencing ( WXS ) datasets ( p-value <0 . 05 ) . In contrast 3/10 whole genome sequencing ( WGS ) datasets showed significant differences between TCGA plates using tissue-derived mtDNA copy number ( 5/10 using blood-derived mtDNA copy number ) . Manual inspection further indicated that the magnitude of the batch effect was smaller in WGS compared to WXS . We also calculated , for each TCGA plate i in a given cancer type , the mean mtDNA copy number in ( 1 ) blood ( mib ) and ( 2 ) tumor/adjacent-normal tissue ( mit ) . We observed a statistically significant positive linear correlation ( Pearson p-value <0 . 1 ) between mt and mb in 17/19 cancer types profiled with WXS , but in 0/7 cancer types profiled with WGS ( analysis restricted to studies with adequate numbers of samples , defined as at least 3 different TCGA plate IDs with at least 3 blood and 3 tissue samples ) . Importantly , in many cancer types , tumor and blood from the same patient were often processed in different TCGA plates ( e . g . 30% in BLCA , 30% in BRCA , and 48% in STAD ) , and there was no reason a priori to expect that blood and tumor mtDNA estimates should be correlated across batches . The existence of such a correlation suggested a TCGA-plate-dependent contribution to the observed copy number of all samples ( both blood and tissue-derived ) processed within that plate . Taken together , the results above suggested that a batch effect associated with TCGA plate ID/sequencing center was present in WXS , and possibly also in WGS . Based on these considerations , the small magnitude of the batch effect in WGS , and the prior literature describing significant variation in mtDNA sequencing depth in exome sequencing due to differences in exonic enrichment ( Ju et al . , 2014 ) , we elected to control for a batch effect in WXS , but not WGS . Importantly , in Supplementary file 1 , we report uncorrected mtDNA copy number , corrected mtDNA copy number , and plate IDs for all samples analyzed , so that future work by others may model the effect in whatever manner they see fit . The set of observed log10 mtDNA copy numbers M are indexed by the TCGA plate i , the tissue type ( either tumor/adjacent-normal or blood ) j , and the sample k . We fit the following linear model to the mtDNA copy numbers Mi⁢j⁢k: ( 4 ) Mi⁢j⁢k=μ+∑p⁢l⁢a⁢t⁢eαi⁢Pi+∑t⁢i⁢s⁢s⁢u⁢eβj⁢Tj+ϵi⁢j⁢k where Pi and Tj are indicator variables corresponding to the plate and tissue that a sample comes from . The parameter μ is the grand mean of mtDNA copy number , α the effect attributable to the plate and β that attributable to the tissue . The residual ϵ is the sum of the true log10 mtDNA copy number and measurement error . We constructed a linear model corresponding to Equation 4 for each cancer type . Then , we corrected the observed log10 mtDNA copy number for each WXS sample according to the contribution α^ from the corresponding TCGA plate ID . Corrected copy number values were used for subsequent survival , gene expression , and somatic alteration analysis . Survival analysis was performed with univariate Cox proportional hazards regression models where the independent and dependent variables were the log10-transformed corrected mtDNA copy number and the overall survival respectively . The p-values for the significance of mtDNA copy number as a predictor of survival were obtained from Wald tests . RSEM normalized RNA-Seq gene expression data were downloaded from the Broad Firehose , using the most recent data as of November 4 , 2014 . Data were filtered to remove genes with average read count less than 16 . We calculated the nonparametric Spearman correlation and associated p-value between the expression of each gene and the copy number of mtDNA . To remove putatively spurious correlations , we identified all genes with a p-value greater than 0 . 05 , and set the correlation coefficient for those genes equal to zero . We then use the geneSetTest function in the limma ( Law et al . , 2014 ) package to test whether particular gene sets showed an enrichment for positive/negative correlations , relative to the distribution of correlations across all genes . Thus , a single p-value was calculated for each alternative hypothesis . All p-values were adjusted using the Benjamini-Hochberg procedure . The method described above was applied to every Reactome pathway in the MSigDB canonical pathways gene set . The enrichment score depicted in Figure 5 is the -log10 corrected p-value of this statistical test , with the color ( red or blue ) indicating the direction of enrichment . The analysis was run separately for tumor and normal tissues . We applied our gene set analysis pipeline to all studies for which we had at least 20 samples of RNA-Seq data ( in order to retain sufficient statistical power ) . Analyses were run for each combination of tumor type and tissue , and ensuing results were then aggregated across all studies . All results from the analyses are provided in the Supplementary file 2 . For each study , Gistic2 and MutSigCV results were downloaded from the Broad Firehose ( most recent data as of Nov 14 , 2014 ) . From Gistic , we retained all arm-level and focal alterations with q-value less than 0 . 1 . For mutations , we obtained the MAF file from the output of MutSig . For each gene , we calculated the number of patients in which this gene exhibited a nonsynonymous , coding mutation ( i . e . , missense , non-sense , frameshift , in-frame insertion/deletions , and splice-site mutations ) , excluding those with greater than 600 non-synonnymous coding mutations ) . We then retained any genes which were mutated in greater than 4% of patients . Non-parametric Mann-Whitney U-tests were used to evaluate whether tumors bearing a particular somatic alteration contained significantly higher/lower amounts of mtDNA in tumor samples . After testing all associations , p-values obtained from the U-tests were corrected using the Benjamini-Hochberg procedure . All tissues were fixed in 10% neutral-buffered formalin and paraffin embedded as part of a routine surgical pathology procedure and 5-micron-thick sections stained with Hematoxylin and eosin ( H&E ) were reviewed . Immunohistochemical ( IHC ) analysis was performed on 5-micron-thick sections by Ventana , Discovery XT immunohistochemical stainer . The sections were deparaffinized and subjected to heat induced antigen retrieval using CC1 at high pH before primary incubation with MITO Ab2 ( mouse monoclonal , clone MTC02 , Neomarkers , 1:50 dilution ) . Slides were then counterstained with hematoxylin , dehydrated and cover-slipped .	Within each cell of your body lie hundreds or thousands of mitochondria . These structures are perhaps best known for making energy , but mitochondria also play roles in processes like the immune response and cell signaling . However , in the mutant cells that form cancerous tumors , these roles can be subverted and altered . Mitochondria contain their own DNA , which is distinct from the DNA stored in the nucleus of the cell , and codes for the proteins that the mitochondria need to produce energy . Reznik et al . used next-generation DNA sequencing data produced by The Cancer Genome Atlas consortium to estimate the number of copies of mitochondrial DNA in tumor cells and the adjacent normal tissue . This revealed that in many types of cancer , tumor cells have fewer copies of mitochondrial DNA than the cells that make up normal tissue . In many cases , the depletion of mitochondrial DNA was accompanied by a reduction of the expression of mitochondrial genes , suggesting that mitochondrial activity may be suppressed in these tumor types . Reznik et al . also found that the number of copies of mitochondrial DNA in certain tumor types is related to the incidence of key 'driver' mutations that cause cells to become cancerous . This knowledge may help to develop new treatments for these tumors .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology", "cancer", "biology" ]	2016	Mitochondrial DNA copy number variation across human cancers
We investigate the in vivo patterns of stem cell divisions in the human hematopoietic system throughout life . In particular , we analyze the shape of telomere length distributions underlying stem cell behavior within individuals . Our mathematical model shows that these distributions contain a fingerprint of the progressive telomere loss and the fraction of symmetric cell proliferations . Our predictions are tested against measured telomere length distributions in humans across all ages , collected from lymphocyte and granulocyte sorted telomere length data of 356 healthy individuals , including 47 cord blood and 28 bone marrow samples . We find an increasing stem cell pool during childhood and adolescence and an approximately maintained stem cell population in adults . Furthermore , our method is able to detect individual differences from a single tissue sample , i . e . a single snapshot . Prospectively , this allows us to compare cell proliferation between individuals and identify abnormal stem cell dynamics , which affects the risk of stem cell related diseases . Our mathematical model recovers the temporal change of telomere length distributions in human hematopoietic cells with a minimal number of required model parameters . Since hematopoietic cells proliferate in a hierarchical organised tissue with slowly dividing stem cells at its root , such a model needs to connect properties of cell proliferation and telomere shortening . Telomere length can be assessed on three different levels of resolution , ( i ) the level of single telomeres , ( ii ) the level of single cells and ( iii ) the level of the tissue . Of course these levels are not independent , for example the knowledge of telomere length in all cells allows to obtain the ( average ) telomere length of a tissue . The processes that drive telomere length dynamics differ at these levels of resolution . Single telomeres are prone to stochastic events such as oxidative stress or recombination and thus may also shorten by effects independent of proliferation associated attrition ( von Zglinicki , 2002; Antal et al . , 2007 ) . Healthy human cells contain 184 telomeres , four on each of the 46 chromosomes . Thus , the noise on the level of single telomeres becomes much smaller on the cell level . We capitalise on this and consider telomere length on the cell level in the following . Thus , the average telomere length of a cell shortens by a constant factor during each division . Such an approach might underestimate the number of senescent cells once telomeres become critically short , since it is the length of the shortest telomere rather then the average telomere length that triggers cell cycle arrest ( Hemann et al . , 2001 ) . Our model is sensitive to the accumulation of cells in the state of cell cycle arrest and we can infer this effect experimentally from population wide telomere length distributions . However , this effect can likely be neglected during adolescence and adulthood , but might have important implications in some tumors , at old age or in conditions associated with abnormal telomere maintenance . We further need to consider properties of a hierarchical tissue organization , where few slowly dividing stem cells give rise to shorter lived progeny . Although some of the progeny , particularly primitive progenitor cells , can be long lived and are able to maintain homeostasis without stem cell turnover for intermediate time intervals , eventually all non hematopoietic stem cells will be depleted without continuous stem cell turn over ( Busch et al . , 2015; Sun et al . , 2014 ) . Age dependent differences in telomere shortening across different lineages of hematopoiesis can only persist in the hematopoietic system if they occur on the level of the maintained self-renewing cell population . Cells leaving the stem cell pool have an approximately constant number of cell divisions before they reach maturation ( Takano , 2004; Werner et al . , 2011 ) . This shifts the distribution to shorter values of telomere length and consequently , the distribution of telomere lengths of mature cells is a good proxy for the distribution of telomere lengths in stem cells ( Rodriguez-Brenes et al . , 2013 ) . We measured telomere length distributions in lymphocytes , granulocytes and bone marrow sections separately . This allows us to investigate the myeloid and lymphoid lineage of hematopoiesis independently . In our model , we assume a population of initially N0 stem cells . In the simplest case , each stem cell would proliferate with the same rate r and the cell cycle time would follow an exponential distribution . However , tissue homeostasis requires continuous stem cell turn over in intermediate time intervals , therefore the proliferation rate of the population of stem cells is adjusted , such that a required constant output of differentiated cells per unit of time is maintained . In the simplest case of a constant stem cell population , the effective proliferation rate becomes r/N0 . However , in more complex scenarios , the number of stem cells could differ with age and the effective proliferation rate of stem cells r/Nt also becomes age dependent ( Rozhok and DeGregori , 2015; Bowie et al . , 2006 ) . This resembles a feedback mechanism and results in an approximately Log-normal distribution of cell cycles , see also Equation S26 in Materials and methods for details . In addition , each stem cell clone is characterised by a certain telomere length ( Antal et al . , 2007; Simon and Derrida , 2008 ) . This telomere length shortens with each stem cell division by a constant length Δc and consequently the remaining proliferation potential is reduced in both daughter cells ( Rufer et al . , 1999; Allsopp et al . , 1992 ) . If the telomeres of a cell reach a critically short length , this cell enters cell cycle arrest and stops proliferation , reflecting a cell’s Hayflick limit ( Hayflick and Moorhead , 1961 ) . This can be modelled by collecting cells with the same proliferation potential in states i . A cell enters the next downstream state i→i+1 after a cell division , see also Figure 1 , as well as Equations S1 , S14 in Materials and methods . Since the next cell to proliferate is chosen at random from the reservoir , cells progressively distribute over all accessible states with time ( Olofsson and Kimmel , 1999 ) . This corresponds to the problem of how many cells are expected in a state i at any given time , which we denote by Nit in the following . 10 . 7554/eLife . 08687 . 003Figure 1 . The combination of telomere length data and mathematical modeling allows to infer individualized stem cell proliferation patterns . ( a–c ) Blood or bone marrow samples were taken from healthy persons with ages between 0 and 85 . Telomere length was measured with Flow-FISH and Q-FISH techniques , resulting in individualized telomere length distributions . ( d–g ) Mathematical framework: Stem cells divide either symmetrically or asymmetrically . Each cell is characterized by an average telomere length . Cells with the same state are collected in compartments . The average of the underlying stochastic process is captured by a system of differential equations . The solution of this equation is a generalised truncated Poisson distribution that gives rise to a traveling wave , see Equation S15 . ( h , i ) The combination of modeling and telomere length distribution measurements allows dynamic predictions for individuals , see Figure 6 . These predictions can be tested on population wide data of telomere length , for example see Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 00310 . 7554/eLife . 08687 . 004Figure 1—figure supplement 1 . Results of the mathematical model on the temporal change of individual telomere length distributions . Compared are analytical results ( lines ) and averages of stochastic computer simulations ( dots ) of our mathematical model , see Materials and methods . ( a ) An example of a population of 100 cells , where each cell has 7 proliferation cycles before it enters cell cycle arrest ( cells accumulating in state ) . ( b ) Expected telomere length distributions at 6 distinct time points ( time increases with decreasing remaining number of cell divisions ) . The telomere length distribution gives rise to a traveling wave that progressively widens and shifts towards shorter telomere length . The maximum of this distribution declines proportional to 1∕time ( black line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 004 The shape of the distribution of cells across cell cycles depends on the patterns of stem cell proliferation , for example the ratio of symmetric versus asymmetric divisions . An asymmetric stem cell division produces one stem and one non-stem cell ( for example a progenitor cell that leaves the stem cell compartment ) . If we restrict the stem cells dynamics to only asymmetric divisions , the process results in a stem cell population of constant size and the number of cells in each state i follows a Poisson distribution ( 1 ) N ( i ) ( t ) = N0i ! rtN0ie-rtN0 . A typical example of this distribution is shown in Figure 1—figure supplement 1 and details on the derivation can be found in Materials and methods , see Equation S1 . Cells with maximum proliferation capacity ( cells in state 0 in our model ) are progressively lost and cells accumulate in the final state of cell cycle arrest by passing through all intermediate states . Inferring the dynamics of distribution ( 1 ) from in vivo measurements requires sequential sampling and complicated cell sorting , which seems challenging in realistic clinical settings . On the other hand , the measured ( observed ) telomere length distribution corresponds to a single sample of the underlying Poisson process . The expected shape of this observed distribution is depicted in Figure 1g . It becomes a traveling wave that starts narrowly distributed around an initial telomere length and shifts towards shorter average telomere length with time . We have measured this distribution , which arises from our theoretical model , experimentally in many samples of granulocytes , lymphocytes and bone marrow sections of healthy adult humans , which we discuss in detail below . In addition to asymmetric divisions , stem cells can undergo symmetric self-renewal , which is a prerequisite for development , as it allows for a growing stem cell population . In our model , stem cells divide symmetrically with probability p and asymmetrically with probability 1-p respectively . In this situation , the number of stem cells is not constant , but increases with each symmetric stem cell self-renewal . As a consequence , the expected distribution also changes and is now described by a generalised Poisson distribution ( see Equation S14 in Materials and methods ) given by ( 2 ) Npit=N0i ! 1+ppilnirpN0t+1rpN0t+1p . This distribution also leads to a traveling wave , but the maximum of the distribution decreases considerably slower compared to the case of purely asymmetric stem cell divisions . In the following , we refer to the model that is restricted to only asymmetric stem cell divisions as model 1 and denote the more general case of symmetric and asymmetric cell divisions as model 2 . Ideally , we would like to follow these traveling waves in individual healthy humans over time and compare this sequential data to the dynamics from our model predictions . Unfortunately , the time required to confirm our model across all ages would exceed the life expectancy of the authors . We therefore explored those properties of our analytical model that are directly testable in population wide data of telomere length . One such property is the change of the average telomere length with age , which we measure in a group of 356 healthy individuals . The average telomere length decreases in most human tissues with age ( Harley et al . , 1990 ) . This is well known and has been confirmed numerous times . Surprisingly , less is known about the detailed dynamics of this decrease . We can derive the dynamics of the average telomere length from the telomere length distributions directly . The average telomere length corresponds to the expected value of the telomere length distribution ( in the following denoted by E[c ( t ) ] ) , see Equation S5 in Materials and methods for details . As the telomere length distribution changes with time , the average telomere length becomes time dependent naturally . In the absence of symmetric stem cell self-renewal ( model 1 ) the average telomere length Ect is expected to decrease linearly ( 3 ) Ect≈c-ΔcrtN0 , with age ( denoted by t in the equation above ) . More specifically , the average telomere length of cells of a particular type , e . g . the population of granulocytes or lymphocytes , shorten by a constant fraction each year . The dynamics changes once a significant fraction of cells enter cell cycle arrest , see Equation S9 . The average telomere length transitions from a linear into a power law decline ( when the average telomere length becomes very short ) and the stem cell pool reaches the state of complete cell cycle exhaustion asymptotically . This transition would enable the identification of an age where a considerable fraction of stem cells enter cell cycle arrest , potentially a mechanism important in aging , carcinogenesis or bone marrow failure syndromes . Furthermore , we calculated the variance of the underlying stochastic process . This gives us a measure for the expected fluctuation of the average telomere length in a population of healthy humans . We expect the variance to increase linearly in time in the absence of symmetric stem cell self-renewal . Consequently , the standard deviation is proportional to the square root of age . Yet again , similar to the average telomere length , the dynamics of the variance changes once a significant fraction of cells enters cell cycle arrest . The variance starts to decrease and would reach zero , if all cells stopped proliferation . The distribution of telomere length changes under the presence of symmetric stem cell self-renewal ( model 2 ) . Accordingly , we expect a different decrease of the average telomere length . We find that the telomere length follows a logarithmic decay with age ( see also Equation S19 ) , given by ( 4 ) Epct≈ c-Δc1+ppln ( rpN0t+1 ) The average telomere length of a cell population shortens less with increasing age under the presence of symmetric self-renewal , although the decrease of telomeric repeats per cell division ( denoted by Δc in Equation 4 ) is constant . This effect emerges naturally in our model due to the increasing number of stem cells with age . In a population with only few cells , each cell proliferation has a considerable impact on the average telomere length , while this impact diminishes in larger populations . If the stem cell population increases progressively , telomere shortening reduces on the tissue level with age . In order to test the predictions of our model experimentally , we have measured telomere length in lymphocytes and granulocytes in a cohort of 356 healthy humans with ages between 0 and 85 years . Our data includes 47 cord blood samples of healthy children and bone marrow biopsies of 28 patients with diagnosed Hodgkin lymphoma without bone marrow involvement . We assessed the average telomere length in all 356 samples with established Flow-FISH protocols ( Aubert et al . , 2012; Baerlocher et al . , 2006; Weidner et al . , 2014; Beier et al . , 2012 ) . This reveals the population wide dynamics of telomere length and contains a significant number of cord blood samples that allow us to investigate differences in cell proliferation during adolescence and homeostasis in adulthood . In addition , we have analyzed 28 blood samples of lymphocytes , 10 blood samples of granulocytes and 28 bone marrow biopsies with quantitative-fluorescence in situ hybridisation ( Q-FISH ) ( Beier et al . , 2015; Varela et al . , 2011; Zijlmans et al . , 1997 ) ( see Figure 2 and experimental methods for details ) . The averages of these samples correspond to the open symbols in Figure 3 . From the full distribution , we obtain the telomere length distributions of single individuals and estimate personalised cell proliferation properties , e . g . the ratio of symmetric to asymmetric cell divisions as well as the rate of telomere shortening for each sample separately . We compare these personalised estimates to population wide telomere length to test the consistency of our results on two independent data sets . 10 . 7554/eLife . 08687 . 005Figure 2 . Representative image of the Q-FISH analysis of a bone marrow section . ( a ) Maximum projection image of a paraffin-embedded bone marrow section of confocal Q-FISH with DAPI and Cy3 . ( b , c ) Single DAPI and Cy3 staining respectively . ( d ) Overlay of image analysis of nucleus and telomere detection . ( e ) Image analysis of the DAPI staining is shown . Detected nuclei are shown in red . ( f ) Image analysis of the Cy3 staining . Detected telomeres marked in red . For details on the Q-FISH analysis please see Materials and methods . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 00510 . 7554/eLife . 08687 . 006Figure 2—figure supplement 1 . Representative image of the Q-FISH analysis of a peripheral blood cytospin . ( a ) Maximum projection image of confocal Q-FISH with staining of DAPI and Cy3 . ( b , c ) Single DAPI and Cy3 staining is shown . ( d ) Image analysis with nucleus detection marked with red lines . ( e ) Image analysis of detected single telomeres marked in red . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 00610 . 7554/eLife . 08687 . 007Figure 2—figure supplement 2 . Representative FACS blot of a flow-FISH analysis . ( a ) Representative flow-FISH blot of healthy individual . Based on LDS 751 staining and forward scatter properties , cow thymocytes , lymphocytes and granulocytes can be identified . ( b ) Telomere intensity of Alexa488 of unstained and stained thymocytes is given . ( c ) Telomere intensity of Alexa488 of unstained and stained lymphocytes . ( d ) Telomere intensity of Alexa488 of unstained and stained granulocytes . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 00710 . 7554/eLife . 08687 . 008Figure 3 . The population wide average telomere length of ( a ) lymphocytes and ( b ) granulocytes . The data from a cohort of 356 individuals ( symbols ) is captured by a logarithmic decrease of the average telomere length ( solid line ) , which is predicted by our model 2 that allows for symmetric stem cell divisions and thus leads to a slowly increasing stem cell pool . Based on the fit of the average , the mathematical model predicts a standard deviation that increases with the square root of the age ( dashed lines ) . This approach does not take the genetic variability of telomere length in newborns into account . The decrease of the average telomere length slows down in children and becomes almost linear in adults , see also Figure 4 . For individuals represented by filled symbols , only information on the average telomere length is available . For individuals represented by open symbols , we additionally analysed the distribution of individually detected telomeres , see Figure 6 . An additional parameter estimation on an independent data set is shown in Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 00810 . 7554/eLife . 08687 . 009Figure 3—figure supplement 1 . Decrease of the average telomere length of ( a ) lymphocytes and ( b ) granulocytes in a population of 835 healthy humans . The data was taken from ( Aubert et al . , 2012 ) and confirms the inferred parameter range in Figure 2 independently . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 009 In order to compare our model with the experimental data , we implemented standard maximum likelihood estimates for a regression analysis . Our experimental finding in adults ( we only consider persons of 20 years or older ) show that telomere length in granulocytes and lymphocytes decreases approximately linearly with age on the population level . In both cell populations the telomere length of adults decreases with 50±5 bp/year ( we state the maximum likelihood estimate and the 95% confidence interval ) . If for example a cell looses on average 50 bp telomeric repeats per cell division ( Rufer et al . , 1999 ) , this implies approximately 1 replication per year for the hematopoietic stem cells . This agrees with the observation of rare stem cell turnover under homeostasis ( Busch et al . , 2015; Sun et al . , 2014; Dingli et al . , 2006 ) . However , the assumption of strictly asymmetric cell divisions ( model 1 ) fails to explain the pronounced loss of telomere repeats in infants ( prediction of model 1 for the initial telomere length in lymphocytes: 9 . 8±0 . 15 kbp , measured average initial telomere length: 10 . 67±0 . 4 kbp , similar results for granulocytes , see also Figure 4 for a comparison of model 1 and model 2 ) . This discrepancy can be resolved by introducing an interplay of symmetric and asymmetric stem cell divisions ( model 2 ) that allows for an increasing number of stem cells . In this situation , the proliferation rate of stem cells becomes age dependent and our model predicts that at the youngest ages , when the number of stem cells is lowest , telomere loss is most pronounced . Maximum likelihood estimates of our general mathematical solution ( Equation 4 ) to the telomere length data on the population level ( see Figure 3 ) reveals for the parameter controlling average loss of telomere length in lymphocytes a value of 75±7 bp/year , an initial telomere length of 10 . 4±0 . 2 kbp and a probability for symmetric stem cell self-renewal of 0 . 35 ± 0 . 07 . In granulocytes we find a value of telomere loss of 68±5 bp/year , an initial telomere length of 10 . 2±0 . 3 kbp and a probability for symmetric stem cell self-renewal of 0 . 44 ± 0 . 2 . This probability accounts for the increased loss of telomere repeats in infants and substantially improves the prediction of the initial average telomere length . In addition to our group of 356 healthy humans , we have tested our hypothesis in an independent data set of 835 healthy humans , previously published by an unrelated group in ( Aubert et al . , 2012 ) , see Figure 3—figure supplement 1 . This set confirms our parameter estimations , in particular the accelerated decrease of average telomere length during adolescence is also observed . 10 . 7554/eLife . 08687 . 010Figure 4 . Comparison of the average telomere length decrease of lymphocytes predicted by Model 1 and Model 2 . Model 1 ( red dashed line , best fit to the data ) predicts a linear decrease of the average telomere length with age . The linear decrease underestimates the initial accelerated telomere loss during adolescence ( the average initial telomere length in newborns is shown by the dark grey rectangle ) . In contrast , model 2 ( black line ) predicts a logarithmic decay of the average telomere length with age and is able to capture the increased loss of telomere length during adolescence , as well as the approximately linear decrease in adults . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 010 Our model suggests that the increased loss of telomere repeats in the first years of human life is a consequence of an expanding stem cell population . This expansion is combined with a reduction in proliferation rates of single stem cells . The loss of telomere repeats during cell replication has a more pronounced impact on the average telomere length within a small cell population and diminishes in large stem cell populations . This explains the increased loss of telomeric repeats during adolescence ( see Figure 4 ) naturally as a consequence of growth by an expanding stem cell population . Similarly , a sudden accelerated loss of telomeric repeats in aged individuals could point towards an insufficient stem cell self-renewal . This might provide a promising direction for further investigations with an extended data set of sufficiently high resolution in aged individuals . Our analytical model is consistent with population wide telomere length data . It shows that symmetric stem cell self-renewals are more frequent in adolescence and their effect on the dynamics of average telomere length reduces with age . However , how robust are our conclusions under variation of model parameters or a change of cell proliferation properties with age ? One possibility to address these problems is the implementation of Bayesian inference methods ( Dempster , 1968 ) . In a nutshell , such methods draw a random set of model parameters either from an uninformed ( objective ) or informed ( subjective ) prior distribution and produce independent realizations of the model . These realizations are compared to some ( appropriate ) data of interest and fits with a predefined statistic significance are retained while unsatisfactory realizations are rejected . Originally developed for phylogenetic tree reconstruction , such methods are increasingly used in other applications ( Marjoram and Tavaré , 2006 ) . Bayesian inference methods allow to quantify the uncertainty in an analysis by providing posterior distributions of model parameters . In the following we implement an Approximate Bayesian Computation ( ABC ) rejection sampling framework ( Csilléry et al . , 2010 ) on the data presented in Figure 3 . We derive posterior distributions for our three free model parameters , the initial telomere length c , the relative decrease of telomere length per time Δcr/N0 and the probability of symmetric stem cell divisions p . We draw these variables independently from uniform ( uninformed ) distributions and test 109 independent realizations of our mathematical model 1 and model 2 . We seek parameter regimes that maximize the coefficient of determination R2 between Equation 3 ( model 1 ) or Equation 4 ( model 2 ) and the average telomere length presented in Figure 3 . We discard any parameter combination below a threshold . We perform the same analysis independently on the data set of granulocytes and lymphocytes . In both cases , we find localized posterior parameter distributions . For lymphocytes , parameters peak at Δcr/N0=0 . 071±0 . 005kbp/year , c=10 . 41 ± 0 . 3kbp and p=0 . 32 ± 0 . 2 , see Figure 5c–e . Only a small parameter range explains the exact patterns of telomere shortening . We find approximately 70% of stem cell divisions are asymmetric and 30% are symmetric self-renewals . This stochastic approach confirms the results of the non-linear model fits using a standard maximum likelihood approach that were discussed in the previous section , but provides further information on the distribution of our parameters . The previous analysis assumes a fixed set of parameters for the dynamics of telomere shortening for all ages . In principle , these parameters could also change with age . To see if we can identify ages with different stem cell proliferation parameters , we investigated a third model that allows for successive phases of stem cell dynamics with independent parameter sets for each phase . We consider an additional parameter tT , which corresponds to a transition time . We perform the above Bayesian approach independently for each random partition of the data set . This approach suggests at most two separate phases , with a transition between the 6th and 7th year of life for lymphocytes , see Figure 5f–i , and a transition between the 10th and 15th years of life for granulocytes , see Figure 5j–m . In infants and the first years of life , the probability of stem cell self-renewal shows a significant variance ( Figure 5 ) . However , the data resolution is insufficient for this short time window to provide reliable parameter estimates . The probability of symmetric stem cell self-renewal in adults however is in the the range of p∈ ( 0 , 0 . 2 ) . This is lower as was predicted by the regression analysis across all ages . This suggests a reduction in the self-renewal probability of stem cells after adolescence and points towards an either slower growing or constant stem cell population in adults . This may reflect selection for an optimal stem cell population size to minimize the risk of cancer initiation as suggested in theoretical studies before ( Michor et al . , 2003 ) . 10 . 7554/eLife . 08687 . 011Figure 5 . Posterior distributions of model parameters from Approximate Bayesian Computation ( ABC ) . ( a , b ) Model fit for only asymmetric stem cell divisions ( model 1 ) to the data of average telomere length on the population level . The expected telomere length decreases linearly and two free model parameters , i . e . initial telomere length and stem cell turn over rate are estimated . ( c–e ) ABC with symmetric and asymmetric stem cell divisions ( model 2 ) . In this case one additional free parameter ( probability of symmetric stem cell divisions ) can be estimated . ( f–i ) ABC for a two phase extension of the model inferred from population wide data of lymphocytes , panels ( j–m ) show the same analysis for granulocytes . A likelihood based model selection favours model 2 and rejects model 1 as well as the multiphase model as more likely explanations for the observed data . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 011 Next , we aimed to test which of the three models explains the data best , considering the complexity of the models . We therefore utilise the likelihood estimates of the former subsection and perform a model selection based on the Akaike information criterion ( AIC ) ( Burnham , 2004 ) . Model 1 scores with an AIC of 2550 , model 2 has an AIC of 2328 and a multiphase model with a minimum of 7 parameters yields an AIC of 2361 . The AIC is minimized by model 2 . Based on this approach , model 1 as well as a multiphase model can be rejected as more likely explanations for the telomere length shortening presented in Figure 3 ( given the above numbers and according to standard procedures , the relative likelihood of model 1 to better explain the data compared to model 2 is assumed to be p≈10-48 , the relative likelihood of the multi-phase model to better explain the data compared to model 2 is assumed to be p≈10-8 ) . This selection is robust under the choice of different statistical methods . For example , a BIC approach selects the models in the same order . The actual stem cell population sizes and their dynamics do not only vary with age , but also between individuals . This has immediate consequences on the susceptibility of individuals towards certain diseases ( Calado and Young , 2009; Brümmendorf and Balabanov , 2006 ) and could potentially be used in individualised treatment strategies . Our model describes the telomere length distributions in individuals and quantifies three parameters , i . e . initial telomere length , increase of stem cell pool size and stem cell replication rates of an individual from a single tissue sample . We therefore extended our experimental protocols to further test our theoretical results . First , we measured single telomere signals of peripheral blood sorted for lymphocytes in 28 individuals and sorted for granulocytes in 10 individuals by quantitative confocal FISH in addition to the average telomere length that is provided by flow FISH . Second , we investigated the telomere length distribution in paraffin-embedded bone marrow sections of an additional cohort of 28 healthy individuals using quantitative confocal FISH ( Beier , 2005 ) , see Figure 2 . We compare our general telomere length distribution that allows for any ratio of symmetric and asymmetric stem cell divisions ( model 2 ) to the data set of all 66 individuals . Cases of four representative individuals are shown in Figure 6 . All cases can be found in Figure 6—figure supplements 1–3 and all individual cell proliferation properties as well as quality of fits are summarised in Supplementary file 1 . The average telomere length of these 66 distributions are shown as open symbols in Figure 3 . 10 . 7554/eLife . 08687 . 012Figure 6 . Telomere length distributions of granulocytes for four representative individuals . Telomere length distributions within the nucleus of individual cells are measured once in single individuals ( symbols ) . This data is fitted with our model 2 ( black line , see Equation S29 for details ) , leading to estimates for the parameters of the theoretical distribution . These parameters can be used to extrapolate the distribution to any other age ( gray lines ) . The dashed line shows the prediction for the maximum of the distribution ( Equation S18 ) . Telomere length distributions differ between individuals and change in different patterns , depending on the exact proliferation parameters in individuals . Additional cases are shown the Figure 6—figure supplements 1–3 . A summary of all fitting parameters can be found in Supplementary file 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 01210 . 7554/eLife . 08687 . 013Figure 6—figure supplement 1 . Nonlinear fits of the expected telomere length distribution to telomere length distributions of granulocytes in peripheral blood of 10 healthy donors . For experimental details see Materials and methods , for details on the nonlinear fitting and individual parameters estimates as well as quality of fits , see Materials and methods and Supplementary file 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 01310 . 7554/eLife . 08687 . 014Figure 6—figure supplement 2 . Nonlinear fits of the expected telomere length distribution to telomere length distributions of lymphocytes in peripheral blood of 28 healthy donors . For experimental details see Materials and methods , for details on the nonlinear fitting and individual parameters estimates as well as quality of fits , see Materials and methods and Supplementary file 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 01410 . 7554/eLife . 08687 . 015Figure 6—figure supplement 3 . Nonlinear fits of the expected telomere length distribution to telomere length distributions in bone marrow biopsies of 28 patients with diagnosed M . Hodgkin without bone marrow affection . For experimental details see Materials and methods , for details on the nonlinear fitting and individual parameters estimates as well as quality of fits , see Materials and methods and Supplementary file 1C . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 015 The fits of our calculated distribution ( see Equation S15 for the distribution and Equation S29 for details on the fitting procedure ) reveal substantial differences in initial telomere length , increase of stem cell pool size and stem cell replication rates between the 66 individuals , but also between granulocytes , lymphocytes and bone marrow samples . We find a low probability of symmetric self-renewal ( p between 0 . 005 to 0 . 03 per cell division ) in all individual samples . This agrees with our results on the average telomere length shortening in adults at the population level and supports our observation of an approximately maintained active stem cell number in individuals after adolescence . Also the average telomere loss per year varies between individuals and ranges from 18 bp/year to 110 bp/year . However , the averages of all individual parameter sets agree with the estimated proliferation properties inferred from the population wide data of telomere length . We find differences between individual samples of lymphocytes and granulocytes . While the loss of telomeric repeats slows down with age in granulocytes , it slightly accelerates in lymphocytes , see Figure 7 . These cells represent the myeloid and lymphoid lineage respectively . In our model , such a reduced rate of telomere loss can be explained with an increased reservoir of myeloid specific stem and progenitor cells and is in agreement with a skewed differentiation potential towards the myeloid lineage of aged hematopoietic stem cells ( Geiger et al . , 2013 ) . 10 . 7554/eLife . 08687 . 016Figure 7 . Rate of telomere loss in 66 individuals . Shown is the rate of telomeric shortening ( bp/year ) of granulocytes ( circles ) , lymphocytes ( triangle ) and bone marrow sections ( rectangle ) , inferred from telomere length distributions of 66 different individuals ( see Figure 5 and supplemental figures and Supplementary file 1 for a summary of all parameters ) . Differences between individuals are large , but the average telomere shortening rate conforms to parameter estimates of population wide data of telomere length , see for example Figure 5 . Cells in the bone marrow show a lower proliferation rate and consequently the rate of telomere loss is reduced ( gray dotted line ) . The rate of telomere loss decreases with age in granulocytes ( −0 . 78 bp/year , dark red line ) and in bone marrow sections ( −0 . 36 bp/year , grey dotted line ) , but increases in lymphocytes ( +0 . 27 bp/year , dark green dashed line ) . This observation agrees with a skewed differentiation potential towards the myeloid lineage of aged hematopoietic stem cells ( Geiger et al . , 2013 ) . The lines are only meant to represent a trend of increase or decrease with age . The change with age is most probably not linear . DOI: http://dx . doi . org/10 . 7554/eLife . 08687 . 016 Our knowledge about the dynamics of tissue specific stem cells comes mostly from lineage tracing experiments in transgenic mouse models . They provided insights into many aspects of tissue formation and maintenance , e . g . the intestinal crypt , but also the hematopoietic system ( Busch et al . , 2015; Sun et al . , 2014; Itzkovitz et al . , 2012 ) . However , there is variation between different transgenic mouse models and their significance for human stem cell properties remains a challenging question . In some cases , clonal lineages can be traced by naturally occurring somatic mutations , e . g . particular mtDNA mutations in human intestinal crypts ( Baker et al . , 2014 ) . However , the in vivo dynamic properties of human hematopoietic stem cells remain poorly characterized . Here , we have utilized telomere length distributions of hematopoietic cells as a biomarker that contains information about the proliferation history of cells . We developed a mathematical model that allows us to infer dynamic properties of stem cell populations from data of telomere length distributions . These properties were analyzed in different cell types , e . g . lymphocytes , granulocytes and bone marrow sections of individuals of different ages . These calculated distributions describe the change of telomere length within the human population . The expected changes with age were confirmed in a representative group of 356 healthy individuals and the conclusions are consistent with our individualized parameter estimations . The population wide data of average telomere length reveals different stem cell properties in adolescence and adulthood . Telomere length decrease is logarithmic and occurs at a faster rate during adolescence , suggesting a stem cell pool expansion in the first years of human life compatible with growth . This decrease becomes almost linear in adults and is in line with an approximately constant stem cell population . It is an interesting question why the number of stem cells would reach a certain targeted size . This could be simply because of spatial constrains in the bone marrow . Yet , from an evolutionary perspective , intermediate stem cell pool sizes were suggested to minimize the risk of cancer initiation ( Rodriguez-Brenes et al . , 2013; Michor et al . , 2003 ) . Such an optimization requires feedback signals that ensures the maintenance of an intermediate sized stem cell population , feedback signals that might be prone to ( epi ) genetic change and potentially are involved in cancer and ageing . It is still a debated question if stem cells in mammals are maintained by predominantly asymmetric divisions , or alternatively by a population strategy of balanced symmetric self-renewal and symmetric differentiation . While the former strategy can be implemented on the single cell level , the latter strategy would require further feedback signals . From a modelling perspective , a population strategy of symmetric self-renewal and symmetric differentiation was suggested to minimize the clonal load within a stem cell population ( Shahriyari et al . , 2013 ) . On the other hand , experimental findings seem to point towards predominantly asymmetric divisions , but this might also differ across tissues ( Morrison and Kimble , 2006 ) . In our model , the stem cell pool is maintained by asymmetric cell divisions . A balance of symmetric and asymmetric cell divisions would on average result in the same telomere length dynamics and thus would be indistinguishable from asymmetric divisions on the population level , only the interpretation of p , the probability of symmetric self-renewal would change in this case . Yet , the variance of the distribution would be expected to increase under the presence of symmetric differentiation and symmetric self-renewal . However , likely this effect is weak compared to the measurement related noise of telomere length . Our method quantifies the parameters of telomere dynamics from a single blood sample or paraffin-embedded tissue samples of an individual . It is independent of any particular tissue organization and thus can be applied , in principle , to any tissue . This general method will be of particular interest to distinguish stem cell dynamics in healthy and sick individuals . We expect characteristic changes in telomere length distributions in certain ( hematopoietic ) stem cell disorders such as chronic leukemias ( Braig et al . , 2014 ) and bone marrow failure syndromes ( Calado and Young , 2009; Beier , 2005 ) . Therefore , our model can serve as a tool to infer stem cell dynamics in vivo retrospectively and prospectively from a single tissue sample . Such an approach can not only increase our understanding of disease dynamics but may also contribute to personalized disease diagnosis and prognosis in the future . Peripheral blood of 309 healthy blood donors was obtained from the blood donor bank in Aachen . Q-FISH of peripheral blood cytospins was performed on 28 healthy blood samples . 47 cord blood and blood samples from healthy children and adolescents were obtained from the Department of Pediatrics and Neonatology of the University Hospital of Aachen . Bone marrow biopsies of 28 patients with diagnosed Hodgkin lymphoma without bone marrow involvement were used for bone marrow analysis . All samples were taken with informed consent and according to the guidelines of the ethics committees at University Hospital Aachen . The Flow-FISH technique provides the mean telomere length per nucleus . Flow-FISH was carried out according to previously published protocols ( Aubert et al . , 2012; Baerlocher et al . , 2006; Weidner et al . , 2014; Beier et al . , 2012 ) . Briefly , after osmotic lysis of erythrocytes with ammonium chloride , white blood cells were mixed with cow thymocytes . Cells were hybridized with FITC labeled , telomere specific ( CCCTAA ) 3- peptide nucleic acid ( PNA ) probe ( Panagene ) and DNA was counterstained with LDS 751 ( Sigma ) . FACS analysis was carried out on Navios or FC-500 ( both Beckman Coulter ) . Thymocytes , lymphocytes and granulocytes subsets were identified based on LDS571 staining and forward scatter . Mean telomere length was calculated by subtracting the unstained autofluorescence value of the respective lymphocyte , granulocyte or thymocyte subpopulation . Cow thymocytes with a determined telomere length were used as an internal control to convert telomere length in kilobase ( kb ) . All measurements were carried out in triplicate . Q-FISH offers the possibility to analyze the distribution pattern of individual telomeres . For cytospins of peripheral blood cells , erythrocytes were lysed using ammonium chloride ( Stem cell Technologies , Vancouver , British Columbia , Canada ) and 50 , 000 cells were centrifuged for cytospin . Cells were fixed with 70% ethanol solution for 30 seconds and air dried for 15 min . Bone marrow sections were deparaffinized with xylol and rehydrated with ethanol following standard protocols . Deparaffinized bone marrow tissue sections , metaphases and peripheral blood cells were processed following previously published protocols ( Beier et al . , 2015; Varela et al . , 2011; Zijlmans et al . , 1997 ) . After initial washing with PBS , slides were fixed in formaldehyde ( Sigma ) ( 4% ) in PBS for 2 min . Slides were further washed ( three times for 5 min ) with PBS followed by dehydration with ethanol and air drying for 30 min . Hybridization mixture containing 70% formamide ( Sigma ) , 0 . 5% Magnesium chloride ( Sigma ) , 0 . 25% ( wt/vol ) blocking reagent ( Boeringer ) 0 . 3 μg/ml Cy-3-conjugated ( C3TA2 ) 3 peptide nucleic acid probe ( Pnagene ) , in 10 mM Tris ( pH 7 . 2 , Sigma ) was added to the slide . After adding a coverslip; DNA was denatured for 3 min at 85°C . Hybridization was carried out for 2 h at room temperature . After washing the slides twice with 70% formamide/10 mM Tris ( pH 7 . 2 ) /0 . 1% bovine serum albumin ( BSA ) , slides were washed again ( three times for 5 min ) with 0 . 05 M Tris/0 . 15 M NaCl ( pH 7 . 5 ) containing 0 . 05% Tween-20 . After dehydration with ethanol slides were air dried and stained with PBS containing 0 . 1 ng/ml of 4’-6-diamidino-2-phenylindole ( DAPI ) for 5 min . After mounting the cells ( Vectashield , Vectorlabs ) , a coverslip was added . Confocal microscopy analysis was carried out at a Leica TCS-sp5 confocal microscope ( Leica ) . Images were acquired at 63x magnification and 1 . 5-2 . 0 digital zoom . Multi-tracking mode was used to acquire images . Stacks of DAPI and Cy3 staining were taken with a step size of 1 μm . Peripheral blood cells and bone marrows were captured including five steps ( z-range 4 μm ) . Maximum projection of the images was carried out and Definiens XD 1 . 5 image analysis software ( Definiens GmbH ) was used for quantitative image analysis . Nucleus and telomere detection was carried out based on DAPI and Cy3 intensity patterns . A valid image analysis was assumed in case of a correct detection of 90% of all visible telomeres . All image analysis was carried out single-blinded . Individual telomere signals were calculated after subtraction of the mean background value per detected nucleus . For bone marrow section and peripheral blood cells , values of all detected telomeres were used for analysis . Paraffin embedded lymphocytes of three healthy donors and granulocytes of a patient with chronic myeloid leukemia with a determined telomere length were used as controls for bone marrow biopsies . Linear regression of the control cells was carried out to convert telomere length from arbitrary units to kb . Telomere length in kb of the Q-FISH analysis of peripheral blood cells was calculated based on the linear regression of the corresponding Flow-FISH values . We assume a finite number of 1+c accessible telomere states of stem cells , where each state i contains cells of equal average telomere length . Initially , N0 cells are in state 0 and cells will progressively enter downstream states after cell divisions . An asymmetric division of a cell in state i leads to one more differentiated cell ( more committed within a hierarchically tissue organization ) and one stem cell . The committed ( progenitor ) cell leaves the pool of stem cells and does not further contribute to dynamics in the stem cell population . The second cell keeps the stem cell properties and enters state i+1 , reflecting the shortening of its telomeres by a length of Δc . Similarly , a symmetric cell division results in two stem cells , both entering the next subsequent state . In our model , stem cells divide symmetrically with probability p and asymmetrically with probability 1-p , respectively . A cell in state c enters cell cycle arrest and cannot reach subsequent states - the next proliferating cell is randomly chosen amongst all cells not yet in state c . The number of cells in each state i follows a Poisson distribution ( S22 ) N ( i ) ( t ) =N0i ! rtN0ie-rtN0 in the case of only asymmetric stem cell divisions , see Equation S2 for details . We introduce x=rtN0 , and upon normalisation ( S22 ) becomes ( S23 ) N ( i ) ( x ) ∝xii ! e-x , where x is a Poisson distributed variable . For x sufficiently large , this random variable is well described by a normal distribution and we have x ∝ Normal distribution . If we allow for symmetric cell divisions , cells in state i followed a generalised Poisson distribution ( S24 ) Np ( i ) ( t ) =N0i ! 1+ppilni rptN0+1rptN0+1-p , see Equation S15 for details . Choosing y=rptN0+1 and neglecting normalisation factors we can write ( S25 ) Np ( i ) ( y ) ∝1i ! lni ( y ) y-p If we change variables again and choose y=ex , Equation S25 becomes ( S26 ) Np ( i ) ( y=ex ) ∝1i ! lniexex-p=xii ! e-xp∝N ( i ) ( x ) . As x=rtN0 is approximately normally distributed , and y=ex , y=rptN0+1 follows a Log-normal distribution . We implement Approximate Bayesian Computation ( ABC ) rejection samplings to derive posterior parameter distributions for the predicted average telomere length under asymmetric ( model 1 , Equation S10 ) and combined symmetric and asymmetric ( model 2 , Equation S20 ) cell proliferations respectively . Utilizing Equation S10 , we have to infer two parameters: ( i ) the average decrease of telomere length per time r/N0 and ( ii ) the initial telomere length c . In the case of Equation S20 a third variable has to be determined: ( iii ) the probability of symmetric cell divisions p . We draw these variables independently from uniform distributions ( prior ) with ranges r/No ∈ [0 , 0 . 2]kbpyear , c∈7 , 15kbp and p∈0 , 1 and produce 5×108 independent realizations of Equation S10 , S20 . We calculate the coefficient of determination R2 between each of these realizations and the average telomere length from a data set of 356 healthy individuals ( see for example Figure 1 in the main text ) via ( S27 ) R2=1-∑iEcti-yti∑iy¯-yti . Here , yti denotes , the measured telomere length of an individual with age ti , y¯ is the average measured telomere length of the population and Ecti the value of a single realization of Equation S10 or Equation S20 at time ti given the random set of parameter values . We seek parameter regimes that maximize R2 and discard any parameter combination below a certain threshold .	Human cells die off regularly due to normal wear and tear , aging or injury . To replace these cells , humans maintain pockets of tissue specific stem cells that can develop into one of several different types of specialized cell . For example , stem cells in the bone marrow can develop into red blood cells , white blood cells or any of the other blood cell types . Unavoidably , over the course of a lifetime stem cells accumulate mutations that may cause them to become cancerous . Researchers have learned a lot about stem cells by studying them under laboratory conditions . However , these studies cannot answer all the questions we have about human stem cells . As a result , human studies are needed; but frequently taking samples of stem cells from humans to assess them is impossible for numerous reasons , most importantly it is invasive and potentially harmful . Instead , researchers are looking for indirect ways to measure how stem cells grow . Each time a cell divides , the protective ends of a chromosome – known as telomeres – get shorter . Now , Werner , Beier et al . have developed a mathematical model to assess human stem cell growth based on the length of the cells’ telomeres . This model can gauge the growth patterns of the stem cell populations in an individual based on a sample taken from a single tissue . Werner , Beier et al . tested the model using telomere measurements from blood and bone marrow samples taken from 356 healthy people of different ages . The results suggest that the stem cell population that gives rise to blood cells ( the hematopoietic stem cells ) increases in size during childhood and adolescence , but levels off during adulthood . The model also revealed that patterns of stem cell growth vary among individuals . Further studies of telomere length differences may help scientists identify the abnormal ( stem cell-like ) growth patterns associated with diseases like cancer .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine", "computational", "and", "systems", "biology" ]	2015	Reconstructing the in vivo dynamics of hematopoietic stem cells from telomere length distributions
Colorectal cancer ( CRC ) is a major cause of human death . Mortality is primarily due to metastatic organ colonization , with the liver being the main organ affected . We modeled metastatic CRC ( mCRC ) liver colonization using patient-derived primary and metastatic tumor xenografts ( PDX ) . Such PDX modeling predicted patient survival outcomes . In vivo selection of multiple PDXs for enhanced metastatic colonization capacity upregulated the gluconeogenic enzyme PCK1 , which enhanced liver metastatic growth by driving pyrimidine nucleotide biosynthesis under hypoxia . Consistently , highly metastatic tumors upregulated multiple pyrimidine biosynthesis intermediary metabolites . Therapeutic inhibition of the pyrimidine biosynthetic enzyme DHODH with leflunomide substantially impaired CRC liver metastatic colonization and hypoxic growth . Our findings provide a potential mechanistic basis for the epidemiologic association of anti-gluconeogenic drugs with improved CRC metastasis outcomes , reveal the exploitation of a gluconeogenesis enzyme for pyrimidine biosynthesis under hypoxia , and implicate DHODH and PCK1 as metabolic therapeutic targets in CRC metastatic progression . Colorectal cancer ( CRC ) is a leading global cause of cancer-related death . An estimated 145 , 000 Americans will be diagnosed with CRC and roughly 51 , 000 will die of it in 2019 ( National Cancer Institute , 2019 ) . The majority of these deaths are due to distant metastatic disease with the liver being the most common distal organ colonized ( National Cancer Institute , 2019 ) . While the prognosis of patients diagnosed with non-metastatic or locoregional disease is relatively better , a significant fraction of such patients will nonetheless experience subsequent metastatic progression . A key clinical need is identifying which patients will develop metastatic disease since few prognostic indicators of disease progression exist . Additionally , novel targeted therapies for the prevention and treatment of metastatic CRC are major needs . While cell lines established from human CRC have provided important insights into the biology of CRC , tissue culture drift , adaptation , and evolution can restrict their relationship to the pathophysiology of patient tumors ( Gillet et al . , 2013 ) . Patient-derived xenograft ( PDX ) models , which allow for the growth of patient tumor samples in immunodeficient mice , can capture the endogenous diversity within a tumor as well as the patient-to-patient variability of metastatic cancer . Prior work has revealed that breast cancer , melanoma and non-small cell lung cancer PDX modeling can predict human clinical outcomes ( Quintana et al . , 2012; John et al . , 2011 ) . These models have revealed the potential of individualized mice models for prognostication and tailoring of therapies . Past CRC PDX models were subcutaneous implantations ( Cho et al . , 2014; Gao et al . , 2015; Oh et al . , 2015 ) . Such subcutaneous PDX tumors are useful for tumor growth studies , but do not metastasize and are not exposed to the pathophysiologically relevant and restrictive conditions of the hepatic microenvironment . While orthotopic PDX implantation serves as a good model for studying metastasis in melanoma and breast cancer , this approach has limited feasibility in CRC , as the orthotopic tumor kills the host due to the obstructive dimensions that implanted tumors reach prior to liver metastatic colonization ( Quintana et al . , 2012 ) . Thus , a clinically relevant PDX model of CRC that recapitulates the critical process of liver metastatic colonization is an important need . Beyond predicting clinical outcomes and therapeutic responses , PDX models of mCRC could allow for identification of molecules that contribute to metastatic colonization through the use of in vivo selection . This process subjects a parental population ( e . g . cancer cells , bacteria , yeast ) to a severe physiologic bottleneck such that only those cells with the requisite gene expression states survive ( Fidler , 1973 ) . Such selection is repeated iteratively to enrich for cells that are best adapted to the new microenvironment . Molecular profiling can then reveal molecular alterations that enable the selected population to colonize the new environment ( Minn et al . , 2005; Kang et al . , 2003; Tavazoie et al . , 2008; Pencheva et al . , 2012; Loo et al . , 2015 ) . By engrafting patient-derived CRC primary and metastatic tumors of diverse mutational backgrounds , we observed that subcutaneous tumor engraftment efficiency and liver colonization capacity , but not subcutaneous tumor growth rate , was associated with patient overall survival . By performing liver-specific in vivo selection with CRC PDXs , we enriched for cells optimized for growth in the liver microenvironment . Gene expression analysis revealed the gluconeogenesis enzyme PCK1 to be a robust driver of liver metastatic colonization that is over-expressed in metastatic CRC . Mechanistic studies revealed that PCK1 enhances pyrimidine nucleotide biosynthesis , which enables cancer cell growth in the context of hypoxia—a key feature of the liver microenvironment . Moreover , highly metastatic CRC PDXs contained higher levels of intermediary metabolite precursors for pyrimidine nucleotide biosynthesis . Consistent with these observations , molecular and pharmacologic inhibition of PCK1 or the pyrimidine biosynthetic gene DHODH inhibited CRC liver metastatic colonization . In order to establish a PDX model of CRC liver metastatic colonization , a small sample of CRC tissue , taken either from a primary or metastatic site , was dissociated and injected subcutaneously into the flanks of NOD . Cg-Prkdcscid Il2rgtm1Wjl/SzJ ( Nod-Scid-Gamma; NSG ) mice within 2 hr of surgical resection at MSKCC . Thirty-one subjects provided 40 tumor samples; 48 . 3% of the subjects’ samples engrafted ( Figure 1 ) . The majority of subjects in this study were classified as Stage IV CRC according to the American Joint Committee on Cancer ( AJCC ) . However , AJCC stage was not associated with increased xenograft engraftment ( p=0 . 35; χ2 test ) . The engraftment rates for tumor tissues that originated from the colon and the liver were similar ( 40% vs 37 . 5%; Figure 1 ) . Most subjects had undergone chemotherapy prior to surgical resection of metastases ( 67 . 7% ) . When categorizing tumors by commonly tested clinical mutations ( KRAS , high microsatellite instability ( MSI-H ) , NRAS , BRAF , PIK3CA , and none ) , MSI-H tumors exhibited the highest engraftment rates ( 83 . 3% ) , while tumors lacking these commonly tested for clinical mutations exhibited the lowest engraftment rates ( 18 . 2% ) . We found that subcutaneous tumor engraftment was associated with worse patient survival ( p=0 . 045; Figure 2A ) . The time from subcutaneous tumor implantation to tumor harvest ranged from 35 to 88 days ( Figure 2B ) . Among those CRC tumors that did grow subcutaneously , the time required to reach the pre-determined tumor size ( 1000 mm3 ) was not significantly associated with patient survival ( p=0 . 27; Figure 2C ) . When the estimated subcutaneous tumor volumes reached 1000 mm3 , the mice were euthanized , and the tumors were removed . For each sample , a portion of the xenografted tumor was set aside for cryopreservation , and the rest of the tumor was dissociated into a single cell suspension for portal circulation injection via the spleen . Portal circulation injection has been demonstrated to be a reliable means of establishing liver growth via hematogenous spread of CRC cells , simulating the entry of cells into the portal circulation which is typical of clinical CRC progression . After injection of cells , we observed the mice until they were deemed ill by increased abdominal girth , slow movement , and pale footpads , at which point we proceeded to euthanization and tumor extractions . Successful liver metastatic colonization was achieved upon injection of 15/17 patient samples . The time to mouse sacrifice for the CRC patient-derived liver xenografts ranged from 51 to 407 days ( Figure 2D ) and did not correlate with subcutaneous tumor growth rates ( R2 = 0 . 046 , p=0 . 50; Figure 2E ) . The mCRC liver PDXs fell into two biologically distinct groups based on their growth rates: one set grew quickly , requiring mouse euthanasia within three months of implantation; the other set grew more slowly , requiring animal sacrifice after 6 months , or even 1-year , post-engraftment ( Figure 2—figure supplement 1A ) . Importantly , these two groups of PDXs exhibited similar growth rates when implanted subcutaneously ( p=0 . 09; Figure 2—figure supplement 1B ) . This suggests distinct selective pressures existing in the liver relative to the subcutaneous microenvironment . We found that the liver colonization model mimicked clinical outcomes , as patients whose xenografts rapidly colonized mouse livers fared poorly relative to patients whose xenografts colonized the liver slowly or not at all ( p=0 . 031; Figure 2F ) . Taken together , these results establish that CRC liver metastasis PDX modeling described above is prognostic of clinical survival outcome . A key objection to cell line xenografts is that the histology of animal tumors is often not representative of clinical sample histology . Contrary to this , we observed that both subcutaneous and liver engrafted tumors re-capitulated the architecture of the primary tumor from which they were derived ( Figure 2—figure supplement 2 ) . CLR4 was established from a poorly differentiated liver metastatic colon adenocarcinoma; it remained poorly differentiated in both the subcutaneous and liver xenografts ( Figure 2—figure supplement 2 ) . Similarly , CLR32 and CLR28 were derived from moderate-to-well differentiated primary colon and peritoneal metastatic adenocarcinomas and retained moderate-to-well differentiated histology when passaged subcutaneously and hepatically ( Figure 2—figure supplement 2 ) . We next performed liver-directed in vivo selection through iterative splenic injections of four distinct CRC PDXs with varying mutational and metastatic backgrounds to obtain derivatives with increased capacity for liver colonization and growth ( Figure 3 ) . Tumors were only passaged in vivo without the use of in vitro culture . When a mouse bearing a liver colonization graft had met its pre-determined endpoint , it was euthanized , and the liver tumor was removed and dissociated into a single cell suspension in a similar manner to that of the subcutaneous tumors described above . Dissociated cells were subsequently injected into the spleen of another mouse to generate a second-generation liver metastatic derivative . This process was repeated multiple times to create a highly metastatic derivative for each of the four distinct CRC PDXs . The number of rounds of in vivo selection varied between tumor samples ( range: 5–13 ) and in general tended to represent the number of rounds required to plateau enhanced metastatic colonization capacity . In the last round of in vivo selection , a cohort of mice was subjected to portal circulation injection with either the parental CRC PDX cells or the liver-metastatic derivative CRC PDX cells in order to assess the relative liver colonization capacities among the liver-metastatic derivatives . In each of the four CRC PDX comparisons , the in vivo-selected CRC PDX liver metastatic derivatives colonized the mouse liver more efficiently than their parental counterparts ( Figure 3 ) . The two extreme isogenic populations of each patient , the parental CRC PDX and its liver-metastatic derivative , were then subjected to transcriptomic and metabolite profiling as described below to identify candidate regulators of metastatic colonization . We first sought to identify candidate mCRC liver colonization promoters through mRNA sequencing and differential gene expression analyses from parental CRC PDXs ( anatomical locations included subcutaneous graft , cecal graft , or first-generation liver graft ) and last-generation liver-metastatic CRC PDXs . Comparisons between liver metastatic derivatives and their parental counterparts allowed for isogenic comparisons . A phylogenetic tree using complete clustering and Euclidian distance function based upon the gene expression profiles demonstrated that isogenic pairs mostly clustered together with one exception ( Figure 3—figure supplement 1A ) . Using gene set enrichment analysis ( GSEA ) ( Subramanian et al . , 2005 ) , we interrogated each of the four pairs of tumors individually and as a composite to identify cancer-related pathways and signatures that were significantly altered in liver metastatic derivatives compared to their isogenic parental xenografts . The hypoxia signature was found to be upregulated in all the liver metastatic derivatives individually and in the composite , where it was the most significantly enriched gene signature ( normalized enrichment score ( NES ) = 2 . 12 , q-value <0 . 001; Figure 3—figure supplement 1B ) . Upregulation of hypoxia genes in the liver metastatic derivatives is consistent with previous reports demonstrating that hypoxia is an important feature of the liver metastatic microenvironment ( Loo et al . , 2015; Nguyen et al . , 2016 ) . With each CRC PDX pair , we identified upregulated genes in each liver-metastatic derivative compared to its parental counterpart through a generalized linear model . The number of upregulated genes ( p<0 . 05 ) in the liver-metastatic derivatives ranged from 200 ( CLR28 ) to 345 ( CLR27 ) out of a possible list of more than 12 , 000 genes . Fisher’s combined probability test was used to construct a list of candidate liver colonization promoting genes that were statistically significantly upregulated across the four pairs of CRC PDXs with an effect size of greater than 1 . 5 log2 fold change ( logFC ) . Using this approach , we identified 24 highly upregulated genes in the liver metastatic derivatives ( Figure 3—figure supplement 1C ) , with the 10 most highly upregulated genes annotated on the volcano plot ( Figure 4A ) . Interestingly , two of the top ten upregulated genes ( IFITM1 , and CKB ) have been previously implicated as promoters of CRC metastasis ( Loo et al . , 2015; Yu et al . , 2015 ) . The most common ‘druggable’ targets for cancer therapeutics are enzymes and cell-surface receptors . In the list of candidate genes , three were enzymes ( ACSL6 , CKB and PCK1 ) and one was a cell-surface receptor ( CDHR1 ) . One of the genes on this list , creatine kinase-brain ( CKB ) , was identified by us in a prior study using established CRC cell lines and shown to act extracellularly to generate phosphocreatine , which is subsequently transported into the cell as a metabolic energetic source for ATP generation in the hypoxic microenvironment of the liver ( Loo et al . , 2015 ) . Of the remaining three enzymes on our list , we focused on PCK1 ( phosphoenolpyruvate carboxykinase 1 ) given the availability of a pharmacological inhibitor and its heightened expression in normal liver ( Uhlén et al . , 2015 ) , suggesting potential mimicry of hepatocytes by CRC cells during adaptation to the liver microenvironment . We next investigated whether our 24-gene CRC liver colonization signature was enriched in liver metastases from patients with CRC by querying a publicly available dataset in which transcriptomes of primary CRC tumors and liver metastases were profiled . Of the 24 genes , 22 were represented in this previously published dataset ( Sheffer et al . , 2009 ) . We binned the patient data based on differential gene expression in primary CRC tumors versus the CRC liver metastatic tumors . The upregulated genes were significantly enriched ( p=0 . 007 ) in the bin with the most upregulated genes in CRC liver metastases ( Figure 4B ) ( Goodarzi et al . , 2009 ) , supporting the clinical relevance of our in vivo-selected CRC PDX liver colonization mouse model . In further support of the clinical relevance of our findings , we found that the gene expression upregulation in our metastatic CRC system significantly correlated ( rho = 0 . 39 , p=0 . 047 ) with the gene expression upregulation in human liver CRC metastases relative to CRC primary tumors ( Figure 4C ) . Interestingly , PCK1 was highly upregulated in human CRC liver metastases relative to primary tumors . QPCR quantification confirmed PCK1 up-regulation in liver metastatic derivatives relative to isogenic parental counterparts ( Figure 4D ) . We analyzed other publicly available CRC gene expression datasets and consistently observed PCK1 to be significantly upregulated ( p=0 . 01 , Student’s t-test; Figure 4E ) in CRC liver metastases relative to primary tumors ( Figure 4E–G ) ( Kim et al . , 2014; Stange et al . , 2010; Ki et al . , 2007 ) . Additionally , PCK1 was upregulated ( p=0 . 01; Figure 3F , p<0 . 0001; Figure 4G ) in CRC liver metastases in datasets containing only paired CRC primary tumors and CRC liver metastases obtained from the same patients ( Figure 4F–G ) . We next performed functional in vivo studies using human CRC cell lines in which PCK1 expression was modulated through stable gene knockdown or overexpression . Depletion of PCK1 in SW480 cells by two independent shRNAs significantly impaired ( p<0 . 0001 in both comparison ) CRC liver metastatic colonization of cells introduced into the portal circulation of NSG mice ( Figure 5A ) . PCK1 depletion in another colorectal cell line ( LS174T ) also significantly decreased ( p<0 . 0001 ) liver metastatic colonization ( Figure 5B ) . Conversely , PCK1 over-expression in SW480 cells significantly increased ( p=0 . 003 ) liver metastatic colonization ( Figure 5C ) . In contrast , PCK1 depletion did not impact subcutaneous tumor growth in the SW480 or LS174T cell lines ( Figure 5D ) . To assess whether Pck1 modulation regulated cancer progression in a fully immunocompetent model as well , we depleted Pck1 in the murine CRC cell line CT26 . Consistent with our observations in human cancer lines , Pck1 depletion decreased murine CRC cell liver colonization in an immune competent model ( p=0 . 039 , p=0 . 005 for shCTRL vs shPck1-64 , shCTRL vs shPck1-66 , respectively ) and did not impair in vitro proliferation under basal cell culture conditions ( Figure 5E ) . We next sought to determine the cellular mechanism by which PCK1 impacts metastatic colonization; that is , whether PCK1 influences initial CRC cell liver colonization , apoptosis , or population growth . To identify whether initial liver colonization was the sole step in the metastatic cascade influenced by PCK1 or whether it could provide continued impact on CRC liver growth , we generated SW480 cells expressing an inducible PCK1 shRNA ( Figure 5—figure supplement 1A ) . Four days after portal systemic injection of cancer cells , at which time CRC cells have extravasated into the liver and begun initial outgrowth , we began administering a doxycycline or a control diet . We found that even after the initial liver colonization phase ( days 0–4 ) , PCK1 depletion continued to impair ( p=0 . 004 ) CRC metastatic liver growth ( Figure 5—figure supplement 1B–C ) . We did not observe increased apoptosis using the caspase 3/7 reporter in both PCK1 depleted cell populations in vivo ( Figure 5—figure supplement 1D ) . Evaluation of the in vivo growth rate through natural log slope calculations demonstrated that in each PCK1 modulation experiment , either knockdown or overexpression , in which a luciferase reporter was used , the rate of growth after the first measured time point ( day 4–7 ) did not equal the rate of growth of the controls ( Figure 5—figure supplement 1E ) . These results reveal that PCK1 promotes the rate of metastatic growth in vivo . Given our identification of PCK1 as a metabolic regulator of CRC liver metastatic colonization as well as the enrichment of a hypoxic signature by GSEA in highly metastatic PDXs , we speculated that perhaps PCK1 promotes metabolic adaptation that enables growth under hypoxia—a key feature of the hepatic microenvironment ( Jungermann and Kietzmann , 2000; Dupuy et al . , 2015 ) ( Figure 6A ) . Consistent with this , PCK1 depletion significantly reduced CRC cell growth under hypoxia , an effect not observed under normoxia ( Figure 6B , Figure 6—figure supplement 1A ) . Hypoxia poses a metabolic challenge for cancer cell growth as metabolites needed for biosynthesis of macromolecules required for cell proliferation can become limiting ( Birsoy et al . , 2015; Garcia-Bermudez et al . , 2018 ) . In vivo selected cancer cells can alter cellular metabolism in order to better respond to the metastatic microenvironment ( Loo et al . , 2015; Nguyen et al . , 2016 ) . To search for such adaptive metastatic metabolic alterations that associate with enhanced PCK1 expression , we performed metabolite profiling of the four highly/poorly metastatic CRC PDX pairs . Unsupervised hierarchical clustering analysis was then performed on the differentially expressed metabolite profiles for each pair . Interestingly , the most salient observation was increased abundance in three out of four PDX pairs of multiple nucleoside base precursors and specifically metabolites in the pyrimidine biosynthetic pathway ( Figure 6C ) . These metabolites comprised orotate , dihydroorotate , and ureidopropionate . These findings reveal that metastatic colonization by human CRC cells selects for induction of multiple metabolites in the pyrimidine biosynthetic pathway . We hypothesized that perhaps enhanced levels of pyrimidine precursors were selected for in metastatic CRC cells to enable adaptation to hypoxia where precursors for pyrimidine biosynthesis such as aspartate are known to become depleted ( Birsoy et al . , 2015; Garcia-Bermudez et al . , 2018 ) . Without such an adaptation , cells would experience deficits in pyrimidine bases and consequently nucleotide pools , which would curb growth . How might PCK1 upregulation contribute to maintenance of nucleotide pools ? Nucleotides contain nitrogenous bases covalently coupled to ribose and phosphate . PCK1 was previously shown to promote ribose generation by CRC cells under pathophysiological levels of glucose via the pentose phosphate pathway ( Montal et al . , 2015 ) . We thus hypothesized that PCK1 depletion , by depleting ribose pools , may reduce pyrimidine and purine nucleotide pools in CRC cells . To test this , we performed metabolite profiling of control and PCK1 depleted CRC cells under hypoxia . While metabolites related to glycolysis and the citric acid ( TCA ) cycle were significantly increased , corroborating previous studies ( Montal et al . , 2015 ) , the most salient finding was a significant depletion of nucleosides and nucleotides including uridine , guanine , UMP , CMP , CDP , IMP , GMP , and AMP in PCK1 depleted cells ( Figure 6D–F ) . Consistent with our observations of selective impairment of PCK1-dependent growth under hypoxia , the observed decreases in nucleoside and nucleotide levels were abrogated under normoxic conditions ( Figure 6—figure supplement 1B–C ) . These findings reveal that increased PCK1 expression is required for nucleotide pool maintenance in CRC cells in the context of hypoxia . The above findings reveal that liver metastatic CRC cells enhance pyrimidine biosynthesis and that PCK1 drives pyrimidine nucleotide synthesis under hypoxia . To better understand how PCK1 maintains nucleotide pools under hypoxia , we performed a 13C glutamine metabolic flux analysis in control and PCK1-depleted cells . Glutamine labeling revealed that the generation of pyrimidine precursors , such as aspartate and orotate , by reductive carboxylation was significantly decreased upon PCK1 depletion ( Figure 6G–H ) . In contrast , the generation of pyrimidine precursors derived from oxidative reactions in the TCA cycle were either significantly increased or unchanged ( Figure 6—figure supplement 1D–E ) . These findings reveal that PCK1 promotes reductive carboxylation under hypoxia to produce pyrimidine precursor aspartate . Aspartate has been shown to be sufficient to enable growth under hypoxia or electron transport chain dysfunction ( Birsoy et al . , 2015; Sullivan et al . , 2015 ) . We speculated that PCK1 might enable hypoxic growth either by promoting aspartate synthesis . To test this , we supplemented PCK1-depleted cells with aspartate or with pyruvate , which can i ) replete NAD+ stores upon its reduction to lactate and ii ) can be converted to the aspartate precursor oxaloacetate by pyruvate carboxylase in mitochondria . Supplementation of PCK1-depleted cells with aspartate partially rescued the observed growth defect , while pyruvate supplementation provided a modest rescue ( Figure 6—figure supplement 1F ) . Interestingly , combined aspartate and pyruvate supplementation nearly fully rescued the hypoxic growth defect of PCK1-depleted cells ( Figure 6—figure supplement 1F ) . These findings reveal that PCK1 promotes growth by maintaining the levels of aspartate , a nucleotide precursor . The increased hypoxic growth observed upon combined aspartate and pyruvate supplementation suggest the possibility that PCK1 may additionally promote hypoxic growth by increasing NAD+ pools . These above findings overall suggest that hypoxia acts as a barrier to growth for metastatic CRC by limiting pyrimidine nucleoside abundance . To directly test this , we determined if the growth defect of PCK1 depletion upon hypoxia could be rescued by the pyrimidine nucleoside uridine . Indeed , supplementation of CRC cells with uridine rescued the hypoxic growth defect caused by PCK1 depletion . In contrast , supplementation with purines or ribose provided a modest effect ( Figure 6I ) . These findings reveal PCK1 induction to be a mechanism employed by CRC cells to generate pyrimidine nucleotide pools under hypoxia to fuel growth . Due to the strong reduction in mCRC liver colonization observed upon PCK1 depletion , we hypothesized that PCK1 inhibition may represent a potential therapeutic strategy for impairing CRC metastatic progression . We performed in vivo proof-of-principle experiments in two independent CRC cell lines with a PCK1-inhibitor , 3-mercaptopicolinic acid ( 3 MPA ) ( DiTullio et al . , 1974 ) . We treated CRC cells in vitro for 24 hr at a dose that did not alter cell proliferation in vitro . The following day , mice were subjected to portal circulation injections with either control or 3-MPA-treated cells . Similar to PCK1 inhibition by shRNA , pre-treatment of cells with 3 MPA significantly reduced ( p=0 . 01 ) mCRC liver colonization in vivo ( Figure 7A , Figure 7—figure supplement 1A ) , Pre-treatment of LS174T cells with 3 MPA did not , however , alter subcutaneous tumor growth ( Figure 7—figure supplement 1B ) . We next sought to determine whether experimental therapeutic delivery of 3 MPA could suppress metastatic colonization . Prior to portal systemic injection of LS174T cells , we began oral gavage treatment of mice with either 200 mg/kg of 3 MPA in aqueous solution or control . On day 1 , we repeated the 3 MPA or control gavage . We found that even such short-term treatment of 3 MPA decreased CRC liver colonization in this model ( Figure 7B , Figure 7—figure supplement 1C ) ( p=0 . 008; p=0 . 005 for Figure 7B and Figure 7—figure supplement 1C respectively ) . Taken together , these results indicate that PCK1 promotes colorectal cancer liver colonization and represents a potential therapeutic target in the prevention of CRC at risk for metastatic disease . Dihydroorotate dehydrogenase is a key enzyme in the metabolic pathway that reduces dihydroorotate to orotate , which is ultimately converted to the pyrimidine nucleotides UTP and CTP ( Figure 7C ) . To further confirm that pyrimidine biosynthesis promotes CRC hypoxic growth , we sought to assess CRC growth upon DHODH inhibition . Leflunomide is an approved , well-tolerated , and high-affinity ( Kd = 12 nM ) small-molecule inhibitor of DHODH used in the treatment of rheumatoid arthritis . Leflunomide treatment significantly impaired CRC growth in the context of hypoxia—an effect that was more significant under hypoxia than normoxia ( Figure 7—figure supplement 1D–E ) . These findings confirm that metastatic CRC cell growth under hypoxia is sensitive to pyrimidine biosynthesis inhibition . Our findings as a whole suggest that metastatic CRC liver metastatic colonization may be sensitive to inhibition of the pyrimidine biosynthetic pathway . To directly test this , we first depleted highly metastatic Lvm3b CRC cells of DHODH ( Figure 7—figure supplement 1F ) . DHODH depletion substantially reduced CRC liver metastatic colonization ( Figure 7D ) , revealing a critical role for DHODH activity and pyrimidine biosynthesis in CRC liver metastatic colonization . To determine if leflunomide can therapeutically inhibit CRC liver metastasis , we treated animals with a dose of this drug similar to that used for rheumatoid arthritis ( 7 . 5 mg/kg body weight ) . Treatment of animals injected with highly metastatic Lvm3b cells with leflunomide caused a ~ 90 fold reduction in CRC liver metastatic colonization ( Figure 7E ) . The leflunomide treated mice experienced significantly longer survival ( p=0 . 006 ) than the control mice ( Figure 7F ) . Importantly , these cells are known to be highly resistant to 5-FU ( Bracht et al . , 2010 ) , the backbone chemotherapeutic used in CRC , revealing that inhibition of DHODH can exert therapeutic benefit despite cellular resistance to an anti-metabolite that targets the pyrimidine pathway . Leflunomide treatment only modestly impacted primary tumor growth by two distinct CRC populations ( Figure 7—figure supplement 1G–H ) , suggesting enhanced sensitivity of CRC cells to leflunomide-mediated DHODH inhibition during liver metastatic colonization . To determine if the metastatic colonization defect caused by leflunomide treatment is caused by pyrimidine depletion , we tested cell growth suppression in the presence or absence of uridine—the downstream metabolic product of the pyrimidine biosynthetic pathway . This revealed that the impaired growth upon hypoxia was rescued upon uridine supplementation ( Figure 7G ) . Importantly , leflunomide treatment impaired proliferation significantly more in the context of hypoxia than under normoxia ( Figure 7—figure supplement 1D–E ) . These observations reveal enhanced dependence of highly metastatic cells on pyrimidine biosynthesis and reveal upregulation of metabolites in this pathway as a selective adaptive trait of highly metastatic CRC cells . Overall , these findings identify DHODH as a therapeutic target in CRC progression and provide proof-of-concept for use of leflunomide in therapeutic inhibition of CRC metastatic progression . CRC remains a challenging disease despite multiple advances over the last six decades . Some patients with metastatic CRC can experience regression responses to current therapies , although most succumb to their disease within 3 years . Given that most CRC deaths occur as a result of complications of metastatic disease , a model that can predict which patients with advanced CRC harbor more aggressive disease could aid in appropriately positioning patients for experimental clinical trials . The objectives of our study were two-fold: to develop a CRC liver metastasis patient-derived xenograft model , and to employ this model to identify candidate genes and biology underlying CRC liver metastatic colonization . Most patient-derived xenograft models consist of subcutaneous tumor tissue implantation . Similar to others , we found that successful subcutaneous tumor engraftment associated with worse patient survival in those with CRC ( Oh et al . , 2015 ) . However , among those tumors in our study that did engraft subcutaneously , the subcutaneous tumor growth rate did not significantly correlate with patient survival . In contrast , we found that liver metastatic growth rate was significantly correlated with patient survival . The reason for this discrepancy in the prognostic power of subcutaneous tumor growth versus liver metastatic growth is likely the greater selective pressure inherent to the liver microenvironment . This collection of clinically predictive CRC liver metastatic PDX models represents a valuable resource for the cancer community . PCK1 is the rate-limiting enzyme in gluconeogenesis and is often upregulated in patients with metabolic syndrome and diabetes mellitus . Epidemiologic data suggests that those patients with diabetes that are on metformin , a gluconeogenic-antagonist , exhibit improved CRC clinical outcomes relative to their metformin-free counterparts ( Zhang et al . , 2011; Spillane et al . , 2013; Fransgaard et al . , 2016 ) . Our observations suggest a potential mechanistic basis for the sensitivity of CRC metastatic progression to inhibition of this metabolic pathway . Metabolic rewiring in cancer has been well-established to provide tumor cells with the necessary nutrients and anabolic components to sustain proliferative and energetic demands ( Chandel , 2015; Vander Heiden and DeBerardinis , 2017; Boroughs and DeBerardinis , 2015 ) . While numerous pathways are involved in metabolic reprogramming , metabolic shunting into pathways including glucose metabolism , the citric acid ( TCA ) cycle , and lipogenesis largely support macromolecule synthesis for cancer cells ( Locasale and Cantley , 2011; Vander Heiden et al . , 2009; DeBerardinis et al . , 2008; Possemato et al . , 2011 ) . In line with these notions , there have been two reports on PCK1 and its role in cancer . In one study , PCK1 was found to enhance melanoma tumor re-initiation ( Li et al . , 2015 ) . In this work , the authors demonstrated that , in tissue culture , melanoma ‘tumor re-initiating cells’ consumed more glucose and produced more lactate and glycerate-3-phosphate; PCK1 silencing elicited the opposite phenotype in culture ( Li et al . , 2015 ) . Using cell culture metabolomics , Montal et al . ( 2015 ) recently described a mechanism by which PCK1 promotes CRC growth through its increased ability to metabolize glutamine into lipids and ribose . PCK1 silencing in a CRC cell line in vitro was shown to decrease glutamine utilization and TCA cycle flux ( Montal et al . , 2015 ) . This group found that cells with increased expression of PCK1 consumed more glucose and produced more lactate . They further performed PCK1 staining on a primary CRC tissue microarray , finding that PCK1 was overexpressed in many primary CRC biopsies but its expression was not associated with tumor grade . Our findings demonstrate a major role for PCK1 in liver metastatic colonization by CRC . While we did not find evidence of PCK2 upregulation in our mCRC model , three recent studies demonstrated that PCK2 upregulation in lung cancer cells in vitro can improve cancer cell survival in glucose-depleted conditions ( Leithner et al . , 2015; Vincent et al . , 2015 ) . Vincent et al . found that in glucose-depleted conditions , lung cancer cells increased consumption of glutamine as an energy source in a PCK2-dependent manner . Zhao et al . ( 2017 ) observed PCK2 upregulation in tumor initiating cells and demonstrated that PCK2 promoted tumor initiation through reducing TCA cycle flux by lowering Acetyl-CoA . Our findings provide three novel insights underlying the role of PCK1 in cancer progression . First , we demonstrate that increased PCK1 strongly drives liver metastatic colonization but minimally impacted primary tumor growth rate in the cells studied . Second , we provide the first reported evidence that PCK1 can promote hypoxic growth . Third , we uncover a key role for PCK1 in pyrimidine biosynthesis under hypoxia via reductive carboxylation . Although we do not provide a precise biochemical mechanism by which PCK1 promotes reductive carboxylation , Zamboni et al . ( 2004 ) reported that in a bacterium with pyruvate kinase impairment , PCK1 could operate in reverse—converting phosphoenolpyruvate to oxaloacetate . We speculate that such a reaction would drive reductive carboxylation by generating oxaloacetate , a precursor for aspartate and consequently pyrimidines . In support of this , our metabolite profiling of PCK1-depleted cells under hypoxia revealed increased abundance of phosphoenolpyruvate . Moreover , we had previously observed that highly metastatic CRC cells upregulate PKLR , which was shown to inhibit pyruvate kinase M2 ( PKM2 ) activity ( Nguyen et al . , 2016 ) —analogous to the observations in the aforementioned bacterial study . Further investigation is warranted to conclusively define the mechanism by which PCK1 regulates pyrimidine synthesis in metastatic CRC cells under hypoxia . Because tumor cells alter their metabolic programs within their given tumor microenvironment , metabolic liabilities that provide therapeutic opportunities emerge ( Li et al . , 2015; Wheaton et al . , 2014; Locasale et al . , 2011 ) . Recent work has implicated DHODH as a regulator of differentiation in myeloid leukemia as well as a promoter of cell cycle progression in some solid cancer types such as melanoma and pancreatic adenocarcinoma ( White et al . , 2011 ) . Bajzikova et al . found that de-novo pyrimidine biosynthesis is essential for mouse breast cancer tumorigenesis in a DHODH-dependent manner ( Bajzikova et al . , 2019 ) . Our work reveals that beyond effects on cell growth in vitro and primary tumor growth , CRC metastatic progression selects for upregulation of pyrimidine biosynthesis . Moreover , the use of leflunomide to therapeutically target DHODH has been implicated under various cancer contexts as a metabolic inhibitor ( Mathur et al . , 2017; Luengo et al . , 2017 ) . Here , we observe that molecular or pharmacological inhibition of this pathway with leflunomide strongly impairs CRC metastatic colonization relative to primary tumor growth . Our work reveals that hypoxia enhances the sensitivity of cells to DHODH inhibition , consistent with enhanced pyrimidine biosynthesis enabling enhanced growth under hypoxia—a key feature of the hepatic tumor microenvironment . 5-Fluorouracil ( 5-FU ) was the first chemotherapeutic to demonstrate efficacy in reducing the risk of CRC recurrence ( Moertel et al . , 1990 ) . This agent remains the backbone of the current FOLFOX regimen , which is administered to patients after surgical resection to reduce the risk of metastatic relapse . Interestingly , 5-FU targets thymidylate synthase , an enzyme downstream of DHODH in the pyrimidine biosynthetic pathway—supporting our premise of dependence and susceptibility to inhibition of this pathway in CRC metastasis . Despite its activity , a large fraction of patients treated with 5-FU relapse . Multiple mechanisms of resistance to 5-FU have been described ( Holohan et al . , 2013 ) . Our findings demonstrate that inhibition of DHODH can suppress metastatic progression of a CRC cell line that is resistant to 5-FU—revealing promise for clinical testing of this agent in patients at high risk for relapse and whose tumors may exhibit resistance to 5-FU . Overall , our work reveals that PDX modeling of CRC can be predictive of clinical survival outcomes; that integration of PDX modeling with in vivo selection can give rise to highly metastatic PDX derivatives which can be profiled transcriptomically and metabolically to identify key drivers of metastatic progression; and that PCK1 and DHODH represent key metabolic drivers of CRC metastasis and therapeutic targets in CRC . SW480 , LS174T , and CT26 cell lines were obtained from ATCC . HEK-293LTV cells were obtained from Cell Biolabs . LS174T , HEK-293LTV , and CT26 cells were grown in Dulbecco's Modified Eagle Medium ( Gibco ) supplemented with 10% v/v fetal bovine serum ( Corning ) , L-glutamine ( 2 mM; Gibco ) , penicillin-streptomycin ( 100 U/ml; Gibco ) , Amphotericin ( 1 μg/ml; Lonza ) , and sodium pyruvate ( 1 mM; Gibco ) . SW480 cells were grown in McCoy’s 5A modified media with L-glutamine ( Corning ) supplemented with 10% v/v fetal bovine serum , penicillin-streptomycin ( 100 U/ml ) , Amphotericin ( 1 μg/ml ) , and sodium pyruvate ( 1 mM ) . All cells were grown at 37°C under 5% CO2 and passaged when the monolayer reached 80% confluency . All cell lines were authenticated by SPR profiling at MSKCC . All cells were regularly checked for mycoplasma contamination and have been negative . CT26 cells that had been stably transduced with PCK1-targetting shRNA hairpins or control hairpins were grown in vitro for 3 days and counted on day three using the Sceptor 2 . 0 automated Cell counter ( Millipore ) . Lvm3b cells or LS174T cells were grown under normoxia for 24 hr followed by incubation for 5 days under 0 . 5% oxygen and then counted using the Sceptor 2 . 0 automated Cell Counter ( Millipore ) . LS174T cells were seeded in 6-well plates . On day 1 , the media was replaced with either control media or media supplemented with 1 mM 3 MPA . On day 2 , all the media was replaced with control media . The experiment was terminated on day 5 . Twenty-four hours exposure to 1 mM 3 MPA in media does not alter LS174T cell growth in vitro . 2 × 104 LS174T cells were seeded in triplicate . On day 1 , the media was replaced with either control media or media supplemented with 1 mM 3 MPA . On day 2 , all the media was replaced with control media . The experiment was terminated on day 5 . Lentiviral particles were created using the ViraSafe lentiviral packaging system ( Cell Biolabs ) . ShRNA oligo sequences were based upon the Sigma-Aldrich MISSION shRNA library and were obtained from Integrated DNA technologies . The following shRNAs were used in this study: PCK1 sh3 ( TRCN0000196706 ) , PCK1 sh4 ( TRCN0000199286 ) , PCK1 sh5 ( TRCN0000199573 ) , shControl ( SHC002 ) , mouse PCK1 sh64 ( TRCN0000025064 ) , mouse PCK1 sh66 ( TRCN0000025066 ) , DHODH sh2 ( TRCN0000221421 ) , and DHODH sh3 ( TRCN0000221422 ) . Forward and reverse complement oligos were annealed , cloned into pLKO , and transformed into XL10-Gold E . coli ( 200314 , Agilent ) . For PCK1 overexpression , PCK1 cDNA ( plasmid ID HsCD00045535 ) was obtained from the PlasmID Repository at Harvard Medical School and cloned into pBabe-puromycin or plx304-blasticidin . For tetracycline-inducible experiments , the seed sequences of shRNA control ( SHC002 ) or PCK1 sh4 ( TRCN0000199286 ) were cloned into pLKO-Tet-On ( Wiederschain et al . , 2009 ) . All plasmids were isolated using the plasmid plus midi kit ( Qiagen ) . Transduction and transfection were performed as described previously ( Yu et al . , 2015 ) . All animal works were conducted in accordance with a protocol approved by the Institutional Animal Care and Use Committee ( IACUC ) at The Rockefeller University and Memorial Sloan Kettering Cancer Center . Either NOD . Cg-Prkdcscid Il2rgtm1Wjl/SzJ ( Nod-Scid-Gamma; NSG ) aged 6–10 weeks or Foxn1nu ( Nu/J; athymic nude ) aged 6–10 weeks were used for all mouse experiments . For functional studies of PCK1 , CRC cells ( SW480 or LS174T ) that had been stably transduced with a luciferase reporter ( Ponomarev et al . , 2004 ) were subjected to portal circulation injection in NSG mice; after two minutes , a splenectomy was performed . For functional and pharmacology studies of DHODH , CRC cells ( Lvm3b ) that had been stably transduced with a luciferase reporter were subjected to portal circulation injection in athymic nude mice; after 2 min , a splenectomy was performed . Mice were imaged weekly; experiments were terminated when the luciferase signal had saturated or the mice were too ill , whichever occurred first . Chow was removed from cages four hours prior to injection of LS174T cells . Gavage with either 3 MPA ( 200 mg/kg in aqueous solution ) or placebo was performed one hour prior to injection of LS174T cells . 1 × 106 LS174T cells were injected into the portal systemic circulation as described Animals section above . Chow was returned to cages after injection . On day 1 , chow was removed from cages; 3 MPA or placebo was administered via gavage 4 hr post-chow removal . Chow was returned to cages four hours after drug administration . Mice were imaged bi-weekly . Leflunomide ( Tocris Cat # 2228 ) 7 . 5 mg/kg mouse body weight was intraperitoneally injected every day . Equivalent volume of DMSO were intraperitoneally injected every day to the control cohort . Patient colorectal tumors were prepared and stained with hemotoxylin and eosin ( H and E ) per standard clinical procedures following surgical resection of the tumor specimen . Subcutaneous and liver xenograft samples were removed from the mice at time of sacrifice and fixed in 4% paraformaldehyde solution for 48 hr at 4° . The xenografts samples were subsequently rinsed in PBS twice followed by one-hour incubations in 50% ethanol , then 70% ethanol . The xenografts samples were stained with maintained in 70% ethanol at 4° . The fixed xenografts samples were embedded in paraffin , sectioned , and stained with H and E ( Histoserv ) . Quantitative RT-PCR qRT-PCR was performed to confirm expression of PCK1 . Total RNA was extracted ( 37500 , Norgen ) from CRC PDXs , SW480 , LS174T , or CT26 cells that had been stably transduced with PCK1-targetting shRNA hairpins , control hairpins , pBabe-PCK1 , pBabe-control , plx-PCK1 , or plx-Empty . cDNA was generated using Superscript III first strand cDNA synthesis kit ( 18080051 , Invitrogen ) per manufacturer’s protocol . For quantification of cDNA , Fast SYBR Green Master Mix ( 4385612 , Applied Biosystems ) was used for sample analysis . Gene expression was normalized to HPRT expression . The following sequences were used as primers for CRC PDXs , SW480 , and LS174T cells: PCK1-F , AAGGTGTTCCCATTGAAGG; PCK1-R , GAAGTTGTAGCCAAAGAAGG; HPRT-F , GACCAGTCAACAGGGGACAT; HPRT-R , CCTGACCAAGGAAAGCAAAG . The following sequences were used as primers for CT26 cells: PCK1-F , CTGCATAACGGTCTGGACTTC; PCK1-R , CAGCAACTGCCCGTACTCC; b-actin-F , GGCTGTATTCCCCTCCATCG; b-actin-R , CCAGTTGGTAACAATGCCATGT . The following primers were used for Lvm3b cells: DHODH-F , CCACGGGAGATGAGCGTTTC; DHODH-R , CAGGGAGGTGAAGCGAACA . GEO data sets GSE41258 , GSE 507060 , GSE14297 , and GSE6988 were used to evaluate for expression of PCK1 as described previously ( Yu et al . , 2015; Kim et al . , 2014 ) . Within 2 hr of surgical resection , CRC tumor tissue that was not needed for diagnosis was implanted subcutaneously into NSG mice at the MSKCC Antitumor Assessment Core facility . When the tumor reached the pre-determined end-point of 1 , 000 mm3 , the tumor was excised and transferred to the Rockefeller University . Xenograft tumor pieces of 20–30 mm3 were re-implanted . When the subcutaneous tumor reached 1 , 000 mm3 , the tumor was excised . Part of the tumor was cryogenically frozen in FBS:DMSO ( 90:10 ) for future use . The rest of the tumor was chopped finely with a scalpel and placed in a 50 ml conical tube with a solution of Dulbecco's Modified Eagle Medium ( Gibco ) supplemented with 10% v/v fetal bovine serum ( Corning ) , L-glutamine ( 2 mM; Gibco ) , penicillin-streptomycin ( 100 U/ml; Gibco ) , Amphotericin ( 1 μg/ml; Lonza ) , sodium pyruvate ( 1 mM; Gibco ) and Collagenase , Type IV ( 200 U/ml; Worthington ) and placed in a 37°C shaker at 220 rpm for 30 min . After centrifugation and removal of supernatant , the sample was subjected to ACK lysis buffer ( Lonza ) for 3 min at room temperature to remove red blood cells . After centrifugation and removal of ACK lysis buffer , the sample was subjected to a density gradient with Optiprep ( 1114542 , Axis-Shield ) to remove dead cells . The sample was washed in media and subjected to a 100-μm cell strainer and followed by a 70-μm cell strainer . Mouse cells were removed from the single-cell suspension via magnetic-associated cell sorting using the Mouse Cell Depletion Kit ( ( 130-104-694 , Miltenyi ) , resulting in a single-cell suspension of predominantly CRC cells of human origin . One million PDX CRC cells were injected into the portal circulation of NSG mice via the spleen . Two minutes after injection , the spleen was removed using electrocautery . When the mouse was deemed ill by increased abdominal girth , slow movement , and pale footpads , it was euthanized and the tumors were removed and sectioned in a manner similar to the subcutaneous implants . For a subset of mice ( CLR4 , CLR27 , CLR28 , CLR32 ) the CRC liver metastatic tumor cells were injected into the spleens of another set of NSG mice in order to obtain metastatic derivatives with enhanced ability to colonize the liver . To ensure minimal contamination from mouse stromal or blood cells during RNA sequencing , we performed flow cytometric cell sorting of the PDX cell suspension after it had been processed through the magnetic-based mouse cell depletion kit ( 130-104-694 , Miltenyi ) . Single cells that bound an APC-conjugated anti-human CD326 antibody ( 324208 , BioLegend ) and did not bind to a FITC-conjugated anti-mouse H-2Kd antibody ( 116606 , BioLegend ) were positively selected and considered to be PDX CRC cells . RNA was isolated from these double-sorted CRC PDX cells ( 37500 , Norgen ) , ribosomal RNA was removed ( MRZH11124 , illumina ) , and the samples were prepared for RNA-sequencing using script-seq V2 ( SSV21124 , illumina ) . RNA sequencing was performed by the RU Genomics Resource Center on an Illumina HiSeq 2000 with 50 basepair single read sequencing . The sequencing data was cleaned of low quality base pairs and trimmed of linker sequences using cutadapt ( v1 . 2 ) and aligned to the reference transcriptome ( Hg19 ) using TopHat ( v2 ) . Cufflinks ( v2 ) was used to estimate transcript abundances . Upon merging assemblies ( Cuffmerge ) , comparison of samples was made using Cuffdiff ( v2 ) to determine genes that were differentially expressed between parental and liver-metastatic derivative xenografts . Fisher’s method was used to determine genes that were differentially expressed across all analyzed gene sets . Correlation matrix of gene expression profiles from RNA sequencing were generated using Spearman’s correlation coefficient . Clustering was performed in R using Euclidean distance and complete agglomeration method . Each isogenic tumor pair ( parental and liver-metastatic derivative ) was evaluated for changes in the Hallmark gene sets using GSEA ( v2 . 2 . 1 , Broad Institute ) . Additionally , a composite gene set using Fisher’s method as described in the section above was analyzed using GSEA . shCTRL or shPCK1 LS174T cells were plated at 3 × 105 cells per well in a 6-well plate and allowed to adhere to the plate for 24 hr . Cells were then pre-treated with RPMI-1640 media containing 6 mM glucose at 0 . 5% O2 for 6 hr , then replaced with RPMI-1640 media containing 2 mM U-13C-glutamine ( Cambridge Isotope Laboratories , #CLM-1822-H ) for 0–6 hr at 0 . 5% O2 . Metabolites were extracted at 0 , 2 , 4 , and 6 hr time points . Metabolite extraction and subsequent Liquid-Chromatography coupled to High-Resolution Mass Spectrometry ( LC-HRMS ) for polar metabolites of cells was carried out using a Q Exactive Plus . shCTRL or shPCK1 LS174T were plated at 3 × 105 cells/well in triplicate with RPMI1640 + dialyzed FBS + 6 mM glucose and remained in 0 . 5% O2 or normoxia for 24 hr . For PDX metabolite profiling , 100 mg of frozen PDXs were used . For all metabolite profiling , cells were washed with ice cold 0 . 9% NaCl and harvested in ice cold 80:20 LC-MS methanol:water ( v/v ) . Samples were vortexed vigorously and centrifuged at 20 , 000 g at maximum speed at 4°C for 10 min . Supernatant was transferred to new tubes . Samples were then dried to completion using a nitrogen dryer . All samples were reconstituted in 30 μl 2:1:1 LC-MS water:methanol:acetonitrile . The injection volume for polar metabolite analysis was 5 μl . A ZIC-pHILIC 150 × 2 . 1 mm ( 5 µm particle size ) column ( EMD Millipore ) was employed on a Vanquish Horizon UHPLC system for compound separation at 40°C . The autosampler tray was held at 4°C . Mobile phase A is water with 20 mM Ammonium Carbonate , 0 . 1% Ammonium Hydroxide , pH 9 . 3 , and mobile phase B is 100% Acetonitrile . The gradient is linear as follows: 0 min , 90% B; 22 min , 40% B; 24 min , 40% B; 24 . 1 min , 90% B; 30 min , 90% B . The follow rate was 0 . 15 ml/min . All solvents are LC-MS grade and purchased from Fisher Scientific . The Q Exactive Plus MS ( Thermo Scientific ) is equipped with a heated electrospray ionization probe ( HESI ) and the relevant parameters are as listed: heated capillary , 250°C; HESI probe , 350°C; sheath gas , 40; auxiliary gas , 15; sweep gas , 0; spray voltage , 3 . 0 kV . A full scan range from 55 to 825 ( m/z ) was used . The resolution was set at 70 , 000 . The maximum injection time was 80 ms . Automated gain control ( AGC ) was targeted at 1 × 106 ions . Maximum injection time was 20 msec . Raw data collected from LC-Q Exactive Plus MS was processed on Skyline ( https://skyline . ms/project/home/software/Skyline/begin . view ) using a five ppm mass tolerance and an input file of m/z and detected retention time of metabolites from an in-house library of chemical standards . The output file including detected m/z and relative intensities in different samples was obtained after data processing . Quantitation and statistics were calculated using Microsoft Excel , GraphPad Prism 8 . 1 , and Rstudio 1 . 0 . 143 . Kaplan-Meier analysis was used to evaluate patient survival based upon PDX parameters . Sample size in mouse experiments was chosen based on the biological variability observed with a given genotype . Non-parametric tests were used when normality could not be assumed . Mann Whitney test and Student’s t test were used when comparing independent shRNAs to shControl . One-tailed tests were used when a difference was predicted to be in one direction; otherwise , a two-tailed test was used . A P value less than or equal to 0 . 05 was considered significant ( :p<0 . 05 , :p<0 . 01 , :p<0 . 001 , and **: p<0 . 0001 ) . Error bars represent SEM unless otherwise indicated .	Colorectal cancer , also known as bowel cancer , is the second most deadly cancer in the United States , where it affects over 140 , 000 people each year . This cancer often spreads to the liver , in a process known as metastasis . To do this , the colorectal cancer cells must survive the low oxygen levels found in the blood that carries them from the gut to the liver , and in the liver itself . The majority of colorectal cancer cells that arrive in the liver die , but some survive leading to secondary tumors . To investigate how the colorectal cancer cells that survive and metastasize to the liver accomplish this feat , Yamaguchi , Weinberg , Nguyen et al . took colorectal tumors from patients and introduced them into mice . This showed that tumors from patients with the worst outcomes tended to metastasize more efficiently in mice . Next , Yamaguchi et al . looked to see which genes were active in the colorectal cancer cells that were able to metastasize and compared them to those that were active in the cells that could not . This analysis revealed that the gene coding for a protein called PCK1 was more active in the cells that could metastasize . In healthy cells , PCK1 promotes the generation of the sugar glucose , and Yamaguchi et al . observed that , in a low oxygen environment , higher levels of PCK1 allowed colorectal cancer cells to proliferate faster . Unexpectedly , this was due to PCK1 increasing the production of a molecule needed to make nucleotides , which are the building blocks for DNA and RNA . Consistent with this , metastasizing colorectal cancer cells generated more nucleotide precursors , and inhibiting the enzyme involved in nucleotide synthesis ( DHODH ) with an arthritis drug called leflunomide stopped colorectal cancer cells from spreading . Metastasizing colorectal cancer cells depend on PCK1 and the nucleotide-synthesizing enzyme to grow , so therapies that target these proteins may help more patients to survive this kind of cancer . The findings also suggest that the arthritis drug leflunomide should be explored further as a potential drug for the treatment of colorectal cancer .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cancer", "biology" ]	2019	PCK1 and DHODH drive colorectal cancer liver metastatic colonization and hypoxic growth by promoting nucleotide synthesis
The biotrophic fungus Ustilago maydis causes smut disease in maize with characteristic tumor formation and anthocyanin induction . Here , we show that anthocyanin biosynthesis is induced by the virulence promoting secreted effector protein Tin2 . Tin2 protein functions inside plant cells where it interacts with maize protein kinase ZmTTK1 . Tin2 masks a ubiquitin–proteasome degradation motif in ZmTTK1 , thus stabilizing the active kinase . Active ZmTTK1 controls activation of genes in the anthocyanin biosynthesis pathway . Without Tin2 , enhanced lignin biosynthesis is observed in infected tissue and vascular bundles show strong lignification . This is presumably limiting access of fungal hyphae to nutrients needed for massive proliferation . Consistent with this assertion , we observe that maize brown midrib mutants affected in lignin biosynthesis are hypersensitive to U . maydis infection . We speculate that Tin2 rewires metabolites into the anthocyanin pathway to lower their availability for other defense responses . The fungus Ustilago maydis causes smut disease in maize . U . maydis is a biotrophic fungus that needs living plant tissue for completion of its sexual cycle and spore formation . After penetrating epidermal cells and establishing an extended interaction zone in which hyphae are encased by the host plasma membrane , fungal hyphae traverse mesophyll tissue ( Doehlemann et al . , 2008 , 2009 ) , and accumulate in and around vascular bundles ( Doehlemann et al . , 2008 ) presumably to obtain nutrients from veins . During infection , U . maydis secretes hundreds of effector proteins to suppress plant defense responses , and to reprogram plant signaling and metabolism ( Doehlemann et al . , 2008; Mueller et al . , 2008; Horst et al . , 2010; Djamei and Kahmann , 2012 ) . However , the function of the vast majority of these novel effectors is unknown ( Djamei et al . , 2011; Djamei and Kahmann , 2012; Hemetsberger et al . , 2012; van der Linde et al . , 2012; Mueller et al . , 2013 ) . At later stages of host colonization , large plant tumors develop that provide the environment for massive fungal proliferation . During biotrophic growth , a high affinity fungal sucrose H+ symporter , Srt1 , is upregulated and plays a crucial role for fungal nutrition ( Wahl et al . , 2010 ) . Visible changes in host metabolism include an accumulation of anthocyanin in infected areas ( Banuett and Herskowitz , 1996; Brefort et al . , in press ) . Up to date , anthocyanin induction was thought to result from unspecific biotic stress caused by microbial infection ( Steyn et al . , 2002 ) , and flavonoid-pathway related compounds like luteolinidin have been shown to function as phytoalexins negatively affecting fungal growth ( Snyder and Nicholson , 1990; Zuther et al . , 2012 ) . So far , no positive link has been established between anthocyanin induction and microbial virulence . We have recently shown that the secreted U . maydis effector Tin2 ( Um05302 ) is specifically responsible for anthocyanin induction in infected seedling tissue ( Brefort et al . , in press ) . The tin2 effector gene resides in the largest effector gene cluster 19A comprising 24 effectors ( Kämper et al . , 2006; Brefort et al . , in press ) . Deletion of the entire cluster 19A largely abolishes tumor formation and anthocyanin induction , although the mutant is still able to grow and complete its life cycle inside the plant tissue ( Brefort et al . , in press ) . Here , we elucidate the molecular function of Tin2 and show that Tin2 is a secreted anthocyanin biosynthesis inducing effector protein functioning inside plant cells . We hypothesize that it may be beneficial for U . maydis to induce anthocyanin biosynthesis as a strategy to compete with tissue lignification and to gain access to veins . Tin2 was previously shown to be responsible for anthocyanin induction during biotrophic growth and was shown to have a weak contribution to virulence ( Brefort et al . , in press ) . To unambiguously establish its contribution to virulence , we introduced tin2 in this study into SG200Δ19A-1 ( Brefort et al . , in press ) , a mutant strain lacking the 14 leftmost effector genes of cluster 19A , including tin2 . This mutant is severely attenuated in virulence , is unable to elicit anthocyanin formation and induces at most small tumors ( Brefort et al . , in press ) . Introduction of tin2 enhanced tumor formation significantly and fully restored anthocyanin biosynthesis ( Figure 1A , B ) . In addition , we could demonstrate that tin2 was induced during biotrophic growth ( Figure 1C ) . Microscopic analysis and fungal biomass determination revealed that a SG200Δtin2 mutant did colonize plants efficiently , but was attenuated during late proliferation ( Figure 2 ) . To get an idea , which processes might be affected by Tin2 , we determined the transcriptional response of Z . mays to infection by SG200 and SG200Δtin2 . At 4 days post inoculation ( dpi ) , RNA was prepared from three biological replicates and analyzed by Affymetrix maize genome arrays as described before for other tin mutants ( Brefort et al . , in press ) . Hierarchical clustering analysis of differentially regulated genes showed that the expression patterns of plant genes in response to SG200Δtin2 was related to the response to SG200 ( Figure 3A ) , consistent with the small contribution of Tin2 to tumor formation . Among the top five genes most highly induced after SG200 infection ( compared to SG200Δtin2 ) , two were related to anthocyanin biosynthesis ( Figure 3—source data 1 ) and one of these was the second most highly induced gene when SG200-infected tissue was compared with mock inoculated tissue ( Figure 3—source data 2 ) . To substantiate these data , we performed quantitative real-time PCR ( qPCR ) for the five genes comprising the anthocyanin pathway ( Figure 3B ) using tissue from maize infected with SG200 and SG200Δtin2 , respectively ( Figure 3C ) . This analysis revealed that all five genes tested were significantly induced in SG200-infected tissue , illustrating that the anthocyanin pathway is transcriptionally induced when U . maydis provides the tin2 effector . Another pathway comprising three differentially regulated genes concerned the biosynthesis of the secondary metabolite 2 , 4-dihydroxy-7-methoxy-1 , 4-benzoxazin-3-one ( DIMBOA , Figure 3—source data 1; Figure 3D ) implicated in aphid and fungal resistance ( Ahmad et al . , 2011 ) . When the expression levels of eight genes comprising the entire DIMBOA biosynthesis pathway were analyzed by qPCR , we detected significant differences in expression for Bx1 , Bx2 , Bx5 and Bx8 ( Figure 3E ) in SG200- and SG200Δtin2-infected tissue while we could not detect differences in expression for Bx3 , Bx4 , Bx6 and Bx7 ( not shown ) . To determine whether the transcriptional changes detected can be linked to the presence or absence of respective metabolites , amounts of total anthocyanin were quantified photometrically at the anthocyanin specific absorption of 515 nm ( Figure 4A , B ) . This revealed an approximately 15-fold increase of anthocyanins in leaf material upon infection with the wild-type strain that was absent in the SG200Δtin2-infected leaves ( Figure 4B ) . By UPLC-PDA-TOF-MS analysis the identity of the three highly abundant anthocyanins in infected maize leaves were determined: cyanidin 3-glucoside ( 449 . 0996 Da , C21H21O11+ ) ; cyanidin 3- ( 6’’-malonylglucoside ) ( 535 . 1088 Da , C24H23O14+ ) ; cyanidin 3- ( 3’’ , 6’’-dimalonylglucoside ) ( 621 . 1092 Da , C27H25O17+ ) and confirmed by MS2 fragmentation studies ( Figure 4—source data 1 ) . These cyanidin derivatives were previously described to be the major anthocyanins in maize leaves ( Fossen et al . , 2001 ) . Relative amounts of these three cyanidins identified confirmed the overall anthocyanin profile ( Figure 4C ) . In the UPLC-PDA-TOF-MS analysis , we also identified flavonoids and DIBOA-glucoside to be reduced in tissue infected with the tin2 mutant relative to the corresponding wild type ( Figure 4D , E ) . Exact mass as well as MS2 spectra information allowed to identify the three flavonoids as: orientin 2’’-O-rhamnoside ( 594 . 1585 Da , C27H30O15 ) ; isoorientin 2’’-O-rhamnoside ( 594 . 1585 Da , C27H30O15 ) ; quercetin-glucoside-rhamnoside ( 610 . 1534 , C27H30O16 ) ( Figure 4—source data 1 ) . The biosynthesis of these flavonoids branches off the anthocyanin pathway and is likely a side effect of the upregulation of this pathway after infection by wild-type U . maydis ( Figure 3B ) . In case of DIMBOA biosynthesis we identified 2 , 4-dihydroxy-1 , 4-benzoxazin-3-one glucoside ( DIBOA-glucoside , 343 . 0903 Da , C14H17NO9 , Figure 4E ) to be regulated by Tin2 . These findings suggest that , besides the influence of Tin2 on anthocyanin , additional pathways may be affected directly or indirectly by tin2 . 10 . 7554/eLife . 01355 . 003Figure 1 . Expression and mapping of functional domains in Tin2 . ( A ) Biological function of truncated Tin2 proteins . Virulence of SG200Δ19A-1 strains expressing either wild-type or C-terminally truncated Tin2 proteins . Numbers indicate total infected plants . Dots indicate that at least one of the symptoms ( ligula swelling , small tumors , normal tumors , heavy tumors ) was significantly changed relative to SG200Δ19A-1 . ( B ) The same infected plants as in ( A ) were scored for anthocyanin pigmentation . Percentage of plants displaying anthocyanin coloration is indicated . Error bars depict standard deviation . ( C ) Quantification of tin2 gene expression during biotrophic development of U . maydis . Total RNA was extracted from leaves infected with SG200 at 2 , 4 , 6 , and 8 days post infection ( dpi ) , and also from cells grown in axenic culture in YEPSL medium . Transcript levels of tin2 during different growth stages of U . maydis strain SG200 were determined by quantitative real-time PCR . Constitutively expressed U . maydis peptidylprolyl isomerase ( ppi ) was used for normalization . Three biological replicates were analyzed , error bars depict standard deviation . tin2 expression in budding cells grown in axenic culture was set to 1 . 0 . ( D ) Amino acid sequence alignment of Tin2 proteins ( Um , U . maydis; Sr , S . reilianum; Us , U . scitaminea ) . Identical amino acids are boxed in yellow . Alpha helices ( orange bars ) and beta sheets ( black arrows ) are indicated . Proline-repeat sequences are marked with red lines . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 00310 . 7554/eLife . 01355 . 004Figure 1—figure supplement 1 . Substitution of the C-terminal 5 amino acids of Tin2 affects effector function . The C-terminal 5 amino acids ( PRFPL ) of Tin2 protein were substituted by alanine ( AAAAA ) and the respective gene ( tin2AAAAA ) was introduced in single copy in SG200Δ19A-1 . ( A ) Virulence of SG200Δ19A-1 and the derived strains expressing Tin2 or Tin2AAAAA was scored at 12 dpi following the scheme developed by Kämper et al . ( 2006 ) . The total number of infected plants is indicated above each panel . The expression of Tin2AAAAA did not complement tumor formation . ( B ) The same infected plants as in ( A ) were scored for anthocyanin induction . Tin2AAAAA very weakly complemented anthocyanin induction . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 00410 . 7554/eLife . 01355 . 005Figure 2 . Biotrophic development of SG200 and SG200Δtin2 . ( A ) Maize seedlings were infected by U . maydis strain SG200 and SG200Δtin2 and observed at 3 days post inoculation by confocal microscopy . Fungal hyphae were visualized by WGA-AF488 staining ( green ) . Plant cell walls were visualized by propidium iodide staining ( red ) . Bar = 100 µm . ( B ) Quantification of fungal biomass in SG200- and SG200Δtin2-infected tissue . Genomic DNA was extracted from leaf segments infected with SG200 and SG200Δtin2 at 2 and 6 days post inoculation . Quantitative real-time PCR was performed on genomic DNA . Relative fungal biomass was calculated from the U . maydis ppi gene and the Z . mays GAPDH gene , after setting the ratio of SG200 at 2 dpi to 1 . 0 . Error bars indicate the standard deviation of three biological replicates . p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 00510 . 7554/eLife . 01355 . 006Figure 3 . Differentially expressed maize genes in leaves infected with SG200 and SG200Δtin2 . ( A ) Plant transcriptome analysis of SG200- and SG200Δtin2-infected tissue . Hierarchical clustering was performed by the Partek Genomics Suite version 6 . 12 to visualize expression of maize genes transcriptionally regulated 4 days after mock inoculation ( left ) infection by U . maydis strain SG200 ( middle ) and infection by SG200Δtin2 ( right ) . The X-axis depicts clustering of the microarray samples for each of the three biological replicates per inoculation . The Y-axis shows clustering of the regulated maize transcripts based on similarity of their expression patterns . Red: upregulated genes; blue: downregulated genes . ( B ) Schematic model of the flavonoid biosynthetic pathway in maize . The enzymes involved in the discrete biosynthetic steps are shown in red . CHS , chalcone synthase; CHI , chalcone isomerase; F3H , flavanone 3-hydroxylase; DFR , dihydroflavonol 4-reductase; ANS , anthocyanin synthase . ( C ) qPCR based quantification of the expression levels of genes from the flavonoid pathway after infection with indicated U . maydis strains and collecting infected leaf material 6 dpi . p<0 . 01 , **p<0 . 001 . ( D ) Schematic model of the DIMBOA biosynthetic pathway in maize . The enzymes involved in the discrete biosynthetic steps are shown in red . Bx1 , indole-3-glycerol phosphate lyase; Bx2 , indole monooxygenase; Bx3 , indolin-2-one monooxygenase; Bx4 , 3-hydroxyindolin-2-one monooxygenase; Bx5 , HBOA monooxygenase; Bx8 , DIBOA UDP-glucosyltransferase; Bx6 , 2-oxoglutarate-dependent dioxygenase; Bx7 , TRIBOA-glucoside methyltransferase . ( E ) qPCR based quantification of the expression levels of genes from the DIMBOA pathway after infection with indicated U . maydis strains and collecting infected leaf material 6 dpi . p<0 . 05 , *p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 00610 . 7554/eLife . 01355 . 007Figure 3—source data 1 . List of upregulated maize genes after SG200 infection ( vs SG200Δtin2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 00710 . 7554/eLife . 01355 . 008Figure 3—source data 2 . List of differentially regulated maize genes after SG200 infection ( vs mock ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 00810 . 7554/eLife . 01355 . 009Figure 4 . Identification of anthocyanins and other metabolites in infected maize leaves . ( A ) Anthocyanin was extracted from leaves syringe-inoculated with H2O ( mock ) , SG200 , and SG200Δtin2#3 . Two cm long leaf segments located 1 cm below the injection holes were excised , frozen in liquid nitrogen , ground and extracted with 100% ethanol . For each sample , 5 leaf segments were combined . The supernatant from SG200-infected leaves showed red color after acidification with concentrated HCl while the other two samples stayed green . ( B–E ) Infected leaf segments from mock-inoculated leaves and leaves infected with SG200 , three independent SG200Δtin2 mutants and SG200Δtin2#3-tin2 were collected at 6 dpi . Excised leaf segments were ground in liquid nitrogen and served as a starting material for all subsequent analyses . ( B ) Measurement of total anthocyanin content . Anthocyanins were extracted and their amounts were measured at 515 nm using cyanidin 3-arabinoside chloride for calibration . Error bars indicate the standard deviation of three biological replicates . p<0 . 001 ( vs SG200 ) . ( C ) Specific accumulation of the three major anthocyanins: cyanidin 3-glucoside , cyanidin 3- ( 6″-malonylglucoside ) , and cyanidin 3- ( 3″ , 6″-dimalonylglucoside ) . Powdered samples were extracted and the polar phase was measured by UPLC-PDA-TOF-MS as described ( Djamei et al . , 2011 ) . Data shown are representative of three biological replicates . p<0 . 001 ( vs SG200 ) . ( D ) Decreased accumulation of the three flavonoids in SG200Δtin2-infected tissue: orientin 2’’-O-rhamnoside , isoorientin 2’’-O-rhamnoside , and quercetin-glucoside-rhamnoside . Powdered samples were extracted and the polar phase was measured by UPLC-PDA-TOF-MS as described ( Djamei et al . , 2011 ) . p<0 . 001 ( vs SG200 ) . ( E ) Decreased accumulation of DIBOA-glucoside in SG200Δtin2-infected tissue . Powdered samples were extracted and the polar phase was measured by UPLC-PDA-TOF-MS as described ( Djamei et al . , 2011 ) . Data shown are representative of three biological replicates . p<0 . 05 , *p<0 . 001 ( vs SG200 ) . ( F ) Decreased accumulation of ferulic acid in SG200Δtin2-infected tissue: powdered samples were extracted and the polar phase was measured by UPLC-PDA-TOF-MS as described ( Djamei et al . , 2011 ) . Data shown are representative of three biological replicates . p<0 . 001 ( vs SG200 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 00910 . 7554/eLife . 01355 . 010Figure 4—source data 1 . Markers identified by metabolite fingerprinting ( UPLC-ESI-TOF-MS analysis ) in leaves of Z . mays 6 days post infection and verified by UHPLC-ESI-QTOF-MS/MS analysis or coelution . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 010 The Tin2 protein does not reveal any conserved InterPro domains . Orthologous proteins are absent in other organisms examined to date but exist in the related smut fungi Sporisorium reilianum and U . scitaminea . An amino acid alignment revealed significant sequence conservation in the C-terminal third of the protein , whereas the N-terminus showed less conservation ( Figure 1D ) . The C-terminal region contains two proline repeat sequences ( PxxP ) , suggesting that this domain might be important for the interaction with other proteins ( Saksela et al . , 1995; Kay et al . , 2000; Dornan et al . , 2003 ) ( Figure 1D ) . Based on the high conservation of the C-terminus ( Figure 1D ) a series of C-terminally truncated Tin2 proteins were expressed in SG200Δ19A-1 ( Brefort et al . , in press ) to test for complementation of tumor formation and anthocyanin induction ( Figure 1A , B ) . Tin21-206 truncated by one amino acid complemented as efficiently as full-length Tin21-207 ( Figure 1A , B ) , whereas Tin21-202 lacking five amino acid residues at the C-terminus could neither complement for tumor formation nor anthocyanin induction ( Figure 1A , B ) . Tin2AAAAA protein , in which the five most C-terminal amino acids were substituted by alanine , was also largely unable to complement ( Figure 1—figure supplement 1 ) . These results indicated that the C-terminal region of Tin2 is crucial for tumor formation as well as anthocyanin induction . To demonstrate that the Tin2 protein can be secreted , we constructed a strain constitutively expressing a biologically active Tin2-HA fusion protein ( Figure 5A , B ) and showed that full-length Tin2-HA was detectable in the culture supernatant ( Figure 5C ) . To visualize protein secretion during biotrophic development a Tin21-25-mCherry-Tin226-207 fusion protein was expressed in SG200Δtin2 under the cmu1 promoter , which confers strong expression during plant colonization ( Djamei et al . , 2011 ) . After infection , mCherry fluorescence could be observed around fungal hyphae ( Figure 6A ) . When subjected to plasmolysis , which enlarges the apoplastic compartment , the fluorescence of Tin21-25-mCherry-Tin226-207 resided in the apoplast ( Figure 6B ) , demonstrating secretion ( Doehlemann et al . , 2009 ) . Strain SG200Δtin2-Tin21-25-mCherry-Tin226-207 elicited anthocyanin induction , but compared to SG200 the appearance was delayed ( Figure 6—figure supplement 1 ) , suggesting that the Tin21-25-mCherry-Tin226-207 fusion protein is only partially functional . In addition , mCherry-Tin226-207 was never detected in the cytosol of infected plants . Since anthocyanin biosynthesis is occurring in the cytosol of plant cells ( Holton and Cornish , 1995 ) , we expected that Tin2 would translocate to plant cells after being secreted . To demonstrate a function of Tin2 in the plant cytosol , we expressed tin2 without its secretion signal under the CaMV 35S promoter in maize leaf epidermal cells after biolistic bombardment . As this did not induce anthocyanin ( Figure 6C , leftmost panel ) , we repeated the experiment with leaves from plants infected with SG200Δtin2 . Here , anthocyanin could be induced by Tin226-207 ( Figure 6C , third panel ) , was absent in infected leaves without tin2 expression ( Figure 6C , second panel ) and could also not be observed when biologically inactive Tin226-202 was expressed ( Figure 6C , rightmost panel ) . To confirm that anthocyanin is produced in SG200Δtin2-infected tissue transiently transformed with tin2 , p35S-Tin226-207 and p35S-mCherry as transformation control were introduced simultaneously in SG200Δtin2-infected tissue . Anthocyanin production , detected by its dark color in bright field microscopy , was observed in the bombarded cell that also expressed cytosolic mCherry ( Figure 6D ) . Interestingly , cells neighboring the transformed cell also accumulated anthocyanin ( Figure 6D , red arrow ) . When biolistic transformation of uninfected maize leaves was carried out with a plasmid expressing mCherry-Tin226-207 , the fusion protein was found to spread into non-transformed neighboring cells ( Figure 6E ) . Symplastic spreading was observed for the Cmu1 effector of U . maydis and for several translocated effectors of Magnaporthe oryzae ( Khang et al . , 2010; Djamei et al . , 2011 ) . Collectively , these results demonstrate that Tin2 functions within the plant cytosol and suggest that Tin2 may prepare maize cells for the infection by U . maydis . 10 . 7554/eLife . 01355 . 011Figure 5 . Biological activity of Tin2-HA and demonstration of its secretion . SG200Δ19A-1-PotefTin2-HA expresses Tin2-HA protein under the constitutive otef promoter ( Spellig et al . , 1996 ) . ( A ) Biological activity of Tin2-HA protein was confirmed by assaying virulence of SG200Δ19A-1-PotefTin2-HA and comparing this to SG200Δ19A-1-Tin2 . ( B ) Complementation of anthocyanin induction by SG200Δ19A-1 strains by expressing Tin2-HA protein . The same infected plants as in ( A ) were scored for anthocyanin induction , which is given in % of all plants infected . ( C ) Detection of Tin2-HA protein in culture supernatants . Cells of SG200Δ19A-1 ( − ) and SG200Δ19A-1-PotefTin2-HA ( s , single integration; m , multiple integration ) were grown in CM liquid medium to an OD600 of 0 . 5 . Proteins in supernatants were collected after TCA/DOC-precipitation and used in western blot analysis . The western blot was developed with anti-HA antibody ( Sigma-Aldrich ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 01110 . 7554/eLife . 01355 . 012Figure 6 . Tin2 protein secretion during biotrophic development and its function inside plant cells . ( A ) Tin2 protein secretion during plant infection . Leaves are infected with SG200Δtin2-Tin21-25-mCherry-Tin226-207 . mCherry fluorescence is observed at 3 dpi . Bar = 10 µm . ( B ) Same as ( A ) but treated with 1 M NaCl to induce plasmolysis . Asterisks mark enlarged apoplastic spaces . Bar = 10 µm . ( C ) Anthocyanin induction by transient expression of Tin226-207 . Plasmid constructs indicated were delivered by biolistic gene transfer in maize leaves or leaves infected with SG200Δtin2 . ( D ) Confocal microscopy of cells co-expressing Tin226-207 and mCherry . Dark staining indicates anthocyanin accumulation . Bar = 20 µm . ( E ) Confocal microscopy of mCherry-Tin226-207 transiently expressed in maize epidermal cells . The initially transformed cell is indicated by an arrow . Bar = 30 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 01210 . 7554/eLife . 01355 . 013Figure 6—figure supplement 1 . Biological activity of secreted mCherry-Tin226-207 . ( A ) Virulence of SG200Δtin2-Pcmu1Tin21-25-mCherry-Tin226-207 ( labeled mCherry-Tin2 ) in comparison to SG200Δtin2 . Disease symptoms were assessed at 12 dpi as described in legend to Figure 1—figure supplement 1 . Tin21-25-mCherry-Tin226-207 was able to complement the virulence phenotype of SG200Δtin2 . ( B and C ) The same infected plants as in ( A ) were scored at 6 dpi ( B ) and 12 dpi ( C ) for anthocyanin induction . Anthocyanin induction indicated in percent of total number of plants infected . Compared to SG200 , SG200Δtin2-Pcmu1Tin21-25-mCherry-Tin226-207 was delayed and less efficient in anthocyanin induction . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 013 To gain insight into the plant processes affected by Tin2 , we identified Tin2 interacting proteins by yeast two-hybrid screening . After rescreening , only one putative cytoplasmic serine/threonine protein kinase from maize , designated ZmTTK1 ( Tin2-targeting kinase 1 ) could be confirmed ( not shown ) . ZmTTK1 is an uncharacterized kinase , with its kinase domain located at the C-terminus ( Figure 7A ) . Maize may contain two putative paralogs of ZmTTK1 ( GRMZM2G088409; GRMZM2G068192 ) with highly conserved kinase domains ( 63 and 65% identity ) , but reduced homology in the N-terminal domains ( 34% and 36% identity , respectively ) . Orthologs are found in monocot plants like Sorghum bicolor , Oryza sativa , Hordeum vulgare , and Brachypodium distachyon and these display high amino acid sequence identity with ZmTTK1 from maize ( between 75% and 94% ) ( Figure 7A , B ) . Arabidopsis thaliana also encodes a putative ortholog of ZmTTK1 ( AT2G40120 ) , but in this case the amino acid identity is significantly lower ( 57% ) ( Figure 7A , B ) . ZmTTK1 orthologs from monocot plants display highly conserved N- and C-terminal domains , which are separated by a more variable central domain ( Figure 7A , C ) . In the related A . thaliana protein , the N-terminal region shows only patchy homology to the monocot proteins and the central domain is missing ( Figure 7A ) . None of these proteins have been functionally characterized to date . 10 . 7554/eLife . 01355 . 014Figure 7 . Comparison of ZmTTK1 orthologs from different grasses . ( A ) Amino acid sequence alignment of ZmTTK1 orthologs . ZmTTK1 orthologs ( Zm , Zea mays ) could be identified in Sorghum bicolor ( Sb ) , Oryza sativa ( Os ) , Hordeum vulgare ( Hv ) , Brachypodium distachyon ( Bd ) and a related protein found in Arabidopsis thaliana ( At ) . Amino acids conserved in all 5 orthologs from monocot plants and the related protein in A . thaliana are highlighted in black , amino acids conserved in 4 or 5 of 6 these proteins are highlighted in dark gray and amino acids conserved in 3 of the 6 proteins are highlighted in light gray . The phosphodegron-like motif is indicated by a blue line . The kinase domain is indicated by a red line . ( B ) Phylogenetic tree of ZmTTK1 orthologs in monocot plants and the related protein from A . thaliana . Amino acid sequences of these proteins were analyzed by Phylogeny . fr to calculate the phylogenetic relationship . Bar indicates evolutionary distance . Accession numbers are indicated . ( C ) Schematic structure of ZmTTK1 . The N-terminal region is labeled in gray and the C-terminal kinase domain labeled in orange . The phosphodegron-like motif is indicated in blue . Amino acid sequence identity of the N-terminal middle and C-terminal regions among orthologs in monocots is indicated above . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 014 In yeast two-hybrid assays , full-length ZmTTK1 was able to interact with Tin226-207 and Tin226-206 , while no interaction was observed with Tin226-202 lacking biological activity ( Figure 8A ) . Physical interaction between recombinant Strep-ZmTTK1 protein and recombinant GST-Tin226-207 protein was demonstrated by co-IP using GST-Tin226-202 as negative control ( Figure 9A ) . Recombinant ZmTTK1 purified from E . coli was expressed as full-length protein ( Figure 9B ) and showed in vitro kinase activity determined with MBP ( myelin basic protein ) as substrate . However , no significant differences in MBP phosphorylation were apparent in the presence or absence of Tin2 ( Figure 9C ) . To examine whether Tin2 protein affects ZmTTK1 activity in vivo , we transiently expressed ZmTTK1 protein in Nicotiana benthamiana . In this system , ZmTTK1 protein proved highly unstable and the N-terminal region could not be detected ( Figure 10A ) . This region contains a phosphodegron-like motif DSGxS , which , when phosphorylated at serine residues , serves as a recognition motif for the ubiquitin ligase complex leading to protein degradation via the ubiquitin-proteasome system ( Ravid and Hochstrasser , 2008; Spoel et al . , 2009 ) . Addition of proteasome inhibitor MG132 to N . benthamiana cells transiently expressing ZmTTK1-HA allowed detection of full-length protein ( Figure 10B ) . In addition , ZmTTK1S279/282A-HA protein carrying alanine substitutions at positions serine 279 and serine 282 in the phosphodegron-like motif could be expressed as full-length protein in the transient N . benthamiana system ( Figure 10B ) . Furthermore , poly-ubiquitinated ZmTTK1-HA protein accumulated in ZmTTK1-HA expressing plant cells treated with MG132 but not in ZmTTK1S279/282A-HA expressing N . benthamiana cells ( Figure 10C ) . These results indicate that the phosphodegron-like motif in ZmTTK1 promotes partial degradation via the ubiquitin-proteasome system . 10 . 7554/eLife . 01355 . 015Figure 8 . Mapping the interaction regions between Tin2 and ZmTTK1 by yeast two-hybrid analysis . ( A ) A series of C-terminally truncated Tin2 proteins were expressed in fusion with GAL4BD ( BD ) protein and their interaction with full-length ZmTTK1 protein fused with GAL4AD ( AD ) were tested by yeast two-hybrid assay . Yeast transformants were grown on SD medium lacking indicated amino acids . Interaction was assessed from growth on SD -Trp -Leu -His + 3 mM 3-AT medium and from SD -Trp -Leu + X-α-Gal medium . Protein expression was verified by western blot with anti-myc antibody ( Sigma-Aldrich ) ( bottom panel ) . ( B ) Mapping of the Tin2-interacting region of ZmTTK1 . A series of truncated ZmTTK1 genes were fused with GAL4AD ( AD ) protein and their interaction with Tin226-207 protein fused with GAL4BD ( BD ) was assessed by yeast two-hybrid assay . Interaction was shown by growth of respective strains on SD -Trp -Leu -His + 3 mM 3-AT medium . Truncated ZmTTK1 versions are drawn schematically , the kinase domain is depicted in orange , and the N-terminal domain is colored in gray . The phosphodegron-like motif DSGxS is indicated in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 01510 . 7554/eLife . 01355 . 016Figure 9 . Physical interaction of Tin2 protein with full-length ZmTTK1 and in-vitro kinase activity of ZmTTK1 . ( A ) Physical interaction of Tin2 and ZmTTK1 demonstrated by in vitro GST-pull down assay using recombinant proteins . Recombinant GST-Tin226-207 or GST-Tin226-202 protein bound to glutathione sepharose beads ( GE healthcare ) , respectively , was incubated with extract from induced BL21 ( DE3 ) /pPRIBA102-ZmTTK1 at 4°C for 1 hr . GST fusion proteins were eluted with reduced glutathione . Strep-ZmTTK1 in eluate was detected by western blot ( WB ) using Strep-tactin-HRP and α-GST antibody for detection , respectively . Top panel detects precipitated Strep-ZmTTK1 , bottom panel detects input GST-Tin2 fusion proteins . ( B ) Full-length recombinant Strep-ZmTTK1 expressed and purified from E . coli was detected by western blot analysis with Strep-tactin-HRP . ( C ) In vitro kinase activity of recombinant Strep-ZmTTK1 protein . Recombinant GST-Tin226-207 and Strep-ZmTTK1 proteins at indicated concentration were pre-incubated for 60 min at 4°C followed by addition of γ-32P ATP and myelin basic protein ( MBP ) . Bovine serum albumin ( BSA ) was used as a control . Incubation continued at 28°C for 30 min . Proteins were separated by SDS-PAGE and phosphorylated protein was detected . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 01610 . 7554/eLife . 01355 . 017Figure 10 . Tin2 protein stabilizes cytoplasmic maize protein kinase ZmTTK1 . ( A ) ZmTTK1-HA and HA-ZmTTK1 expression in N . benthamiana . HA-ZmTTK1 or ZmTTK1-HA protein was transiently expressed in N . benthamiana after infiltration with the respective A . tumefaciens strains GV3101 carrying pBIN19AN-HA-ZmTTK1 and pBIN19AN-ZmTTK1-HA . These plasmids express the indicated fusion proteins with a C-terminal IgG binding site . Expression was shown by western blot ( WB ) using indicated antibody . Asterisk labels a non-specific band . Rubisco large subunit ( RBCL ) stained with coomassie brilliant blue ( CBB ) served as a loading control . ZmTTK1-HA transcripts were analyzed by RT-PCR , EF1α served as a control . ( B ) Protein expression of ZmTTK1-HA with proteasome inhibitor MG132 ( 100 µM ) and ZmTTK1S279/282A-HA . ( C ) Detection of poly-ubiquitinated ZmTTK1 . After immunoprecipitation with human IgG-agarose , proteins were subjected to western blot to detect poly-ubiquitinated ZmTTK1 protein ( arrow ) . The western blot was developed with monoclonal anti-ubiquitin antibody ( Sigma-Aldrich ) . ( D ) Co-expression of ZmTTK1-HA with myc-Tin2 . Total protein was analyzed by western blot using indicated antibodies . ( E ) Immunoprecipitated ZmTTK1-HA protein from ( D ) was analyzed by western blot . Kinase activity of immunoprecipitated samples shown on top was analyzed using MBP as a substrate ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 017 The region in ZmTTK1 that interacts with Tin2 was mapped by yeast two-hybrid assays and shown to coincide with the variable region containing the phosphodegron-motif ( Figure 8B ) . Next , when ZmTTK1-HA was co-expressed in N . benthamiana together with functional Tin226-207 , we detected full-length ZmTTK1 , but not when non-functional truncated Tin226-202 was co-expressed with ZmTTK1 ( Figure 10D , E ) . Furthermore , immunoprecipitated ZmTTK1-HA co-expressed with Tin226-207 showed MBP phosphorylation activity in vitro while expression of ZmTTK1-HA alone or co-expression of ZmTTK1-HA with truncated Tin226-202 did not yield an immunoprecipitate with phosphorylation activity ( Figure 10E , bottom panel ) . This illustrates that Tin2 protein targets the phosphodegron-containing region of ZmTTK1 , stabilizing full-length , active ZmTTK1 . To rule out that these effects are specific to the heterologous N . benthamiana system , ZmTTK1-YFP was transiently expressed in maize with and without co-expression of mCherry-Tin226-207 . When ZmTTK1-YFP and mCherry-Tin226-207 were co-expressed , the number of cells showing a YFP signal was increased significantly compared to the situation when ZmTTK1-YFP was expressed alone , and in addition , a significant percentage of cells showed strong YFP fluorescence in the co-expression conditions ( Figure 11A , B ) . This suggests that the ZmTTK1-YFP fusion protein is also unstable in maize epidermal cells . Co-expression of Tin2 stabilizes the fusion protein , similar to what was observed in N . benthamiana . 10 . 7554/eLife . 01355 . 018Figure 11 . ZmTTK1 effects on localization of maize transcription factors and anthocyanin biosynthesis . ( A ) Fluorescence intensity of ZmTTK1-YFP co-expressed with mCherry-Tin226-207 or with mCherry , respectively , in transiently transformed maize cells . Transformed maize cells were first identified by their mCherry fluorescence ( not shown ) and these cells were then screened for YFP fluorescence . The three examples depict the range of ZmTTK1-YFP expression patterns seen . ( B ) Quantification of the three YFP fluorescence patterns indicated in ( A ) . Expressed proteins are listed below panels . In total , 90 transformed cells were scored for YFP fluorescence from three independent experiments . A statistically significant increase of weak and strong fluorescence patterns is observed when ZmTTK1-YFP is co-expressed with mCherry-Tin226-207 compared to co-expression of ZmTTK1-YFP with mCherry . ( C ) Anthocyanin induction after transient expression in maize . p35S-ZmTTK1S279/282A and p35S-mCherry were co-expressed in SG200Δtin2-infected leaves . Anthocyanin was visualized by bright field microscopy 3 days after particle bombardment . Bar = 20 µm . ( D ) mCherry-ZmR1 or mCherry-ZmC1 protein localization pattern . mCherry fluorescence in 60 cells from three different leaves was classified into cytoplasmic and nuclear localization or exclusive nuclear localization . ( E ) Confocal microscopy of mCherry-ZmR1 . p35S-mCherry-ZmR1 was transiently introduced into maize either alone ( − ) or together with p35S-Tin226-207 and p35S-ZmTTK1 . mCherry-ZmR1 was visualized . Bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 018 To elucidate whether there is a direct link between ZmTTK1 activity and anthocyanin induction , the stabilized form of the kinase , ZmTTK1S279/282A , was transiently co-expressed with cytoplasmic mCherry in leaves infected with SG200Δtin2 . In transiently transformed cells weak anthocyanin induction could be observed ( Figure 11C ) . In maize anthocyanin biosynthesis genes are positively regulated via ZmR1 and ZmC1 transcription factors ( Ludwig and Wessler , 1990; Dooner et al . , 1991 ) . Therefore , we co-expressed Tin226-207 and ZmTTK1 in uninfected maize leaves together with mCherry-ZmR1 or mCherry-ZmC1 protein , respectively . mCherry-ZmC1 localized to the nucleus , and this localization did not change significantly in the presence of Tin226-207 and ZmTTK1 ( Figure 11D ) . When mCherry-ZmR1 was expressed alone , the fusion protein showed weak cytoplasmic and nuclear localization ( Figure 11E ) . However , when co-expressed with Tin226-207 and ZmTTK1 , mCherry-ZmR1 protein showed almost exclusively intense nuclear localization ( Figure 11D , E ) . These results suggest that ZmTTK1 protein kinase stabilized by Tin2 might phosphorylate the ZmR1 protein , which is then imported into the nucleus ( or is retained there ) and induces anthocyanin biosynthetic genes . Since anthocyanin precursors accumulate in the plant cytosol and anthocyanin is transported to the vacuole , two compartments U . maydis is not in direct contact with due to its biotrophic life style , we next considered that induction of the anthocyanin pathway might deplete precursors for other pathways , which could be harmful for U . maydis development . In the general phenylpropanoid pathway , phenylalanine is converted to 4-coumaroyl CoA , which then serves as a precursor for both , anthocyanins and lignin ( Figure 12A ) . Since we had shown that all genes in the anthocyanin branch of the pathway are upregulated in tissue infected with SG200 , but not in tissue infected with SG200Δtin2 ( Figure 3 ) , we determined abundance of transcripts encoding key enzymes in the general and lignin-specific branches of the phenylpropanoid pathway by qPCR in leaves infected with strains SG200 and SG200Δtin2 . qPCR analysis demonstrated that the phenylalanine ammonia lyase ( PAL ) gene in the general part of the pathway was significantly upregulated in both SG200- and SG200Δtin2-infected tissue compared to mock inoculation ( Figure 12B ) . One of the two ‘cinnamoyl-CoA reductase’ ( CCR ) genes of maize one was not induced , while the other one ( GRMZM2G131836 ) was clearly induced after U . maydis infection , but there was no difference in induction in SG200- and SG200Δtin2-infected leaves ( Figure 12 ) . The major 4-coumarate-CoA ligase ( 4CL ) gene in the lignin-specific branch ( GRMZM2G075333; based on homology with sorghum; Saballos et al . , 2012 ) was not differentially regulated , while the flavonoid-related 4CL ( GRMZM2G054013 ) was downregulated in SG200Δtin2-infected tissue compared to SG200-infected tissue ( Figure 12B ) . Of the six cinnamyl alcohol dehydrogenase ( CAD ) genes , four genes were significantly upregulated in SG200Δtin2-infected tissue ( Figure 12B ) . The enhanced expression of several lignin biosynthesis genes after infection with the tin2 mutant coincided with an transcriptional upregulation of several plant cell wall remodeling genes ( Figure 12—source data 1 ) and a decrease in ferulic acid ( Figure 4F ) , which is involved in crosslinking hemicellulosic polysaccharides and provides a nucleation site for lignification ( Grabber et al . , 2002 ) , suggesting a connection between these processes . Furthermore , in a maize chalcone synthase mutant ( c2 ) , in which the biosynthetic pathway to anthocyanins is blocked , and in which there could conceivably be more precursor going into the lignin branch , virulence of U . maydis SG200 was attenuated ( Figure 13 ) . 10 . 7554/eLife . 01355 . 019Figure 12 . Phenylpropanoid pathway genes are differentially expressed after infection with SG200 and SG200Δtin2 . ( A ) Schematic model of the phenylpropanoid pathway in maize . The enzymes involved in each biosynthetic step are shown in red . PAL , phenylalanine ammonia lyase; C4H , cinnamate 4-hydroxylase; 4CL , 4-coumarate-CoA ligase; HCT , hydroxycinnamoyl transferase; C3′H , p-coumaroyl shikimate 3’-hydroxylase; CCoAOMT , caffeoyl CoA 3-O-methyltransferase; F5H , ferulate 5-hydroxylase; COMT , caffeic acid 3-O-methyltransferase; CCR , cinnamoyl-CoA reductase; CAD , cinnamyl alcohol dehydrogenase; CHS , chalcone synthase . Dashed lines indicate routes , which may occur in maize under certain conditions , but are not considered major routes based on experimental evidence ( Humphreys and Chapple , 2002 ) . ( B ) qPCR based quantification of the expression levels of genes from the phenylpropanoid pathway after infection with indicated U . maydis strains and collecting infected leaf material 6 dpi . p < 0 . 05 , *p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 01910 . 7554/eLife . 01355 . 020Figure 12—source data 1 . List of upregulated maize genes after SG200Δtin2 infection ( vs SG200 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 02010 . 7554/eLife . 01355 . 021Figure 13 . Pathogenicity of U . maydis in maize lacking chalcone synthase . ( A ) Macroscopic SG200 disease symptoms on C2 ( control ) and c2 ( chalcone synthase mutant ) after 12 dpi are shown . ( B ) Disease symptoms were scored in two independent SG200 infections using the scoring scheme depicted on the right and the data were combined . ( C ) From the plants scored in ( B ) the percentage of plants showing anthocyanin induction is given . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 021 Quantification of total lignin content from isolated cell walls in wild-type maize ( cv . Early Golden Bantam ) revealed a small increase in leaf tissue infected with tin2 mutant strains relative to SG200-infected material that was statistically significant in one of three analyzed mutants ( Figure 14A ) . Collectively , these results suggest that anthocyanin induction by Tin2 protein may negatively affect the lignin pathway . To determine whether lignin content affects virulence of U . maydis , we infected single and double brown midrib ( bm ) mutants of maize with SG200 ( Figure 14B–D ) . The bm1 mutation reduces CAD activity ( Halpin et al . , 1998 ) as a result of a mutation in the ZmCAD2 gene ( Chen et al . , 2012 ) , the major CAD involved in lignification of vascular tissue in leaves and stems . The leaf midribs of the bm2 mutant contain less lignin , and the tissue-specific pattern of lignification in the leaves is altered ( Vermerris and Boon , 2001 ) . Lignin content in the leaves and stems of the bm3 mutant is reduced , lignin subunit composition has shifted towards a higher guaiacyl/syringyl ( G/S ) ratio , and novel benzodioxane structures are present as a result of the incorporation of 5-hydroxyconiferyl alcohol ( Lapierre et al . , 1988; Chabbert et al . , 1994; Marita et al . , 2003 ) . This is the result of a mutation in the gene encoding caffeic acid O-methyl transferase ( COMT ) ( Vignols et al . , 1995 ) . The bm4 mutation has not yet been mapped to a specific gene , but affects the cell wall composition of leaves and lowers the lignin content of leaf midribs ( Vermerris et al . , 2010 ) . Except for the bm2 mutant , all other single bm mutants displayed elevated virulence symptoms after U . maydis infection , with bm3 showing the strongest effect ( Figure 14C ) . Of the double mutants , the bm3-bm4 mutant showed dramatically enhanced disease symptoms , while disease symptoms in the bm2-bm3 double mutant was lower than in the single bm3 mutant ( Figure 14C ) , suggesting a negative effect of the bm2 mutation on virulence in this combination . When bm2 was combined with bm1 , virulence was comparable to the single bm1 mutant ( Figure 14C ) . The alterations in virulence in the bm mutants were not only visible in symptom severity but also with respect to symptom distribution ( Figure 14D ) . After infection with SG200 , more than 80% of leaf tumors stayed locally confined , while in the bm mutants ( with the exception of the single bm2 mutant ) leaf tumors were much more spread and covered the entire leaf area from the injection hole to the stem ( Figure 14B , D ) , suggesting that lignification restricts U . maydis growth and spread in the infected tissue . In line with this assertion , we observed strongly lignified cells in vascular bundle tissue in cross sections of maize leaves infected with SG200Δtin2 and could only rarely observe fungal hyphae in bundle sheath cells ( Figure 14E ) . By contrast , lignification was significantly less pronounced in SG200-infected tissue ( Figure 14F ) and in tissue infected by SG200Δtin3 ( Figure 14G ) , a mutant lacking another effector from cluster 19A that does not affect anthocyanin induction ( Brefort et al . , in press ) . Additionally , SG200 and SG200Δtin3 hyphae were readily detected inside bundle sheath cells ( Figure 14F , G ) . This suggests that anthocyanin induction by the Tin2 effector attenuates lignification of vascular bundle cells , conceivably facilitating fungal access to nutrients in vascular bundle tissue . 10 . 7554/eLife . 01355 . 022Figure 14 . Virulence of U . maydis in hosts with altered lignin . ( A ) Measurement of total lignin content in leaves infected with U . maydis . Infected leaf segments from mock inoculated leaves and leaves infected with three independent SG200Δtin2 strains as well as SG200 were collected at 6 dpi . Excised leaf segments were ground in liquid nitrogen and extracted with methanol . Isolated lignin was hydrolyzed under alkaline conditions and amount of monomers was measured at 280 nm using coniferyl alcohol for calibration . Error bars indicate the standard deviation of three biological replicates . p<0 . 01 ( vs SG200 ) . ( B ) Macroscopic disease symptoms of U . maydis SG200 on maize A619 and the double brown midrib mutant bm3-bm4 . Symptoms were scored at 12 dpi . ( C ) Pathogenicity of U . maydis SG200 on indicated bm mutants . Virulence was scored at 12 dpi . The total number of infected plants is indicated above each panel and data were combined from two independent experiments . ( D ) Symptoms of SG200 on bm mutants shown in ( C ) were classified into local tumors ( tumors are restricted in one area on the leaf blade ) , and spreading tumors ( tumors extend from infected leaf area all the way into the stem area ) . ( E–G ) Lignin deposition in cross-sections of plants infected with the indicated strains was visualized by Safranin O staining ( red ) . Fungal hyphae inside vascular bundles are indicated by arrows . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 022 Comprehensive genetic , biochemical and transcriptome analyses allowed us to uncover a connection between anthocyanin and lignin biosynthesis that is targeted by a novel secreted effector of U . maydis , Tin2 . Our results are summarized in the model depicted in Figure 15 . The link between Tin2 and anthocyanin biosynthesis appears direct and occurs through the stabilization of the kinase ZmTTK1 via Tin2 . We have shown that after U . maydis infection , when Tin2 is present , the stabilization of ZmTTK1 will increase biosynthesis of anthocyanin . Conversely , when U . maydis is lacking tin2 , more of the common precursor for anthocyanin and lignin , 4-coumaric acid , is likely to be available for lignin biosynthesis ( Figure 15 ) resulting in cell wall fortification in vascular tissue . The deposition of lignin in the plant cell wall is considered to provide an undegradable physical barrier to infection ( Vance et al . , 1980 ) . The observed decrease in ferulic acid in SG200Δtin2-infected leaves compared to SG200-infected tissue is consistent with increased lignification where ferulic acid acts as a nucleation site for lignin polymerization ( Grabber et al . , 2002 ) . Tin2 affected all steps of the anthocyanin pathway , but did not affect transcription of all genes coding for lignin pathway components . While none of the early genes implicated in lignin biosynthesis were upregulated after infection with U . maydis wild type , we observed transcriptional effects for four of the six CAD genes acting at a late step of the lignin pathway ( Figure 12A ) . At present , there is only a limited understanding of how the lignin biosynthetic pathway in maize is regulated . Several Myb and NAC transcriptional activators involved in secondary cell wall formation have been identified ( Fornalé et al . , 2010; Zhong et al . , 2011 ) , but the mechanisms behind tissue-specific and developmentally regulated gene expression , and the role of posttranscriptional regulation of gene expression are yet to be fully elucidated . In tin2 mutant infected tissue total lignin levels appear increased , although statistically relevant effects were seen in only one of the mutants analyzed ( Figure 14A ) . This may reflect that increased lignification does not affect the entire leaf but , as we have shown , is restricted to the vascular tissue colonized by the tin2 mutant late in infection . Therefore , when total tissue is analyzed , differences in total lignin content may not be so evident in wild-type and tin2 mutant infected tissue . However , the increased susceptibility of maize bm mutants to U . maydis and the increased leaf areas displaying disease symptoms , provide strong evidence for a role of lignification in disease resistance . Several mechanisms may form the basis for this observation . A less lignified cell wall will be easier to penetrate by the fungus . The plant may also be less effective at producing defense-related lignin that could slow down the migration of the fungus . Lastly , since the fungus spreads via the vascular tissue and may rely on nutrients obtained from this tissue for massive proliferation , the altered physico-chemical characteristics of the vascular tissue in the bm mutants , combined with altered dimensions of the xylem vessels ( Vermerris et al . , 2010 ) may influence entry and the rate of migration of the fungus . The bm2 mutant appears to be an exception in that it does not show increased susceptibility to U . maydis , and even appears to attenuate the effect of infection in bm1-bm2 and bm2-bm3 double mutants relative to the bm1 and bm3 single mutants . This effect may be due to the accumulation of a fungitoxic ( phenolic ) compound in this mutant , as has been proposed for the sorghum brown midrib12 mutant , which was shown to be less susceptible to colonization by Fusarium spp . ( Funnell-Harris et al . , 2010 ) . Also in N . attenuata lines in which two CAD genes were silenced , compensatory processes for the structural deficiencies were observed when plants were grown in the field , and in this case such plants accumulated various phenylpropanoids ( Kaur et al . , 2012 ) . Therefore , while we consider direct effects of Tin2 on lignification to be the simplest model to explain the reduced virulence on the tin2 mutant strain ( Figure 15 ) , the situation may in reality be much more complex . For example we cannot formally rule out that in the absence of anthocyanin biosynthesis it is not actually the enhanced lignification that affects fungal spread but an altered composition of the plant cell walls . Conversely , the enhanced disease symptoms observed in the bm mutants could also be attributed to an lower incorporation of phenolics rather than to the reduction in lignin , that is in wild-type plants cell wall phenolics may negatively affect fungal growth and restrict pathogen spread . 10 . 7554/eLife . 01355 . 023Figure 15 . Hypothetical model for Tin2 function . In the upper part the hypothetical model for infection of maize with wild-type U . maydis is depicted . Tin2 effector ( red triangles ) is secreted , taken up by maize cells and binding to the ZmTTK1 maize kinase ( orange ellipsoid ) , which leads to its stabilization and renders it active . Active kinase is proposed to positively affect anthocyanin biosynthesis ( dark pink ) via the R transcription factor . Arrows and their thickness depict the activity of the anthocyanin and lignin pathway , respectively . The lower part depicts the hypothetical model for infection with an U . maydis tin2 mutant . In the absence of Tin2 , ZmTTK1 is degraded via the ubiquitin-proteasome system . As a result , there is no anthocyanin biosynthesis , which could conceivably allow more precursor to enter the lignin branch , leading to fortified cell walls ( brown lining ) that limit fungal access . DOI: http://dx . doi . org/10 . 7554/eLife . 01355 . 023 Reduction in lignin content and changes in lignin subunit composition have been shown to improve biomass conversion in a number of species ( Chen and Dixon , 2007; Studer et al . , 2011 ) , including maize , with the bm1 and bm3 mutations leading to 50% higher yields of fermentable sugars on a per-gram-biomass-basis ( Vermerris et al . , 2007 ) . Based on the experiments described here , it is clear that modifying lignin composition in maize has the risk of increased susceptibility to U . maydis , and the associated reduction in yield . Whether Tin2 is actively diverting the flux from lignin to anthocyanin remains speculative as flux measurements would require the absolute quantification of lignin and anthocyanin in the infected tissue and this is hampered by the technical difficulties in lignin extraction and quantification due to limited amounts of infected tissue material . In addition , the direct or indirect effects of Tin2 on other pathways like DIMBOA and flavonoid biosynthesis might complicate such analyses . However , precedence for a connection between lignin and anthocyanin biosynthesis comes from A . thaliana , where the silencing of HCT ( hydroxycinnamoyl transferase ) , the first committed enzyme from the lignin pathway , redirected the metabolic flux into flavonoids through chalcone synthase and this resulted in elevated levels of flavonols and anthocyanin ( Besseau et al . , 2007 ) . In addition , transgenic A . thaliana plants expressing the lignin biosynthesis repressing ZmMYB31 transcription factor from maize redirected carbon flux towards the biosynthesis of anthocyanins ( Fornalé et al . , 2010 ) . Recently , a competition between anthocyanin and lignin pathways for their common precursor has also been detected in strawberry ( Ring et al . , 2013 ) . In that system it was demonstrated that plant class III peroxidase ( FaPRX27 ) , a gene connected with lignin biosynthesis , is linked to a region implied in fruit color decrease . In addition , it was demonstrated that the downregulation of chalcone synthase lead to an induction of FaPRX27 and this diverted the flux from anthocyanins to lignin ( Ring et al . , 2013 ) . Thus , there appears to be a metabolic connection between anthocyanin and lignin biosynthesis pathways in monocot as well as dicot systems and U . maydis may be altering this with the help of Tin2 . Future work will be needed to address whether this connection is direct with lignin functioning as barrier or indirect with anthocyanin induction negatively affecting the accumulation of other defense compounds . Transient expression of the ZmR1 transcription factor known to positively regulate the anthocyanin pathway in maize revealed strong nuclear localization when ZmR1 was co-expressed with ZmTTK1 and Tin2 , but not when Tin2 was omitted . We suggest that the difference in ZmR1 localization may be caused by its phosphorylation , although we cannot exclude other possibilities . In A . thaliana MYB75 , which is related to maize ZmC1 , but is not the closest ortholog , acts as a repressor for the lignin-specific branch of the phenylpropanoid pathway presumably through an interaction with TT8 , a bHLH protein related to ZmR1 ( Bhargava et al . , 2010 ) . If such a scenario also exists in maize , the observed Tin2-mediated nuclear localization of ZmR1 might actually promote the formation of such a repressive complex for lignin biosynthesis . How Tin2 affects the expression of DIMBOA is currently unclear . It is conceivable that this could reflect a co-regulation with anthocyanin biosynthesis genes . Alternatively , the activity of another transcription factor might be modulated by the Tin2-stabilized ZmTTK1 protein . An upregulation of Bx1 and elevated amounts of DIMBOA had previously been detected after colonization of maize by U . maydis ( Basse , 2005 ) . In this context , it had also been demonstrated that U . maydis is resistant to DIMBOA ( Basse , 2005 ) while several other fungal pathogens of maize are susceptible ( Glenn et al . , 2002 ) . This could indicate that the induction of DIMBOA might deter infection by other pathogens in U . maydis infected tissue . However , in a macro-array based expression analysis of phenylpropanoid and related genes in brown midrib mutants , two genes from the DIMBOA pathway , Bx4 and Bx5 , were significantly upregulated in the bm3 mutant , that is the mutant with the strongest effect in lignin biosynthesis ( Guillaumie et al . , 2007 ) . The authors suggest that this may be a compensatory reaction for the lack of COMT activity in the bm3 mutant ( Guillaumie et al . , 2007 ) . If true , the upregulation of DIMBOA after SG200 infection could also be the consequence of the locally reduced lignin content in the infected areas . For its function , the Tin2 protein must translocate into plant cells after being secreted . Although an mCherry-Tin2 fusion protein produced by U . maydis was partially biologically active , we could never see the fusion protein accumulate inside maize cells . Based on similar results with Cmu1 , another transferred effector ( Djamei et al . , 2011 ) , we speculate that large fluorescent tags fused to the respective effector prevent uptake in this pathosystem . Cmu1 and Tin2 do not share obvious common motifs that can be recognized in their primary amino acid sequence that could serve as translocation signal , and it remains to be determined how they cross the plant plasma membrane . Anthocyanin induction after transient transformation of Tin2 could be observed only when plants were pre-infected with a tin2 mutant strain . We consider that it may be necessary to pre-infect plants to obtain high levels of phenylalanine or tyrosine for anthocyanin biosynthesis . These amino acids supply the flavonoid pathway and U . maydis infection was shown to increase their amounts in infected leaves through Cmu1 and possibly other effectors ( Doehlemann et al . , 2008; Djamei et al . , 2011 ) . In maize , anthocyanin induction is connected to kernel development ( Holton and Cornish , 1995 ) and is also observed when coleoptiles are exposed to UV-B ( Beggs and Wellman , 1985 ) . We hypothesize that there will be a native regulatory system stabilizing the ZmTTK1 protein kinase in such conditions , and Tin2 may mimic this regulator . Interestingly , ZmTTK1 orthologs in barley and B . distachyon have substitutions in critical serine residues of the phosphodegron motif DSGxS , and in the related kinase from A . thaliana this region is missing altogether ( Figure 7A ) . Since the barley smut U . hordei and the B . distachyon smut U . bromivora ( Barbieri et al . , 2011 ) both lack an ortholog of Tin2 ( Laurie et al . , 2012; AD and RK , unpublished data ) , this could suggest that the ongoing co-evolution between host and pathogen has already generated stable kinases in barley and B . distachyon , respectively , that no longer need respective orthologs of the Tin2 effector for activity . Anthocyanin serves for pollinator attraction of flowers ( Sheehan et al . , 2012 ) . In addition , its accumulation is frequently observed under abiotic and biotic stress conditions . Anthocyanin is transported and accumulates in the vacuole ( Goodman et al . , 2004 ) , provides UV-B stress protection and allows scavenging of reactive oxygen species ( Steyn et al . , 2002 ) . So far , the observed accumulation of anthocyanin after biotic stress was discussed in the frame of defense responses of resistant cultivars ( Hammerschmidt and Nicholson , 1977; Hipskind et al . , 1996 ) . In addition , some other examples , not connected to classical resistance responses , have been described: in a recent report it has been demonstrated that tomatoes enriched in anthocyanins are less susceptible to the necrotrophic fungus Botrytis cinerea . The enhanced resistance occurs through reduced spreading of the ROS burst , which limits the induction of cell death necessary for growth of this necrotroph ( Zhang et al . , 2013 ) . In potato it was shown that ectopic overexpression of an anthocyanin 5-O-glycosyl transferase conferred enhanced resistance to the bacterial pathogen Erwinia carotovora . In this system , it is not clear , however , whether the effects are direct or indirect ( Lorenc-Kukuła et al . , 2005 ) . In perennial ryegrass infected with the fungal endophyte Epichloae festucae , prominent anthocyanin induction is observed at the base of tillers , while fungal Δsak mutants , lacking a stress activated MAP kinase , fail to induce anthocyanin , display alterations in the vascular tissue and accumulate less fungal biomass ( Eaton et al . , 2010 ) . In this system it has been speculated that the observed downregulation of key enzymes in the shikimate and anthocyanin pathway may divert precursors into the catechin/tannin pathway with negative consequences for fungal development ( Eaton et al . , 2010 ) . In the present study , we have uncovered a new positive connection between anthocyanin induction and development of a biotrophic pathogen . We suggest that examples where anthocyanin induction is observed in response to biotic stress should be revisited to see , whether metabolic rewiring of the phenylpropanoid pathway is adopted as a common strategy by biotrophic microbes colonizing plants . Zea mays cv . Early Golden Bantam ( Olds Seeds , Madison , WI , USA ) was grown in a temperature-controlled greenhouse ( 14 hr/10 hr light/dark cycle; 28°C/20°C ) and used for infection by U . maydis . Z . mays chalcone synthase mutant M541J ( A1 A2 C1 c2 R1-nj ) and cognate control strain M142V ( A1 A2 C1 C2 R1-nj ) were obtained from the Maize Genetics Stock Center ( http://maizecoop . cropsci . uiuc . edu/ ) and are near-isogenic lines . The brown midrib maize mutants bm1 , bm2 , bm3 and bm4 mutant are near-isogenic lines generated in the background of inbred line A619 ( Vermerris and McIntyre , 1999; Marita et al . , 2003 ) , following eight backcrosses . The double mutants bm1-bm2 , bm1-bm4 , bm2-bm3 and bm3-bm4 were created by crossing the respective single mutant lines , followed by self-pollination of the resulting F1 plants , selection of double mutants among the F2 progeny , and confirmed with test crosses to the individual single mutants ( Vermerris et al . , 2010 ) . U . maydis strains were grown in YEPSL ( 0 . 4% yeast extract , 0 . 4% peptone , 2% sucrose ) and cell suspensions in H2O adjusted to OD600 = 1 . 0 were injected into the stem of 7-day-old maize seedlings with a syringe as described ( Kämper et al . , 2006 ) . Disease symptoms were scored at 12 days post infection using a previously developed scoring scheme ( Kämper et al . , 2006 ) and statistical analysis was performed as described ( Brefort et al . , in press ) . N . benthamiana was grown under the same conditions as maize . For plasmid constructions , standard molecular cloning strategies and techniques were applied ( Sambrook et al . , 1989 ) . All plasmids used and generated are listed in Supplementary file 1 . To generate truncated tin2 genes , the plasmid pIP-pum05302 ( Brefort et al . , in press ) was digested with XbaI-AscI . Truncated tin2 genes were amplified by respective primer pairs ( Supplementary file 2 ) and introduced into the XbaI-AscI site . tin2AAAAA was generated by amplification with respective primers ( Supplementary file 2 ) and used to replace um05302 in pIP-pum05302 . For constitutive overexpression of tin2 in U . maydis , p123Potef containing the constitutive otef promoter ( Wang et al . , 2011 ) was digested with NcoI-NotI and full-length tin2 amplified by respective primer pairs ( Supplementary file 2 ) was inserted . For transient expression in maize cells , p35S-mCherry ( Djamei et al . , 2011 ) was digested by EcoRI-XbaI and tin2 or ZmTTK1 genes amplified by respective primers ( Supplementary file 2 ) were introduced . For yeast two-hybrid constructs , pGBKT7 or pGADT7 were digested with NdeI-BamHI or EcoRI-XhoI , respectively and tin2 and truncated tin2 genes or ZmTTK1 and truncated versions thereof were inserted after amplification with respective primer pairs ( Supplementary file 2 ) . For protein expression in N . benthamiana , tin2 , and ZmTTK1 were amplified by respective primer pairs ( Supplementary file 2 ) and inserted into pBIN19AN ( Hirota et al . , 2007 ) digested with ApaI-NotI . Point-mutations in ZmTTK1 were introduced using the QuikChange Site-Directed Mutagenesis Kit ( Agilent Technologies , Santa Clara , USA ) following the manufacturer’s protocol using primers ( Supplementary file 2 ) . All plasmid constructs containing amplified gene fragments were sequenced . The haploid solopathogenic strain SG200 of U . maydis ( Kämper et al . , 2006 ) was used as a reference strain throughout this study . SG200Δ19A-1 , SG200Δtin2 , and SG200Δtin3 mutants were generated in a previous study ( Brefort et al . , in press ) . All U . maydis strains ( Supplementary file 3 ) are derived from SG200 and were generated by insertion of p123 derivatives into the ip locus as described ( Loubradou et al . , 2001 ) . Isolated U . maydis transformants were tested by southern analysis to assess single or multiple integration events in the ip locus . Maize seedlings inoculated with H2O , SG200 , SG200Δtin2 and SG200Δtin2-tin2 were harvested 6 days after infection . 2–3 cm segments below the injection hole showing anthocyanin induction in leaves infected with SG200 and SG200Δtin2-tin2 were excised from 20 plants , pooled and ground in liquid nitrogen . From plants infected with the tin2 mutant strains corresponding leaf segments , which did not show anthocyanin pigmentation , were excised . Total anthocyanins were quantified according to the method described by Ray et al . ( 2003 ) . 100 mg homogenized plant material was heated for 10 min in 1 ml of 2 N HCl at 55°C , cooled down and incubated overnight in the dark at room temperature . The extract was centrifuged and the clear supernatants were analyzed at 515 nm at a Cary 100 Bio Spectrophotometer ( Varian , Darmstadt , Germany ) . Commercial cyanidin 3-arabinoside chloride ( PhytoLab , Vestenbergsgreuth , Germany ) was used for quantification . For further identification of anthocyanins , flavonoids , and DIBOA-glucoside , as well as for recording their profiles , powdered samples were extracted and the polar phase was measured by UPLC-PDA-TOF-MS as described ( Djamei et al . , 2011 ) . For unambiguous identification fragmentation studies were performed by Ultra High Performance Liquid Chromatography ( 1290 Infinity , Agilent Technologies ) coupled with a 6540 UHD Accurate-Mass QTOF LC MS instrument with Agilent Jet Stream Technology as ESI source ( Agilent Technologies ) as described ( Koch et al . , 2013 ) . For cell wall isolation , 200 mg of plant material from the same batch was used for the anthocyanin quantification , was extracted with 1 . 6 ml 80% methanol ( vol/vol ) for 1 hr in a 2-ml plastic tube and centrifuged for 20 min at 16 , 000×g . The pellet was washed two times with 1 . 6 ml of 1 M NaCl and 0 . 5% Triton X-100 , for two times with 1 . 8 ml water , two times with 1 . 7 ml of methanol , and two times with 1 . 7 ml of acetone . Next alkaline hydrolysis was performed by adding 1 . 7 ml 2 M NaOH followed by incubation for 1 hr under vigorous shaking . Then , the pellet was washed twice with 1 . 8 ml water and dried at 80°C overnight . For lignin determination , 2 mg of the cell wall material was used and 250 µl of 25% ( vol/vol ) acetylbromide in glacial acetic acid were added and incubated for 30 min at 70°C . Afterwards , the sample was cooled down immediately on ice and 250 µl 2 M NaOH were added before centrifuging for 5 min at 16 , 000×g . For UV analysis ( 280 nm ) , 139 ml of the supernatant were taken and 2 . 8 µl 50% ( vol/vol ) hydroxylamine and 1 . 25 ml glacial acetic acid were added . A calibration curve was generated with coniferyl alcohol as standard compound in the range from 5 to 20 µg . Signal peptide prediction was performed with the online program SignalP 4 . 1 ( http://www . cbs . dtu . dk/services/SignalP/ ) . Protein secondary structure prediction was performed with the online program Jpred3 ( http://www . compbio . dundee . ac . uk/www-jpred/ ) . Phylogenic analysis was performed with the online program Phylogeny . fr ( http://www . phylogeny . fr/ ) . Protein conserved domain search was performed with the online program Pfam ( http://pfam . sanger . ac . uk/search/ ) . For the microarray experiments , maize plants ( Early Golden Bantam ) were grown in a phytochamber in a 15 hr/9 hr light/dark cycle; light period started/ended with 1 hr ramping of light intensity . Temperature was 28°C and 20°C , relative humidity 40% and 60% during light and dark periods , respectively , with 1 hr ramping for both values . Plants were inoculated with H2O ( mock ) , SG200 , and SG200Δtin2 as described above . Infected or water-inoculated tissue from 20 plants per experiment was harvested at 4 dpi by excising a section of the third leaf between 1 and 3 cm below the injection holes . For RNA extraction , material was pooled , ground to powder under liquid nitrogen and RNA was extracted with Trizol reagent ( Invitrogen , Karlsruhe , Germany ) . RNA was purified applying the RNeasy kit ( Qiagen , Hilden , Germany ) . Affymetrix maize genome arrays were done in three biological replicates , using standard Affymetrix protocols ( Midi_Euk2V3 protocol on GeneChip Fluidics Station 450; scanning on Affymetrix GSC3000G ) . Expression data were submitted to Gene Expression Omnibus ( http://www . ncbi . nlm . nih . gov/geo/ ) ( Accession number: GSE48536 ) . The microarray data obtained in this study were analyzed using the Partek Genomics Suite version 6 . 12 . Expression values were normalized using the RMA method . Criteria for significance were a corrected p-value ( per sample ) with an FDR of 0 . 05 and a fold-change of >2 . Differentially expressed genes were calculated by a 1-way ANOVA . Total RNA was extracted from U . maydis cell grown in YEPSL medium and from infected leaves by excising 2–3 cm segments from below the injection holes . At least 20 leaf segments were pooled for each of the indicated time points . RNA was isolated with the Trizol reagent ( Invitrogen ) following manufacturer’s protocol . cDNA was synthesized with Superscript III First Strand Synthesis SuperMix ( Invitrogen ) . Quantitative real-time PCR reactions were performed as previously described ( Berndt et al . , 2010 ) . All reactions were performed in at least three biological replicates . Statistical analysis was performed with Student’s t-test . Relative expression levels of fungal or maize genes were calculated in relation to the values obtained for the constitutively expressed peptidylprolyl isomerase gene ( ppi ) of U . maydis or glyceraldehyde 3-phosphate dehydrogenase ( GAPDH ) of Z . mays . Primers used for PCR are listed in Supplementary file 2 . For fungal biomass assessment , samples from infected plants were prepared as above . DNA was extracted and fungal biomass was determined as previously described ( Brefort et al . , in press ) . To screen for interactors of Tin2 , Tin2 lacking the signal peptide for secretion was expressed from pGBKT7-Tin226-207 and screened against a pAD_Gal4-library constructed from cDNAs generated from RNA of U . maydis FB1 × FB2 infected tissue . Material was collected 2 and 5 days after infection ( Farfsing , 2004 ) . Transformation of Saccharomyces cerevisiae strain AH109 ( Clontech , Mountain View , USA ) was done as described in the DUAL membrane starter kit manual ( Dualsystems Biotech AG , Schlieren , Switzerland ) . The yeast two-hybrid screen was performed following the instructions of the matchmaker yeast two-hybrid manual ( Clontech ) using 1 mg of bait-DNA ( pGBKT7-Tin226-207 ) and 0 . 5 mg of library-DNA . All resulting yeast clones were tested by immunoblotting for expression of the respective fusion proteins . Interacting clones were verified by empty vector controls excluding self-activating clones as well as by vector swap experiments . Full-length cDNAs of putative interaction partners were generated from SG200-infected tissue collected 6 days post infection , cloned in pGADT7 and tested for interaction with Tin226-207 expressed from pGBKT7-Tin226-207 . For yeast protein interaction assays , the genes encoding the proteins tested for interaction were cloned into pGBKT7 or pGADT7 vectors ( Clontech ) , respectively , to express fusion proteins with the yeast GAL4 binding ( BD ) and activation domain ( AD ) , respectively , in strain AH109 ( Clontech ) . Yeast transformants were screened on selective dropout media ( SD ) lacking only tryptophan ( Trp ) and leucine ( Leu ) to select yeast cells , which contained the desired plasmids ( Supplementary file 4 ) . Protein interactions were assessed on SD selection medium lacking tryptophan , leucine , and histidine ( His ) containing 3 mM 3-Amino-1 , 2 , 4-triazole ( 3-AT ) . Yeast MEL1p activity was measured by addition of 5-bromo-4-chloro-3-indolyl α-D-galactopyranoside ( X-α-gal ) . Fusion protein expression was verified by western blot . For protein expression in Escherichia coli , pRSET-GST-PP ( Mueller et al . , 2013 ) and pPR-IBA102 ( IBA Lifesciences , Göttingen , Germany ) vectors were used for the preparation of GST-Tin226-207 or Strep-ZmTTK1 , respectively . E . coli BL21 ( DE3 ) strain was used as host for protein expression . E . coli transformants carrying respective expression plasmids were grown in YT medium ( 1 . 6% Bacto Tryptone , 1 . 0% Yeast Extract , 0 . 5% NaCl ) containing ampicillin ( 50 µg/ml ) . Protein induction was performed by addition of 1 mM isopropyl β-D-1-thiogalactopyranoside ( IPTG ) at 28°C for 5 hr . Cell pellets were lysed by BugBuster master mix ( Merck , Schwalbach , Germany ) . Soluble supernatant was incubated with Glutathione Sepharose 4 Fast Flow ( GE Healthcare , München , Germany ) or Strep-tactin sepharose ( IBA Lifesciences ) for GST- or Strep-tagged protein purification , respectively . Protein enrichment and purity was assessed by western blot . For in vitro co-IP , recombinant GST-Tin226-207 or GST-Tin226-202 protein ( 200 µg ) bound to glutathione sepharose beads ( GE healthcare ) , respectively , was incubated with 500 µl E . coli soluble extract in PBS buffer ( 10% glycerol , 0 . 1% Tween20 and 0 . 5 mg/ml bovine serum albumin ) from induced BL21 ( DE3 ) /pPRIBA102-ZmTTK1 at 4°C for 1 hr . GST fusion proteins were eluted with reduced glutathione . Strep-ZmTTK1 in eluate was detected by western blot . Detection of Tin2-HA protein in culture supernatants of U . maydis was performed as previously described ( Doehlemann et al . , 2009 ) . For western blot analysis , protein was separated by 10% or 13% SDS-polyacrylamide gel electrophoresis . Protein was visualized by staining with Coomassie Brilliant Blue ( CBB ) . Monoclonal anti-c-myc antibody produced in mouse , anti-HA antibody produced in rabbit or monoclonal anti-ubiquitin antibody produced in mouse were purchased from Sigma-Aldrich ( Taufkirchen , Germany ) and used as a primary antibody . Anti-mouse IgG HRP-linked or anti-rabbit IgG HRP-linked antibodies were purchased from Cell Signaling ( Danvers , MA , USA ) and used as a secondary antibody . Strep-tactin-HRP ( IBA Lifesciences ) was used for the detection of Strep-tagged protein . For immunoprecipitation of ZmTTK1 from plant cell , total plant protein extract was incubated with human-IgG agarose ( Sigma-Aldrich ) at 4°C for 2 hr on a rotary shaker . For in vitro kinase assay , recombinant proteins or human-IgG agarose carrying immunoprecipitates were incubated at 28°C for 30 min in a volume of 25 µl reaction buffer containing 0 . 5 mg/ml myelin basic protein ( Sigma-Aldrich ) , 20 µM ATP , and 2 . 5 µCi [γ-32P]ATP . Phosphorylated myelin basic protein was detected using a STORM phosphoimager ( GE Healthcare ) . For protein expression in N . benthamiana , Agrobacterium tumefaciens GV3101 strains carrying expression plasmids indicated in each experiment were grown in YT medium containing kanamycin ( 50 µg/ml ) . Cells were resuspended in 10 mM MES ( pH 5 . 6 ) and the concentration was adjusted as OD600 = 0 . 5 . A . tumefaciens strains were mixed at an equal ratio with the A . tumefaciens strain LBA2566 expressing the 35S:p19 construct encoding the p19 silencing suppressor ( Voinnet et al . , 2003 ) . Bacterial mixtures were infiltrated with needleless syringe to the upper leaves of 3-weeks-old N . benthamiana plants . Protein was extracted from the leaves at 3 days after infiltration with extraction buffer ( 100 mM Tris-HCl pH 7 . 5 , 100 mM NaCl , 5 mM EDTA , 5 mM EGTA , 10 mM NaF , 0 . 1 mM Na3VO4 , 10 mM β-glycerophosphate , 1 mM β-mercaptoethanol , 0 . 1% Triton-X100 , 10% Glycerol , protease inhibitor cocktail [Roche , Mannheim , Germany] ) under liquid nitrogen . For proteasome inhibition , 100 µM MG132 ( Merck ) solution was infiltrated into N . benthamiana leaves at 1 day before protein extraction . For transient protein expression in maize by biolistic gene transfer , 1 . 6 µm gold particles ( Bio-Rad , München , Germany ) were coated with the plasmids coding for the indicated genes driven under CaMV 35S promoter ( Supplementary file 1 ) . Bombardment was performed using a PDS-1000/He system ( Bio-Rad ) onto 2-weeks-old maize leaves or SG200Δtin2-infected leaves ( at 4–5 dpi ) . To visualize fungal proliferation in infected tissue , the area 1–3 cm below the injection site was excised at 3 days post infection . Fungal hyphae were stained with WGA-AF488 ( Invitrogen ) , the plant cell wall was stained with propidium iodide ( Sigma-Aldrich ) . Leaf samples were incubated in staining solution ( 1 µg/ml propidium iodide , 10 µg/ml WGA-AF488 ) and observed with a TCS-SP5 confocal laser-scanning microscope ( Leica Microsystems , Wetzlar , Germany ) under the following conditions . WGA-AF488: excitation at 488 nm and detection at 500–540 nm , propidium iodide: excitation at 561 nm and detection at 580–660 nm . For fluorescent protein detection , leaf samples were directly observed by confocal microscopy under the following conditions . mCherry: excitation at 561 nm and detection at 580–630 nm , YFP: excitation at 495 nm and detection at 510–550 nm . For observation of vascular bundles , infected leaf segments collected at 6 dpi were cut into small pieces and fixed with 4% glutaraldehyde . The segments were embedded in epoxy resin and sliced with a microtome ( 0 . 4 µm thickness ) . Cross-sections were stained with 1% ( wt/vol ) Safranin O ( Sigma-Aldrich ) for 2–3 min and observed by light microscopy . U . maydis genes and encoding protein sequences are available at the MIPS database ( http://mips . helmholtz-muenchen . de/genre/proj/ustilago/ ) ; tin2 , um05302; ppi , um03726 . For Z . mays genes , sequence data can be found at MaizeSequence . org ( http://www . maizesequence . org ) under accession numbers; ZmGAPDH , GRMZM2G046804; ZmTTK1 , GRMZM2G448633; ZmR1 , GRMZM5G822829; ZmC1 , GRMZM2G005066; ZmPAL , GRMZM2G074604; ZmCHS , GRMZM2G422750; ZmCHI , GRMZM2G155329; ZmF3H , GRMZM2G062396; ZmDFR , GRMZM2G026930; ZmANS , GRMZM2G345717; ZmBx1 , GRMZM2G085381; ZmBx2 , GRMZM2G085661; ZmBx5 , GRMZM2G063756; ZmBx8 , GRMZM2G085054; Zm4CL , GRMZM2G075333 , GRMZM2G054013; ZmCCR , GRMZM2G131205 , GRMZM2G131836; ZmCAD , GRMZM2G046070 , GRMZM5G844562 , GRMZM2G167613 , AC234163 . 1_FGP002 , GRMZM2G700188 , GRMZM2G118610 .	The production of agricultural crop plants is severely hindered by bacteria , viruses , and fungi that have developed their own strategies to colonize these plants and obtain nutrients from them . Some pathogens kill the plants they colonize , but ‘biotrophic pathogens’ employ sophisticated strategies to manipulate the host plant without killing it . During the past decade it has been recognized that the interactions between plants and biotrophic pathogens are largely governed by effector proteins—which are typically secreted by the pathogen after it makes contact with the host . These effector proteins can either stay in the space between the plant cells and pathogen cells , or actually enter inside the plant cells . The fungus Ustilago maydis is one such biotrophic pathogen that colonizes maize plants and causes a disease called corn smut . Hallmarks of this infection are the formation of large plant tumors and the production of a red pigment , called anthocyanin , in infected plant tissues . Now , Tanaka et al . reveal that an effector called Tin2 , which is secreted by the corn smut fungus , causes the production of this anthocyanin pigment . Tin2 moves inside plant cells , where it blocks the breakdown of a protein-modifying enzyme that is necessary to ‘switch on’ the production of anthocyanin . When a mutant fungus that lacks Tin2 infects a maize plant , no anthocyanin is induced and the pathogen fails to reach the vascular tissue , where it would normally get most of its nutrients . Tanaka et al . revealed that , in these infections , this vascular tissue was strongly reinforced with a compound , called lignin , suggesting that the plant fortifies these cell walls to block access by the mutant fungus . Since the building blocks for making lignin are also required for making anthocyanins , Tanaka et al . suggest a model whereby Tin2 compromises the ability of plants to protect themselves by diverting resources away from making lignin . In line with this speculation , the corn smut fungus was shown to cause stronger disease symptoms in maize plants with mutations that prevent them from producing lignin . The Tin2 effector of the corn smut fungus appears to target a critical protein in maize that can shift the balance of the plant’s metabolic pathways in favor of the pathogen . Further , since anthocyanin production is also observed after infections of plants with other microbes , these findings may have uncovered a microbial strategy to enhance virulence that is also employed by other plant pathogens .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology", "microbiology", "and", "infectious", "disease" ]	2014	A secreted Ustilago maydis effector promotes virulence by targeting anthocyanin biosynthesis in maize
The adaptive immune system responds to pathogens by selecting clones of cells with specific receptors . While clonal selection in response to particular antigens has been studied in detail , it is unknown how a lifetime of exposures to many antigens collectively shape the immune repertoire . Here , using mathematical modeling and statistical analyses of T cell receptor sequencing data , we develop a quantitative theory of human T cell dynamics compatible with the statistical laws of repertoire organization . We find that clonal expansions during a perinatal time window leave a long-lasting imprint on the human T cell repertoire , which is only slowly reshaped by fluctuating clonal selection during adult life . Our work provides a mechanism for how early clonal dynamics imprint the hierarchy of T cell clone sizes with implications for pathogen defense and autoimmunity . The hallmark of adaptive immunity is the generation of diversity through genetic recombination and clonal selection . Their interplay balances the breadth and specificity of the ~1012 T cells in the human body ( Figure 1A; Arstila et al . , 1999; Farber et al . , 2014 ) : The genetic recombination of the T cell receptor ( TCR ) locus , termed VDJ recombination , generates an enormous potential diversity of receptors ranging from early estimates of ~1015 ( Davis and Bjorkman , 1988 ) to more recent estimates of ~1061 ( Mora and Walczak , 2016 ) different possible receptor TCRαβ heterodimers . Clonal selection expands the number of specific cells during an infection for effector functions , a fraction of which is retained over prolonged periods of time as immune memory ( Ahmed and Gray , 1996; Farber et al . , 2014 ) . Much progress has been made deciphering the mechanisms of regulation and control of T cell dynamics over the last few decades ( Antia et al . , 2005; Sallusto et al . , 2010; Farber et al . , 2014 ) . However , much of that progress has focused on the dynamics of subsets of T cells specific to a particular antigen and has come from experiments in mice . An important open question is how exposures to many antigens over a human lifetime collectively shape our T cell repertoire ( Farber et al . , 2014; Davis and Brodin , 2018 ) . High-throughput repertoire sequencing enables direct surveys of the diversity and clonal composition of T cells from human blood or tissue samples and thus promises to provide quantitative answers to this question ( Robins et al . , 2009; Thomas et al . , 2014; Britanova et al . , 2016; Emerson et al . , 2017; Oakes et al . , 2017; Thome et al . , 2016; Robins et al . , 2010; Qi et al . , 2014; Lindau et al . , 2019; TRACERx consortium et al . , 2019 ) . However , while the TCR locus provides a natural barcode for clonal lineages due to its large diversity , this same diversity also makes inferring past clonal dynamics a challenging inverse problem , in particular given practical limitations on sequencing depth and temporal resolution in longitudinal studies . Mathematical modeling can help address this challenge by solving the forward problem of linking clonal dynamics to emergent statistical patterns ( Desponds et al . , 2016; Lythe et al . , 2016; Dessalles et al . , 2019; Altan-Bonnet et al . , 2020; de Greef et al . , 2020 ) . Comparing patterns to data can provide insights about dynamics from static snapshots of repertoire organization in different individuals . A particularly striking such pattern has been the observation of power-law scaling of clone sizes spanning several orders of magnitude ( Robins et al . , 2009; Thomas et al . , 2014; Britanova et al . , 2016; Emerson et al . , 2017; Oakes et al . , 2017 ) . In a typical sample of T cells from peripheral blood , a large fraction of clones is only seen once within 105–107 sampled sequences , while the most abundant clones account for more than 1% of all sequencing reads . Such power-law scaling of clone sizes has been shown to arise at steady state in models of fluctuating clonal selection driven by different antigen encounters ( Desponds et al . , 2016 ) . However , it is unclear whether this mechanism alone is sufficient to explain how clone size scaling is established , and more broadly how variable the clonal hierarchy is over time . Here , we develop a theory of T cell dynamics throughout the human lifespan based on the statistical laws of repertoire organization and dynamics revealed by cohort and longitudinal human TCR repertoire sequencing ( Britanova et al . , 2016; Emerson et al . , 2017; Chu et al . , 2019; Lindau et al . , 2019 ) . We find that clonal expansions during repertoire formation play an important role in establishing the clone size hierarchy , which is only slowly reshaped by fluctuating clonal selection during adult life . An important statistic to summarize repertoire organization is the clone size distribution , which tabulates the number of clones found at different multiplicities within a repertoire or sample . Multiple previous studies have shown that these distributions are heavy-tailed ( Robins et al . , 2009; Thomas et al . , 2014; Britanova et al . , 2016; Emerson et al . , 2017; Oakes et al . , 2017 ) . However , potential confounding by noise introduced during the sequencing process has remain debated ( Altan-Bonnet et al . , 2020 ) and systematic analyses of how variable these distributions are across healthy individuals have been lacking . To fill these gaps , we reanalyzed data from two large-scale cohort repertoire sequencing studies of human blood samples , which used fundamentally different sequencing pipelines and thus have different sources of noise ( Materials and methods – Experimental data sources ) . Both studies sequenced the locus coding for the hypervariable TCR CDR3 β-chain from peripheral blood T cells of healthy human volunteers spanning a large range of ages ( Figure 1—figure supplement 1 ) . In both studies , T cells were sequenced without regard to their phenotypic characteristics . A clone thus represents the full lineage of cells derived from a common ancestral cell irrespective of differentiation status ( see Appendix 5 – Relation between clone size and cellular phenotypes and Figure 1—figure supplement 2 for a phenotypically resolved analysis ) . After normalizing clone sizes to account for variations in sampling depth and for the increasing fraction of T cells of memory phenotype with age ( Figure 1—figure supplement 3 ) , we found that the tails of the clone size distributions collapsed to the same statistical law across individuals and cohorts ( Figure 1B , C ) : Ranking clones by decreasing size , the rank of the largest clones approximately scales with their size C as a power law , ( 1 ) rank∼C-α , where α is a scaling exponent . To quantify the apparent similarity of the scaling relationship , we determined α for each sample by maximum likelihood estimation . Only a small fraction of all T cells are sampled , which poses a challenge because subsampling a power law leads to deviations from scaling at small clone sizes ( Stumpf et al . , 2005 ) . To overcome this challenge , we used a trimming procedure and excluded clones smaller than a minimal size from the fitting , which decreases bias arising from subsampling ( Appendix 1 ) . Determined in this subsampling-robust manner the fitted power-law exponents agree remarkably well within the range of ages covered by both cohorts ( Figure 1D ) : with α = 1 . 17 ± 0 . 03 ( mean ± standard error [SE] ) and α = 1 . 18 ± 0 . 01 in the Britanova cohort and Emerson cohort , respectively . Moreover , the fitted exponents varied little between individuals in both cohorts; with a sample standard deviation of fitted exponents of 0 . 14 and 0 . 21 , respectively . The agreement of the mean exponents is noteworthy given the different sequencing pipelines and provides strong evidence that the scaling relationship ( Equation 1 ) is a true feature of the clone size distribution and not of the measurement process . What drives the emergence of a power-law distributed hierarchy of clone sizes ? Given the reproducibility of the scaling law across individuals , we might hope for a statistical explanation independent of the precise antigenic history that has driven the expansion of specific cells in an individual . To test hypotheses about mechanisms underlying scaling , we describe repertoire dynamics using a general mathematical framework based on effective stochastic rate equations for the recruitment of new clones , and the proliferation and death of already existing clones within a T cell compartment ( Materials and methods – Mathematical framework ) . In macroecology , where such reductionist approaches have a long history , simple neutral models within this framework have had surprising success in describing species abundance distributions only accounting for demographic stochasticity ( Volkov et al . , 2003 ) , but this source of variability is insufficient to account for the observed breadth of T cell clone sizes ( Desponds et al . , 2016; de Greef et al . , 2020; for a detailed discussion see Appendix 2 ) . The failure of this null model has prompted a search for other mechanisms that explain scaling . To constrain this search , we analyzed how fitted exponents varied with age . In particular , based on a finite time solution we derived for a previously proposed model ( Desponds et al . , 2016 ) of how power-law scaling can emerge from the cumulative effect of temporal fluctuations in clonal growth rates ( Materials and methods – Slow convergence to steady–state scaling ) , we expected a substantially steeper tail in young individuals . While exponents overall decreased slightly with age , the dependence on age accounted for surprisingly little variation in both cohorts ( Figure 1D ) , including when controlling for sex and cytomegalovirus ( CMV ) exposure status ( Figure 1—figure supplement 4 ) . Notably , scaling is established within the first decade of life , with significant clone size variability existing as early as at birth ( Figure 1—figure supplement 5 ) , defying previous model predictions . We hypothesized that scaling might result from clonal expansions during repertoire formation , which would naturally explain the early onset of scaling . Our hypothesis is based on experimental evidence in mice ( Le Campion et al . , 2002; Min et al . , 2003; Haluszczak et al . , 2009; Kawabe et al . , 2017 ) and human ( Rufer et al . , 1999; Schönland et al . , 2003 ) that repertoire formation is driven not only by increased thymic output , but also by large proliferative expansion of some T cell clones . Additionally , multiple studies Hammarlund et al . , 2003; Pogorelyy et al . , 2017; Tanno et al . , 2020 have shown that some T cell clones can persist over multiple decades , which suggests that clonal turnover might be sufficiently slow for transient expansionary dynamics early in life to shape repertoire organization over a prolonged time period ( see also Appendix 2 –Relaxation time scale in a neutral model ) . To test our hypothesis , we constructed a minimal model of repertoire formation based on known T cell biology ( Figure 2A ) . Following previous work ( Bains et al . , 2009; Lythe et al . , 2016 ) , we assume that T cells proliferate at a rate inversely proportional to the total number of cells N already present in the repertoire . This assumption leads to increased proliferation early in life before the repertoire has reached its homeostatic size , for which there is experimental evidence Le Campion et al . , 2002; Min et al . , 2003; Haluszczak et al . , 2009; Kawabe et al . , 2017; Rufer et al . , 1999; Schönland et al . , 2003 . This assumption is also compatible with a simple mechanistic model of T cell competition ( Materials and methods – Mechanistic motivation for the competition function ) . We further assume that cells die at a rate d , and that new clones are recruited at rate θ with an initial size C0 . For simplicity , we set C0 equal to one in the following and we assume constant rates d and θ . Importantly , recruitment of new clones and total expansion of already existing clones maintain a constant ratio throughout development under these assumptions in line with findings that the fraction of cells with TCR excision circles , which are diluted during peripheral division , is constant during fetal development ( Schönland et al . , 2003 ) and infancy ( Douek et al . , 2001; Bains et al . , 2009 ) . We simulated the model starting from an empty repertoire and found that large clones displayed power-law scaling ( Figure 2B blue lines ) . The simulation results contrast with steady-state predictions ( Figure 2B black line ) , where the model effectively reduces to the neutral null model introduced earlier ( Materials and methods – Steady-state distribution ) . The power-law tail persisted over multiple decades of aging , much beyond the timescale of cellular turnover , 1/d = 5 years , assumed in the simulations . A mathematical analysis shows that relative timescales of clonal and cellular turnover are controlled by the control parameter γ=θC0/b0 , which is the ratio of the contribution of recruitment and proliferation to overall compartment maintenance ( Appendix 2 ) . The long timescale of clonal turnover emerges because the chosen parameters are in the biological parameter regime γ<1 , where most cell death is balanced by proliferation ( den Braber et al . , 2012; Macallan et al . , 2017 ) . Thus , we find that repertoire formation can produce transient but long-lasting power-law scaling of clone sizes . To obtain intuitive insight into how scaling is established , we developed a continuum theory of clonal dynamics during repertoire growth ( Materials and methods – Continuum theory of clonal growth ) . We find that the clone size Ci of the i-th clone recruited at time ti follows a subexponential growth law Ci ( t ) =C0 ( t/ti ) 1/ ( 1+γ ) . Clones recruited early grow large deterministically until competition lowers proliferation rates below the death rate ( Figure 2C , lower panel ) . Different clones are recruited at different times and thus have more or less time to grow ( Figure 2C , upper panel ) , which leads to a clone size distribution that follows power-law scaling with an exponent α=1+γ . We note that this origin of the power-law scaling is closely related to a well-known generative mechanism for power-laws first studied by Yule , 1924 ( for a detailed discussion see Appendix 4 – A unified view on mechanisms generating power laws in different growth processes ) . The predicted exponent closely matches simulation results for different values of γ ( Figure 2D dashed lines ) . Intuitively , when recruitment rates are higher , clones founded early have less time to outgrow later competitors , and thus the power law is steeper ( α is larger ) . Importantly , in the biological parameter regime in which proliferation dominates , γ<1 , the exponent is compatible with experiments ( Figure 1B–D ) . We thus find , that the model – without fine tuning of parameters – reproduces the observed scaling exponent . To expose a basic mechanism capable of producing broad clone size distributions , we have kept the model deliberately simple . More detailed models demonstrate the conditions and limits on the generalizability of this mechanism ( Appendix 3 ) . Variable recruitment sizes only affect the distribution of small clones ( Figure 2—figure supplement 1 ) ; while a saturation of proliferation rates , or competition between subsets of T cells for specific resources maintain distributions at small and intermediate sizes while leading to cutoffs for the largest clones ( Figure 2—figure supplement 2 and Figure 2—figure supplement 3 , respectively ) . Our proposed theory for the rapid emergence of scaling predicts that large clones have expanded massively during repertoire formation . To test this prediction , we need to trace the dynamics of early founded clones . To this end , we exploit a change in the recombination statistics taking place during fetal development ( Feeney , 1991; Rechavi et al . , 2015; Park et al . , 2020; Figure 3A ) . While T cells are produced by the thymus from the late first trimester the enzyme terminal deoxynucleotidyl transferase ( TdT ) , which inserts non-templated nucleotides during VDJ recombination , is not expressed until the mid second trimester ( Park et al . , 2020 ) . Therefore , many more T cells in fetal and neonatal blood have zero insertions than expected from the adult recombination statistics ( Rechavi et al . , 2015 ) . This enables a statistical dating of individual clones in a repertoire based on their sequence ( Sethna et al . , 2017; Pogorelyy et al . , 2017 ) . If our model is correct , we expect abundant clones to be more likely to have zero insertions than smaller clones . Analyzing data from the Emerson cohort , we find that zero insertion clones are indeed highly enriched within the most abundant clones ( Figure 3B ) . This generalizes a previous report of such an enrichment within the naive compartment ( Pogorelyy et al . , 2017 ) . The large cohort size allows us to perform a fine-grained analysis of how the fraction of zero-insertion clones depends on clonal abundance and age . We find that enrichment is particularly pronounced in the young and decreases with age at different speeds depending on clone size . Among the largest clones many more still have zero insertions than expected from the adult recombination statistics even multiple decades after repertoire formation . This suggests that the incumbent large clones created during repertoire formation are only slowly replaced by clones expanding later in life , similarly to what has been observed in mice ( Gossel et al . , 2017; Hogan et al . , 2019 ) . Additional analyses rule out other potential explanations for the relation between insertion statistics and clonal abundance . First , sequences with zero insertions are similarly enriched among the largest clones in productive and unproductive sequences ( Figure 3—figure supplement 1 ) demonstrating that convergent selection pressures during adult life are not a primary source of the higher abundance of these clones . Second , while abundant clones are also enriched for sequences with known antigen specificity ( Figure 3—figure supplement 2 ) and sequences likely to be convergently recombined ( Figure 3—figure supplement 3 ) , these enrichments do not show the same striking dependence on age . Furthermore , we find that zero insertion clones are consistently less enriched in individuals infected by CMV ( Figure 3—figure supplement 4 ) , in contrast to the hypothesis that this infection might drive their expansion ( Pogorelyy et al . , 2017 ) . Taken together , these analyses support the conclusion that dynamics during the perinatal time window of repertoire formation leave a long-lasting imprint on the T cell clonal hierarchy well into adulthood . Building on this successful validation of a core prediction of our theory , we asked whether we could leverage the detailed pattern of enrichments at different ranks and ages to quantify how much being part of the wave of early expansions determines the fate of a clone relative to other sources of clone size variability . To this end , we extended our model beyond repertoire formation and allowed clonal proliferation rates to fluctuate over time to model the net effect of clonal selection by changing antigenic stimuli during adult life ( Desponds et al . , 2016 ) . Specifically , we added a stochastic term to the growth rate of each clone to provide an effective Langevin description of the long-term dynamics induced by fast-changing antigen stimuli , ( 2 ) bi⁢ ( t ) =b0/N+2⁢σ⁢ηi⁢ ( t ) , where ⟨ηi⁢ ( t ) ⁢ηj⁢ ( t′ ) ⟩=δi⁢j⁢δ⁢ ( t-t′ ) . To determine a biologically plausible fluctuation strength σ , we analyzed the variability of clone sizes over time in a longitudinal repertoire sequencing study ( Chu et al . , 2019 ) . We first analyzed how much recently expanded clones contribute to the tail of the clone sizes , and found that only a small fraction of the largest clones in any sample were not already large at the earliest time point ( Figure 4A and Figure 4—figure supplement 1 ) . To minimize confounding by transient dynamics affecting these clones , we excluded them from further analysis . We found that large clones had remarkably stable abundances over time , which we quantified by calculating the variance of log-foldchanges in clone size between the second and every subsequent time point ( Figure 4B and Figure 4—figure supplement 2 ) . The variability of clone sizes increased linearly over time as expected theoretically , from which we determined a magnitude of net growth rate fluctuations σ compatible with the slope of increase ( Materials and methods – Modeling long-term repertoire dynamics with fluctuating clonal growth rates ) . Using the fitted fluctuation strength , we constructed an in silico cohort of individuals of different ages according to the extended model ( Materials and methods – Simulated cohort ) . In short , we computationally assigned each newly recruited clone to have zero insertions in a way that mimics the change in fetal recombination statistics , and we simulated memory repertoire dynamics based on the combined effect of early expansion and fluctuating clonal selection ( Equation 2 ) . The enrichment of zero insertion clones in the simulated cohort ( Figure 4C ) closely recapitulated the empirical findings using plausible parameter values . Notably , the longer lasting enrichment of zero insertion clones among the very largest clones is also found in the simulated cohort , and the timescales over which the enrichment decays agree remarkably well . To obtain analytical insight into the enrichment dynamics , we studied how fluctuating growth rates reorder the clone size hierarchy established during repertoire formation ( Materials and methods – Relaxation of the zero insertion distribution ) . The early clone size hierarchy in our model is dominated by the time of recruitment , which leads to a steep gradient in the zero insertion probabilities as a function of rank in the first decade of life ( Figure 4C ) . We thus calculated how the zero insertion probabilities P0⁢ ( r , t ) for clones of rank r at time t change due to fluctuating clonal growth starting from an idealized initial condition resembling the early clone size hierarchy , in which the clone sizes follow power-law scaling and the r⋆ largest clones have a zero insertion probability p0 , - and all others a probability p0 , + . We find that the probability of zero insertion clones then follows a sigmoidal shape as a function of log clone size rank , which changes with age as follows , ( 3 ) P0 ( r , t ) =Δp02erfc ( log⁡ ( r/r⋆ ) +t/τd2t/τd ) +p0 , + , where Δ⁢p0=p0 , --p0 , + is the difference of zero insertion probabilities , τd a characteristic timescale , and erfc ( x ) the complementary error function . These analytical results suggest a two-parameter rescaling of the enrichment of zero insertion clones as a general test of our theory . To demonstrate the feasibility of fitting these parameters from the enrichment dynamics , we determined them from the simulated data using weighted least squares fits setting r and t in Equation 3 to the mid-value of each bin . Rescaling the data with the fitted r⋆ and τd lead to a collapse of all simulated datapoints onto the predicted sigmoidal curve ( Figure 4—figure supplement 3 ) . We then applied the same fitting and rescaling procedure to the experimental data , and found that it also leads to a remarkably good data collapse ( Figure 3C ) . The fitted parameters quantify key features of long-term repertoire dynamics , with τd characterizing the timescale over which fluctuations change the clone size hierarchy , and r⋆ being related to the number of clones recruited during early repertoire growth . In line with the long-lived enrichment of zero insertion clones , the fitting reveals a remarkably slow timescale of about a decade over which the clone size hierarchy is reordered during healthy aging . The fitted r⋆ indicates that early repertoire formation involves the expansion of a large number of different clones . Overall , the agreement between theory and data demonstrates that our model quantitatively captures how early expansions and ongoing fluctuating selection together shape the clone size hierarchy . The evolution of the adaptive immune system has endowed vertebrates with the ability to adapt to pathogens that evolve on a timescale faster than host reproduction ( Mayer et al . , 2016 ) . However , this ability comes with a cost: every generation needs to rebuild immune memory anew . As the organism first comes into contact with the outside world , it quickly needs to train its adaptive immune system to tolerate innocuous antigens and build up immune memory against pathogens . Here , we have shown that this process of rapid adaptation leaves a long-lasting imprint on the organization of the human T cell repertoire . More broadly , we propose a theory of repertoire dynamics that quantitatively describes how early expansions during repertoire formation combine with a lifetime of exposures to cumulatively shape the T cell hierarchy . Notably , we find that the T cell repertoire is remarkably stable over time in adult individuals outside of the punctuated expansions and contractions of specific clones in acute responses . Our study demonstrates that despite its vast complexity , repertoire dynamics is partially predictable by quantitative models . The model predictions can help guide future longitudinal studies , which in turn will allow refinements of modeling assumptions . The current work thus provides a stepping stone toward a detailed quantitative understanding of T cell dynamics that we hope will ultimately power the rational development of immunodiagnostics and therapeutics . The general mechanism we describe for imprinting in the adaptive immune system provides a unified lens through which to view a number of converging lines of evidence about how a developmental time window shapes adaptive immunity ( Guerau-de-Arellano et al . , 2009; Farber et al . , 2014; Gostic et al . , 2016; Constantinides et al . , 2019; Li et al . , 2019; Davenport et al . , 2020; Hong et al . , 2020 ) . In our model , overall repertoire growth early in life amplifies the effect of any early exposures , as the responding clones continue to proliferate as memory cells and thus are much larger than memory produced from similar exposures after the homeostatic repertoire size is reached . We thus expect early pathogen exposures to be particularly potent , as has been observed in influenza , where disease severity across age cohorts for different strains depends on the first exposure ( Gostic et al . , 2016 ) . Conversely , we expect the presence of tolerizing factors early in life to be particularly crucial during repertoire formation to avoid autoimmunity , as has been observed for the autoimmune regulator gene AIRE , for which expression is only essential during a perinatal time window ( Guerau-de-Arellano et al . , 2009 ) . A limitation of datasets used in this study is that they do not provide direct information about the phenotypic characteristics of cells belonging to different clones . Repertoire sequencing of phenotypically sorted blood samples shows that the largest clones predominantly consist of cells with memory phenotype ( Appendix 5 ) . This indirectly suggests that the clonal expansions during repertoire formation produce memory cells as we have assumed in our simulated cohort ( Figure 4C ) . Supporting this interpretation , a substantial number of memory cells circulate in the blood quickly following birth ( Pediatric AIDS Clinical Trials Group et al . , 2003 ) and recent evidence suggests that memory-like T cells are already generated in particular tissue sites such as the intestine even before birth ( Zhang et al . , 2014; Li et al . , 2019 ) . However alternatively , early expansions could also set up a broad distribution of naive T cell clone sizes ( de Greef et al . , 2020 ) , whose hierarchy would then need to be roughly maintained during the transition into memory to be compatible with the observed impact of early expansions on the hierarchy of the most abundant clones . Advances in repertoire sequencing of T cells sorted with increasing granularity using cell surface markers ( Oakes et al . , 2017; Soto et al . , 2020 ) along with advances in single-cell technologies linking TCR sequencing and cellular phenotyping could help differentiate between these scenarios in the future . An important question raised by our work is which antigens drive the expansion of early T cell clones . To address this question , it will be necessary to determine the exposures that imprint the abundance of these clones , as has been done recently for mucosal-associated invariant T cells ( Constantinides et al . , 2019 ) , a subset of non-conventional T cells . Going forward , the highly abundant clones with sequences close to the genetically inherited gene templates resulting from the absence of TdT expression during early fetal development are a particularly interesting target of study . They might constitute an evolutionarily controlled set of innate-like defenses within the adaptive immune system . Determining what imprints their abundances will help resolve the question of whether their large abundances are simply a byproduct of rapid repertoire formation or whether these clones serve particular functions . We analyzed T cell repertoire sequencing data from two large published cohort studies of healthy human volunteers by Britanova et al . , 2016 and Emerson et al . , 2017 and from a longitudinal study by Chu et al . , 2019 , detailed descriptions of which we provide in the following . In short , Britanova et al . , 2016 sequenced reverse transcribed mRNA with added unique molecular identifiers ( UMIs ) , while Emerson et al . , 2017 and Chu et al . , 2019 sequenced genomic DNA coding for this region without the addition of UMIs . These approaches have complementary strengths: The addition of UMIs allows to correct for stochasticity during PCR amplification and sequencing artifacts , while DNA sequencing removes the influence of cell-to-cell gene expression heterogeneity . The Britanova cohort comprises 71 individuals spanning ages 6–103 years , as well as eight cord blood samples . The Emerson cohort spans ages 1–74 years and consists of a training and validation set of 666 and 120 individuals , respectively . From the training set , we excluded 111 samples with missing age information and 62 samples with a conflicting data format . We used only samples from the training set to analyze how the scaling law of repertoire organization changes with age ( Figure 1 ) . For the zero insertion enrichment analyses ( Figure 3B , C ) , we combined both the training and validation set together with separately published repertoire sequencing data from eight elderly individuals Lindau et al . , 2019 generated using the same experimental pipeline ( immunoSEQ , Adaptive Biotechnologies , Seattle ) to achieve the broadest possible coverage of all age groups . The longitudinal study by Chu et al . , 2019 performed repertoire sequencing of peripheral blood from three healthy female volunteers ( using the immunoSEQ pipeline ) over eight time points spanning a ~1 year time frame . One individual in the study was in mid-adulthood ( 24–45 years , Subject 3 in the original study ) , while two were in early adulthood ( 18–24 years , Subjects 1 and 2 in the original study ) . In Figure 4A , B , we display data from the older individual as we expect dynamics of large clones to be masked less by measurement noise as the large clones increase in relative abundance with age . All studies from which we analyzed data sequenced the locus coding for the TCR CDR3 β-chain only , and we thus define clones as collections of cells sharing the same CDR3 β-chain . Clone sizes are defined as the number of distinct unique molecular identifiers ( UMIs ) sequenced ( Britanova cohort ) , or based on sequencing reads ( Emerson cohort ) . The definition of a clone solely based on the CDR3 β-chain neglects convergent recombination of the most easily produced receptors with different CDR3 α-chains , but we expect convergent recombination to be sufficiently rare overall for this distinction not to qualitatively affect clone size distributions . For all studies we used data , which was preprocessed as described in the original study . This data is publicly available using the links provided in Table 1 . We also used flow cytometry data on the fraction of naive cells from Britanova et al . , 2014 ( available at https://doi . org/10 . 1371/journal . pcbi . 1005572 . s016 ) and from Pediatric AIDS Clinical Trials Group et al . , 2003 . We describe T cell dynamics using the following general set of stochastic rate equations . The class of models we consider are known in the mathematical literature as birth-death-immigration models . The number of cells Ci , i=1 , … , M of each of the M clones in the repertoire changes according to ( 5 ) proliferation:Ci→bi ( C , X , t ) CiCi+1 , ( 6 ) death:Ci→di ( C , X , t ) CiCi−1 , where the rate of proliferation bi⁢ ( 𝑪 , 𝑿 , t ) or cell death di⁢ ( 𝑪 , 𝑿 , t ) generally can depend on the repertoire composition 𝑪 , on the time t , and on the state of the environment 𝑿⁢ ( t ) representing for example the levels of different antigens and cytokines in the organism at a given time . We furthermore consider that new clones are added at rate θ⁢ ( 𝑿 , t ) at a size C0 , ( 7 ) recruitment:→θ ( X , t ) CM+1=C0 . This recruitment represents thymic output and antigen-driven differentiation of naive cells for the naive and memory compartment , respectively . bi ( C , X , t ) =b0/N , di ( C , X , t ) =d , θ ( X , t ) =θwhere N⁢ ( t ) =∑j=1M⁢ ( t ) Cj⁢ ( t ) is the total repertoire size . In the results section 'Long-lived incumbency advantage shows early expansions imprint clone size hierarchy' w modify this model by adding a noise term that describes the effective influence of environmental variations 𝑿⁢ ( t ) on clonal proliferation , ( 9 ) bi⁢ ( 𝑪 , 𝑿 , t ) =b0/N+2⁢σ⁢ηi⁢ ( t ) , where ⟨ηi⁢ ( t ) ⁢ηj⁢ ( t′ ) ⟩=δi⁢j⁢δ⁢ ( t-t′ ) .	The human immune system develops a memory of pathogens that it encounters over its lifetime , allowing it to respond quickly to future infections . It does this partly through T cells , white blood cells that can recognize different pathogens . During an infection , the T cells that recognize the specific pathogen attacking the body will divide until a large number of clones of these T cells is available to help in the fight . After the infection clears , the immune system ‘keeps’ some of these cells so it can recognize the pathogen in the future , and respond quicker to an infection . Over the course of their lives , people will be infected by many different pathogens , leading to a wide variety of T cells that each respond to one of these pathogens . However , it is not well understood how various infections throughout the human lifespan shape the overall population of different T cells . Gaimann et al . used mathematical modelling to study how the composition of the immune system changes in people of different ages . Different populations of T cells – each specialized against a specific antigen – had been previously identified through genetic sequencing . Gaimann et al . analyzed their dynamics to show that many of the largest populations originate around birth , during the formation of the immune system . These findings suggest a potential mechanism for how exposure to pathogens in infancy can influence the immune system much later in life . The results may also explain variations in how people respond to infections and in their risk of developing autoimmune conditions . This understanding could help develop new treatments or interventions to guide the immune system as it develops .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "physics", "of", "living", "systems", "immunology", "and", "inflammation" ]	2020	Early life imprints the hierarchy of T cell clone sizes
Stable mutualism between a host and its resident bacteria requires a moderated immune response to control bacterial population size without eliciting excessive inflammation that could harm both partners . Little is known about the specific molecular mechanisms utilized by bacterial mutualists to temper their hosts’ responses and protect themselves from aggressive immune attack . Using a gnotobiotic larval zebrafish model , we identified an Aeromonas secreted immunomodulatory protein , AimA . AimA is required during colonization to prevent intestinal inflammation that simultaneously compromises both bacterial and host survival . Administration of exogenous AimA prevents excessive intestinal neutrophil accumulation and protects against septic shock in models of both bacterially and chemically induced intestinal inflammation . We determined the molecular structure of AimA , which revealed two related calycin-like domains with structural similarity to the mammalian immune modulatory protein , lipocalin-2 . As a secreted bacterial protein required by both partners for optimal fitness , AimA is an exemplar bacterial mutualism factor . The vertebrate gastrointestinal tract harbors the highest density of microbes recorded for any microbial habitat ( Whitman et al . , 1998 ) . These microbial communities are not only tolerated by their hosts , but are also required for normal host development and physiology , highlighting the intimacy of these mutualisms ( McFall-Ngai et al . , 2013 ) . How these host-microbe mutualisms are established and maintained is unresolved . Understanding these processes will require identifying the molecular mechanisms by which resident microbes modulate the host immune system and defining the benefit these microbes derive from investing in such interactions . This knowledge is critical to dissect the microbial etiology of a range of diseases associated with excessive inflammation , including inflammatory bowel disease ( IBD ) , diabetes , metabolic syndrome , and sepsis ( Huttenhower et al . , 2014; Turnbaugh et al . , 2006; Wen et al . , 2008 ) . Furthermore , a thorough understanding of these host-microbe mutualisms will accelerate the development of effective treatments for these inflammatory diseases . Here , we explore the host-bacterial mutualism between zebrafish and their prominent colonizer , Aeromonas , and identify a secreted bacterial protein that promotes both bacterial colonization and host survival by preventing excessive inflammation . Pathogenic bacterial species are often defined by unique virulence factors that enable host colonization and determine the severity of host disease symptoms . Much less well known are the unique factors produced by anti-inflammatory bacterial species , which likely play a key role in establishing host-bacterial mutualism and maintaining intestinal homeostasis . One such factor is the zwitterionic polysaccharide , PSA , produced by the human symbiont Bacteroidetes fragilis , that induces a toleragenic T cell profile and protects against intestinal inflammation ( Mazmanian et al . , 2008; Mazmanian et al . , 2005; Round and Mazmanian , 2010 ) . The anti-inflammatory properties of Lactobacillus rhamnosus have been attributed to secreted proteins p75 and p40 ( Yan et al . , 2007 ) , and Faecalibacterium prausnitzii secretes an anti-inflammatory 15 kDa protein called microbial anti-inflammatory molecule or MAM ( Yan et al . , 2007; Quévrain et al . , 2016; Sokol et al . , 2008; Carlsson et al . , 2013; Martín et al . , 2014 ) . These secreted proteins reduce inflammation in mouse models of colitis ( Yan et al . , 2011 ) , which suggests their therapeutic potential , but we do not yet understand how these proteins aid the bacteria that produce them . Investigating the bacterial fitness benefits for producing anti-inflammatory activities is critical for developing approaches to promote membership of these immunoregulatory bacterial species in chronic inflammatory diseases like IBD and to use bacterial immune modulators as therapeutics to shape microbiota composition . The investigation of microbiota-derived immunomodulatory factors is challenging because their effects on the host are subtler then those of disease-causing toxins and they are produced by complex and genetically intractable microbial consortia . We used the zebrafish , Danio rerio , as a high-throughput , gnotobiotic model vertebrate system to identify individual bacterial products with immunomodulatory properties and understand how those products promote host-bacterial mutualism . Zebrafish have long been used as a vertebrate model for developmental biology because of their optical transparency , high fecundity , and rapid early development ( Grunwald and Eisen , 2002 ) . Zebrafish embryos develop ex-utero within their protective chorions and first encounter environmental microbes when they hatch as larvae between 2 and 3 days post fertilization ( dpf ) . By 5 dpf the animals have a fully functional digestive tract that is colonized with bacteria ( Bates et al . , 2006; Stephens et al . , 2016; Wallace et al . , 2005 ) . The larvae are also equipped with a functional innate immune system with highly dynamic neutrophils and macrophages that serve as the primary cellular defenses before the emergence of adaptive immune cells at approximately 3 weeks of age ( Renshaw and Trede , 2012 ) . The zebrafish is an excellent model for studying the establishment of host-bacterial symbioses ( Burns and Guillemin , 2017 ) because hundreds of zebrafish can be easily derived and maintained in a gnotobiotic state with defined microbial isolates during larval development ( Melancon et al . , 2017 ) . The large population sizes of animals allow for the analysis of the subtle effects of resident microbes on host phenotypes . For example , we have shown that intestinal neutrophil populations in larval zebrafish range from an average of 3 cells in germ free ( GF ) animals to 7 cells in conventionally reared ( CV ) animals , a difference that can be readily detected in well-powered experiments despite extensive inter-host variation ( Rolig et al . , 2015 ) . Our use of gnotobiology to assay the effect of individual microbes and their products on the host has been aided by the fact that many zebrafish bacterial isolates are culturable , have known genome sequences ( Stephens et al . , 2016 ) , and are genetically tractable ( Hill et al . , 2016; Wiles et al . , 2016 ) . This gnotobiotic approach , however , fails to capture the complexity of naturally occurring vertebrate microbiota . Additionally , in the zebrafish we are largely limited to using gnotobiology to study the larval period because we have yet to develop adequate nutritional and husbandry approaches to promote the growth of GF zebrafish to adulthood . In this study we focused on the mutualism established between larval zebrafish and their resident Aeromonas species . Aeromonas is an important symbiont of the zebrafish and the sole bacterial genus that we found present in 100% of individuals at all developmental time points in a comprehensive longitudinal study of zebrafish gut microbiomes across their lifespan ( Stephens et al . , 2016 ) . Furthermore , we have shown that in mono-associations , representatives of the Aeromonas genus can fully rescue defects in glycan expression ( Bates et al . , 2006 ) , epithelial cell proliferation ( Cheesman et al . , 2011 ) , innate immune response ( Rolig et al . , 2015 ) , and pancreatic beta cell expansion ( Hill et al . , 2016 ) that are characteristic phenotypes of GF fish . We have examined the dynamics of the Aeromonas-zebrafish mutualism , taking advantage of the optical transparency of the larvae to monitor both bacterial and host cells in vivo ( Taormina et al . , 2012; Jemielita et al . , 2014 ) . Here we specifically quantified the innate immune response against Aeromonas bacterial genetic variants , using GFP-expressing intestinal neutrophils as a metric of the host response ( Renshaw et al . , 2006 ) . Neutrophils are a primary component of the initial inflammatory response and are responsive to both infecting pathogens and resident intestinal microbes ( Harvie and Huttenlocher , 2015; Bates et al . , 2007; Kanther et al . , 2014 ) . This responsiveness makes the intestinal neutrophil population an ideal readout of the host-microbe mutualism between zebrafish and Aeromonas . Using the gnotobiotic zebrafish model , we identified a potent immune modulatory protein secreted by Aeromonas , which we named Aeromonas immune modulator A ( AimA ) because it reduces intestinal inflammation . To understand the molecular mechanism of AimA’s function , we determined the protein crystal structure , which revealed two related calycin-like domains that are structurally similar to the immune modulatory factor found in mammals , lipocalin-2 ( LCN2 ) . Consistent with AimA targeting the host immune response , as opposed to counteracting a specific bacterial product , we showed that purified AimA , and each of its subdomains , prevents chemically induced intestinal inflammation . Furthermore , we found that this activity is blocked by exogenous mouse LCN2 ( mLCN2 ) , implying a shared mechanism of action between these structurally similar bacterial and mammalian proteins . However , AimA does not appear to function like LCN2 as a binding protein for bacterial siderophores . To understand AimA’s function for Aeromonas , we generated mutants lacking aimA and its homologue aimB . In mono-associations with aim mutants , we found that both the host and bacterial partners suffered reduced viability , which could be rescued with exogenous AimA protein . We showed that the fitness cost to the bacteria was due to excessive inflammation because the aim mutant survival deficit was erased in an immunocompromised host . In parallel , we showed AimA’s benefit to the host extended to protection from lipopolysaccharide ( LPS ) intoxication . Together , these data provide an exemplar of a bacteria immune modulatory factor that is beneficial for both partners in a bacterial mutualism . Based on the mutualism between Aeromonas and zebrafish ( Stephens et al . , 2016 ) , we hypothesized that Aeromonas would use specific secreted factors to co-exist with its host . We tested this hypothesis by comparing the zebrafish intestinal neutrophil response to mono-association with an Aeromonas isolate , A . veronii strain Hm21 ( Maltz and Graf , 2011 ) , versus mono-association with an isogenic mutant lacking the type two secretion system ( T2SS ) ( ΔT2 ) ( Table 1 ) , a major system for secretion in Gram-negative bacteria ( Maltz and Graf , 2011 ) . We colonized each bacterial strain in zebrafish from 4 dpf to 6 dpf and measured the number of intestinal neutrophils that accumulated in response to each colonizing bacterial strain . In these experiments we compared the ΔT2 strain to a genetically complemented strain , ΔT2C , previously shown to restore T2SS function ( Maltz and Graf , 2011 ) , which allowed us to rule out the possibility that phenotypes associated with the ΔT2 strain were due to polar or second site mutations . The ΔT2C and ΔT2 colonized the zebrafish intestine to similar levels in mono-associations ( Figure 1—figure supplement 1 ) , however the ΔT2 mutant induced a greater intestinal neutrophil response than the ΔT2C ( Figure 1A ) . This result suggests that A . veronii strain Hm21 secretes a product that decreases the neutrophil response to its colonization . The increased neutrophil response to ΔT2 was rescued by concurrent treatment with cell-free supernatant ( CFS ) collected from the ΔT2C strain , but not CFS collected from ΔT2 ( Figure 1A ) . To identify the factor produced by ΔT2C and not from the ΔT2 strain that resulted in fewer intestinal neutrophils , we performed mass-spectrophotometry on the CFS from these two strains and determined which proteins were enriched in the ΔT2C compared to the ΔT2 strain . This analysis resulted in a list of 22 proteins that were enriched by greater than 10 counts in the ΔT2C CFS ( Supplemental file 1 ) . We narrowed down the potential candidates further by performing ammonium sulfate fractionation on the ΔT2C CFS and adding fractions to fish mono-associated with ΔT2 . We found neutrophil modulating activity primarily in the 40–60% salt fraction , but also saw some activity in the 20–40% salt fraction ( Figure 1A ) . An SDS PAGE gel of these fractions revealed a similar banding pattern for each fraction , except the bands in the 20–40% fraction were considerably weaker . We observed two prominent bands at 33 kDa and 55 kDa . These sizes corresponded to two proteins on our list of proteins enriched in the ΔT2C CFS , an uncharacterized protein ( UP ) and a chitin binding protein ( CBP ) , respectively ( Supplemental file 1 ) . To test whether either of these proteins modulated the neutrophil response to ΔT2 colonization , we cloned the sequence for each gene into an overexpression vector and induced expression in Escherichia coli , resulting in E . coli CFS that was heavily dominated by the protein of interest ( Figure 1B ) . We treated zebrafish mono-associated with ΔT2 with the CFS resulting from overexpression of each protein of interest and found neutrophil modulating activity in the E . coli CFS with the 33 kDa UP and not with CBP or control CFS from E . coli containing an empty expression vector ( Figure 1C ) . We named this neutrophil-modulating protein Aeromonas immune modulator A ( AimA ) . DNA and amino acid homology searches revealed that no homologues to AimA exist outside of the Aeromonas genera . Further , analysis of the amino acid sequence of AimA with the N terminal amino acid secretion signal removed revealed a lack of sequence identity with known domains or structurally characterized proteins and therefore offered little insight into the structure or function of AimA . Thus , to gain insight into the mechanism of AimA , we constructed a His-tagged version that was purified from E . coli , and we crystallized the protein . We determined the molecular structure of AimA using a heavy atom derivative and Single-wavelength Anomalous Dispersion ( SAD ) phasing to a resolution of 2 . 3 Å ( PDB 6B7L; Table 2 ) . The structure of AimA revealed two domains connected by a short linker ( Figure 2A , B ) . β-strands dominate each domain with the carboxy terminal ( C-term ) domain forming a complete β-barrel and the amino terminal ( N-term ) domain containing a curved β-sheet . Structural homology searches of full length AimA against all Protein Data Bank-deposited structures using PDBeFold resulted in structures that aligned to only one domain or the other , with the majority aligning with the C-term domain ( Krissinel et al . , 2004; Berman et al . , 2000 ) . Therefore , we performed structural homology searches against each domain separately , which revealed that both domains had similarity to proteins in the calycin superfamily ( Supplemental file 2 and Figure 2—figure supplement 1 ) . The calycin superfamily of proteins contains members from all domains of life and includes lipocalins , fatty acid binding proteins , and avidins . This superfamily is defined by an anti-parallel β-barrel with a repeated +1 topology ( Flower , 1996 ) . Notably , the calycin superfamily is known for structural conservation without high amino acid sequence conservation ( Flower , 1996; Lakshmi et al . , 2015 ) , consistent with a lack of sequence homology between AimA and other calycin family members . The N-term domain of AimA has an incomplete β-barrel , but maintains some structural homology to streptavidin . Streptavidin binds biotin tightly , but we found no evidence that AimA binds biotin ( Figure 2—figure supplement 1 ) . The C-term domain of AimA has structural homology both to avidins and lipocalins ( Supplemental file 2 ) . AimA’s structural similarity to lipocalin proteins was intriguing because of these proteins’ known roles in modulating neutrophil behavior ( Moschen et al . , 2017 ) . Based on this structural similarity , we wondered whether AimA’s influence on intestinal neutrophil numbers was specific to Aeromonas colonization or whether it acts as a more general immunoregulatory molecule to influence neutrophil behavior . To test whether AimA modulates neutrophil numbers in response to stimuli other than Aeromonas mono-association , we employed the zebrafish model of soysaponin-induced inflammation . Farmed fish , such as salmon and carp , fed soybean meal as a protein source are well known to develop intestinal inflammation , and zebrafish are a good model of this irritation ( Hedrera et al . , 2013 ) . We fed conventionally raised zebrafish larvae Zeiglers fish food with 0 . 3% soysaponin from 4 dpf to 6 dpf and saw a significant increase in the number of intestinal neutrophils in response to the soysaponin , as expected ( Figure 3A ) . When we treated the fish with 100 ng/mL purified AimA concurrently with soysaponin , AimA prevented the increase in neutrophil influx in response to soysaponin ( Figure 3A ) . This result demonstrates that AimA alone can inhibit pro-inflammatory signaling pathways elicited by stimuli other than Aeromonas . Because AimA reduces neutrophil influx in a model of intestinal inflammation and the C-term domain has structural homology to lipocalin proteins ( Figure 3A , Supplementary file 2 ) , we hypothesized that the neutrophil reducing capacity of AimA was related to its lipocalin-like structure . Lipocalins are defined by three key structurally conserved regions , SCR1 , SCR2 , and SCR3; of these , SCR1 contains four key residues and is the most conserved ( Figure 3B ) ( Flower , 1996; Lakshmi et al . , 2015 ) . The SCR1 in the C-term domain of AimA overlaps structurally with the SCR1 of mouse lipocalin-2 ( mLCN2 ) and shares some related amino acids ( Figure 3B ) . Given this structural similarity , we hypothesized that AimA may function in the same pathways as mLCN2 . Thus , we asked whether the presence of mLCN2 protein would alter the capacity of AimA to reduce the neutrophil response to soysaponin . We found that the addition of mLCN2 concurrently with soysaponin did not influence the neutrophil response to soysaponin ( Figure 3C ) ; however , addition of mLCN2 in conjunction with AimA blocked the protective effect of AimA against soysaponin-induced inflammation . This result suggests that mLCN2 may interfere with AimA’s activity by direct binding , competing for the same host receptor , or acting in a competing pathway . LCN2’s immunomodulatory activity is mediated , at least in part , by binding iron-sequestering siderophores , including enterobactin from E . coli ( Xiao et al . , 2017 ) , thus we explored whether AimA had a similar enterobactin-binding activity . The original evidence for LCN2’s enterobactin binding came from the protein crystal structure of human LCN2 , in which enterobactin was co-crystalized ( Goetz et al . , 2002 ) . We did not find enterobactin in the AimA structure , which was expected because the E . coli strain , BL21 DE3 , which we used to express AimA , does not produce enterobactin . However , in comparison to the wide enterobactin-binding cleft ( calyx ) of LCN2 , the homologous regions of the N- and C-term domains of AimA are much narrower , are occluded by loops or a helix , and lack the enterobactin-interacting residues of LCN ( Figure 3—figure supplement 1A–F ) . To test whether AimA binds enterobactin , we grew E . coli K12 strain ( MG1655 ) under iron limiting conditions that make it dependent on the siderophore function of its secreted enterobactin , which can be sequestered by exogenously added LCN2 ( Goetz et al . , 2002 ) . As predicted , when we grew E . coli K12 in the presence of the iron chelator dipyridyl to limit available iron , we observed a dose dependent inhibition of E . coli growth upon addition of purified mouse LCN2 protein ( Figure 3—figure supplement 1G ) . In contrast , when we added the same range of concentrations of purified AimA protein to the E . coli cultures , we saw no growth inhibition ( Figure 3—figure supplement 1G ) , suggesting that AimA does not bind and sequester enterobactin in a similar manner to LCN2 . Human LCN2 exists both as a monomer and a homodimer , and while the functional distinction between the two forms is unknown , the homodimer is the major molecular form secreted by neutrophils ( Cai et al . , 2010 ) . Both the C- and N-terminal domains of AimA have structural homology to proteins in the calycin superfamily . Although they share only 17% amino acid identity , a structural overlay between the two domains reveals some structural conservation ( Figure 4A and Figure 4—figure supplement 1A ) . Because human LCN2 exists as both a monomer and homodimer and the two domains of AimA each have a lipocalin-like fold , we asked if each domain alone was sufficient to alter the intestinal neutrophil response . The individual domains were not as soluble as full-length AimA , but we were able to crudely purify each domain . Concurrently with soysaponin , we added approximately 100 ng/mL of the domain of interest to zebrafish from 4 to 6 dpf . We found that both the N- and C-term domains alone were sufficient to reduce the neutrophil response in the soysaponin model of intestinal inflammation ( Figure 4B ) . An in depth analysis of the structural overlay between the N- and C-term domains revealed that seven out of eight β-strands in the barrels align well; however , only seven total residues are in analogous functional positions across the two domains ( Figure 4—figure supplement 1 ) . Of these seven residues , two—Val 60/201 and Thr 117/265—stand out as possible candidates to interact with a hydrophobic ligand that could bind inside the barrel cavity . The other five residues may overlap by coincidence or they may be positioned to interact with a promiscuous protein or ligand partner . Furthermore , given the overall structural similarity between the N- and C-term domains , it is possible that critical residues are in a flexible loop region that could become structured upon binding ( Figure 4A and Figure 4—figure supplement 1A ) . Interestingly , both AimA domains have comparable structural similarity with mLCN2 ( Figure 4C ) . While the mLCN2 binding cleft for enterobactin is not conserved in the domains of AimA , the majority of the β-strands in the barrels overlap in all three domains , which suggests these domains could interact with the same binding partner ( s ) ( Figure 4C ) . With a clear understanding that AimA controls the host intestinal neutrophil response , we next asked whether the activity of AimA also increases Aeromonas fitness , thus facilitating a mutualistic relationship with the host . We began by asking how prevalent AimA was across bacterial genomes . Knowing that proteins in the calycin superfamily have low sequence conservation , we were not surprised to find AimA homologues by sequence similarity only within the Aeromonas genus ( Figure 5—figure supplement 1A ) . We found an almost identical homologue ( 99% ) to AimA in a zebrafish intestinal Aeromonas isolate , ZOR0001 , referred to here as ZF Aer ( Table 1 ) ( Stephens et al . , 2016 ) . Some Aeromonas species , including A . veronii strain Hm21 ( Table 1 ) , have both AimA and a second copy , which we named AimB . AimB is distantly related to AimA by amino acid sequence conservation ( 27% ) , yet a model generated by the Iterative Threading ASSEmbly Refinement ( I-TASSER ) program of the structure of AimB overlays directly on the structure of AimA ( Figure 5A ) ( Yang et al . , 2015; Roy et al . , 2010; Zhang , 2008 ) . A reexamination of our mass spectrometry analysis of the ΔT2C CFS uncovered peptides corresponding to AimB that were enriched in the sample from ΔT2C compared to the ΔT2 mutant , although the difference between the two samples was less than our original threshold ( Supplemental file 1 ) . To determine whether the bioactive protein AimA and its homologue AimB benefit Aeromonas , we constructed deletion strains of each gene individually and of both genes in A . veronii strain Hm21 ( Aer ΔaimA , Aer ΔaimB , and Aer ΔAΔB ) and an AimA deletion in the ZF Aer background ( ZF Aer ΔaimA ) . All of these mutants displayed normal growth in vitro in rich media ( Figure 5—figure supplement 1B ) . Because AimA has structural homology to LCN2 , which binds siderophores , and Aeromonas species make several siderophores , including enterobactin ( Maltz et al . , 2015 ) , we asked whether AimA functions in iron acquisition for Aeromonas by testing the ability of wild type Hm21 and the Aer ΔAΔB mutant to grow in iron depleted media containing dipyridyl . We found that the two strains had identical growth curves when grown under iron starvation ( Figure 5—figure supplement 1C ) , suggesting that AimA and AimB do not function in iron acquisition for Aeromonas . We next tested whether the strains lacking aim genes exhibited growth defects in mono-associations in larval zebrafish . We found that when inoculated into GF larvae at 3 dpf , the ZF Aer ΔaimA mutant colonized gnotobiotic zebrafish to lower levels than its wild type counterpart ( Figure 5B ) , with the colonization defect becoming apparent at 7 dpf ( Figure 5—figure supplement 2 ) . We observed a similar colonization defect at 7 dpf with the Aer ΔAΔB mutant relative to the wild-type strain ( Figure 5C ) . However , the single gene deletion strains colonized to wild-type levels , suggesting that AimA and AimB provide redundant functions for Aeromonas growth within the larval intestine . We note that Aer ΔT2 did not exhibit a colonization defect ( Figure 1—figure supplement 1 ) similar to the Aer ΔAΔB mutant , which could be due to other phenotypic consequences of its altered secretion profile . We next tested whether colonization defects of the aim deficient strains could be rescued by supplementation of the flask media with purified AimA protein . Remarkably , addition of purified AimA protein at a concentration of 100 ng/mL to the flask water from 4 to 7 dpf completely restored the colonization efficiency of the aim deficient strains ( Figure 5B , C ) . Because purified AimA reduces the intestinal neutrophil response ( Figure 3A , C ) , we hypothesized that the aim deletion strains would induce more inflammation , and thus upon mono-association could create a less hospitable intestinal environment for Aeromonas colonization . To test our hypothesis , we quantified intestinal neutrophil numbers in response to Δaim mono-associations at 7 dpf , corresponding to the time point when the bacterial colonization deficit for ZF Aer ΔaimA was apparent ( Figure 5—figure supplement 2 ) . We found that zebrafish mono-associated with ZF Aer ΔaimA exhibited significantly higher intestinal neutrophil counts as compared with those colonized with wild-type ZF Aer ( Figure 5D ) , despite carrying a lower bacterial load ( Figure 5B ) . Similarly , zebrafish colonized with the double mutant , Aer ΔAΔB , had significantly higher neutrophil counts than fish colonized with wild-type Aer ( Figure 5E ) , despite being colonized at a lower level ( Figure 5C ) . Notably , the Aer ΔaimA and Aer ΔaimB single mutants , which did not exhibit decreased colonization in mono-associations , also did not induce higher intestinal neutrophil numbers ( Figure 5E ) . For both Aeromonas lineages , when exogenous AimA was added to the flasks of fish mono-associated with inflammation-inducing Δaim strains , the protein reduced neutrophil numbers back to normal levels ( Figure 5D , E ) and at the same time restored colonization levels ( Figure 5B , C ) . We previously showed that resident zebrafish bacterial species can vary dramatically in their per capita impact on host response to bacterial colonization ( Rolig et al . , 2015 ) . By extrapolating the number of neutrophils elicited per 104 colonizing bacteria , we calculated that Aer ΔAΔB is much more immuno-stimulatory that the wild-type strain , recruiting nearly twice the number of neutrophils , on average , for the same number of bacteria ( Figure 6A ) . This suggests that the Aim proteins function to reduce Aeromonas’ immune stimulation , allowing it to reach high colonization density without eliciting a strong inflammatory response . Knowing that loss of the Aim proteins resulted in both a significantly increased per capita effect and intestinal neutrophil response , we explored whether the increased neutrophil response led to other downstream consequences for the host by monitoring the survival rate of mono-associated fish . By 72 hr post-inoculation , the survival rate of zebrafish colonized with wild-type Aer was 92% ( n = 163 ) . We observed a significant decline in the survival of fish mono-associated with Aeromonas lacking Aim genes [Aer ΔaimA ( n = 139 ) , Aer ΔaimB ( n = 174 ) , and Aer ΔAΔB ( n = 148 ) ] , with survival rates of 64% , 73% , and 55% , respectively ( Figure 6B ) . Remarkably , this decreased survival rate was rescued back to 90% by the presence of purified AimA protein ( n = 60; Figure 6B ) . These data demonstrate that the Aim proteins act to promote both bacterial colonization and host survival , identifying AimA as a key mediator of host-bacterial mutualism . To test whether AimA benefits both the bacteria and the host by controlling the neutrophil response to bacterial colonization , we evaluated the importance of Aim proteins in a colonization model lacking intestinal neutrophils . We reasoned that in the absence of intestinal inflammation , bacteria possessing or lacking Aim proteins would have a similar capacity to survive in the intestine . For these experiments , we used myd88-/- mutant fish lacking the adaptor protein for inducing many toll-like receptor and IL-1 dependent pro-inflammatory pathways ( Larsson et al . , 2012 ) ; these fish have a severely attenuated intestinal neutrophil response ( Figure 6C ) that is indistinguishable from GF wild-type zebrafish ( Figure 1A ) , as we have reported previously ( Bates et al . , 2007; Burns et al . , 2017 ) . We found that both Aer ΔAΔB and wild-type Aer colonized myd88-/- fish to equivalent levels , which were significantly higher than in wild-type fish ( Figure 6D ) . These results suggest that the decreased colonization of Aer ΔAΔB compared to wild-type Aeromonas in wild-type fish is due to the increased inflammatory response to the mutant bacteria , and moreover suggest that the hosts’ innate immune response contributes to limiting bacterial colonization . We further found that myd88-/- fish inoculated with wild-type Aeromonas or Aer ΔAΔB died at the same rate ( Figure 6E ) , suggesting that the increased inflammatory response of wild-type fish colonized with Aer ΔAΔB compared to wild-type Aeromonas accounts for the decreased fish survival rate with the mutant bacteria . Notably , the myd88-/- fish mono-associated with either wild-type or Aer ΔAΔB perished at higher rates than wild type fish , indicating that there is a host fitness cost to failing to control Aeromonas levels . Given the increase in bacterial colonization and fish death in the myd88-/- fish , regardless of the colonizing strain , we hypothesized that these fish may be dying of septic shock as a result of bacterial overgrowth . Thus we asked whether AimA could delay death as a result of another zebrafish model of bacterial septic shock , which is induced by LPS intoxication ( Bates et al . , 2007; Philip and Wang , 2017 ) . To test this , we treated fish with 100 ng/mL AimA on 4 dpf and then challenged them with 600 μg/mL LPS on 5 dpf and tracked their survival over the subsequent 60 hr . The median survival time post LPS treatment was 32 hr , which we found was extended to 47 hr in the presence of AimA ( Figure 6F ) . Using the same treatment paradigm , we found that purified AimB also extended the median survival of LPS-exposed zebrafish to 47 hr ( Figure 6F ) . In contrast , treatment with mLCN2 had no effect on zebrafish survival ( median survival 34 hr , Figure 6F ) . Thus , AimA and AimB , but not mLCN2 , significantly extend the survival of zebrafish in an LPS intoxication model . Appropriate modulation of host immune responses to resident microbes is critical for maintaining health , with too muted a response leaving hosts susceptible to microbial overgrowth ( Dukowicz et al . , 2007; Rolig et al . , 2017 ) and too aggressive a response wreaking havoc on both host tissue and resident microbiota ( Huttenhower et al . , 2014 ) . Fitness effects of host-bacterial interactions are often considered from the host perspective , yet it is essential to examine both sides of these partnerships to understand how these interactions evolve and persist . Here we describe the discovery of AimA , a bacterial immunomodulatory protein secreted by Aeromonas that is required during bacterial colonization of larval zebrafish to prevent an intestinal inflammatory response that is detrimental to both partners . We determined the molecular structure of AimA and show that it consists of two structurally similar lipocalin-like domains , each of which acts to suppress intestinal inflammation . We demonstrate that purified AimA protein is protective against host intestinal inflammation and LPS intoxication and simultaneously rescues Aeromonas growth in the context of inflammation . When the inflammatory response is attenuated in myd88 deficient hosts , then the bacterial colonization advantage associated with AimA production is eliminated , indicating that it is AimA’s anti-inflammatory activity that benefits Aeromonas . All resident bacteria produce microbial associated molecular patterns ( MAMPs ) that are inherently pro-inflammatory , as we have shown in zebrafish larvae for LPS , which elicits intestinal neutrophil influx , proinflammatory cytokine gene expression , and endotoxemia in a dose-dependent manner ( Bates et al . , 2007 ) . However , our analysis of dose response curves to resident bacteria revealed that few zebrafish microbiota members elicit dose-dependent inflammation ( Rolig et al . , 2015; Rolig et al . , 2017 ) . We did identify a few examples of bacteria with potent and dose-dependent inhibitory effects on intestinal neutrophil influx . These dose-dependent inhibitory effects were exploited to identify an immunoregulatory Shewanella strain that produces a secreted , heat-labile , anti-inflammatory factor ( Rolig et al . , 2015 ) and an Escherichia strain that reduces intestinal inflammation through a secretion-independent mechanism ( Rolig et al . , 2017 ) . However , the majority of zebrafish gut bacteria we surveyed , including Aeromonas species , showed no correlation between their bacterial load and the number of neutrophils they elicited ( Rolig et al . , 2015; Rolig et al . , 2017 ) , suggesting that most gut bacteria posses both pro-inflammatory products , like LPS , along with neutralizing mechanisms to counteract the effects on their hosts . AimA , which decreases intestinal neutrophil influx and septic shock elicited by both Aeromonas and LPS , represents such a neutralizing activity . Future studies will determine whether AimA acts primarily by reducing the production of pro-inflammatory cytokines that recruit neutrophils to the intestine or by inhibiting the responsiveness of neutrophils to such cues . Sequence-based searches for AimA-related proteins recovered only other Aeromonas homologues , including the highly divergent but functionally redundant AimB . Structural determination of AimA , however , revealed two linked domains of the lipocalin superfamily . Lipocalin domains are found across the tree of life , with bacterial lipocalin fold-containing proteins being only more recently identified ( Bishop , 2000 ) . In eukaryotes , proteins with lipocalin folds commonly bind small hydrophobic molecules such as odorant molecules , retinol , and fatty acids ( Flower , 1996 ) , but proteins with a lipocalin fold , especially those found in prokaryotes , can also take on a variety of other functions , such as inducing apoptosis ( Tavares and Pathak , 2017 ) , binding extracellular matrix ( Parker et al . , 2016 ) , and binding hydrophobic antibiotics and antimicrobials ( El-Halfawy et al . , 2017; Sisinni et al . , 2010 ) . We were particularly intrigued by AimA’s structural similarity to mammalian LCN2 , a known regulator of intestinal inflammation ( Moschen et al . , 2017; Xiao et al . , 2017 ) . LCN2 was originally identified and purified from neutrophil granules and shown to play a critical role in neutrophil extravasation and migration—processes that AimA may also modulate . LCN2 was subsequently shown to be a sensitive fecal biomarker of intestinal inflammation in humans and mice ( Chassaing et al . , 2012 ) . Inflammation-associated induction of LCN2 appears to be protective against pathology because Lcn2 deficient mice are sensitized to developing colitis ( Toyonaga et al . , 2016; Moschen et al . , 2016; Singh et al . , 2016 ) . Further , Lactococcus lactis engineered to overexpress mLCN2 confers greater protection against colitis than a control strain of L . lactis in a mouse model of acute colitis ( Saha et al . , 2017 ) . While it is clear that LCN2 protects against inflammation , the mechanisms by which it does so are not well understood . One possibility is that mLCN2’s capacity to sequester bacterial enterobactin-like siderophores may cause beneficial shifts in intestinal microbiota composition and reduce the toxic effects of iron in the inflamed environment ( Moschen et al . , 2016 ) . Our data , however , suggest that AimA confers protection against inflammation by a distinct mechanism , because we found no evidence that AimA binds enterobactin or contributes to Aeromonas survival in iron limiting conditions . To test for a possible overlap between LCN2 and AimA function , we asked whether addition of mLCN2 to a model of soysaponin-induced intestinal inflammation would alter AimA’s efficacy . Indeed , the presence of mLCN2 inhibited the ability of AimA to reduce the neutrophil response . This result may indicate that the two proteins compete by acting on the same pathway governing neutrophil behavior . For example , AimA and mLCN2 may compete directly for binding to the same host receptor . LCN2 binds at least two known mammalian receptors , megalin and LCN2R , which have homologues in zebrafish ( Kim et al . , 2011 ) , although the binding to these receptors has not been directly connected to the influence of LCN2 on neutrophil behavior ( Moschen et al . , 2017; Cabedo Martinez et al . , 2016 ) . Another possibility is that mLCN2 interferes with AimA’s activity by binding directly to AimA . mLCN2 is known to exist both as monomers and as dimers ( Axelsson et al . , 1995; Perduca et al . , 2001; Niemi et al . , 2015 ) , and while little is known about the conditions under which each form is active , the dimeric form is associated with neutrophils ( Cai et al . , 2010 ) . In AimA , the two lipocalin-like domains are tethered , but we found that each structural half can function independently to inhibit inflammation . Future biochemical studies will determine whether either of AimA’s lipocalin domains can homodimerize or bind to other lipocalin proteins and which oligomeric states of AimA confer its immune modulatory function . Our analysis of aimA and aimB deficient Aeromonas revealed that the proteins’ anti-inflammatory activity on the host is inextricable from the fitness advantage it confers to the bacteria . In this regard , AimA represents a new class of bacterial effector proteins , which we refer to as mutualism factors . Whereas virulence factors promote the fitness of pathogens to the detriment of their hosts and are revealed by the decreased fitness of the pathogen mutants coupled to the increased fitness of the host , a mutualism factor like AimA confers mutual benefits , and both partners suffer when it is removed from the host-microbe symbiosis . Another example of a mutualism factor is the zwitterionic polysaccharide PSA produced by Bacteroidetes fragilis , which colonizes the murine colon . PSA induces T regulatory cells ( T regs ) , which benefit its murine host by preventing excessive inflammation . The T regs in turn contribute to B . fragilis colonization , as demonstrated by the fact that in the absence of T regs , due either to genetic ablation or absence of PSA-induction , B . fragilis suffers a colonization defect ( Round et al . , 2011 ) . Given the constraints of bacterial-host coexistence in the vertebrate intestine , we speculate that intestinal microbiota are rich sources of novel anti-inflammatory factors that enable resident microbes to colonize this tissue at high density without eliciting excessive inflammation . Such anti-inflammatory factors offer promise for useful therapeutic applications , and the high throughput gnotobiotic larval zebrafish system provides an experimentally tractable platform for their discovery . These factors , however , may represent novel biological molecules that defy easy functional classification based purely on their gene sequences or the gene sequences of their biosynthetic machinery , thus additional approaches , such as protein structure determination , will be needed to further advance our understanding of their mechanisms of action . In the case of Aeromonas AimA , the crystal structure revealed it to belong to a large family of lipocalin proteins , which includes the mammalian immunomodulatory protein LCN2 . Our discovery of AimA raises the possibility that AimA and LCN2 have overlapping functions as mutualism factors in the vertebrate intestine even though they are provisioned by different members of the partnership . Given the diversity of the microbial communities that are part of the complex host-microbe ecosystem , AimA is likely only the first of many unique bacterial proteins that facilitate mutualistic interactions critical for both host and resident microbes to thrive . All zebrafish experiments were performed following protocols approved by the University of Oregon Institutional Animal Care and Use Committee and followed standard zebrafish protocols ( Westerfield , 2000 ) . Conventionally-raised ( CV ) wild-type ( AB x Tu strain ) , Tg ( BACmpx:GFP ) i114 ( referred to as mpx:GFP ) ( Renshaw et al . , 2006 ) and myd88-/- ( Burns et al . , 2017 ) were maintained as described ( Westerfield , 2000 ) . Zebrafish embryos were derived germ free ( GF ) as previously described ( Melancon et al . , 2017 ) . Subsequently , 15 GF embryos were transferred to sterile tissue culture flasks ( 25 cm2 , Techno Plastic Products , Trasadingen , Switzerland ) with 15 mL embryo medium ( EM ) ( Melancon et al . , 2017 ) . Monoassociated zebrafish were generated by inoculating flasks with 4 dpf GF zebrafish with 106 colony forming units ( CFU ) /mL of each bacterial strain . GF flasks were chosen at random for their respective treatment , and the researcher was blinded to the treatment group until after data collection . Bacteria used for monoassociations were zebrafish isolate Aeromonas ZOR0001 ( Stephens et al . , 2016 ) , Aeromonas ZOR0001ΔaimA , Aeromonas ZOR0001 ΔaimA complement , Aeromonas HE-1095 ( ΔT2 ) ( Maltz and Graf , 2011 ) , Aeromonas HEC-1344 ( ΔT2C ) ( Maltz and Graf , 2011 ) , Aeromonas Hm21S ( Graf , 1999 ) , Hm21S ΔaimA , Hm21S ΔaimB , and HM21S ΔaimAΔaimB . Control CV fish were prepared GF , as above , and inoculated with 1 mL of fish facility water on 4 dpf . All manipulations to the GF flasks were performed under a class II A/B3 biological safety cabinet . The flasks were kept at 28° C until analysis of fluorescent myeloperoxidase positive ( MPX+ ) cells on 7 dpf . Aeromonas mutants were created using plasmids and protocols as described in Wiles et al . ( 2018 ) . Briefly , Aeromonas mutants were constructed using homologous recombination of 1 kb regions upstream and downstream of the aimA gene to replace aimA with a chloramphenicol resistance gene . The aimB mutant was constructed in a similar manner , except as a markerless deletion by first generating a merodiploid strain . Zebrafish larvae were fixed in 4% paraformaldehyde ( PFA ) overnight . Whole larvae were stained with Myeloperoxidase kit ( Sigma ) following the manufacturer’s protocol and processed and analyzed as previously described ( Bates et al . , 2007 ) , except that MPO +cells were quantified in whole dissected intestines rather than tissue sections . For analysis of neutrophils in mpx:GFP fish , GFP +cells in the intestine were quantified as previously described ( Rolig et al . , 2015 ) . Briefly , the mpx:GFP zebrafish were anesthetized in Tricaine ( Western Chemical , Inc . , Ferndale , WA ) and mounted in 4% methylcellulose ( Fisher , Fair Lawn , NJ ) , and their intestines were sterilely dissected . The number of GFP-positive cells was quantified visually for each fish using a fluorescent microscope ( SteREO Discovery . V8 , Zeiss ) . To determine the CFU/intestine , dissected zebrafish intestines were placed in 100 μL sterile EM , homogenized in a bullet blender ( Next Advance ) , diluted , and cultured on tryptic soy agar ( TSA , BD , Sparks MD ) . The TSA plates were incubated at 30° C overnight and then colonies were counted . Aeromonas HEC-1344 or E . coli BL21 were grown overnight to stationary phase . Then 500 μL of the overnight culture was used to inoculate a 50 mL culture , which was kept shaking at 30° C for 2 hr , until the bacteria had grown through exponential and into stationary phase ( OD of approximate 0 . 6 ) . To overexpress proteins of interest , 1 mM of IPTG was added to the cultures during exponential growth phase and allowed to grow for an additional 2 hr at 30° C . To prepare the CFS , the 50 mL cultures were centrifuged at 7000 x g for 10 min at 4° C . Subsequently , the supernatant was filtered through a 0 . 22 μm sterile tube top filter ( Corning Inc . , Corning , NY ) . The sterile supernatant was concentrated at 4° C for 1 hr at 3000 x g with a centrifugal device that has a 10 kDa weight cut off ( Pall Life Sciences , Ann Arbor , MI ) . The concentration of the supernatant was determined with a Nanodrop and inoculated into the flasks at a final concentration of 500 ng/mL . Ammonium sulfate fractionation experiments were done as previously described ( Hill et al . , 2016 ) . Briefly , unconcentrated CFS from a 50 mL overnight culture of Aeromonas HEC-1344 was fractionated by slowly adding cold 100% ammonium sulfate until solutions reached 20% , 40% , and 60% ammonium sulfate . Precipitated proteins were collected by centrifugation at 4° C , 15000 g for 15 min . Recovered proteins were resuspended in cold EM , dialyzed for 2–3 hr at 4° C , then added to mono-associated fish flasks at a concentration of 500 ng/mL . The protein constituents of concentrated cell free supernatants from Aeromonas HE-1095 and Aeromonas HEC-1344 ( Maltz and Graf , 2011 ) were determined by analysis of peptide MS/MS spectra at the Proteomics Shared Resource Facility at Oregon Health and Sciences University in Portland , Oregon . Soysaponin ( Sigma ) was mixed with Zieglers fish food at a concentration of 0 . 3% . Ten CV zebrafish were maintained 10 mL EM in 60 × 15 mm petri dishes . Larval fish were fed once daily from 4 dpf to 6 dpf . During each feeding , the larvae had access to the food for 3–4 hr before being washed into fresh EM . For experiments with mLCN2 , recombinant mouse LCN2 ( Biolegend ) was added to the fish EM at a concentration of 100 ng/mL after the soysaponin feeding on 4 dpf and 5 dpf after the fish were moved into fresh EM . To induce LPS intoxication , CV zebrafish were treated with 600 μg/mL LPS ( Sigma ) on 5 dpf and monitored for survival for the following 2 days . In additional treatment groups , 100 ng/mL AimA , AimB , or mLCN ( purified as described below ) were added on 4 dpf . The aimA gene was PCR amplified from gDNA excluding the 5’ 66 nucleotide ( 22 amino acid ) secretion signal and cloned into pET21b using NdeI and XhoI restriction sites . The resultant gene product ( NCBI hypothetical protein WP_021230730 . 1 ) expressed well in E . coli BL21 DE3 as a C-terminal 6X His tagged protein of 300 amino acids long ( including His tag and linker ) . The E . coli culture was grown at 37° C until OD600 0 . 4–0 . 6 , then moved to 30° C and induced with 1 mM IPTG for 3–4 hr . All subsequent steps were performed at 4° C . 1–2 L of pelleted E . coli cells were lysed in lysis buffer ( 50 mM HEPES pH 7 . 9 , 300 mM NaCl , 10 mM imidazole and 5 mM BME ) , sonicated , and debris pelleted . The supernatant was washed over 5 mL Ni-NTA resin in a gravity column that was pre-washed with lysis buffer . The resin was washed with 15X bed volume of lysis buffer , then 10X bed volume of lysis buffer with 30 mM imidazole , 10X bed volume of lysis buffer with 50 mM imidazole and finally eluted with 3x - 5x bed volume of lysis buffer plus 100–300 mM imidazole . The high absorbance ( 280 nm wavelength ) fractions were pooled and dialyzed overnight into 150 mM NaCl , 50 mM HEPES pH 7 . 9 and 5 mM BME , concentrated , flash frozen in liquid nitrogen , and stored at −80° C . The aimB gene , excluding the 5’ 57 nucleotide ( 19 amino acid ) secretion signal , was cloned into pET21b using NdeI and XhoI restriction sites ( GenScript ) . The resulting gene product ( NCBI hypothetical protein WP_021230165 . 1 ) expressed well in E . coli BL21 DE3 as a C-terminal 6X His tagged protein of 321 amino acids long ( including His tag and linker ) . AimB protein was expressed and purified similarly to AimA , above , with some changes . The lysate was run over HisTrap HP 5 mL column ( GE healthcare ) using an AKTA FPLC ( GE healthcare ) and eluted with a gradient from 10 to 500 mM imidazole . Fractions containing the purest AimB were pooled , dialyzed ( 150 mM NaCl , 50 mM HEPES pH 7 . 9 and 5 mM BME ) and concentrated . Due to the difficulty of separating away contaminating E . coli proteins , AimB was only crudely purified to >50% pure by SDS PAGE . The mouse lipocalin2 gene ( NCBI gene NM_008491 . 1 ) , excluding the 5’ 63 nucleotide ( 21 amino acids ) secretion signal , was cloned into pET21b ( GenScript ) , using Nde1 and Xho1 restriction sites . Mouse LCN2 with a C-term 6X His tag expressed well in E . coli BL21 DE3 using 0 . 25 mM IPTG at 18° C overnight . mLCN2 was purified using Ni-NTA agarose resin ( Qiagen ) in a gravity column with increasing concentrations of imidazole from 10 to 30 mM , and eluted with 50 mM imidazole . The resultant 22 kDa protein ( 189 amino acids including His tag and linker ) was dialyzed into 300 mM NaCl , 50 mM potassium phosphate pH 7 and 5 mM BME . The dialyzed protein was concentrated and confirmed to be >95% pure by SDS PAGE . Purified AimA with C-terminal 6x His tag was concentrated to 10 . 9 mg/mL and set up in hanging drops as 1 μL protein: 1 uL well solution at room temperature . AimA crystallized in thick hexagons in 3 . 5 M sodium formate , 75 mM NaCl , 25 mM HEPES pH 7 . 9 and 2 . 5 mM BME . One to two weeks after the crystals grew , they were transferred to wells containing 3 . 8 M sodium formate to toughen them up for approximately one week . The heavy atom derivative crystals were then transferred to drops with 3 . 8 M sodium formate , 0 . 5 M NaI ( for iodide data set ) and 15% glycerol ( as a cryoprotectant ) for several hours before being scooped and flash frozen in liquid nitrogen for data collection at the Advanced Light Source in Berkeley , CA , beamline 5 . 0 . 2 using the Pilatus detector at a wavelength of 1 . 0 Å . The native crystals were transferred to 3 . 8 M sodium formate and 15% PEG 200 briefly , then flash frozen in liquid nitrogen for data collection as described for the iodide soaked crystals . Manual model building was performed using Coot 0 . 8 . 1 . 6 ( Emsley and Cowtan , 2004 ) and refinement was carried out using PHENIX 1 . 12–2829 ( Adams et al . , 2010 ) . Initial rigid body refinement resulted in R/Rfree values of 26 . 5/28 . 3% . Using the extended resolution improved the electron density maps and allowed placement of additional water molecules , two formate molecules ( present at 3 . 5 M in the crystallization buffer ) , N-terminal residues 1–8 , and an alternate chain path for residues 153–165 , improving R/Rfree to 20 . 9/24 . 8% . Residues 180–181 are at the tip of a disordered loop and were not modeled , and residues 293–294 and the C-terminal His-tag beyond it are not visible in the electron density . Electron density is weak in several regions including the N-terminus and several loops , but the chain path was clear enough to build at least the backbone atoms for these residues . In late stages of refinement , TLS was implemented using one group per chain , dropping R/Rfree to 17 . 9/20 . 9% . B-factor weights were optimized in the final refinement step , yielding final R/Rfree of 17 . 2/20 . 4% for the final AimA model ( Table 2 ) . E . coli and Aeromonas cultures were grown overnight ( 37° C and 30° C , respectively ) from a glycerol freezer stock in Luria Bertani ( LB ) broth to stationary phase . The cultures were back-diluted 1:100 the next morning in LB and loaded into a transparent 96-well plate . We used a total volume of 200 μL per well . 2 , 2’-dipyridyl ( Sigma-Aldrich ) titrations in DMSO were administered in 10 μL treatments for the Aeromonas growth curves . For the E . coli growth curves , AimA and mLCN titrations were administered in 20 μL treatments , with dipyridyl added to 300 μM in the LB 1:100 back-dilution . Controls for the Aeromonas and E . coli growth curves were treatments of DMSO or protein buffer ( 300 mM NaCl , 50 mM potassium phosphate pH 7 , 5 mM BME ) , respectively . Growth ( OD 600 nm ) at 30° C was monitored using a FLUOStar Omega ( BMG LABTECH ) plate reader . Readings were taken every 1 hr for up to 36 hr . Replicate curves ( three or four ) were plotted with standard deviation using Prism ( GraphPad ) . The appropriate sample size for experiments quantifying intestinal neutrophils and bacterial colonization level was estimated a priori using a power of 84% and a significance level of 0 . 05 . From previously published data on intestinal neutrophil quantification ( Rolig et al . , 2015 ) and zebrafish gut bacterial mono-association colonization ( Rolig et al . , 2015; Hill et al . , 2016 ) we estimated a effect size of 0 . 24 for neutrophil influx and 0 . 35 for bacterial colonization . These parameters suggested using an n of 40 and 23 in order to detect significant changes between treatment groups for either neutrophil influx or bacterial colonization , respectively . Each experiment described herein contains about 10–15 biological replicates ( individual fish per treatment group ) . These experiments of 10–15 fish were repeated multiple times ( technical replicates ) , resulting in pooled data sets ranging from 20 to 70 biological replicates . Grubbs statistical outlier test was applied to the data and data points that met the criteria for a statistical outlier were removed . In the figures , these data are presented as box and whisker plots , which display the data median ( line within the box ) , first and third quartiles ( top and bottom of the box ) , and minimum and maximum values ( whiskers ) . All data points that generate the box and whisker plots are presented as individual points within the plot . These pooled data were analyzed through the statistical software Prism Graphpad software . For experiments measuring a single variable with multiple treatment groups , a single factor ANOVA with post hoc means testing ( Tukey ) was utilized . A p-value of less than 0 . 05 was required to reject the null hypothesis that no difference existed between groups of data . For survival curves a Mantel-Cox test was used to determine significance .	Animals , including humans , harbor vast numbers of bacteria inside our digestive tracts . But rather than wage constant war , we have learned to coexist peacefully , and many of these bacteria are important to keep us healthy . Our immune system controls the number of bacteria , but in some diseases , this balance fails , and immune cells called neutrophils start a defense response . However , such attacks also cause inflammation in our guts , which can damage the tissues and organs . Understanding the delicate balance between immune cells and individual bacteria in humans remains a challenge . Our gut bacteria live in complex communities with many species of microorganisms . Using animals like zebrafish can help to find out if gut bacteria are able to prevent such inflammation . These fish can grow under sterile conditions in the laboratory , allowing the study of added individual bacteria and the molecules they release . They are also transparent , making it easy to capture images of their organs under the microscope . In the wild , they live side-by-side with a type of bacteria called Aeromonas . Here , Rolig et al . added Aeromonas bacteria to sterile zebrafish and observed how they interacted with the fish immune system . This revealed that the bacteria produce a protein to pacify immune cells , named 'Aeromonas immune modulator' ( AimA ) . The sequence of this new protein did not look like any known molecules , but its 3D structure resembled a protein found in animals called lipocalin . AimA dampened inflammation in the gut . When Aeromonas without AimA colonized the zebrafish , their neutrophils went to war . Their guts became inflamed and the bacteria started to die . But when the zebrafish had no neutrophils in their gut , nothing could stop the AimA-deficient bacteria and they grew too much . They produced toxic products , triggering septic shock and killing the fish . Adding AimA back into the fish stopped the inflammation and prevented septic shock , restoring balance . Both partners need AimA to survive . The bacteria need it to shield them from the immune response and the fish need it for protection against toxic products made by the bacteria . Understanding proteins like AimA could help us to control bacteria and gut inflammation . Further work could reveal ways to use similar molecules to treat inflammation or to boost the growth of friendly bacteria .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease", "immunology", "and", "inflammation" ]	2018	A bacterial immunomodulatory protein with lipocalin-like domains facilitates host–bacteria mutualism in larval zebrafish
Caenorhabditis elegans is a powerful model for studying gene regulation , as it has a compact genome and a wealth of genomic tools . However , identification of regulatory elements has been limited , as DNA-binding motifs are known for only 71 of the estimated 763 sequence-specific transcription factors ( TFs ) . To address this problem , we performed protein binding microarray experiments on representatives of canonical TF families in C . elegans , obtaining motifs for 129 TFs . Additionally , we predict motifs for many TFs that have DNA-binding domains similar to those already characterized , increasing coverage of binding specificities to 292 C . elegans TFs ( ∼40% ) . These data highlight the diversification of binding motifs for the nuclear hormone receptor and C2H2 zinc finger families and reveal unexpected diversity of motifs for T-box and DM families . Motif enrichment in promoters of functionally related genes is consistent with known biology and also identifies putative regulatory roles for unstudied TFs . Transcription factors ( TFs ) are sequence-specific DNA-binding proteins that control gene expression , often regulating specific biological processes such as pluripotency and differentiation ( Takahashi and Yamanaka , 2006 ) , tissue patterning ( Lemons and McGinnis , 2006 ) , the cell cycle ( Evan et al . , 1994 ) , metabolic pathways ( Blanchet et al . , 2011 ) , and responses to environmental stimuli ( Benizri et al . , 2008 ) . The nematode Caenorhabditis elegans is a powerful model for studying gene regulation as it is a complex and motile animal , yet has a compact genome ( ∼100 Mbp ) ( C . elegans Sequencing Consortium , 1998 ) featuring relatively short intergenic regions ( mean 1389 bp; median 662 bp ) . Indeed , the observation that proximal promoter sequence is often sufficient to produce complex tissue-specific gene expression patterns ( Dupuy et al . , 2004; Zhao et al . , 2007; Grove et al . , 2009; Sleumer et al . , 2009; Niu et al . , 2011 ) indicates that long-range gene regulation through enhancers is not as abundant in C . elegans as it is in flies or mammals ( Gaudet and McGhee , 2010; Reinke et al . , 2013 ) . C . elegans has 934 annotated TFs ( Reece-Hoyes et al . , 2005 ) , and 744 proteins that possess a well-characterized sequence-specific DNA-binding domain ( DBD ) ( Weirauch and Hughes , 2011; Weirauch et al . , 2014 ) . C . elegans contains major expansions of several specific TF families , with nuclear hormone receptor ( NHR ) , Cys2His2 ( C2H2 ) zinc finger , homeodomain , bHLH , bZIP , and T-box together comprising 74% of the TF repertoire ( Reece-Hoyes et al . , 2005; Haerty et al . , 2008 ) . The lineage-specific expansion of C2H2 zinc finger ( ZF ) TFs is similar to that observed in many animals , including diversification of DNA-contacting ‘specificity residues’ , suggesting diversification in DNA-binding specificity ( Stubbs et al . , 2011 ) . The C . elegans genome encodes an unusually large number of NHRs ( 274 members ) , more than five times the number in human ( 48 members ) ( Enmark and Gustafsson , 2001; Reece-Hoyes et al . , 2005 ) . It is speculated that the NHRs may serve as environmental sensors ( Enmark and Gustafsson , 2001; Arda et al . , 2010 ) , providing a possible explanation for their variety and numbers . Five of the six major NHR sub-families found across metazoa are also found in C . elegans ( NR3 is lacking ) , but the vast majority of C . elegans NHRs define novel sub-families that are not present in other metazoans ( Van Gilst et al . , 2002 ) and which are derived from an ancestral gene most closely resembling HNF4 ( aka NR2A ) ( Robinson-Rechavi et al . , 2005 ) . Extensive variation in the DNA-contacting recognition helix ( RH ) or ‘P-box’ suggests that C . elegans NHRs , like C2H2 and bHLH families , have diversified DNA-sequence specificities , and that many will recognize novel motifs ( Van Gilst et al . , 2002 ) . The T-box gene family presents another example of a nematode-specific expansion , with 22 members in C . elegans , of which 18 lack one-to-one orthologs in other metazoan lineages ( Minguillon and Logan , 2003 ) . Only four have known binding motifs , and unlike most other TFs , T-box binding motifs are virtually identical across the metazoa ( Sebé-Pedrós et al . , 2013; Weirauch et al . , 2014 ) ; the diversification of TFs is often associated not only with changes in DNA-sequence specificity but also alteration in protein–protein interactions and expression of the TF gene itself ( Grove et al . , 2009; Reece-Hoyes et al . , 2013 ) . Despite extensive study of gene regulation , including several large-scale efforts ( Deplancke et al . , 2006; Grove et al . , 2009; Lesch et al . , 2009; Gerstein et al . , 2010; Niu et al . , 2011; Sarov et al . , 2012; Reece-Hoyes et al . , 2013; Araya et al . , 2014 ) , the landscape of C . elegans TF-sequence specificities remains largely unknown . To our knowledge , motifs are currently known for only 71 C . elegans TFs , including those determined in single-gene studies , previous protein binding microarray ( PBM ) analyses , and modENCODE TF ChIP-seq data ( Matys et al . , 2006; Araya et al . , 2014; Mathelier et al . , 2014; Weirauch et al . , 2014 ) . It has been surprisingly difficult to obtain motifs from ChIP-seq data ( Niu et al . , 2011; Araya et al . , 2014 ) , possibly due to indirect binding , or a dominant role of chromatin structure in either determining in vivo binding sites ( Song et al . , 2011 ) or in the purification of chromatin fragments ( Teytelman et al . , 2013 ) . Yeast one-hybrid ( Y1H ) assays ( Reece-Hoyes et al . , 2011 ) cannot be used easily to derive TF motifs , because the DNA sequences tested are too large ( ∼2 kb on average ) . However , there is a strong statistical correspondence between motifs determined by PBMs and Y1H data ( Reece-Hoyes et al . , 2013 ) . Computational approaches coupling promoter sequence conservation and/or gene expression data to identify TF motifs de novo have collectively produced many more motifs than there are TFs ( Beer and Tavazoie , 2004; Sleumer et al . , 2009; Zhao et al . , 2012 ) and also do not inherently reveal the cognate TFs that correspond to each putative motif . Multimeric binding represents one possible complication in the analysis of in vivo TF-binding data ( Ao et al . , 2004 ) . Indeed , TF co-associations were identified based on ChIP-seq peak-binding overlaps in C . elegans modENCODE studies ( Araya et al . , 2014 ) , but the underlying sequence recognition mechanisms were not apparent . Here , we use PBMs to systematically identify C . elegans TF DNA-binding motifs . We selected a diverse set of TFs to assay , ultimately obtaining 129 motifs from different TF families and subclasses . The data show that the expansion of most major TF families is associated with diversification of DNA-binding motifs . Motif enrichment in promoters reveals that our motif collection readily associates individual TFs with putative regulated processes and pathways . The key goal of this project was to expand our knowledge of DNA-sequence specificities of C . elegans TFs . To do this , we analyzed a diverse set of TF DBDs ( see below ) with PBM assays ( Berger et al . , 2006; Weirauch et al . , 2014 ) . Briefly , the PBM method works by ‘hybridizing’ a glutathione S-transferase ( GST ) -tagged DNA-binding protein ( in our assays , the DBD of a TF plus 50 flanking amino acids ) to an array of ∼41 , 000 defined 35-mer double-stranded DNA probes . The probes are designed such that all 10-mer sequences are present once , and all non-palindromic 8-mers are , thus , present 32 times in difference sequence contexts ( palindromic 8-mers occur 16 times ) . A fluorescently labelled anti-GST antibody illuminates the extent to which each probe is bound by the assayed TF . Using the signal intensity for each probe , the specificity of the TF is derived . For each individual 8-mer , we derive both E-scores ( which represent the relative rank of microarray spot intensities , and range from −0 . 5 to +0 . 5 [Berger et al . , 2006] ) and Z-scores ( which scale approximately with binding affinity [Badis et al . , 2009] ) . PBMs also allow derivation of position weight matrices ( PWMs ) up to 14 bases long ( Berger et al . , 2006; Mintseris and Eisen , 2006; Badis et al . , 2009; Weirauch et al . , 2013 ) ( hereafter , we take ‘motif’ to mean PWM ) . To determine PWMs , we used the data from PBM assays performed on two different array designs to score the performance of PWMs obtained from different algorithms , as previously described ( Weirauch et al . , 2013 , 2014 ) . In this study , we selected TFs to analyze on the basis of their DBD sequence , aiming to examine at least one TF from each group of paralogous TFs , and biasing against TFs that have known PBM motifs , or close orthologs or paralogs with known motifs ( see ‘Materials and methods’ for full description of selection scheme ) . The selections were guided by previous PBM analyses that determined sequence identity thresholds for each DBD class that correspond to motif identity ( Weirauch et al . , 2014 ) . To identify TFs , we used the Cis-BP definition of DBDs ( Weirauch et al . , 2014 ) , which employs a list of well-characterized eukaryotic DBDs and a distinct significance threshold for each DBD class . CisBP identified 744 C . elegans proteins , encompassing 52 domain types ( listed in Figure 1—source data 1 ) . 689 ( 93% ) of these 744 are present in the wTF catalog of 934 annotated TFs ( Reece-Hoyes et al . , 2005 ) ; thus , these sets are largely overlapping . We manually examined the differences between the two TF lists ( see Supplementary file 1 ) and found that most of them can be accounted for by ( i ) changes to the C . elegans protein catalog over time , ( ii ) differences in domain classes included , ( iii ) differences in domain score threshold , ( iv ) fewer manual annotations in Cis-BP , and ( v ) ambiguity in classifying C2H2 zinc fingers as TFs . Overall , wTF2 . 0 contains only 19 proteins that are not in Cis-BP and that are very likely bona fide sequence-specific TFs . wTF2 . 0 also contains 83 C2H2 proteins that fall below the CisBP score threshold , 52 of which have only a single C2H2 domain . DNA recognition typically requires multiple C2H2 domains; however , some fungal TFs do bind DNA with a single C2H2 , employing additional structural elements ( Wolfe et al . , 2000 ) . Thus , these proteins have an ambiguous status . In general , Cis-BP excludes proteins with lower domain scores and those with little or no evidence for sequence-specific DNA binding , and we , therefore , refer to the 744 in CisBP plus the 19 additional bona fide TFs as the 763 ‘high confidence’ C . elegans TFs . We attempted to clone DBDs from 552 unique high confidence TFs , ultimately obtaining clones for 449 , all of which we assayed by PBMs . After employing stringent success criteria ( see ‘Materials and methods’ ) we obtained sequence-specificity data ( 8-mer scores and motifs ) for 129 DBDs . PBM ‘failures’ may be due to any of several causes , including protein misfolding , requirement for cofactors or protein modifications ( e . g . , phosphorylation ) , or bona fide lack of sequence-specific DNA-binding activity . The overall success rate ( 29% ) is comparable to that we have observed from the analysis of thousands of DBDs from diverse species ( 35% ) ( Weirauch et al . , 2014 ) . A summary of our results is presented in Figure 1 , broken down by motif numbers and percent coverage for individual DBD classes . Our motif collection encompasses 26 different DBD classes , and greatly increases the number and proportion of C . elegans TFs for which motifs have been identified experimentally , from 71 ( 10% ) to 195 ( 26% ) ( five of the 129 had previously-known motifs ) . The new data encompass all of the large TF families , including C2H2 zinc fingers , NHRs , bZIPs , homeodomains , DM domains , and GATA proteins . 10 . 7554/eLife . 06967 . 003Figure 1 . Motif status by DBD class . Stacked bar plot depicting the number of unique C . elegans Transcription factors ( TFs ) for which a motif has been derived using PBM ( this study ) , previous literature ( including PBMs ) , or by homology-based prediction rules ( see main text ) . The y-axis is displayed on a log2 scale for values greater than zero . See Figure 1—source data 1 for DNA-binding domain ( DBD ) abbreviations . Correspondence between motifs identified in current study and previously reported motifs is shown in Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 00310 . 7554/eLife . 06967 . 004Figure 1—source data 1 . Table of C . elegans TF repertoire motif coverage and list of TF DBDs present in C . elegans . The number of unique C . elegans TFs by DNA-binding domain family for which a motif has been derived using PBM ( this study ) , previous literature ( including PBMs ) , or by homology-based prediction rules and the list of C . elegans TFs by DNA-binding domain family type . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 00410 . 7554/eLife . 06967 . 005Figure 1—figure supplement 1 . Correspondence between TF motifs identified from our PBM study and previously reported motifs from several types of experimental data . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 005 We next asked whether our new data are consistent with previous knowledge . Of the 129 TFs , only five have previously known motifs , all of which we recapitulated ( Figure 1—figure supplement 1 ) . The sequence preferences for most of the 129 TFs were different from those of any previously assayed TF , however . The boxplots in Figure 2A–C and Figure 2—figure supplements 1–4 show that , on average , the new TFs we analyzed bound a set of 8-mers that was largely non-overlapping with that of the most similar protein that had been analyzed previously by PBM ( red circles indicate the 8-mer overlap between individual TFs analyzed by PBM in our study , and the most similar TF analyzed by PBM in any study ) . Nonetheless , some pairs of TFs have DBDs that are highly similar and bind highly overlapping 8-mers . These observations are quantitatively consistent with the prior study we used for guidance in selecting TFs ( black box plots ) ( Weirauch et al . , 2014 ) , and thus , we expect that the scheme for predicting sequence specificity via amino acid identity that was proposed in the prior study can also be used in C . elegans . In this scheme , TFs without DNA-binding data are simply assigned the motifs and 8-mer data for other TFs with DBD amino acid similarity above a threshold , if those data exist . These TFs can be from C . elegans or from other species . If we include these predicted motifs , then the number of C . elegans TFs with an associated motif increases to 292 ( 39% ) , including TFs with motifs predicted from other C . elegans TFs ( 24 ) and those with motifs predicted from other species ( 79 ) . 10 . 7554/eLife . 06967 . 006Figure 2 . Motif prediction , motif clustering , and identification of representative motifs . ( A–C ) Boxplots depict the relationship between the %ID of aligned AAs and % of shared 8-mer DNA sequences with E-scores exceeding 0 . 45 , for the three DBD classes , as indicated . %ID bins range from 0 to 100 , of size 10 , in increments of five . Red dots indicate individual TFs in this study , vs the next closest TF with PBM data . Vertical lines indicate AA %ID threshold above which motifs can be predicted using homology , taken from ( Weirauch et al . , 2014 ) . Boxplots for all other DBDs in current study are shown in Figure 2—figure supplements 1–4 . ( D ) Clustering analysis of motifs of bZIP domains using position-weight matrices ( PWM ) clus ( Jiang and Singh , 2014 ) . Colored gridlines indicate clusters . Cluster centroids are shown along the diagonal; expert curated motifs are shown within the box at right . ‘E’ indicates experimentally determined motifs; ‘P’ indicates predicted motifs . Source of motif is also indicated . Results of motif curation for GATA family TFs is displayed in Figure 2—figure supplement 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 00610 . 7554/eLife . 06967 . 007Figure 2—figure supplement 1 . C . elegans TFs adhere to established thresholds for motif inference . ( A–F ) Boxplots depict the relationship between the %ID of aligned AAs and % of shared 8-mer DNA sequences with E-scores exceeding 0 . 45 , for the DBD classes of TFs with PBMs from this study . %ID bins range from 0 to 100 , of size 10 , in increments of five . Red dots indicate individual proteins in this study , vs the next closest protein with PBM data . Vertical blue lines indicate AA %ID threshold above which motifs can be predicted using homology . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 00710 . 7554/eLife . 06967 . 008Figure 2—figure supplement 2 . C . elegans TFs adhere to established thresholds for motif inference ( continued ) . ( A–F ) Boxplots depict the relationship between the %ID of aligned AAs and % of shared 8-mer DNA sequences with E-scores exceeding 0 . 45 , for the DBD classes of TFs with PBMs from this study . %ID bins range from 0 to 100 , of size 10 , in increments of five . Red dots indicate individual proteins in this study , vs the next closest protein with PBM data . Vertical blue lines indicate AA %ID threshold above which motifs can be predicted using homology . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 00810 . 7554/eLife . 06967 . 009Figure 2—figure supplement 3 . C . elegans TFs adhere to established thresholds for motif inference ( continued ) . ( A–F ) Boxplots depict the relationship between the %ID of aligned AAs and % of shared 8-mer DNA sequences with E-scores exceeding 0 . 45 , for the DBD classes of TFs with PBMs from this study . %ID bins range from 0 to 100 , of size 10 , in increments of five . Red dots indicate individual proteins in this study , vs the next closest protein with PBM data . Vertical blue lines indicate AA %ID threshold above which motifs can be predicted using homology . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 00910 . 7554/eLife . 06967 . 010Figure 2—figure supplement 4 . C . elegans TFs adhere to established thresholds for motif inference ( continued ) . ( A–E ) Boxplots depict the relationship between the %ID of aligned AAs and % of shared 8-mer DNA sequences with E-scores exceeding 0 . 45 , for the DBD classes of TFs with PBMs from this study . %ID bins range from 0 to 100 , of size 10 , in increments of five . Red dots indicate individual proteins in this study , vs the next closest protein with PBM data . Vertical blue lines indicate AA %ID threshold above which motifs can be predicted using homology . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 01010 . 7554/eLife . 06967 . 011Figure 2—figure supplement 5 . GATA TF motif clustering and identification of representative motifs . Clustering analysis of C . elegans GATA TF's motifs using PWMclus ( Jiang and Singh 2014 ) . Colored gridlines indicate clusters . Cluster centroids are shown along the diagonal , while manually curated motifs are shown within the box at right . Bolded row names represent motifs obtained from this study . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 011 The entire C . elegans motif collection , including our new data , previously published motifs , and those predicted by homology from other TFs in C . elegans and other species , encompasses 1769 unique motifs representing only 292 TFs . About half ( 157 , or 54% ) of the 292 TFs with motifs are represented by only a single motif , as there were no data prior to our study for these TFs or their close homologs . Some TFs ( e . g . , homeodomains , PAX , and forkheads ) , however , are highly conserved and thus have many orthologs above the prediction threshold . In addition , TFs that are known developmental regulators tend to be well studied , and often possess multiple associated motifs . To gain an overview of the full motif collection and to compare among the multiple motifs for each protein , we used the PWMclus tool ( Jiang and Singh , 2014 ) , with default settings , to obtain groups of highly related motifs from all TFs within each DBD class . This tool uses an information-content weighted Pearson correlation between aligned PWM columns as a similarity measure for hierarchical clustering , and then selects branches within which the average internal correlation exceeds R ≥ 0 . 8 . This procedure collapsed the 1769 motifs into a set of 424 clusters . This number is still larger than the number of TFs with either known or predicted motifs ( 292 ) , since there are many cases in which motifs for a single TF are distributed across multiple clusters , although in 67% of cases in which there are multiple known and predicted motifs for a given protein , the majority of them do form a single cluster . There appear to be several explanations for this phenomenon , as exemplified by the bZIP family shown in Figure 2D . First , different studies and different experimental ( or computational ) techniques often yield motifs for the same protein that are clearly related by visual examination , but score as different from each other using PWMclus . For example , there are four different motifs for SKN-1 ( from PBM , Chip-seq , and Transfac ) that all contain the same half-site , ATGA , but have different flanking sequence preferences . Similarly , for Forkhead TFs FKH-1 and UNC-130 , different methods produce variants with differences in the sequences flanking the core TGTTT Forkhead binding site . A related explanation is that a single motif may not adequately capture all aspects of TF sequence preferences , such as the ability of many TFs to bind as both a monomer and a homodimer ( or multimer ) with preferred spacing and orientation , variability in the preferred spacing , changes to the preferred monomeric sites that are associated with dimerization , and effects of base stacking that result in preferred polynucleotides at some positions ( Jolma et al . , 2013 ) . In addition , different experimental methods may capture some aspects of DNA-binding complexity better than others . It is inconvenient to have a large number of motifs for a single protein for several reasons . First , it is difficult to peruse the full motif collection . In addition , comprehensive motif scanning is slower with a large number of motifs , and the motif scans produce partially redundant results that require deconvolution and reduce statistical power . We , therefore , sought to identify a single motif or set of motifs for each protein that are minimally redundant and are best supported by existing data . We used a semi-automated scheme that considers all data available ( similar to that described in ( de Boer and Hughes , 2012 ) ; see ‘Materials and methods’ ) . Briefly , we prioritized motifs that are ( a ) measured experimentally , rather than predicted; ( b ) more similar to other motifs for the same TF , or highly similar TFs , especially if they are derived from in vitro data , which would be free of confounding effects present in vivo; ( c ) assigned to the cluster that contains the majority of motifs for that TF; ( d ) most consistent with the type of sequences that a given DBD class typically binds; ( e ) best supported by ChIP-seq or Y1H data , if available ( see below ) . This procedure resulted in a set of 284 motifs representing the 292 C . elegans TFs with experimentally determined or predicted motifs ( Supplementary file 2 ) . The outcome for the bZIP family is shown on the right of Figure 2D , which illustrates that the motif curation procedure produces motifs that are consistent with known bZIP class binding sites . The curated set also contains 16 cases in which the same protein is represented by multiple motifs ( exemplified by the GATA family TF ELT-1 , which binds as both a monomer and a homodimer , Figure 2—figure supplement 5 ) , and 11 cases in which more than one protein is represented by the same motif ( e . g . , GATA family TFs MED-1 and MED-2 , Figure 2—figure supplement 5; in all of these cases , the TFs are highly similar proteins ) . We also note that PWMclus subdivides the 284 curated motifs into only 127 different clusters ( data not shown ) , because the motif ( s ) contained in many of the 424 original clusters met few or none of the selection criteria above . The majority of the expert curated motifs ( 237 , or 84% ) are derived from the PBM data described in this study or from previous studies ( compiled in [Weirauch et al . , 2014] ) , which are the only data available for the majority of the 292 TFs with motifs . We reasoned that the PBM data should facilitate direct comparison among TF sequence preferences , as they were generated using identical methodology . In addition , PBMs facilitate comparisons because they produce scores for individual DNA 8-mers . Thus , to complement the PWM analysis above , we examined as a composite 8-mer E-score data for all of the TFs analyzed in this study by using PBMs . Figure 3 illustrates that the 8-mers recognized by each individual protein are in general distinct and further highlights the distinctiveness of the sequences preferred by different TFs that share the same type of DBD . For example , C . elegans homeodomain and Sox TFs display different sequence preferences that largely reflect the known subclasses ( Figure 3 and data not shown; all data and motifs are available in the Cis-BP database ( see ‘Data Access’ section below ) ) . We also observed subtle differences in Forkhead DNA-sequence preferences: despite the motifs having similar appearance , the proteins prefer slightly different sets of 8-mers , as previously observed using PBM data ( Badis et al . , 2009; Nakagawa et al . , 2013 ) . Other large C . elegans TF families display undocumented and unexpected diversity in their DNA-sequence preferences , which we next examined in greater detail . 10 . 7554/eLife . 06967 . 012Figure 3 . Overview of 8-mer sequences preferences for the 129 C . elegans TFs analyzed by PBM in this study . 2-D Hierarchical agglomerative clustering analysis of E-scores performed on all 5728 8-mers bound by at least one TF ( average E > 0 . 45 between ME and HK replicate PBMs ) . Colored boxes represent DBD classes for each TF . Average E-score data is available in Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 01210 . 7554/eLife . 06967 . 013Figure 3—source data 1 . Table showing 8-mers bound by at least one TF with an average E-score ≥0 . 45 for all the 129 C . elegans TFs analyzed by PBMs in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 013 Previously , the literature contained motifs for only eight of the 271 C . elegans NHRs , while motifs for an additional 13 could be predicted from orthologs and paralogs ( Hochbaum et al . , 2011; Weirauch et al . , 2014 ) . It has also been reported that additional C . elegans NHRs bind sequences similar to those bound by their counterparts in other vertebrates ( Van Gilst et al . , 2002 ) , but the data available do not lend itself to motif models that can be used for scanning . We obtained new PBM data for 20 C . elegans NHRs ( Figure 4 ) , among which only one had a previously known motif ( DAF-12 , which yielded a motif identical to one found by ChIP–chip [Hochbaum et al . , 2011] ) . None of the remaining 19 could have been predicted by simple homology; due to their widespread divergence , and absence of motifs for most NHRs , few motifs can be predicted by homology among the C . elegans NHR class at our threshold for motif prediction ( 70% identity for NHRs ) . However , these 19 new NHR motifs do lead to predicted motifs for eight additional C . elegans NHRs . 10 . 7554/eLife . 06967 . 014Figure 4 . 8-mer binding profiles of NHR family reveal distinct sequence preferences . Left , ClustalW phylogram of nuclear hormone receptor ( NHR ) DBD amino acid sequences with corresponding motifs . TF labels are shaded according to motif similarity groups identified by PWMclus . Center , heatmap showing E-scores . NHRs are ordered according to the phylogram at left . The 1406 8-mers with E-score > 0 . 45 for at least one family member on at least one PBM array were ordered using hierarchical agglomerative clustering . Each TF has one row for each of two-replicate PBM experiments ( ME or HK array designs ) . Right; recognition helix ( RH ) sequences for the corresponding proteins , with identical RH sequence types highlighted by colored asterisks . Variant RH residues are underlined at bottom . Right , matrix indicates cluster membership according to PWMclus . Top and bottom , pullouts show re-clustered data including only the union of the top ten most highly scoring 8-mers ( taking the average E-score from the ME and HK arrays ) for each of the selected proteins . E-scores for k-mers in all three heatmaps are available in Figure 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 01410 . 7554/eLife . 06967 . 015Figure 4—source data 1 . Table showing 8-mer E-score profiles of NHRs analyzed by PBMs . 8-mers bound by at least one NHR with an E-score ≥0 . 45 for all the C . elegans NHRs that have been analyzed by PBMs ( center panel ) and a table of pullouts ( top and bottom panel ) showing average ( ME and HK ) E-scores of the union of the top ten highly scoring 8-mers bound by at least one NHR within the selected motif cluster . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 015 The most striking feature of the NHR motifs is their diversity , but an equally surprising observation is that very different NHRs can bind very similar sets of sequences . Data from the 27 NHRs that have been analyzed by PBMs in our study or others are shown in Figure 4 . We obtained 13 different groups of motifs , using the PWMclus methodology described above ( indicated by shading of dendrogram labels in Figure 4 ) . We expected that all 27 of these NHRs might have yielded a distinct motif , as no two are more than 70% identical to each other . In several cases , however , NHRs with very different overall DBD sequences ( below the threshold for predicting motif identity ) in fact display similar sequence preferences , while more similar NHR TFs often bind different motifs , as the shading on the labels in Figure 4 does not strictly reflect the dendrogram . We also note that the data for individual 8-mers appear more complex than the motif groups capture ( see heatmaps in Figure 4 ) . For example , the individual 8-mer scores for TFs represented by the two largest groups of motifs—sets binding sequences related to G ( A/T ) CACA and ( A/T ) GATCA , respectively—indicate that they may in fact possess distinct DNA-sequence preferences ( Figure 4 , top and bottom ) . These subtle and complex differences are presumably obscured by the motif derivation process , which tends to produce degenerate ( i . e . , low information content ) motifs for most of these TFs . In addition , or possibly as a consequence , the default correlation threshold used by the PWMclus algorithm groups these TFs together . To examine the determinants of NHR sequence preferences more closely , we considered NHR RH sequences ( Figure 4 , middle ) . Of the 95 unique RH sequences found in C . elegans NHRs , 15 are found in our data , including multiple representatives of most of the populous RHs ( our data contain ten of the 75 with RA-AA; 3 of the 19 with NG-KT; 2 of the 10 with NG-KG; and one of the seven with AA–AA ) . It is believed that identity in the RH corresponds to identity in sequence preference ( Van Gilst et al . , 2002 ) ; surprisingly , however , we found that TFs with identical RH sequences can bind very different DNA sequences . For example , NHR-177 shares the RA-AA RH with nine other NHRs examined in our study , yet binds a completely different set of sequences ( resembling CGAGA , unlike the CACA-containing motifs of the others ) . Conversely , NHRs with different RH sequences can have very similar DNA-sequence preferences . NHR-66 and NHR-70 , for example , differ at two of the four variable residues in the RH ( AA-SA vs RA-AA ) , and share only ∼49% amino acid identity ( and NHR-66 contains a three-residue insertion ) . Yet , they bind highly overlapping sets of 8-mers and produce motifs featuring CTACA . Thus , there is an imperfect correspondence between identity in the RH and identity in DNA-binding sequence preferences , suggesting that additional residues within ( or flanking ) the DBD contribute to the specificity of C . elegans NHR proteins . These observations also show that , when NHRs with very different overall DBD sequences bind similar motifs , it is typically not due to the two proteins sharing the same RH . Only one NHR , SEX-1 , produced a motif strongly resembling the canonical steroid hormone response element ( SHRE ) ( GGTCA ) ; SEX-1 shares three of four variable residues in the RH with canonical SHRE binding TFs such as the estrogen receptor ( SEX-1: EG-KG; ER: EG-KA ) . Moreover , none of the NHRs examined produced a motif matching that of HNF4 , the presumed ancestor of most C . elegans NHRs . We obtained new PBM data for 42 C2H2 ZF TFs ( Figure 5 ) , only one of which was previously known ( Figure 1—figure supplement 1 ) . Previously there were only six experimentally-determined C . elegans C2H2 motifs in the literature , and 11 that could be predicted by homology , all of which are well conserved in distant metazoans ( members of KLF , SP1 , EGR , SNAIL , OSR , SQZ , and FEZF families ) ; seven of these are among our data and have PBM motifs consistent with those predicted ( data not shown ) . Only two additional TFs ( ZTF-25 and ZTF-30 ) can be assigned motifs by homology using our new data . Together , the new data and predictions bring the total number of C . elegans C2H2 TFs with motifs to 53 ( ∼50% of the 107 C2H2s in our list of 763 TFs ) . 10 . 7554/eLife . 06967 . 016Figure 5 . C2H2 motifs relate to DBD similarity and to the recognition code . Left , ClustalW phylogram of C2H2 zinc finger ( ZF ) amino acid sequences with corresponding motifs . Right , examples in which motifs predicted by the ZF recognition code are compared to changes in DNA sequences preferred by paralogous C2H2 ZF TFs . Cartoon shows individual C2H2 ZFs and their specificity residues . Dashed lines correspond to 4-base subsites predicted from the recognition code . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 01610 . 7554/eLife . 06967 . 017Figure 5—figure supplement 1 . Comparison of C2H2 ZF recognition model with motifs derived PBM . Motif correlations between PBM derived motifs and ZF-model based predictions for TFs with both typical and atypical ( A ) linker lengths between ZF modules that are longer than 6 amino acids or shorter than 4 amino acids ( B ) zinc coordinating cysteine or histidine structural motifs and ( C ) differing length of the ZF array . Examples of recognition code predictions ( sequence logos ) for both typical and atypical TFs are compared with PBM motifs for each case . The p-values shown are estimated from Student's t-test . The number of TFs in each boxplot is shown above in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 017 The C2H2 motifs are diverse ( Figure 5 ) , but unlike the NHR family , the molecular determinants of C2H2 DNA sequence specificities are more readily understood . The motifs we obtained are broadly consistent with previously determined relationships between DNA contacting residues and preferred bases ( the so-called ‘recognition code’ ) ( Wolfe et al . , 2000 ) , although the motifs predicted by the recognition code are not sufficiently accurate to be used in motif scans ( median R2 = 0 . 21 vs predictions made by an updated recognition code that surpasses all previous recognition codes when compared against gold standards ( Najafabadi et al . , 2015 ) ) . While most of the motifs are similar to those predicted by the recognition code ( Figure 5—figure supplement 1 ) , lower similarity is observed for TFs with unusual inter-C2H2 linker lengths and atypical zinc-coordinating residues ( Figure 5—figure supplement 1 ) . In some cases , differences in the motifs obtained from related C2H2 TFs can be rationalized: Figure 5 ( right ) shows the example of paralogs EGRH-1 and EGRH-3 , in which the motifs obtained by PBM closely reflect those predicted by the recognition code , which differ at several positions . Figure 5 also shows the example of Snail homologs CES-1 and K02D7 . 2 , in which a short linker between fingers 2 and 3 may explain the truncated motif in K02D7 . 2 , and may also explain the differences previously observed between these two proteins in Y1H assays ( Reece-Hoyes et al . , 2009 ) . We obtained motifs for four nematode-specific T-box TFs ( i . e . , lacking one-to-one orthologs in other phyla ) : TBX-33 , TBX-38 , TBX-39 , and TBX-43 . In addition , TBX-40 was previously analyzed by PBM , and our motif for the related protein TBX-39 ( 93% identical ) is very similar . T-box TFs can bind to dimeric sites , with the characteristic spacing and orientation varying among different T-box proteins ( Jolma et al . , 2013 ) . The monomeric sequence preference ( resembling ‘GGTGTG’ ) is thought to be constant , however , as it is observed across different T-box classes and in distant phyla ( Sebé-Pedrós et al . , 2013; Weirauch et al . , 2014 ) . Strikingly , our new PBM data indicate that monomeric T-box sites can also vary considerably ( Figure 6A ) . While the motifs for TBX-38 and TBX-43 are highly similar to the canonical ‘GGTGTG’ motif , TBX-33 , TBX-39 , and TBX-40 exhibit novel recognition motifs . 10 . 7554/eLife . 06967 . 018Figure 6 . Nematode-specific sequence preferences in T-box and DM TFs . PBM data heatmaps of preferred 8-mers for T-box ( A ) , and DM ( B ) TFs . TFs are clustered using ClustalW; 8-mers were selected ( at least one instance of E > 0 . 45 ) and clustered using hierarchical agglomerative clustering , as in Figures 4 , 5 . Ten representative 8-mers ( those with highest E-scores ) are shown below for each of the clusters indicated in cyan . C . elegans TFs with data from this study are bolded . E-score data for T-box and DM TFs available in Figure 6—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 01810 . 7554/eLife . 06967 . 019Figure 6—source data 1 . Table showing 8-mer E-score profiles of T-box and DM TFs from C . elegans and other metazoans that have been analyzed by PBMs . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 01910 . 7554/eLife . 06967 . 020Figure 6—figure supplement 1 . T-box sequence alignments and the crystal structure of mTBX3 illustrate C . elegans specific variations . ( A ) Multiple-sequence alignment of T-box DBDs from C . elegans , the protist C . owczarzaki CoBra and mouse Eomes , and TBX3 . Key DNA-binding residues identified from crystal structure of mTBX3 are highlighted in red . Sequence insertions ( for TBX-33 ) , changes in the variable region ( for TBX-39/40 ) , and significant sequence changes in the key 310C RH ( for TBX- 39/40 ) are highlighted in blue frames . ( B ) Crystal structure model of mTBX3 is used as a prototype to illustrate C . elegans specific sequence variations . The primary recognition helix , 310C , is highlighted in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 020 The primary determinants of sequence specificity of T-box TFs are believed to reside in amino-acid residues located in α-helix 3 and the 310-helixC , which contact the major and minor groove , respectively ( Muller and Herrmann , 1997; Coll et al . , 2002; Stirnimann et al . , 2002 ) , and indeed , the DNA contacting residues in TBX-33 , TBX-39 , and TBX-40 are different from those in T-box TFs that bind the canonical motif ( Figure 6—figure supplement 1 ) . In addition , TBX-39 and TBX-40 exhibit sequence deletion in the ‘variable region’ , and TBX-33 has an 18 amino acid insertion in the region leading up to the β-strand e′ , which could also potentially alter sequence preferences via structural rearrangements . DM TFs are well studied because of their established roles in sex determination , and previous analyses established that different DM TFs often bind distinct motifs that typically contain a TGTAT core , including Drosophila doublesex , for which the family is named ( Gamble and Zarkower , 2012 ) . C . elegans and other nematodes encode several lineage-specific DM TFs in addition to orthologs shared across metazoans , with eight of the eleven C . elegans DM domains having less than 85% identity ( our threshold for DM motif prediction ) to any DM domain in insects and vertebrates ( Weirauch et al . , 2014 ) . Accordingly , most of the C . elegans DM domains have highest preference for sequences that are different from TGTAT , although in all but two cases the motifs do contain a TGT ( Figure 6B ) . DM domains encode intertwined CCHC and HCCC-zinc binding sites and are hypothesized to bind primarily in the minor groove ( Zhu et al . , 2000; Narendra et al . , 2002 ) . A DNA-protein structure has not yet been described for any DM protein , however; mapping the determinants of their variable DNA sequence preferences will therefore require further study . We next examined whether motifs from our collection correspond to modENCODE TF ChIP-seq data ( Araya et al . , 2014 ) , and to TF prey—promoter bait interactions from Y1H experiments ( [Reece-Hoyes et al . , 2013] and JF-B and AJMW , unpublished data ) . Among the 40 TFs analyzed by ChIP-seq and present in our motif collection , peaks for 20 TFs displayed central enrichment of motif scores ( q-value < 0 . 05 ) using the CentriMo algorithm on the top 250 peaks ( Bailey and Machanick , 2012 ) ( Figure 7 ) . Similarly , among 145 TFs both analyzed by Y1H and present in our motif collection , motif affinity scores for 103 were significantly enriched ( Mann–Whitney U test; q-value < 0 . 05 ) among promoter sequences scoring as positive by Y1H , relative to those scoring as negative by Y1H ( Figure 7—figure supplement 1 ) . The correspondence among these data sets is presumably imperfect due to indirect DNA binding in vivo , and/or the impact of chromatin and cofactors on binding site selection ( Liu et al . , 2006 ) , both of which occur in C . elegans and yeast . We note that , among the 25 TFs that are present in Y1H data , ChIP-seq data , and our motif collection , 11 are only significantly enriched in Y1H ( using the cutoff above ) , five are only significantly enriched in ChIP-seq , and only five are significantly enriched in both . Thus , most motifs ( 21/25; 84% ) can be supported by independent assays , although the in vivo assays appear to capture different aspects of TF binding . Overall , the clear relationship between our motifs and independent data sets strongly supports direct in vivo relevance of the motifs . 10 . 7554/eLife . 06967 . 021Figure 7 . The C . elegans curated motif collection explains ChIP-seq and Y1H TF binding data . Heatmap of CentriMo −log10 ( q-values ) for central enrichment of TF motifs in the top 250 peaks for each ChIP experiment . Motif enrichment in Y1H data is presented in Figure 7—figure supplement 1 . Both ChIP-seq and Y1H heatmaps use a common colour and annotation scale . Heatmaps are symmetric with duplicate rows to ensure the diagonal represents TF-motif enrichment in it's matching data set ( s ) . Red and blue coloring depicts statistically significant enrichments and depletions ( q ≤ 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 02110 . 7554/eLife . 06967 . 022Figure 7—figure supplement 1 . The C . elegans curated motif collection explains Y1H TF binding data . Heatmap depicting enrichment or depletion ( Mann–Whitney U test ) of TF motifs in the interactions of TF's with a collection of promoter bait sequences in Y1H experiments compared to all non-interacting bait sequences . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 022 We also examined whether we could detect multimeric or composite motifs ( CMs ) in existing ChIP-seq data sets by searching for the enrichment of patterns in which there is fixed spacing and orientation between two or more motifs within the peaks , one of which corresponds to the TF that was ChIPed . We identified 185 significantly enriched CMs ( see ‘Materials and methods’ ) involving 11/40 ChIPed TFs , and 14 different TF families ( including partner motifs ) ( Figure 8 , Figure 8—source data 1 ) . As an example , the most highly significant result involves NHR-28 in which the six-base core sequence ‘ACTACA’ ( which could correspond to NHR-28 or NHR-70 ) is found repeated in both dimeric and trimeric patterns ( Figure 8A , top ) . We also identified a CM involving LSY-2 ( a C2H2 ZF protein ) and NHR-232 with a spacing of one base between the core motifs ( Figure 8A , middle ) . A subset of these instances included a ZIP-6 ( bZIP ) motif at a one base distance 3′ of the NHR-232 motif , yielding a multi-family trimeric CM ( Figure 8A , bottom ) . 10 . 7554/eLife . 06967 . 023Figure 8 . Composite motifs enriched in C . elegans ChIP-seq peaks . ( A ) Stereospecificity plots showing enriched CM configurations for pairs of TF motifs . The identical ‘1F2F’ and ‘2F1F’ results in A ( top row ) demonstrate homodimer and homotrimer CMs , while those involving LSY-2 , NHR-232 , and R07H5 . 10 demonstrate heterodimer and heterotrimer CMs ( middle and bottom rows , respectively ) . Black arrows represent orientation of the motif within CMs , while gray dashed arrows designate shadow motifs within trimeric CMs . Error bars are ± S . D . , corrected p < 0 . 05 . ( B ) Forest plot of odds ratios for TF family enrichment in CMs vs input TF list . ( C ) Venn diagram showing overlap of significant CMs identified by null model 1 ( dinucleotide shuffled sequence ) and null model 2 ( motif shuffling ) . ( D ) Number of significant CMs identified relative to dinucleotide scrambled sequences using shuffled and non-shuffled non-ChIPed motifs , as a function of motif pair distance . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 02310 . 7554/eLife . 06967 . 024Figure 8—source data 1 . Table displaying enrichment statistics , spacing and orientations between PWMs for CMs identified in modENCODE ChIP-seq data . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 02410 . 7554/eLife . 06967 . 025Figure 8—figure supplement 1 . Summary of clustered CMs enriched in C . elegans ChIP-seq peaks . CM-cluster centroids are shown for the enriched motifs . For each cluster , the ChIPed TF ( s ) and potential partner TF ( s ) are listed along with information about motif overlap ( OLAP ) , spacing ( GAP ) , and enrichment . Colored arrows over motif indicate high-information content portions of either factor . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 02510 . 7554/eLife . 06967 . 026Figure 8—figure supplement 2 . Summary of clustered CMs enriched in C . elegans ChIP-seq peaks ( continued ) . CM-cluster centroids are shown for the enriched motifs . For each cluster , the ChIPed TF ( s ) and potential partner TF ( s ) are listed along with information about motif overlap ( OLAP ) , spacing ( GAP ) , and enrichment . Colored arrows over motif indicate high-information content portions of either factor . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 02610 . 7554/eLife . 06967 . 027Figure 8—figure supplement 3 . Summary of clustered CMs enriched in C . elegans ChIP-seq peaks ( continued ) . CM-cluster centroids are shown for the enriched motifs . For each cluster , the ChIPed TF ( s ) and potential partner TF ( s ) are listed along with information about motif overlap ( OLAP ) , spacing ( GAP ) , and enrichment . Colored arrows over motif indicate high-information content portions of either factor . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 02710 . 7554/eLife . 06967 . 028Figure 8—figure supplement 4 . Summary of clustered CMs enriched in C . elegans ChIP-seq peaks ( continued ) . CM-cluster centroids are shown for the enriched motifs . For each cluster , the ChIPed TF ( s ) and potential partner TF ( s ) are listed along with information about motif overlap ( OLAP ) , spacing ( GAP ) , and enrichment . Colored arrows over motif indicate high-information content portions of either factor . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 02810 . 7554/eLife . 06967 . 029Figure 8—figure supplement 5 . Summary of clustered CMs enriched in C . elegans ChIP-seq peaks ( continued ) . CM-cluster centroids are shown for the enriched motifs . For each cluster , the ChIPed TF ( s ) and potential partner TF ( s ) are listed along with information about motif overlap ( OLAP ) , spacing ( GAP ) , and enrichment . Colored arrows over motif indicate high-information content portions of either factor . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 029 PWMclus grouped the 185 CMs into 37 clusters ( Figure 8—figure supplements 1–5 ) . Most of the CMs were identified repeatedly for the same TF ChIPped in different developmental stages; these instances were considered separately in the analysis above and highlight the robustness of the observations . Some of the clusters also correspond to CMs containing motifs for related TFs , demonstrating robustness to the exact motif employed . Our methodology allowed the individual motifs to overlap , and half of the 37 CM clusters represent such overlaps . However , the majority of overlaps occur in the flanking low-information-content sections of motifs , such that most the 37 CM clusters resemble a concatenation of two motif ‘cores’ with or without a small gap ( 1–4 bases ) . Surprisingly , 17 of the 37 clusters ( ∼46% ) were obtained from the embryonic ChIP-seq data for the poorly-characterized , essential bZIP protein F23F12 . 9 ( ZIP-8 ) , which is most similar to human ATF TFs and binds both the ATF site and the CREB site ( Figure 8—figure supplements 1–3 ) . In total , these results suggest that multimeric interactions within and between TF families may be a prevalent phenomenon in C . elegans . To identify potential roles for TFs in the regulation of specific groups of functionally related genes , we asked whether the set of promoters containing a strong motif match to each TF ( FIMO p-value < 10−4 in the region −500 to +100 relative to TSS ) overlapped significantly with any tissue expression ( Spencer et al . , 2010 ) , Gene Ontology ( GO ) categories ( Ashburner et al . , 2000 ) , or Kyoto Encyclopedia of Genes and Genomes ( KEGG ) pathways ( Kanehisa et al . , 2014 ) ( Fisher's exact test , one-sided probability , false discovery rate ( FDR ) corrected q-value < 0 . 05 ) . We obtained dozens of significant relationships ( Figure 9 ) , including known roles for GATA TFs in the regulation of intestinal gene expression ( and related GO categories ) ( Pauli et al . , 2006; McGhee , 2007 ) , HLH-1 in the regulation of muscle gene expression ( Fukushige et al . , 2006 ) , DAF-19 ( an RFX TF ) in the regulation of ciliary genes ( Swoboda et al . , 2000 ) , and PHA-4 in development of the pharynx ( Gaudet and Mango , 2002 ) ( boxed in Figure 9 ) . We also note that the association of the motif for ZTF-19 ( PAT-9 ) , a C2H2 ZF protein , with genes expressed in L2 body wall muscle tissue is consistent with observed expression patterns for this gene in body wall muscle , as well as defective muscle development in a mutant ( Liu et al . , 2012 ) . The ZTF-19 binding motif may , therefore , enable the identification of specific downstream targets . Most of the associations in Figure 9 , however , appear to represent putative regulatory interactions , suggesting that the motif collection can be used to gain new biological insight . 10 . 7554/eLife . 06967 . 030Figure 9 . Enrichment of motifs upstream of gene sets . Each row of the heatmap represents a motif from our curated collection that is enriched ( q < 0 . 05 ) in at least one gene set category . Known regulatory interactions between TFs and gene sets are highlighted ( black outlines ) . ‘E’ indicates experimentally determined motifs; ‘P’ indicates predicted motifs . Source of motif is also indicated . Enrichment q-values are available in Figure 9—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 03010 . 7554/eLife . 06967 . 031Figure 9—source data 1 . Table of motif enrichments −log10 ( p-values ) in the promoters of gene set categories identified from KEGG pathway modules , Gene Ontology processes , and tissue/developmental stage specific expression lists . DOI: http://dx . doi . org/10 . 7554/eLife . 06967 . 031 The collection of motifs described here will further advance C . elegans as a major model system for the study of gene regulation . TF DNA-binding motifs enable dissection of promoters , prediction of new targets of TFs , and identification of putative new regulatory mechanisms . Statistical associations between motif matches in promoters and expression patterns or functional categories of genes also provide a ready starting point for directed experimentation , for example , analysis of gene expression in mutants . Apparent position and orientation constraints between motif matches also suggest functional relationships . Our observation that the largely unstudied bZIP TF F23F12 . 9 ( ZIP-8 ) was involved in almost half of all CMs identified in this study suggests that it may function as a cofactor for targeting to open chromatin: pioneer TF activity and partnering with other TFs has previously been proposed for other Creb/ATF proteins in mouse embryonic stem cells ( Sherwood et al . , 2014 ) . A key observation in this study is that all the large groups of TFs in C . elegans are malleable in their DNA-binding sequence preferences . The NHR is a striking case , even more so when we consider that our motifs encompass only monomeric binding sites . A previous analysis classified the C . elegans NHRs into four subtypes , on the basis of their RH sequences , and predicted that all of those in Class I ( those most similar to canonical NHRs such as the estrogen receptor ) would likely bind canonical SHRE GGTCA subsites ( Van Gilst et al . , 2002 ) . Instead , we find that those in Class I ( the entire lower half in Figure 4 ) bind a wide range of sequences , and that the RH cannot be the only determinant of sequence specificity . Our observations are consistent with the previous demonstration that mutation of one or a few residues in the NHR RH can result in dramatic changes in sequence preferences , but that mutations elsewhere in the DBD play a role in sequence selectivity ( e . g . , McKeown et al . , 2014 ) . Recent analyses of other DBD classes ( e . g . , C2H2 and Forkhead ) also highlight the importance of residues beside the canonical specificity residues ( Nakagawa et al . , 2013; Siggers et al . , 2014 ) . Together , these analyses strongly confirm that alteration of binding motifs is widespread among TF classes throughout evolution . Our study experimentally determined motifs for 129 TFs , all but five of which were previously unstudied , bringing the total number of C . elegans TFs with motifs to 292 ( including predicted motifs and data already in the literature ) . We estimate that the remaining 453 C . elegans TFs encode as many as 409 different DNA-binding motifs , most of which correspond to NHRs , C2H2 ZFs , bHLHs , and homeodomains ( Supplementary file 3 ) . Additional effort will thus be required to obtain a complete motif collection . For instance , even with our new motifs , and including motifs predicted by homology , coverage for the C . elegans NHR family is only 17% . For some classes of DBDs , most of the PBM assays yielded negative data . The NHRs in particular yielded only 20% success ( 27/135 ) . In addition , of the 108 that failed by PBM , we have tested 100 by Y1H , of which only 17 succeeded ( three or more detected interactions; data not shown ) . We observed no obvious property of their DNA-contacting residues that strongly predicts success or failure and hypothesize that requirement for ligand binding , dimerization , cofactors , or protein modifications may represent other potential explanations for failures in heterologous assays . Like human NHRs , the C . elegans NHRs have a ligand-binding domain that is distinct from the DNA-binding domain and is thought to primarily regulate interactions with coactivators and corepressors ( Sonoda et al . , 2008 ) . Thus , ligand-dependent DNA binding seems unlikely , especially for an in vitro assay . Yeast two-hybrid screens have identified several interactions between different NHRs ( Simonis et al . , 2009; Reece-Hoyes et al . , 2013 ) , suggesting that heterodimerization may be prevalent . If DNA binding is often dependent on heterodimerization , then ChIP-seq should often succeed where PBMs and Y1H fail , and heterodimeric motifs should be identified . To the best of our knowledge , however , there is only one published ChIP-seq data set for a C . elegans NHR that has yielded a motif de novo ( NHR-25 ) ( Araya et al . , 2014; Boyle et al . , 2014 ) . We did not test NHR-25 by PBM , although our predicted monomeric motif ( from Drosophila Ftz-f1 ) resembles the motif identified by ChIP-seq . As noted above , our motif for DAF-12 is also consistent with the motif obtained by ChIP–chip ( Hochbaum et al . , 2011 ) . In support of heteromeric binding , however , CMs involving NHRs and other TF families were prevalent in modENCODE ChIP-seq data ( Figure 8—source data 1 ) . Further analysis will be required to explore the role of multimerization in C . elegans TF DNA binding and gene regulation . Finally , we note that there are numerous similarities between the TF collections of human , Drosophila , and C . elegans . First , the total number TFs containing a canonical DBD varies only by a factor of ∼2 ( ∼1734 , 701 , and 744 for human , Drosophila , and C . elegans , respectively [Weirauch et al . , 2014] ) . The total number of groups of closely related paralogs ( taken using the thresholds established in Weirauch et al . ( 2014 ) ) , which should approximate the number of distinct motifs that can eventually be expected , also varies by only a factor of ∼2 ( the TFs fall into 1339 , 656 , and 632 groups of proteins expected to have very similar sequence specificity , respectively ) . Our study also brings the proportion of C . elegans TFs with a known or predicted motif much closer to the proportion in human and Drosophila , ( 56 . 1% , 54 . 1% , and 39 . 2% for human , Drosophila , and C . elegans , respectively ) . In addition , all three species possess a large number of diverse lineage-specific TFs , which are known or expected to bind different motifs: in human , ∼700 C2H2 ZF TFs; in Drosophila , 230 C2H2 ZF TFs; and in C . elegans , 266 NHRs , 99 C2H2 ZF TFs and—as we show here—roughly a dozen T-box and DM TFs . Thus , despite widespread conservation in TF number and gene expression ( Stuart et al . , 2003 ) among metazoans , extensive rewiring of the metazoan trans regulatory network is apparently common . We compiled a list of 874 known and putative C . elegans TFs . We took 740 from build 0 . 90 of the Cis-BP database ( Weirauch et al . , 2014 ) , plus additional candidate TFs that lack canonical DBDs ( Reece-Hoyes et al . , 2005 ) . We considered selecting each TF for characterization using the following criteria:Include the TF if its characterization would provide multiple-motif predictions for other TFs based on established prediction thresholds for the given DBD class ( Weirauch et al . , 2014 ) ;Include the TF if it is a member of a DBD class with relatively little available motif information;Include the TF if it has a known important biological role;Exclude the TF if it has already been characterized by PBM in another study;Exclude the TF if it is a member of a DBD class with a low-PBM success rate;Exclude the TF if the resulting construct would be excessively long ( for example , exclude C2H2 ZF TFs with many DBDs ) . We identified putative DBDs for all TFs by scanning their protein sequences using the HMMER tool ( Eddy , 2009 ) , and a collection of 81 Pfam ( Finn et al . , 2010 ) models taken from ( Weirauch and Hughes , 2011 ) , as described previously ( Weirauch et al . , 2014 ) . For some TFs , we could not identify DBDs using this procedure . In such cases , DBDs were manually detected by lowering HMMER scanning thresholds , using DBDs annotated in the SMART database ( Letunic et al . , 2012 ) , or performing literature searches . Using the above criteria for selection of TFs , we initially chose 398 TFs from the Walhout clone collection for characterization . We designed primers ( Supplementary file 4 ) to clone open reading frames ( ORFs ) comprising the DBDs plus additional flanking sequences ( 50 endogenous amino acid flanking residues , or until the end of the protein ) . We inserted the resulting sequences using AscI and SbfI restriction sites into a modified T7-driven expression vector ( pTH6838 ) that expresses N-terminal GST fusion proteins ( Supplementary file 5 ) . In the first round of cloning , we attempted cloning using both individual plasmids and pooled mRNA ( by RT-PCR ) or cDNA . After PBM analysis with the resulting clones , we then considered remaining uncharacterized TFs and selected an additional 154 TFs using the same criteria as above . The DBDs and flanking bases of these TFs were created using gene synthesis ( BioBasic , Canada ) and inserted into vectors as described above . Primers and insert sequences are provided on our project Web site . All clones were sequence verified . PBM laboratory methods were identical to those described previously ( Lam et al . , 2011; Weirauch et al . , 2013 ) . Each plasmid was analyzed in duplicate on two different arrays with differing probe sequences . Microarray data were processed by removing spots flagged as ‘bad’ or ‘suspect’ and employing spatial de-trending ( using a 7 × 7 window centered on each spot ) ( Weirauch et al . , 2013 ) . Calculation of 8-mer Z- and E-scores was performed as previously described ( Berger et al . , 2006 ) . Z-scores are derived by taking the average spot intensity for each probe containing the 8-mer , then subtracting the median value for each 8-mer , and dividing by the standard deviation , thus yielding a distribution with a median of zero and a standard deviation of one . E-scores are a modified version of the AUROC statistic , which consider the relative ranking of probes containing a given 8-mer , and range from −0 . 5 to +0 . 5 , with E > 0 . 45 taken as highly statistically significant ( Berger et al . , 2008 ) . We deemed experiments successful if at least one 8-mer had an E-score > 0 . 45 on both arrays , the complimentary arrays produced highly correlated E- and Z-scores , and the complimentary arrays yielded similar PWMs based on the PWM_align algorithm ( Weirauch et al . , 2013 ) . Motif derivation followed steps as outlined previously ( Weirauch et al . , 2014 ) . Briefly , to obtain a single representative motif for each protein , we generated motifs for each array using four different algorithms: BEEML-PBM ( Zhao and Stormo , 2011 ) , FeatureREDUCE ( manuscript in prep , source code available at http://rileylab . bio . umb . edu/content/software ) , PWM_align ( Weirauch et al . , 2013 ) , and PWM_align_Z ( Ray et al . , 2013 ) . We scored each motif on the complimentary array using the energy scoring system utilized by the BEEML-PBM algorithm ( Zhao and Stormo , 2011 ) . We then compared these PWM-based probe score predictions with the actual probe intensities using ( 1 ) the Pearson correlation coefficient and ( 2 ) the AUROC of ‘bright probes’ ( defined by transforming all probe intensities to Z-scores , and selecting probes with Z-scores ≥ 4 ) , following ( Weirauch et al . , 2013 ) . Finally , we chose a single PWM for each DBD construct using these two criteria , as previously described ( Weirauch et al . , 2014 ) . For every TF with motif information , we selected a representative motif , or motifs if that TF appears to have multiple-binding modes ( e . g . , multimers ) , using the following scheme . If the TF has an experimentally derived motif , it is selected as the primary motif . If there are multiple such motifs , we selected one that was derived in vitro , if any . If the TF had multiple in vitro motifs , then we ranked PBM > B1H > SELEX , to maximize comparability among motifs , and excluded motifs that are inconsistent with known motifs for the same or highly related proteins . If the TF had only predicted motifs , we selected a motif from a highly similar TF that is: preferably derived from an in vitro method ( PBM > B1H > SELEX ) ; assigned to the cluster that contains the majority of motifs for that TF in our PWMclus analysis; consistent with known DBD preferences; and best supported by ChIP-seq or Y1H data , if available . We calculated motif enrichment in ChIP-seq peaks using CentriMo ( Bailey and Machanick , 2012 ) , which uses TF motifs to look for central enrichment of motifs in ChIP-seq peaks , as an indication of direct binding by that TF . We obtained ChIP-seq peaks from the C . elegans modENCODE consortium ( Araya et al . , 2014 ) . We used the top 250 peaks ranked by Irreproducible Discovery Rate ( Landt et al . , 2012 ) as the input data sets . We scored the curated set of motifs for TFs with peak data sets across all the peaks . We report false discovery rate ( FDR ) -adjusted p-values for a motif's central enrichment in TF-peak data sets . For yeast one-hybrid ( Y1H ) data , we assigned motif scores to promoter bait sequences using the BEEML scoring system ( Zhao et al . , 2009 ) . We included TFs in the analysis only if they bound five or more promoters in Y1H ( those with 3 or 4 promoters bound were excluded to minimize sampling error in Mann–Whitney tests ) . We scored only the promoter-proximal 500 bp of Y1H bait sequences , as activating TF-binding sites are mainly effective within a few hundred bases of TSS in Saccharomyces cerevisiae ( Dobi and Winston , 2007 ) . We calculated motif enrichment or depletion for motifs using a two-tailed Mann–Whitney U test and reported with FDR-corrected p-values , with Y1H interactors as positives and the remaining non-interacting baits as the background . We performed composite motif ( CM ) analysis by scanning 77 C . elegans ChIP-seq top 250 peak sequences for all pairwise combinations between the 40 ChIPed TFs ( using the curated list of PWMs ) and 129 PBM-derived PWMs from this study . Relative PWM spacing was restricted to −5 ( overlapping ) to +10 ( gapped ) bp separation , with four possible stereospecific arrangements of TFs: TF-1 forward TF-2 forward ( 1F2F ) , TF-1 forward TF-2 reverse ( 1F2R ) , 2R1F , and 2F1F , yielding 64 stereospecific combinations . We identified sequence matches using the standard log-likelihood scoring framework ( Stormo , 2000 ) , with a threshold of 0 . 50max_score for each PWM , where max_score is the highest possible score for the given PWM . We created 10 sets of background sequences by scrambling the input sequences ( maintaining dinucleotide frequencies ) . We calculated sample Z-scores and p-values by comparing the number of sequence matches observed in the ‘real’ sequence to the number observed in the random sequences and applied a Bonferroni correction to each p-value . To identify significant composite motifs , we filtered to retain only results with sequence match counts ≥10% of the number of input peak sequences and Bonferroni-adjusted p-values ≤ 0 . 05 ( alpha = 0 . 05 ) . We also considered an alternative null model , in which we shuffled the non-ChIPed motif , and counted matches in the original DNA sequences ( this procedure was repeated 10× ) . Overall , we found very good agreement using this approach and our original null model . Out of the 635 , 712 possible patterns we tested , both methods call 635 , 483 insignificant , both call 49 positive , and they disagree on 180 ( Figure 8C ) . Figure 8D plots the number of significant hits identified relative to dinucleotide scrambled sequences using shuffled ( blue ) and non-shuffled ( red ) non-ChIPed motifs . This plot indicates , however , that the shuffled motif null model over-estimates the significance of CMs as the overlap of their constituent motifs increases , presumably due to dispersal of high-information content ‘core’ positions , which are typically adjacent in the real motifs . We , therefore , use and report results based only on null model 1 . Sequence logos were constructed using the actual matches obtained in the ChIP-seq peak sequences , and the WebLogo 3 . 4 tool ( Crooks et al . , 2004 ) . For each TF family F , we calculated an odds ratio ( OR ) comparing the ratio of families in CMs to the ratio of families in the motif list . We define OR as ( a/b ) / ( c/d ) , where a is the number of TFs of family F involved in a CM; b is the total number of unique TF pairs involved in a CM , minus a; c is the number of TFs of family F in the motif list; and d is the total number of TFs in the motif list , minus c . We calculated the standard error as √ ( 1/a + 1/b + 1/c + 1/d ) , and the 95% confidence interval as eln ( OR ) ± 1 . 96SE . We obtained selectively enriched gene sets for each tissue from ( http://www . vanderbilt . edu/wormdoc/wormmap/ ) , GO annotations from ( http://www . geneontology . org/ ) , and KEGG pathway modules ( http://www . genome . jp/kegg/module . html ) . We ran FIMO ( Grant et al . , 2011 ) with default parameters . PBM-microarray data are available at GEO ( www . ncbi . nlm . nih . gov/geo/ ) under accession number GSE65719 . Motifs and 8-mer data ( E- and Z-scores ) are available at www . cisbp . ccbr . utoronto . ca . Supplementary data files , including plasmid and primer information , motifs , source data for figure heatmaps , and lists of TFs are found on our project web site http://hugheslab . ccbr . utoronto . ca/supplementary-data/CeMotifs/ .	Many scientists use ‘model’ species—such as the fruit fly or a nematode worm called Caenorhabditis elegans—in their research because these organisms have useful features that make it easier to carry out many experiments . For example , C . elegans has a smaller genome compared to many other animals , which is useful for studying the roles of individual genes or stretches of DNA . Transcription factors are a type of protein that can bind to specific stretches of DNA and help to switch certain genes on or off . These ‘motifs’ may be close to the gene or further away in the genome , and therefore , must stand out amongst the rest of the DNA , like lights on a landing strip . However , the motifs for only 10% of the estimated 763 transcription factors in C . elegans have been identified so far . In this study , Narasimhan , Lambert , Yang et al . used a technique called a ‘protein binding microarray’ to identify the motifs for many more of the C . elegans transcription factors . These findings were then used to predict motifs for other transcription factors . Together , these methods increased the proportion of C . elegans transcription factors with known DNA-binding motifs from 10% to around 40% . Now that more DNA motifs have been identified , it is possible to look for similarities and differences between them . For example , Narasimhan , Lambert , Yang et al . found that transcription factors with similar sequences can bind to very varied motifs . On the other hand , some transcription factors that are very different are able to recognize very similar motifs . The experiments also indicate that motifs found very close to genes—in sequences known as ‘promoters’—may be able to interact with many proteins to influence the activity of genes . Narasimhan , Lambert , Yang et al . 's findings increase the number of C . elegans transcription factors with a motif , bringing the knowledge of these proteins more in line with the better-studied transcription factors of humans and fruit flies . The next challenge is to identify DNA motifs for the remaining 60% of transcription factors .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology", "genetics", "and", "genomics" ]	2015	Mapping and analysis of Caenorhabditis elegans transcription factor sequence specificities
Purified microtubules have been shown to align along the static magnetic field ( SMF ) in vitro because of their diamagnetic anisotropy . However , whether mitotic spindle in mammalian cells can be aligned by magnetic field has not been experimentally proved . In particular , the biological effects of SMF of above 20 T ( Tesla ) on mammalian cells have never been reported . Here we found that in both CNE-2Z and RPE1 human cells spindle orients in 27 T SMF . The direction of spindle alignment depended on the extent to which chromosomes were aligned to form a planar metaphase plate . Our results show that the magnetic torque acts on both microtubules and chromosomes , and the preferred direction of spindle alignment relative to the field depends more on chromosome alignment than microtubules . In addition , spindle morphology was also perturbed by 27 T SMF . This is the first reported study that investigated the mammalian cellular responses to ultra-high magnetic field of above 20 T . Our study not only found that ultra-high magnetic field can change the orientation and morphology of mitotic spindles , but also provided a tool to probe the role of spindle orientation and perturbation in developmental and cancer biology . The mitotic spindle is a highly dynamic microtubule assembly responsible for chromosome segregation and cleavage furrow positioning during cell division . Spindle orientation is central to cell fate determination and tissue architecture , and mounting evidences show that astral microtubules , multiple cortical proteins and their interactions with plasma membrane are all critical to achieve correct orientation ( Toyoshima et al . , 2007; Lancaster and Knoblich , 2012; Lu and Johnston , 2013; Bergstralh and St Johnston , 2014; Nestor-Bergmann et al . , 2014 ) . Tissues may have additional control mechanisms to minimize spindle orientation errors , such as apoptosis in the wing disc ( Bergstralh and St Johnston , 2014 ) . The connection between spindle orientation and tumorigenesis has also been an active field in recent years ( Gonzalez , 2007; Knoblich , 2010; Pease and Tirnauer , 2011 ) . Most biological materials are diamagnetic , such as proteins , DNA and lipids . SMFs can align large biological objects that have diamagnetic anisotropy , such as microtubule polymers and nucleic acid chains , as well as some types of cells and organisms ( Maret and Dransfeld , 1977; Torbet and Ronzière , 1984; Higashi et al . , 1993; Emura et al . , 2001; Takeuchi et al . , 2002; Teodori et al . , 2006; Stern-Straeter et al . , 2011 ) . The degree of alignment with the externally applied magnetic field is proportional to the product of the molecular magnetic susceptibility and the magnetic field strength . For proteins , the diamagnetic anisotropy is mainly due to the alpha helix , beta sheet and aromatic rings ( Chabre , 1978; Worcester , 1978; Pauling , 1979 ) . Even individual peptide bonds , which have weak diamagnetic anisotropy can contribute when linked together in a fixed and organized orientation in alpha helix or beta sheet . Most proteins have very weak diamagnetic anisotropy , but their response to a magnetic field can be amplified by ordered polymerization , as in microtubules , where the additive diamagnetic anisotropy can be significant . The DNA chain is another example of a biopolymer with relatively large diamagnetic anisotropy ( Maret et al . , 1975 ) , mainly due to its stacked aromatic bases . Theoretical predictions suggested that mitotic chromosome arms , where DNA is highly compacted , might generate electromagnetic fields along the chromosome arm direction ( Zhao and Zhan , 2012 ) and , less speculatively , that chromosomes should be fully aligned by SMFs of around 1 . 4 T ( Maret , 1990 ) . Mitotic spindles comprise both microtubules and chromosomes , but most studies of their potential response to magnetic fields have focused on microtubules . Multiple studies have shown that purified microtubules can be aligned by moderate and high SMFs and the alignment effect increases significantly with magnetic field strength ( Vassilev et al . , 1982; Bras et al . , 1998; Glade and Tabony , 2005 ) . This is due to the diamagnetic anisotropy of tubulin and microtubules ( Bras et al . , 2014 ) . In addition , tubulin assembly in vitro was disordered by a 10–100 nT hypogeomagnetic field ( natural geomagnetic field is usually around 50000 nT/0 . 5 Gauss ) ( Wang et al . , 2008 ) . However , whether SMF can change the mitotic spindle orientation in a cell has never been reported . Denegre et al found that 16 . 7 T large gradient SMFs can affect the division orientation of Xenopus eggs ( Denegre et al . , 1998 ) . They proposed that SMF may affect the orientation of astral microtubules and/or spindles , which was theoretically and experimentally proved later ( Valles , 2002; Valles et al . , 2002 ) . Although the importance of aster microtubules in spindle orientation determination in some types of cells is well known ( Palmer et al . , 1992; Shaw et al . , 1997 ) , the metaphase spindles still can orient themselves parallel to the substrate in the absence of aster microtubules in both Madin-Darby Canine Kidney ( MDCK ) and human cervical cancer HeLa cells ( Lazaro-Dieguez et al . , 2015 ) . In addition , the spindle orientation is controlled by multiple signaling proteins , microtubules and associated proteins , as well as actin and associated proteins ( Toyoshima and Nishida , 2007a , 2007b; Woolner et al . , 2008; Thaiparambil et al . , 2012 ) . Although a few reports in recent years showed that the microtubule and actin cytoskeleton in interphase cells could be affected by 7–17 T ultra-high SMFs in some cell types ( Valiron et al . , 2005 ) , information about the mitotic spindle was not provided . Therefore , although the theoretical prediction of the SMF effect on spindle orientation in cells has been experimentally tested in Xenopus embryo , how the spindles in mammalian cells , which have very different structure than Xenopus embryo , are affected is still unknown . In our previous study , we found that prolonged treatment of 1 T moderate intensity SMF can cause multipolar spindles in cells ( Luo et al . , 2016 ) and we predict that stronger SMFs are likely to produce more obvious effects on spindles . To accommodate cells in the ultra-high field magnet we constructed a custom cell incubation system using two sample holders that could fit inside the new 32 mm bore ultra-high field magnets in the Chinese High Magnetic Field Laboratory ( CHMFL , China ) . This platform provides accurate temperature and gas control for animal and human cells as well as some small model animals . The WM4 magnet we use ( Figure 1A ) provides vertical ultra-high homogenous SMF at the center of the magnet . The magnetic field direction is upward . Bright field microscopic observation and flow cytometry did not reveal obvious changes of cell morphology or cell death after 4 hr of 27 T SMF exposure in human nasopharyngeal carcinoma CNE-2Z cell ( Figure 2A–D; Figure 2—source data 1 ) . Moreover , immunostaining analysis did not reveal obvious microtubule or actin cytoskeleton abnormalities in interphase cells after 4 hr of 27 T exposure either ( Figure 2E ) . Therefore 27 T SMF treatment for 4 hr is not acutely toxic to CNE-2Z cells . In addition , we found that the cell morphology was not obviously changed ( Figure 2—figure supplement 1A ) three days post-exposure but the cell number decreased by ~40% ( Figure 2—figure supplement 1B ) . Flow cytometry analysis showed that the cell cycle was only slightly changed ( Figure 2—figure supplement 1C ) while the cell death was not obviously affected ( Figure 2—figure supplement 1D ) . 10 . 7554/eLife . 22911 . 003Figure 1 . 27 T ultra-high water-cooled magnet and the biological sample incubation system . ( A ) The WM4 ( water-cooled magnet#4 in the Chinese High Magnetic Field Lab ) . ( B , C ) The design and picture of the biological sample incubation system . Two identical sets were made . One was used in the magnet while the other was placed outside of the magnet to serve as the ‘sham’ control . ( D ) The top view of the magnet , where the biological sample incubation tube was inserted . ( E ) The magnetic field was maintained at 27 T ( total of 4 hr , 3 hr stable maintenance at 27 T with half hour increase and half hour decrease ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 00310 . 7554/eLife . 22911 . 004Figure 1—figure supplement 1 . The design and actual picture of the 18 mm custom made cell culture plate . ( A ) The design of the 18 mm plate . ( B ) The picture of the cell culture plates of different sizes . 100 mm , 60 mm and 35 mm plates were placed side-by-side in the bottom picture for size comparison . The smallest plate in the picture was the custom made 18 mm cell plate . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 00410 . 7554/eLife . 22911 . 005Figure 2 . 27 T Ultra-high magnetic field does not have immediate cytotoxicity effects in CNE-2Z cells . CNE-2Z cells were plated directly on 18 mm tissue culture plate or coverslips in the18 mm tissue culture plate one night ahead to allow the cells to attach . On the day of experiment , they were placed in regular full-sized cell incubator ( control ) or the sample incubators in sham or in 27 T magnet for 4 hr before they were taken out and subjected to the following analysis . ( A ) Representative bright field images and of control , sham and 27 T SMF treated CNE-2Z cells . Scale bar: 20 μm . ( B ) Quantification of cell numbers in control , sham and 27 T SMF treated CNE-2Z cells from three independent experiments ( n = 3 ) . Data is mean ± SD . ‘ns’ , not significant . ( C ) Flow cytometry results of CNE-2Z cells treated with control , sham or 27 T for 4 hr and dual staining with annexin V and PI . Bottom left leaflet shows the live cells that have intact cell membrane and have negative staining for both dyes . Top left leaflet shows necrotic cells . Right parts show apoptotic cells . ( D ) Quantification of cell numbers in each population . ( E ) Immunofluorescence of CNE-2Z cells shows that 27 T SMF does not have obvious effects on microtubule and actin cytoskeleton in interphase cells . CNE-2Z cells were fixed and stained with anti-tubulin antibody and fluorescently labeled phalloidin for microtubules ( green ) and actin ( red ) cytoskeleton . Experiments have been repeated for three times and representative images are shown in the figure . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 00510 . 7554/eLife . 22911 . 006Figure 2—source data 1 . Quantification of cell numbers in control , sham and 27 T SMF treated CNE-2Z cells . This is the source data for Figure 2B . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 00610 . 7554/eLife . 22911 . 007Figure 2—figure supplement 1 . 27 T SMF reduced CNE-2Z cell number three days post-exposure . CNE-2Z cells were plated one night ahead to allow the cells to attach , exposed to 27 T SMF for 4 hr before they were taken out and returned back to the regular full sized cell incubator for another 3 days before they were subjected to the following analysis . ( A ) Representative bright field images and of control , sham and 27T SMF treated CNE-2Z cells . ( B ) Quantification of cell numbers in control , sham and 27 T SMF treated CNE-2Z cells from three independent experiments ( n = 3 ) . Data is mean ± SD . p<0 . 05 . ( C ) Flow cytometry cell cycle analysis shows the cell distribution in each cell cycle . ( D ) Flow cytometry results of CNE-2Z cells treated with control , sham or 27T for 4 hr and dual staining with annexin V and PI for cell death . Quantification of cell numbers in each cell population from three independent experiments ( n = 3 ) . Data is mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 007 Tubulin and microtubules have been shown to have diamagnetic anisotropy and purified microtubules can align along the magnetic field in vitro . However , the spindle orientation after strong SMF exposure had never been experimentally investigated . Using immunofluorescence analysis of cells fixed immediately after they were taken out of the magnet , we found that mitotic spindle orientation was perturbed by 27 T SMF ( Figure 3A ) . In sham incubated or control cells , the mitotic spindle axis is usually parallel to the tissue culture plate or coverslip ( ‘lateral’ ) while the 27 T magnetic field increased the percentage of spindles that were not parallel to the coverslip ( ‘non-lateral’ ) ( Figure 3A , B ) . We separately measured metaphase spindles ( Figure 3C; Figure 3—source data 1 ) as well as prometaphase and metaphase spindles together ( Figure 3—figure supplement 1 ) and both results showed that non-lateral spindles were increased by 27 T SMF . We also examined lower magnetic field intensities for their effect on spindle orientation . The 0 . 05 T and 1 T SMFs were provided by permanent magnets placed in regular incubators ( Figure 4A ) . We found that 4 hr of exposure does not increase the non-lateral spindles ( Figure 4B; Figure 4—source data 1 ) as 27 T SMF . We also tested 9 T SMF provided by a superconducting magnet ( Zhang et al . , 2016 ) and found that 4 hr of 9 T exposure does not increase the non-lateral spindles either ( Figure 4C; Figure 4—source data 2 ) . However , although exposure for a prolonged time of 3 days to 0 . 05 T or 1 T SMFs still had no effect ( Figure 4D; Figure 4—source data 3 ) , 9 T SMF treatment for 3 days could perturb spindle orientation ( Figure 4E; Figure 4—source data 4 ) . In contrast , 0 . 05 T and 1 T SMFs had no effect on spindle orientation even after 7 days of exposure ( Figure 4F; Figure 4—source data 5 ) . Therefore the effect of SMF on spindle orientation was field intensity-dependent and time-dependent . Ultra-high SMF of 27 T could change spindle orientation in 4 hr but 9 T SMF need 3 days to show effects . 10 . 7554/eLife . 22911 . 008Figure 3 . 27 T SMF changed spindle orientation . CNE-2Z cells were plated on coverslips in the 18 mm tissue culture plate one night ahead to allow the cells to attach . On the day of experiment , they were placed in regular full-sized cell incubator ( control ) or the sample incubators in sham or in 27 T magnet for 4 hr before they were taken out , fixed and stained with anti-tubulin antibody ( for microtubules ) , phalloidin ( for F-actin ) . ( A ) Representative immunofluorescence images of CNE-2Z cells show that 27 T SMF changes spindle orientation . Multistack images were taken and individual vertical image planes ( Z ) were displayed to show the spindle orientation . Microtubules are shown in green and F-actin filaments are shown in red . One cell in sham and two cells in 27 T SMF treated group are shown . ( B ) Illustration of spindles with different orientations . ‘B’ shows the magnetic field direction and ‘g’ shows the gravity direction . ( C ) Quantification of spindle orientations in control , sham or 27 T treated cells from four independent experiments ( n = 4 ) . Data is presented as mean ± SD . p<0 . 05; p<0 . 01 . Total of 921 metaphase spindles were counted . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 00810 . 7554/eLife . 22911 . 009Figure 3—source data 1 . Quantification of spindle orientations in control , sham or 27 T treated cells from four independent experiments . This is the source data for Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 00910 . 7554/eLife . 22911 . 010Figure 3—figure supplement 1 . Spindle orientation in prometaphase and metaphase CNE-2Z cells were changed by 27 T SMF . Quantification of spindle orientations in control , sham or 27 T treated cells from four independent experiments ( n = 4 ) . Data is presented as mean ± SD . Both prometaphase and metaphase cells were quantified . Total of 1447 spindles were measured . p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01010 . 7554/eLife . 22911 . 011Figure 4 . SMFs affected spindle orientation in a field intensity dependent manner . CNE-2Z cells were plated on coverslips in the 35 mm or 24 well tissue culture plate one night ahead to allow the cells to attach . On the day of experiment , they were exposed to different intensity SMFs for different time before they were taken out , fixed and stained with anti-tubulin antibody ( for microtubules ) and DAPI ( for DNA ) . ( A ) 0 . 05 T and 1 T moderate intensity SMF exposure experimental set-up . Permanent magnets were placed inside a regular full-sized cell incubator to ensure proper culture conditions . Cell culture plate was placed on the top surface center of the magnet . ( B–F ) Quantification of spindle orientations in control , 0 . 05 T , 1 T or 9 T treated cells . Total of 200–300 metaphase spindles from 3–4 independent coverslips were examined for each condition . Data is presented as mean ± SD . ‘ns’ , not significant; p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01110 . 7554/eLife . 22911 . 012Figure 4—source data 1 . Quantification of spindle orientations in control , 0 . 05 T or 1 T treated cells ( 4 hr treatment ) . This is the source data for Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01210 . 7554/eLife . 22911 . 013Figure 4—source data 2 . Quantification of spindle orientations in control , or 9 T treated cells ( 4 hr treatment ) . This is the source data for Figure 4C . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01310 . 7554/eLife . 22911 . 014Figure 4—source data 3 . Quantification of spindle orientations in control , 0 . 05 T or 1 T treated cells ( 3d treatment ) . This is the source data for Figure 4D . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01410 . 7554/eLife . 22911 . 015Figure 4—source data 4 . Quantification of spindle orientations in control , or 9 T treated cells ( 3d treatment ) . This is the source data for Figure 4E . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01510 . 7554/eLife . 22911 . 016Figure 4—source data 5 . Quantification of spindle orientations in control , 0 . 05 T or 1 T treated cells ( 7d treatment ) . This is the source data for Figure 4F . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 016 Both microtubules and chromatin exhibit diamagnetic anisotropy , but prior works about spindle orientation had only assumed microtubules are the target of strong magnetic fields and did not consider the chromatin . As a start to distinguish these effects , we aligned coverslips in the same axis as the magnetic field direction , so the field should only change orientation in the plane of the coverslip ( Figure 5A–C ) . We also added another non-transformed human retinal pigment epithelial cell line RPE1 to the experiment to see whether the effects we observed are cell specific ( Figure 5C ) . Synchronization was also used to enrich mitotic cells ( Figure 5—figure supplement 1 ) . We grouped the spindles into three different groups based on their orientation relative to the field direction: parallel to the magnetic field , normal to the magnetic field and other ( Figure 5D ) . Interestingly , spindle orientation was different in prometaphase cells vs . metaphase cells ( Figure 5E ) . In the 27 T SMF treated group , prometaphase cells tended to orient with their spindle long axis in parallel with the field direction , while metaphase cells tended to orient with the spindle long axis normal to the field direction ( Figure 5E ) . Synchronized and unsynchronized cells responded similarly ( Figure 5—figure supplement 2A , B ) so we combined data from both synchronized and unsynchronized cells for statistical analysis . As shown in Figure 5F ( Figure 5—source data 1 ) and Figure 5G ( Figure 5—source data 2 ) , 27 T SMF had similar orientation effects on CNE-2Z and RPE1 spindles . Figure 5—figure supplement 3 shows the Cosinus of spindle angle . The main difference between cells in prometaphase and metaphase is the organization of chromosomes , which are distributed in prometaphase and tightly aligned in metaphase . We suspect this difference accounts for the differential response to magnetic fields . 10 . 7554/eLife . 22911 . 017Figure 5 . Prometaphase and metaphase spindles have different orientations in 27 T SMFs . ( A–C ) Schematic illustration of the experimental set-up . ( A ) CNE-2Z and RPE1 cells were plated on pre-cut coverslips one night ahead to allow the cells to attach . ( B ) On the day of experiment , the coverslips were inserted onto agarose gel in the 18 mm plates . ( C ) Cells were treated with or without synchronization , and with or without 27 T magnetic field for 4 hr before they were fixed and stained with anti-tubulin antibody ( for microtubules ) and fluorescently labeled phalloidin ( for actin polymer ) and DAPI ( for DNA ) . ‘B’ shows the magnetic field direction and ‘g’ shows the gravity direction . ( D ) The orientation of the spindle long axis was measured and characterized into ‘parallel’ ( green ) , ‘normal’ ( blue ) and ‘others’ ( grey ) according to the angle between spindle long axis and the magnetic field direction . ( E ) Representative immunofluorescence images of prometaphase and metaphase RPE1 cells that have different orientation when they were exposed to 27 T SMF for 4 hr . Scale bar: 10 μm . ( F , G ) Quantification of prometaphase and metaphase spindle orientations in control , sham or 27 T treated CNE-2Z ( F ) , and RPE1 ( G ) cells . One experiment was done in synchronized cells and the other was done with unsynchronized . Total of 1575 spindles were measured from four independent coverslips from two independent experiments . The histograms were created in excel ( mean ± SD ) . Scatter plots were created in GraphPad ( mean ± SEM ) . p<0 . 05 , p<0 . 01 , p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01710 . 7554/eLife . 22911 . 018Figure 5—source data 1 . Quantification of prometaphase and metaphase spindle orientations in control , sham or 27 T treated CNE-2Z cells . This is the source data for Figure 5F . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01810 . 7554/eLife . 22911 . 019Figure 5—source data 2 . Quantification of prometaphase and metaphase spindle orientations in control , sham or 27 T treated RPE1 cells . This is the source data for Figure 5G . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 01910 . 7554/eLife . 22911 . 020Figure 5—figure supplement 1 . Synchronization procedure to enrich mitotic cells and spindle orientation measurement . ( A ) Schematic illustration of the experimental procedure . ( B ) Representative immunofluorescence images of CNE-2Z and RPE1 cells show that 27 T SMF changes spindle orientation . Spindles with different orientations were labeled by boxes with different colors . Orange indicates parallel to the field direction; Red indicates normal to the field directions; Gray indicates others that are not parallel or vertical to the field direction . Scale bar: 10 μm . ( C ) The angle between spindles and the magnetic field direction was measured . The angles were measured by drawing a line between the two spindle poles and a line for the field direction . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 02010 . 7554/eLife . 22911 . 021Figure 5—figure supplement 2 . Prometaphase and metaphase cells show different orientation in both synchronized and unsynchronized CNE-2Z and RPE1 cells after 27 T SMF exposure for 4 hr . Quantification of prometaphase and metaphase spindle orientations in 27 T treated CNE-2Z ( A ) and RPE1 ( B ) cells from four independent coverslips from an unsynchronized or synchronized experiment . One dot represents one spindle . Data is mean ± SEM . *p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 02110 . 7554/eLife . 22911 . 022Figure 5—figure supplement 3 . Cosinus of angles between spindle long axis and the 27 T magnetic field direction . The calculation was based on Figure 5F and G . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 022 To further investigate the relative contribution of chromosomes vs . microtubules in spindle orientation , we used a CENP-E inhibitor to disrupt metaphase chromosome alignment without damaging the bipolar spindle ( Figure 6A ) . In this drug , many mitotic cells exhibited highly disorganized chromosomes ( Figure 6B ) . Then we measured the spindle length , chromosome distribution , metaphase plate width ( Figure 6C ) as well as the angle between spindle long axis and the magnetic field/gravity direction in Adobe Photoshop . After quantifying four independent coverslips for both CNE-2Z and RPE1 cells , we found that the chromosome distribution in spindles seemed to affect the angle between spindle long axis and the magnetic field/gravity direction ( Figure 6D , Figure 6—source data 1 and Figure 6—figure supplement 1 ) . The preferred spindle orientation relative to the field differed between cells that have well aligned chromosomes and misaligned chromosomes , and that this difference was significant ( p<0 . 01 in RPE1 cells and p<0 . 05 in CNE-2Z cells ) ( Figure 6E , Figure 6—source data 2 , Figure 6F , Figure 6—source data 3 ) . This indicates that chromosomes play a major role in spindle orientation in response to SMFs . Spindles with misaligned chromosomes tend to align with the spindle long axis in parallel with the field direction while spindles with well aligned chromosomes tend to align their metaphase plate in parallel with the field direction ( Figure 6G ) . Not surprisingly , lower magnetic fields of 1 T or 9 T treatment for 4 hr did not affect the spindle orientation ( Figure 6—figure supplement 2 ) . 10 . 7554/eLife . 22911 . 023Figure 6 . Chromosome alignment affects spindle orientation . CNE-2Z and RPE1 cells were treated with RO-3306 and the CENP-E inhibitor ( GSK923295 ) before they were exposed to 27 T SMF for 4 hr . Cells were then harvested for Immunofluorescence experiment . ( A ) Schematic illustration of the experimental procedure . ( B ) Representative immunofluorescence images show that CNE-2Z cells with misaligned chromosomes prefer to align with their spindle long axis in parallel with the magnetic field direction ( upward ) but the cells with well aligned chromosomes prefer to align their chromosome plate in parallel , with the spindle long axis normal to , the magnetic field direction . Scale bar: 10 μm . ( C ) Schematic illustration of the spindle and chromosome measurement . ‘a’ is the spindle length , defined as the distance between two spindle poles . ‘b’ is the chromosome distribution , defined by the maximum distance between chromosomes along the spindle long axis . ‘c’ is the metaphase plate width . Measurements were done in Adobe Photoshop . ( D ) The 27 T SMF affects the angle between spindle long axis with magnetic field direction in CNE-2Z and RPE1 cells , which is determined by chromosome distribution . Sham and 27 T groups are shown in the figure . Control groups are shown in the figure supplement due to space limitation . ( E , F ) Quantification of the angle between spindle long axis with the magnetic field direction in CNE-2Z ( E ) or RPE1 ( F ) cells in control , sham control or 27 T SMF treated group to compare the difference between spindles with misaligned vs . aligned chromosomes . ‘Misaligned chromosomes’ were defined as c/b <1 . 5 ( CNE-2Z cells ) or 1 . 8 ( RPE1 cells ) and ‘Aligned chromosomes’ were defined as c/b >1 . 5 ( CNE-2Z cells ) or 1 . 8 ( RPE1 cells ) . Quantifications for D-F were from total of 618 CNE-2Z spindles and 452 RPE1 spindles . Spindles for each cell type were from four independent coverslips in two independent days . Data is mean ± SEM . ‘ns’ , not significant; p<0 . 05; *p<0 . 01 . ( G ) Cartoon illustrates that spindles with misaligned chromosomes tend to align along the magnetic field direction ( B , upward ) while spindles with compact metaphase plate tend to align normal to the field direction . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 02310 . 7554/eLife . 22911 . 024Figure 6—source data 1 . The 27 T SMF affects the angle between spindle long axis with magnetic field direction in CNE-2Z and RPE1 cells , which is determined by chromosome distribution . This is the source data for Figure 6D . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 02410 . 7554/eLife . 22911 . 025Figure 6—source data 2 . Quantification of the angle between spindle long axis with the magnetic field direction in CNE-2Z cells in control , sham control or 27 T SMF treated group to compare the difference between spindles with misaligned vs . aligned chromosomes . This is the source data for Figure 6E . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 02510 . 7554/eLife . 22911 . 026Figure 6—source data 3 . Quantification of the angle between spindle long axis with the magnetic field direction in RPE1 cells in control , sham control or 27 T SMF treated group to compare the difference between spindles with misaligned vs . aligned chromosomes . This is the source data for Figure 6F . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 02610 . 7554/eLife . 22911 . 027Figure 6—figure supplement 1 . The chromosome distribution and angle between spindle long axis and the magnetic field/gravity direction in control CNE-2Z and RPE1 cells . These are the control groups of Figure 6D . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 02710 . 7554/eLife . 22911 . 028Figure 6—figure supplement 2 . SMFs of 1 T and 9 T did not affect spindle orientation in CNE-2Z cells . Similar to Figure 6 , ‘a’ is the spindle length , defined as the distance between two spindle poles . ‘b’ is the chromosome distribution , defined by the maximum distance between chromosomes along the spindle long axis . a/b value reflects the chromosome distribution within the spindle . ( A ) The angle between spindle long axis with the magnetic field direction was not affected by 1 T and 9 T SMFs or the chromosome distribution within the spindle; ( B ) The angle between spindle long axis with the magnetic field direction was not affected by chromosome alignment in 1 T and 9 T SMFs . Quantifications were from total of 466 CNE-2Z spindles from four independent coverslips in two independent days . Data is mean ± SEM . ‘ns’ , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 028 We also noted that spindles aligned normal to the magnetic field direction were wider than those parallel to the field direction ( Figure 7A ) . We carefully compared it in control and 27 T SMF treated cells ( Figure 7B ) . In both synchronized and unsynchronized CNE-2Z and RPE1 cells , the spindles in 27 T SMF-treated cells that were normal to the field direction has bigger spindle pole angles , which makes them look ‘wider’ ( Figure 7C , Figure 7—source data 1 and Figure 7—figure supplement 1 ) compare to the ones parallel to the field direction . In control cells that were not treated with SMFs , the spindle pole angles are similar in both directions ( Figure 7C ) . To get a more quantitative measurement , we quantified the spindle length ( a ) and width ( w ) and found that the spindle length was not much affected by the 27 T magnetic field ( Figure 7—figure supplement 2A , B ) but the spindle width was increased in both CNE-2Z and RPE1 cells , when the spindle was normal to the field direction ( Figure 7D , Figure 7—source data 2 and Figure 7—figure supplement 2C , D ) , which confirms that the spindle is ‘wider’ . 10 . 7554/eLife . 22911 . 029Figure 7 . 27 T SMF changes spindle morphology in both CNE-2Z and RPE1 cells . ( A ) Representative immunofluorescence images of CNE-2Z and RPE1 cells with or without 27 T SMF treatment for 4 hr . Scale bar: 5 μm . ( B ) Illustration of the pole angle measurement of the metaphase spindles in CNE-2Z and RPE1 cells with or without 27 T SMF treatment for 4 hr . ‘1’ measures the pole angle of metaphase spindles in parallel to the magnetic field/gravity direction ( green ) and ‘2’ measures the pole angle of metaphase spindles normal to the magnetic field/gravity direction ( blue ) . Scale bar: 5 μm . ( C ) Quantification of the metaphase spindle pole angle measurement for synchronized CNE-2Z and RPE1 cells with or without 27 T SMF . RO-3306 and MG132 synchronization was used to increase the percentage of mitotic cells . Total of 295 metaphase spindles were measured from four independent coverslips . Data is mean ± SEM . ( D ) Quantification of the spindle width for RO-3306 and CENP-E inhibitor treated CNE-2Z and RPE1 cells . Experimental procedure was as shown in Figure 5A . Sham and 27 T treated groups are shown here and the control groups are shown in Figure 7—figure supplement 2C , D . ( E ) Illustration of the spindle and chromosome measurement . ( F ) Quantification of the relationship between spindle morphology and chromosome alignment in CENP-E inhibitor treated CNE-2Z cells that have spindle axis normal to the magnetic field/gravity direction ( angle of 80–90 degree ) . Misaligned chromosomes ( black ) vs . aligned chromosomes ( red ) were classified by different c/b ratio values . Measurement was done on spindles from four independent coverslips from two independent experiments . Data is mean ± SEM . ‘ns’ , not significant; p<0 . 05; **p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 02910 . 7554/eLife . 22911 . 030Figure 7—source data 1 . Quantification of the metaphase spindle pole angle measurement for synchronized CNE-2Z and RPE1 cells with or without 27 T SMF . This is the source data for Figure 7C . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 03010 . 7554/eLife . 22911 . 031Figure 7—source data 2 . Quantification of the spindle width for RO-3306 and CENP-E inhibitor treated CNE-2Z and RPE1 cells . This is the source data for Figure 7D . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 03110 . 7554/eLife . 22911 . 032Figure 7—source data 3 . Quantification of the relationship between spindle morphology and chromosome alignment in CENP-E inhibitor treated CNE-2Z cells that have spindle axis normal to the magnetic field/gravity direction ( angle of 80–90 degree ) . This is the source data for Figure 7F . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 03210 . 7554/eLife . 22911 . 033Figure 7—figure supplement 1 . Quantification of the spindle pole angle measurement in unsynchronized CNE-2Z and RPE1 cells . Green dots represent spindles that are in parallel to the field direction . Blue dots represent spindles that are normal to the field direction . Total of 166 metaphase spindles were measured from four independent coverslips . Data is mean ± SEM . ‘ns’ , not significant; p<0 . 05; ***p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 03310 . 7554/eLife . 22911 . 034Figure 7—figure supplement 2 . Spindle length was not affected by 27 T SMF when spindle axis was normal to the field direction . Spindle length and width , as well as their orientation were measured in CNE-2Z and RPE1 cells . ( A , B ) Spindle length of control , sham or 27 T treated CNE-2Z ( A ) and RPE1 ( B ) cells . ( C , D ) Spindle width in control CNE-2Z ( C ) and RPE1 ( D ) cells . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 034 To find out whether chromosome alignment affected the spindle morphology change in the presence of 27 T field , we analyzed the spindle dimension and chromosome distribution ( Figure 7E ) for CENP-E inhibitor treated spindles that are normal to the field direction ( the angle between spindle long axis to the magnetic field direction within 80–90 degree ) . It is interesting that the chromosome alignment or misalignment did actually affect the spindle ellipticity ( Figure 7F , Figure 7—source data 3 ) . Our results show that spindles with well aligned chromosomes became obviously wider in 27 T ( p<0 . 005 ) , but the morphology of spindles with misaligned chromosomes was not much affected ( Figure 7F ) . We also noticed that spindles with well aligned chromosomes themselves are slightly wider in the sham control group compared to spindles with misaligned chromosomes ( Figure 7F ) , which may have their microtubules relative easier to be aligned by a vertical magnetic field . We propose that the SMF-induced spindle orientation and morphology changes are due to the combined alignment effects of both microtubules and chromosomes in the magnetic field ( Figure 8 ) . When the field is oriented normal to the substrate , magnetic torques on both microtubules and chromatin may combine to re-orient spindles away from the surface plane , opposing torques on astral microtubules that promote the normal orientation . Application of the magnetic field parallel to the coverslip allowed us to discriminate torques on chromatin vs . microtubules , and in this case it appears that torques on well aligned chromatin dominated , aligning spindles preferentially with their microtubules normal to the field , and their metaphase plate parallel to the field ( Figures 5 and 6 ) . Our result is opposite to that proposed by Denegre et al , who studied Xenopus egg cleavage in 16 . 7 T high magnetic fields and theoretically proposed that the orientation of spindle in magnetic field is a result of the balance between aster microtubules and spindle microtubules ( Denegre et al . , 1998 ) . They did not directly image spindles , but we note opposite results in the two systems might depend on huge size-scaling differences between the two systems we used . The cleavage plane in Xenopus eggs is oriented by huge microtubule asters with almost mm dimensions , while the amount of chromatin is the same as a mitotic cell . In contrast , the ratio of chromatin to microtubules is much larger in human somatic cells . Theoretical calculation predicted that the highly compacted mitotic chromosomes could be fully aligned by magnetic fields at around 1 . 4 T ( Maret , 1990 ) . Therefore the metaphase plate composed of chromosomes likely dominated the SMF-induced orientation in normal-sized somatic cells . 10 . 7554/eLife . 22911 . 035Figure 8 . Models show that ultra-high SMFs align microtubules and chromosomes to change spindle orientation and morphology . Blue upward arrows show magnetic field direction . Cells were plated on coverslips , which were placed in the ultra-high magnetic field either normal to or in parallel with the field direction . DOI: http://dx . doi . org/10 . 7554/eLife . 22911 . 035 This is the first reported study that investigated the mammalian cellular responses to ultra-high magnetic field of above 20 T . The torque of a substance is equal to the product of magnetic field intensity and the magnetic susceptibility of the object , in which the susceptibility could also be field-dependent and can be Taylor expanded . This means that the torque could be parabolically proportional to the magnetic field strength . Thus , a high field has much more severe impact than a low field as far as field-induced alignment is concerned , and the relationship is not linear . Although there are various potential concerns for the safety issue of the high magnetic fields , the known biological effects of high fields of above 10 T are limited . There are only a few studies that have investigated the animal or human cells at ≥10 T . Nakahara et al and Zhao et al showed that 10 T or 13 T SMFs did not have obvious effects on the human-hamster hybrid ( AL ) cells , Chinese Hamster Ovary ( CHO ) cells or human primary skin fibroblasts ( AG1522 ) cells ( Nakahara et al . , 2002; Zhao et al . , 2010 ) ; Denegre et al found that 16 . 7 T large gradient SMFs can affect the division of Xenopus eggs , which is presumed to act through astral microtubules and spindles ( Denegre et al . , 1998 ) ; the Shang group found that 12-16-12T large gradient SMFs can affect microtubule actin crosslinking factor 1 ( MACF1 ) expression and its association with cytoskeleton ( Qian et al . , 2009 ) as well as osteoblast ultrastructure and cell viability by disrupting collagen I or Febronectin/αβ1 integrin in human osteoblast-like cell lines ( Qian et al . , 2013 ) ; Valiron et al used 17 T , the highest SMF strength applied on cells so far and they did not observe strong cell killing effects ( Valiron et al . , 2005 ) . Instead , they found that the cytoskeleton of interphase mouse embryo fibroblast 3T3 cells and human cervical cancer HeLa cells can be affected by 17 T SMF . Here we found that 27 T SMF does not have an immediate cell killing effects on human CNE-2Z and RPE1 cells but it changes the mitotic spindle orientation and morphology . The spindle orientation change induced by ultra-high SMF is likely conserved between different cell types , although there may be variations among them . The spindles align with their long axis vertical to the field direction and their metaphase plate parallel to the field direction are different from our prediction . We thought the microtubules within the spindle would align with the magnetic field direction so that the spindle would also align with the field . However , it has been shown that the mitotic chromosomes have electromagnetic properties ( Zhao and Zhan , 2012 ) . Although the chromosomes alignment in the magnetic field has not been experimentally shown , calculation predicted that the chromosomes could be fully aligned with the magnetic field as long as the field is above 1 . 4 T ( Maret , 1990 ) . Based on theoretical calculation and experimental validation , Valles and his colleagues proposed that the direction of mitotic spindle alignment in the magnetic field depends on how the diamagnetic anisotropies of its individual components , including microtubules and chromosomes , align within it ( Valles , 2002; Valles et al . , 2002 ) . Denegre et al’s results about Xenopus egg division shows that the cleavage plane , which is perpendicular to the spindle long axis , tend to parallel with the magnetic field direction ( Denegre et al . , 1998 ) . Xenopus eggs have very different long astral microtubules , which composed a large portion of the whole cell . In contrast , the cells used in our study as well as most human somatic cells have much less astral microtubules and much bigger spindle microtubules compared to Xenopus eggs . Since both astral and spindle microtubules , as well as chromosomes respond to magnetic field , their relative contribution will be critical to determine the final outcome in a given cell type . In conclusion , although it is well accepted that microtubules can be affected by magnetic fields , the effect of high magnetic field on mitotic spindles in a cell has never been investigated . Here we examined the mitotic spindles in 27 T ultra-high SMF treated human CNE-2Z and RPE1 cells and found that the spindle orientation is indeed affected and the direction is dependent on the chromosome alignment . This provides a powerful tool to study spindle orientation related questions in developmental biology , such as cell fate , tissue architecture , and cancer biology . The 18 mm cell culture plates were custom made by Guangzhou Jet Bio-Filtration Company , China . The antibody for β-tubulin ( #HC101 ) was from Beijing TransGen Biotech . The Alexa Fluor 488 and 594 Phalloidin ( RRID: AB_2315147 and RRID: AB_2315633 ) , secondary antibodies for immunofluorescence and anti-fade ProLong Gold with DAPI were all from Invitrogen ( RRID: SCR_008410 ) . The 1 T magnets ( Neodymium magnet , N38 , dimension of 5 cm x 5 cm x 5 cm ) were from China Dafeng Zhongxin Permanent Magnet Material . The CENP-E inhibitor ( GSK923295 , # S7090 ) was from Selleck . FITC-Annexin V Apoptosis Detection Kit was from BD Pharmingen . The 27 T water-cooled magnet ( WM ) in Chinese High Magnetic Field Laboratory ( CHMFL , China ) facility ( WM#4 ) has a 32 mm diameter room temperature bore . Figure 1 shows the device that can fit the 27 T WM . The device consists of two coaxial non-magnetic stainless steel tubes . The outer diameter of the outer tube ( OT ) is 30 mm and the inner diameter of the inner tube ( IT ) is 22 . 4 mm . To investigate the biological effect of the magnetic field provided by The WM4 ( Figure 1A ) , we designed ( Figure 1B ) and constructed a set of biological sample incubation system ( Figure 1C , and Figure 1—figure supplement 1 ) , with accurate temperature , gas and humidity control . A non-magnetic stainless steel tube with 10 mm outer diameter was used as a shaft and was inserted into the inner space of the IT . We have used a rubber O-ring to hold the shaft and sealed the inner space of the IT . A sample house and a Teflon concentricity spacer were fixed on the shaft . The concentricity spacer was used to make sure that the shaft and the sample house are coaxial with the IT . Because the shaft was held by the rubber O-ring , it can be moved on axial easily to adjust the position of the sample house in the WM . A PT100 near the samples was used as a temperature sensor and connected to a temperature display to monitor the temperature of the samples . The temperature of the samples can be controlled by thermal conduction from the temperature controlled water , which flow through the space between the IT and OT . By adjusting the temperature of the water , the temperature of the samples can be controlled precisely . To adjust the atmosphere of the sample house , the air with 5% CO2 with temperature and humidity control was introduced by the shaft . To control for possible effect of the incubation system we made two identical sets , one for ‘sham’ and the other for the experimental 27 T SMF exposure group ( Figure 1C ) . The experimental group was placed in the tube and inserted into the WM4 ( Figure 1D ) , while the sham group was placed in the other identical tube and left outside of the magnet . The control group was kept in the regular full-size CO2 cell incubator in the lab . Although the maximal magnetic field intensity that WM4 can reach is 27 . 5 T , here we used 27 T to ensure the stability during the whole experiment process ( Figure 1E ) . The biological sample incubation system we constructed can hold samples up to 18 mm in diameter and incubate at temperature range of 4–100 degree . We custom made small plates to fit in this biological sample incubation system . Most experiments in this study were carried out in these custom made 18 mm plates , including full size incubator control , sham control and the 27 T experimental groups . Since the gas , humidity and temperature can all be well controlled , the platform we built is suitable to study a wide range of biological samples , such as various cell cultures , small model animals such as fruit flies , C elegans , zebrafish and mouse tissues . CNE-2Z cells ( RRID:CVCL_6890 ) were cultured in RPMI-1640 ( #10–040-CVR , CORNING Life Sciences ) supplemented with 10% FBS and 1% P/S ( penicillin/streptomycin ) . RPE1 cells ( RRID: CVCL_4388 ) were maintained in DMEM ( #10–017-CV , CORNING Life Sciences ) supplemented with 10% FBS and 1% P/S . Both cells were maintained in cell incubator with 5% CO2 , at 37°C . Both cell lines were from ATCC and have been confirmed by STR profiling . Mycoplasma has been tested to be negative . CNE-2Z and RPE1 cells were placed on coverslips in 18 mm , 24-well plates or 35 mm cell culture plates and treated with 27 T , 9 T , 1 T or 0 . 05 T magnetic fields . Cells were washed once with PBS and fixed by 4% ( vol/vol ) formaldehyde at room temperature for 20 min . Then the coverslips were washed with TBS-Tx ( TBS supplemented with 0 . 1% Triton X-100 ) and blocked by AbDil-Tx ( TBS-Tx with 2% ( wt/vol ) BSA and 0 . 05% sodium azide ) at room temperature for at least 30 min . Coverslips were stained with anti-β-tubulin antibody at room temperature for 2 hr , followed by fluorescently conjugated secondary antibodies at room temperature for 1 hr . Then the cells were directly stained with fluorescently labeled Phalloidin for 1 hr at room temperature . After washing with TBS-Tx , coverslips were mounted in anti-fade ProLong Gold mounting medium with DAPI ( Invitrogen ) . The antibodies and reagents used in immunofluorescence experiments include β-tubulin ( used at 1:1000 dilution ) , the secondary fluorescently conjugated antibodies ( used at 1:250 dilution ) , as well as Alexa 488 or 594 Phalloidin ( used at 1:40 dilution ) . All attached and floating cells were collected for cell counting , apoptosis and cell cycle analysis . Bright field images were taken before the cells were harvested by trypsinization . An aliquot of the cells were counted by hemocytometer and the rest cells were used for flow cytometry analysis ( cell death and cell cycle ) . Most experiments were repeated for at least three independent times by two researchers . The results were gathered together for analysis . Cells were trypsinized and washed three times with PBS before they were fixed in 70% ice-cold ethanol overnight at 4°C . Then they were washed with PBS again , and incubated in PI ( propidium iodide ) solution ( BD Pharmingen ) for 30 min in the dark at room temperature . Samples were then analyzed on a BD Flow Cytometry ( RRID: SCR_013311 ) ( BD bioscience , Calibur ) . 1 × 104 cells per sample were collected for each condition . Data were analyzed by ModFit LT . Experiments were done for at least three times . Cells were trypsinized and washed twice with ice-cold PBS before they were resuspended in binding buffer at 1 × 106 cells/ml . Then 100 μl of them was transferred to a 1 . 5 ml culture tube . FITC-Annexin V ( 5 μl ) and PI ( 5 μl ) were added to the tube , mixed , and incubated for 15 min at room temperature in the dark . Then 400 μl of binding buffer was added to the stained cells and mixed before they were analyzed by flow cytometry within 1 hr . Approximately 1 × 104 cells were collected by flow cytometer . Data were analyzed by FlowJo . Experiments were done for at least three times and representative results were shown in the figures . The bright field images were taken by a DSZ2000 microscope equipped with ISH300 3 . 0MP camera ( UOP ) . Most Immunofluorescence images shown in the figures were using a DeltaVision microscope ( GE Healthcare ) equipped with a 60× objective lens , 0 . 5 μm step size . Deconvolved images were projected into a single picture using Image J software ( RRID:SCR_003070 ) and maximum projection images are shown in the figures . Some low magnifications immunofluorescence images were taken by a Leica DMI4000B microscope ( RRID:SCR_000011 ) for spindle measurement . Since most high field MRIs for preclinical or research uses are applied in the head region , including the highest field MRI ( 21 . 1 T ) used in mice ( Schepkin et al . , 2010 ) , we chose a human nasopharyngeal carcinoma CNE-2Z cell line for its potential clinical relevance . Cells were plated on round or pre-cut coverslips one night before to allow the cells to attach . They were then exposed to different magnetic fields for indicated time points and then removed for analysis . Bright field microscopic and annexin V/PI double stain in flow cytometry assays were used to analyze apoptosis and necrosis For 0 . 05 T and 1 T magnetic field exposure perpendicular to coverslips , CNE-2Z cells on coverslips were exposed to magnetic fields for 4 hr , 3 days or 7 days in regular full-sized CO2 cell incubator ( Shanghai Boxun , BC-J160S ) that has accurate control of temperature ( 37°C ) , humidity and CO2 ( 5% ) . The cell plates were placed right on the top center of the magnets ( Neodymium magnet N38 , dimension: 5 cm x 5 cm x 5 cm ) and the magnetic field intensity measured by a Gauss meter ( LakeShore 475 DSP Gaussmeter ) showed the magnetic field intensity of 1 . 07 T ( 10700 Gs ) for the 1 T magnet . The sham control group was in the same incubator , around 30–40 cm away from the magnets , where the magnetic field is 0 . 925 Gs ( background magnetic field in the lab was measured to be 0 . 875 ± 0 . 171 Gs and in a separate CO2 cell incubator with no magnets was 0 . 875 ± 0 . 096 Gs ) . For 9 T magnetic field ( upward direction ) exposure perpendicular to coverslips , 2 ml of 5 × 105 cells /ml CNE-2Z cells for 4 hr assays , or 2 ml of 1 . 25 × 105 cells /ml CNE-2Z cells for 3 days assays were placed on coverslips in each 35 mm cell culture plates one night ahead to allow the cells to attach . On the second day , they were placed in 5% CO2 at 37°C under 9 T SMF for 4 hr ( 3 hr stable maintenance at 9 T with half hour increase and half hour decrease ) or 3 days before they were fixed and stained for analysis . For 27 T magnetic field exposure perpendicular to coverslips , 400 μl 6 . 25 × 105/ml CNE-2Z cells were placed on coverslips in each 18 mm cell culture plates one night ahead to allow the cells to attach . On the second day , they were placed in regular full-sized cell incubator ( control group ) or in two high magnetic field biological sample incubation systems . One was used as sham group and the other was 27 T group . Then the 27 T group was placed in the center of the 27 T water-cooled magnet ( WM4 ) and the whole cell culture plate was exposed homogeneously to the 27 T magnetic field for 4 hr in total ( increasing field for 30 min , constant 27 T field for 3 hr and reducing field for 30 min ) . For the 3 days post-exposure experiments , cells taken out of the magnet were then maintained in regular full-sized cell incubator for another 3 days before they were taken out and subjected to further analyses . The sham group was processed identically . For 1 T , 9 T or 27 T magnetic field exposure parallel to coverslips , coverslips were cut in half using a glass cutter and 1 × 107 CNE-2Z or RPE1 cells were plated in 100 mm plate one night ahead to allow the cells to attach . On the second day , the half coverslips were inserted vertically into the agarose gel pre-plated and solidified on the bottom of 35 mm ( 1 T and 9 T groups ) or 18 mm ( 27 T group ) plates . 1 ml ( 1 T and 9 T groups ) or 400 μl ( 27 T group ) of 1 . 5% agarose was used for each plate . They were placed in regular full-sized cell incubator ( control group ) , in 1 T , 9 T or in two high magnetic field biological sample incubation systems . One was sham group and the other was 27 T group . The rest of the experiment was identical to the above mentioned experiments . To measure the spindle angles with the coverslips ( lateral or non-lateral ) in Figures 3–4 , we used the Leica DMI4000B fluorescent microscope and observed the spindle poles under the 100 x objective lens . We consider the spindle lateral with the coverslips if the two poles of spindle apparatus were on one focal plane . When the two poles were not on one focal plane , we defined it as ‘non-lateral’ with the coverslips . To measure the spindle orientation or the spindle pole angles when they were exposed to 27 T magnetic fields that were parallel to the coverslips in Figures 5–7 , we used the Leica DMI4000B fluorescent microscope to get low magnification images . Then we used the Picpick software to measure the angles between spindle long axis and the magnetic field lines on these images for Figure 5 as well as the spindle pole angles in Figure 7C . The spindle length , width , chromosome distribution , metaphase plate width and the angles between spindle long axis and the magnetic field lines in Figures 6 and 7 were measured by Adobe Photoshop ( RRID:SCR_014199 ) . Spindles dimension and chromosome distributions were measured from four independent coverslips of CNE-2Z cells and four independent coverslips of RPE1 cells that were treated with RO-3306 and CENP-E inhibitors from two independent assays . To increase the sample size and ensure strong statistics , we applied a standard drug synchronization protocol by using RO-3306 to arrest cells in G2 ( Vassilev , 2006; Vassilev et al . , 2006 ) , then washout into MG132 to enrich cells in mitosis with intact spindles . We then collected many low magnification images to provide reliable statistics . RPE1 and CNE-2Z cells were plated onto the pre-cut coverslips 24 hr prior to the assay . Cells were then treated with RO-3306 ( 10 μM ) for 21 hr to arrest cells at late G2 phase . After washing with warm PBS for three times , cells were subsequently treated with MG-132 ( 20 μM ) or CENP-E inhibitor ( GSK923295 , 1 μM ) for another 6 hr to arrest cells in metaphase or to produce misaligned chromosomes . Magnetic field was applied during the last 4 hr . For quantifications in the manuscript , cells and spindles were counted for each condition from at least three to four independent coverslips or cell culture plates . Comparisons between treatments were analyzed by a two-tailed Student t test in GraphPad Prism software ( RRID:SCR_002798 ) . P values are labeled in the figures for where data were compared .	Nowadays , a number of methods can be used to ‘look’ inside the body to investigate potential health problems . One of these is a technique called magnetic resonance imaging ( MRI ) that uses magnetic fields that are several hundred times stronger than a fridge magnet ( or over 10 , 000 times stronger than the Earth’s natural magnetic field ) to generate images of the inside of the body . In general , stronger magnetic fields enable higher quality images to be obtained . However , the effects of exposing the body’s cells to these magnetic fields have not been fully determined . Like most other biological materials , protein polymers called microtubules can respond to high magnetic fields – for example , by aligning with the field . Microtubules play a number of roles inside cells . This includes forming the mitotic spindle that separates copies of chromosomes – the structures in which the majority of a cell’s genetic material is stored – equally between dividing cells . The orientation of the mitotic spindle determines the direction in which a cell will divide . This direction is important for generating different types of cells and tissues . Furthermore , many cancerous cells have incorrectly oriented spindles . Zhang , Hou et al . have now exposed cancerous and normal human cells to magnetic fields of varying strengths . The maximum magnetic field strength tested ( 27 Tesla – or around 10 times the highest field strengths produced by standard hospital MRI scanners ) did not kill the cells after four hours of exposure , but the orientation of the spindles inside the cells did change . In addition , the 27 Tesla magnetic field caused spindles that were perpendicular to the direction of the field to widen . At an intermediate field strength ( 9 Tesla – a magnetic field strength that has been used in some experimental MRI scanners ) , the orientation of the spindle only changed after three days of continuous exposure to the magnetic field . Lower field strengths ( such as those currently used in hospital MRI scanners ) did not alter the orientation of the spindle even after seven days of exposure . Zhang , Hou et al . also observed that the magnetic field acts on both the microtubules and chromosomes . However , the alignment of the chromosomes in the cell was the greatest determinant of the direction in which the spindle would align itself in response to the magnetic field . The next step is to analyze the consequences of magnetic field-induced spindle orientation changes – can these lead to cancer or reduce cancer growth , or change how animal tissues develop ? Understanding how to control the position of the spindle could also ultimately make it possible to use ultra-high magnetic fields to engineer tissues or stimulate their regeneration .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "physics", "of", "living", "systems" ]	2017	27 T ultra-high static magnetic field changes orientation and morphology of mitotic spindles in human cells
Pyomyositis is a severe bacterial infection of skeletal muscle , commonly affecting children in tropical regions , predominantly caused by Staphylococcus aureus . To understand the contribution of bacterial genomic factors to pyomyositis , we conducted a genome-wide association study of S . aureus cultured from 101 children with pyomyositis and 417 children with asymptomatic nasal carriage attending the Angkor Hospital for Children , Cambodia . We found a strong relationship between bacterial genetic variation and pyomyositis , with estimated heritability 63 . 8% ( 95% CI 49 . 2–78 . 4% ) . The presence of the Panton–Valentine leucocidin ( PVL ) locus increased the odds of pyomyositis 130-fold ( p=10-17 . 9 ) . The signal of association mapped both to the PVL-coding sequence and to the sequence immediately upstream . Together these regions explained over 99 . 9% of heritability ( 95% CI 93 . 5–100% ) . Our results establish staphylococcal pyomyositis , like tetanus and diphtheria , as critically dependent on a single toxin and demonstrate the potential for association studies to identify specific bacterial genes promoting severe human disease . Microbial genome sequencing and bacterial genome-wide association studies ( GWAS ) present new opportunities to discover bacterial genes involved in the pathogenesis of serious infections ( Sheppard et al . , 2013; Chewapreecha et al . , 2014; Earle et al . , 2016; Lees et al . , 2016; Falush , 2016; Power et al . , 2017 ) . Pyomyositis is a severe infection of skeletal muscle most commonly seen in children in the tropics ( Chauhan et al . , 2004; Verma , 2016; Bickels et al . , 2002 ) . In up to 90% of cases , it is caused by a single bacterial pathogen , Staphylococcus aureus ( S . aureus ) ( Chauhan et al . , 2004; Verma , 2016; Bickels et al . , 2002; Moriarty et al . , 2015 ) . Unlike infections of the skin and superficial soft tissues , the skin and subcutaneous tissues are not usually involved in pyomyositis , by contrast to intense inflammation in the infected muscles ( Chauhan et al . , 2004; Verma , 2016 ) . Pyomyositis is thought to arise from haematogenous seeding of bacteria from blood to muscle ( Verma , 2016 ) . There is evidence that some S . aureus strains have heightened propensity to cause pyomyositis – the incidence in the USA doubled during an epidemic of community-associated methicillin resistant S . aureus ( CA-MRSA ) ( Pannaraj et al . , 2006 ) – but molecular genetic investigation of S . aureus from pyomyositis has been limited ( Borges et al . , 2012 ) . Panton–Valentine leucocidin ( PVL ) , a well-known staphylococcal toxin causing purulent skin infections and found in epidemics caused by CA-MRSA , has been implicated in pyomyositis , pneumonia and other S . aureus disease manifestations , but its role in these invasive infections is disputed ( Shallcross et al . , 2013; Vandenesch et al . , 2003; Labandeira-Rey et al . , 2007; Villaruz et al . , 2009 ) . PVL is a bipartite pore-forming toxin comprising the co-expressed LukF-PV and LukS-PV proteins ( Löffler et al . , 2010; Boakes et al . , 2011 ) . The coding sequence for PVL , lukSF-PV , is usually carried on bacteriophages , ( Shallcross et al . , 2013; Löffler et al . , 2010 ) , which facilitate lukSF-PV exchange between lineages ( McCarthy et al . , 2012 ) . The mechanism of PVL toxicity has been shown to involve cell lysis in human myeloid cells , particularly neutrophils , by insertion into the cellular membrane , ( Niemann et al . , 2018 ) leading the tissue to release inflammatory neutrophil products ( Spaan et al . , 2013 ) . Neutrophil lysis is mediated by PVL binding to target complement receptors C5aR; in binding , PVL has both toxic and immunomodulatory effects , as it also inhibits C5a mediated immune activation ( Sina et al . , 2013 ) . Although small case series testing for candidate genes have reported a high prevalence of PVL among pyomyositis-causing S . aureus , ( Pannaraj et al . , 2006; Sina et al . , 2013; García et al . , 2013 ) a detailed meta-analysis found no evidence for an increased rate of musculoskeletal infection ( or other invasive disease ) in PVL-positive bacteria versus controls ( Shallcross et al . , 2013 ) . These conflicting results may reflect insufficiently powered studies , and some case series lack comparative control strains ( Sina et al . , 2013 ) . A further problem with the use of candidate gene studies in studying pathogenesis is that they may miss important variation elsewhere in the genome . One study reporting a critical role for PVL in the causation of severe pneumonia ( Labandeira-Rey et al . , 2007 ) was later found to have overlooked mutations in key regulatory genes , capable of producing the virulence that had been attributed to PVL by the original study ( Villaruz et al . , 2009 ) . Thus , while some evidence suggests an association between pyomyositis and PVL , there remains significant uncertainty regarding the bacterial genetic predisposition of S . aureus to pyomyositis , and whether PVL is an important virulence factor , or merely an epiphenomenon , carried by bacteria alongside unidentified genetic determinants ( Otto , 2011; Day , 2013 ) . GWAS offer a means to screen entire bacterial genomes to discover genes and genetic variants associated with disease risk . They are particularly appealing because they enable the investigation of traits not readily studied in the laboratory , and do not require the nomination of specific candidate genes ( Falush , 2016 ) . Proof-of-principle GWAS in bacteria have demonstrated the successful rediscovery of known antimicrobial resistance ( AMR ) determinants ( Chewapreecha et al . , 2014; Earle et al . , 2016; Lees et al . , 2016 ) . However , AMR is under extraordinarily intense selection in bacteria . More subtle traits , including host specificity ( Sheppard et al . , 2013 ) and the duration of pneumococcal carriage , ( Lees et al . , 2017 ) have also been demonstrated using GWAS . Promising results for GWAS in human infecting bacteria include identifying possible loci for invasive infection with Streptococcus pyogenes ( Lees et al . , 2016 ) and Staphylococcus epidermidis ( Méric et al . , 2018 ) , and the identification of virulence-associated genes corresponding with regional differences in disease manifestations of melioidosis ( Chewapreecha et al . , 2017 ) . Within species , lineage-specific variants have been shown to predict mortality following S . aureus bacteraemia ( Recker et al . , 2017 ) . These studies support the potential GWAS has to precisely pinpoint genes and genetic variants underlying the propensity to cause specific human infections , making it a promising tool to investigate the possible contribution of bacterial genomic variation to pyomyositis . To understand the bacterial genetic basis of pyomyositis , we sampled and whole-genome sequenced S . aureus from 101 pyomyositis infections and 417 asymptomatic nasal carriage episodes in 518 children attending Angkor Hospital for Children in Siem Reap , Cambodia between 2008 and 2012 ( Supplementary file 1 ) . As expected , we observed representatives of multiple globally common lineages in Cambodia , together with some globally less common lineages at high frequency , in particular clonal complex ( CC ) 121 , identified by multi-locus sequence typing ( MLST ) . There were no major changes in lineage frequency over time ( Figure 1—figure supplement 1 ) . In our study , some S . aureus lineages were strongly overrepresented among cases of pyomyositis compared with asymptomatic , nasally-carried controls over the same time period . Notably , 86/101 ( 85% ) of pyomyositis cases were caused by CC-121 bacteria , whereas no pyomyositis cases were caused by the next two most commonly carried lineages , sequence type ( ST ) −834 and CC-45 ( Figure 1 ) . We estimated the overall heritability of case/control status to be 63 . 8% ( 95% CI 49 . 2–78 . 4% ) in the sample , reflecting the strong relationship between bacterial genetic variation and case/control status . We used bugwas ( Power et al . , 2017 ) to decompose this heritability into the principal components ( PCs ) of bacterial genetic variation . PC 1 , which distinguished CC-121 ( the most common pyomyositis lineage ) from ST-834 ( which was only found in carriage ) , showed the strongest association with case/control status ( p=10-29 . 6 , Wald test ) . The strongest association was with PC 20 , which differentiated a sub-lineage of CC-121 within which no cases were seen ( p=10-13 . 9 ) , and PC 2 , which distinguished CC-45 from the rest of the species ( p=10-4 . 9 ) . We conducted a GWAS to identify bacterial genetic variants associated with pyomyositis , controlling for differences in pyomyositis prevalence between S . aureus lineages . We used a kmer-based approach ( Sheppard et al . , 2013 ) in which every variably present 31 bp DNA sequence observed among the 518 genomes was tested for association with pyomyositis versus asymptomatic nasal carriage , controlling for population structure using GEMMA ( Zhou and Stephens , 2012 ) . These kmers captured bacterial genetic variation including single nucleotide polymorphisms ( SNPs ) , insertions or deletions ( indels ) , and presence or absence of entire accessory genes . We found 10 . 7 million unique kmers variably present across the bacterial genomes . In total , 9175 kmers were significantly associated with case/control status after correction for multiple testing ( 10-6 . 8 ≤ p ≤ 10-21 . 4; Figure 2A ) . When mapped to the de novo assembly of a CC121 isolate from pyomyositis ( PYO2134 ) , the vast majority of these kmers ( 9 , 074/9 , 175; 98 . 9% ) localised to a 45 . 7 kb region spanning an integrated prophage with 95% nucleotide identity to φSLT ( Figure 2B ) . Most ( 9 , 173/9 , 175; 99 . 98% ) significant kmers were found at an increased frequently in pyomyositis , with odds ratios ( OR ) ranging from 2 . 7 to 139 . 8 , indicating that the presence of each was associated with increased risk of disease . The presence of bacteriophage φSLT was thus strongly associated with pyomyositis . We were able to localise the most statistically significant signal of association to kmers that mapped within φSLT to the lukS-PV and lukF-PV cargo genes . These genes encode the subunits of PVL , which multimerise into a pore-forming toxin capable of rapidly lysing the membranes of human neutrophils ( Löffler et al . , 2010; Otto , 2011 ) . 1630 kmers tagging the presence of the lukSF-PV coding sequences ( CDS ) were highly significantly associated with disease , being present in 98/101 ( 97% ) pyomyositis cases and 84/417 ( 20% ) carriage controls ( unadjusted OR 129 . 5 , p=10-17 . 9 ) . Kmers tagging variation in the 389 bp region immediately upstream of the CDS were also strongly associated with disease ( p=10-21 . 4 ) . The most significant of these kmers were co-present with the CDS in the same cases ( 98/101 , 97 . 0% ) , but present in fewer controls ( 79/417 , 18 . 9% ) , producing an OR of 140 . Closer examination of this ~400 bp upstream region in genomes assembled from short-read Illumina sequencing showed that assembly of the region was problematic , with breaks or gaps in the assembly ( Figure 2—figure supplement 2 ) . To improve the accuracy of this region of the assembled genomes we performed long-read Oxford Nanopore sequencing on the 37 genomes with incomplete or discontinuous assembly upstream of the PVL CDS . By integrating long-read and short-read data we were able to assemble a single contig spanning this region in all isolates ( Figure 2—figure supplement 2 ) . When these improved assemblies were introduced , the signal of association upstream of the PVL CDS was no more significant than within the CDS ( Figure 2C ) . Therefore , the presence of genomic sequence spanning the PVL toxin-coding sequences and the upstream , presumed regulatory , region exhibited the strongest association with pyomyositis in the S . aureus genome . All isolates with kmers mapping to the PVL CDS had 98% or more coverage for the PVL CDS genes in de novo assembly ( Figure 2—figure supplement 3 ) . The signal of association in the earlier and later periods of the study were examined and found to be consistent ( Figure 2—figure supplement 4 ) . Out of 9175 kmers significantly associated with pyomyositis , we only found 101 kmers related to regions outside the PVL-carrying prophage ( Figure 2A , Supplementary file 2 ) . Two kmers mapping at a position near 0 . 2 Mb in the PYO2014 reference genome showed homology to platelet adhesin sraP by BLAST . Thirty-five kmers mapping to a 50 bp non-coding fragment at 0 . 6 Mb and two kmers mapping to 2 . 8 Mb showed homology to an MSSA476 intergenic sequence between adhesin-encoding sdrC and sdrD by BLAST . One kmer mapping to position 2 . 0 Mb showed no sequence homology by BLAST . Sixty-one kmers mapping to a 61 bp non-coding region at 2 . 7 Mb showed homology to an MSSA476 intergenic sequence between acetyltransferase-encoding genes SAS2453 and SAS2454 by BLAST . In conclusion , these other signals were short , fragmentary and mostly non-coding so we did not investigate them further . The presence of high-risk kmers mapping to the PVL region explained the vast majority of observed heritability . When the presence or absence of the most significant kmer pattern , a set of kmers with an identical pattern of presence in the population , all of which mapped to the PVL region , was included as a covariate in GEMMA , the remaining heritability not explained by other factors was estimated and found to be 0 . 0% ( 95% CI 0–2 . 5% ) . Thus , the point estimate for heritability ( not explained by the inclusion of PVL-tagging kmers ) is reduced by 100% ( 95% CI 93 . 5–100% ) , meaning we have little evidence for any remaining heritability in case/control status . Presence or absence of the PVL region accounted for the differences in pyomyositis rates between lineages . It was common in pyomyositis-associated lineages including CC-121 and absent from carriage/non-pyomyositis-associated lineages including ST-834 and CC-45 ( Figure 1 ) , explaining over 99 . 9% of observed heritability in case-control status . It was infrequent in the non-pyomyositis-associated sub-lineage of CC-121 ( 2/36 , 5 . 6% ) , and sporadically present in pyomyositis cases in otherwise non-pyomyositis-associated , PVL-negative strains CC-1 and CC-88 . Its absence from only three cases ( in lineages CC-88 , CC-1 and CC-121 ) suggested that the PVL region approached necessity for development of pyomyositis in the current setting in Cambodian children , while its presence in 20% of controls indicated that PVL-associated pyomyositis is incompletely penetrant , that is presence of the PVL region does not always lead to disease . PVL genes were carried on multiple genetic backgrounds in this population , and the phage backgrounds vary by clonal complex . We examined all assemblies for sequence similarity to six known bacteriophages that carry the PVL genes , ( Boakes et al . , 2011 ) as well as the 45 . 7 kb region in PYO2134 , a hypothesised integrated prophage identified in the reference genome prepared for this study , which we have called φCC121 ( Figure 2—figure supplement 3 ) . The finding of PVL genes on BLAST corresponded completely with the presence of kmers mapping across the PVL locus . We find sequences with >95% homology to four of these seven phages in the population . Regions in some assemblies showed homology for multiple phages , reflecting the similarity between jSLT , jSa2USA and jCC121 rather than the presence of multiple phages , and resolution of phages was limited by fragmented assemblies from short reads ( Supplementary File 2B ) . Phage types were restricted within most lineages , with jSLT found in ST-3206 , jSa2USA in ST-1232 and jPVL in CC-1 . jCC121 was the phage most often identified in the dominant pyomyositis strain CC-121 , but it was absent in all but one isolate from the low risk subclade within CC-121 . Strikingly , PVL-negative isolates in CC-121 and ST-834 strains frequently retained sequence homologous to >95% of jSLT , suggesting that some PVL-negative CC-121 isolates lost PVL secondarily by gene deletion rather than prophage excision . In this study , we found a strong association between pyomyositis , a highly distinctive tropical infection of skeletal muscle in children , and Panton-Valentine leukocidin , a bacterial toxin commonly carried by bacteriophages . We found that a single coding region together with the upstream sequence are all but necessary for the development of pyomyositis: its sporadic presence is associated with pyomyositis in otherwise low-frequency strains , and its absence is associated with asymptomatic carriage in a high-propensity strain . PVL appears to be carried on multiple phage backgrounds in this population , but PVL-positive lineages generally carry a single phage type , as expected given the observed strain restriction of phages in S . aureus ( Xia and Wolz , 2014; Stegger et al . , 2014 ) . The locally common PVL-positive CC-121 lineage contributes most strongly to the prevalence of pyomyositis in Cambodian children . While PVL has long been thought an important S . aureus virulence factor , ( Diep et al . , 2008; Bocchini et al . , 2006; Kurt et al . , 2013 ) its role in invasive disease has been controversial , ( Otto , 2011; Day , 2013 ) with conflicting results in case-control studies and an absence of supporting evidence on meta-analysis ( Shallcross et al . , 2013 ) . In previous studies , the PVL-positive USA300 lineage was associated with musculoskeletal infection ( both pyomyositis and osteomyelitis ) , however in these studies almost all such infections were caused by the USA300 strain , so the role of PVL was almost completely confounded by both methicillin resistance and strain background ( Pannaraj et al . , 2006; Bocchini et al . , 2006 ) . In our study , this confounding influence is broken down by the movement of PVL on mobile genetic elements ( MGEs ) . Despite the emergence of CA-MRSA in carriage in Cambodia ( Nickerson et al . , 2011 ) , all the pyomyositis cases were MSSA ( Figure 1—figure supplement 1 ) . By applying a GWAS method to a well-powered cohort , our study resolves the controversy around pyomyositis and PVL , demonstrating strong heritability which localises to a single region , even when the full bacterial genome is considered . Bacterial GWAS can pinpoint virulence variants when MGEs act to unravel linkage disequilibrium , if effect sizes are sufficiently strong . There is strong biological plausibility for the association demonstrated in this study . PVL is a well characterised S . aureus toxin , toxic to the myeloid cells that form a first line of defence against bacterial infection , ( Oliveira et al . , 2018 ) and , in binding to myeloid cells by a complement receptor ( C5aR ) , exerts immunomodulatory effects ( Spaan et al . , 2013 ) . The establishment of muscle abscesses is a critical step in the pathogenesis of pyomyositis , but unlike renal , hepatic and splenic abscesses , skeletal muscle abscesses are rarely observed in experimental models of bacteraemia , unless there is preceding muscle trauma ( Miyake , 1904 ) . S . aureus strains containing PVL show increased duration of bacteraemia in a rabbit model of sepsis , ( Xia and Wolz , 2014 ) and result in larger muscle abscesses ( Tseng et al . , 2009 ) . PVL has been found strongly bound to necrotic muscle in an individual with myositis associated with necrotizing fasciitis ( Lehman et al . , 2010 ) . These observations support the hypothesis that PVL may facilitate bacterial seeding to muscles via the bloodstream and tropism for muscular infection . These results establish that , for children in Cambodia , staphylococcal pyomyositis is a disease whose pathogenesis depends crucially on a single toxin . This property is shared by toxin-driven , vaccine-preventable diseases such as tetanus and diphtheria . Therefore , vaccines that generate neutralising anti-toxin antibodies against PVL ( Landrum et al . , 2017 ) might protect human populations specifically against this common tropical disease . These results also raise the hypothesis that antibiotics which decrease toxin expression , and have been recommended in some PVL-associated infections , ( Saeed et al . , 2018 ) may offer specific clinical benefit in treating pyomyositis . More generally , our study provides an example of how microbial GWAS can be used to elucidate the pathogenesis of bacterial infections and identify specific virulence genes associated with human disease . We retrospectively identified pyomyositis cases from the Angkor Hospital for Children in Siem Reap , Cambodia , between January 2007 and November 2011 . We screened all attendances in children ( ≤16 years ) using clinical coding ( ICD-10 code M60 ( myositis ) ) and isolation of S . aureus from skeletal muscle abscess pus . We reviewed clinical notes to confirm a clinical diagnosis of pyomyositis was made by the medical staff , and bacterial strains cultured by routine clinical microbiology laboratory were retrieved from the local microbiology biobank . 106 clinical episodes of pyomyositis were identified , in 101 individuals , and we included the earliest episode from each individual . We identified S . aureus nasal colonisation from two cohort studies undertaken at Angkor Hospital for Children . The first were selected from a collection characterising nasal colonization in the region between September and October , 2008 , which has previously been described using multi-locus sequence typing ( Nickerson et al . , 2011 ) . The swabs had been saved at −80°C since the study , these samples were re-examined for the presence of S . aureus using selective agar , confirmed using Staphaurex ( Remel , Lenexa , USA ) and the DNAse agar test ( Oxoid , Hampshire , UK ) . Antimicrobial susceptibility testing was performed according to the 2014 Clinical and Laboratory Standards Institute guidelines ( M100-24 ) ( CLSI , 2014 ) . We undertook a second cohort study in 2012 . Inclusion criteria were children ( ≤16 years ) attending as an outpatient at Angkor Hospital for Children with informed consent . There were no exclusion criteria . Children were swabbed between the 2-7th July 2012 , using sterile cotton tipped swabs pre-moistened ( with phosphate buffered saline , PBS ) using three full rotations of the swab within the anterior portion of each nostril with one swab being used for both nostrils , the ends were broken into bottles containing sterile PBS and kept cool until plated in the laboratory ( within the hour ) . The swabs were plated onto Mannitol Salt agar to select for S . aureus . The M100-24 CLSI ( CLSI , 2014 ) standards were followed for susceptibility testing and bacteria stored in tryptone soya broth and glycerol at −80°C . We selected controls from carriers in these two cohorts using the excel randomization function: 222 of 519 from the 2008 cohort and 195 of 261 from the 2012 cohort . Approval for this study was provided by the AHC institutional review board and the Oxford Tropical Ethics Committee ( 507-12 ) . For each bacterial culture , a single colony was sub-cultured and DNA was extracted from the sub-cultured plate using a mechanical lysis step ( FastPrep; MPBiomedicals , Santa Ana , CA ) followed by a commercial kit ( QuickGene; Autogen Inc , Holliston , MA ) , and sequenced at the Wellcome Centre for Human Genetics , Oxford on the Illumina ( San Diego , California , USA ) HiSeq 2500 platform , with paired-end reads 150 base pairs long . A subset of samples were sequenced using long-read sequencing technology . We selected 37 isolates with incomplete assembly upstream of the PVL locus , 22 with ambiguous base calls in the assembly , and 15 where the region was assembled over two contigs . DNA was extracted using Genomic Tip 100/G ( Qiagen , Manchester , UK ) and DNA libraries prepared using Oxford Nanopore Technologies ( ONT ) SQK-LSK108 library kit ( ONT , Oxford , UK ) according to manufacturer instructions . These were then sequenced on ONT GridION device integrated with a FLO-MIN106 flow cell ( ONT , Oxford , UK ) . ONT base calling was performed using Guppy v . 1 . 6 . For short-read sequencing , we used Velvet ( Zerbino and Birney , 2008 ) v1 . 0 . 18 to assemble reads into contigs de novo . Velvet Optimiser v2 . 1 . 7 was used to choose the kmer lengths on a per sequence basis . The median kmer length was 123 bp ( IQR 119–123 ) . To obtain multilocus sequence types we used BLAST ( Altschul et al . , 1990 ) to find the relevant loci , and looked up the nucleotide sequences in the online database at http://saureus . mlst . net/ . Strains that shared 6 of 7 MLST loci were considered to be in the same Clonal Complex ( Feil et al . , 2003 ) . Antibiotic sensitivity was predicted by interrogating the assemblies for a panel of resistance determinants as previously described ( Gordon et al . , 2014 ) . We used Stampy ( Lunter and Goodson , 2011 ) v1 . 0 . 22 to map reads against reference genomes ( USA300_FPR3757 , Genbank accession number CP000255 . 1 ) ( Diep et al . , 2006 ) . Repetitive regions , defined by BLAST ( Altschul et al . , 1990 ) comparison of the reference genome against itself , were masked prior to variant calling . Bases were called at each position using previously described quality filters ( Didelot et al . , 2012; Young et al . , 2012; Golubchik et al . , 2013 ) . After filtering ONT reads with filtlong v . 0 . 2 . 0 ( with settings filtlong -- min_length 1000 --keep_percent 90 --target_bases 500000000 --trim --split 500 ) , hybrid assembly of short ( Illumina ) and long ( ONT ) reads were made , using Unicycler v0 . 4 . 5 ( Wick et al . , 2017 ) ( default settings ) . The workflow for these assemblies is available at https://gitlab . com/ModernisingMedicalMicrobiology/MOHAWK ) We constructed a maximum likelihood phylogeny of mapped genomes for visualization using RAxML ( Stamatakis , 2014 ) assuming a general time reversible ( GTR ) model . To overcome a limitation in the presence of divergent sequences whereby RAxML fixes a minimum branch length that may be longer than a single substitution event , we fine-tuned the estimates of branch lengths using ClonalFrameML ( Didelot and Wilson , 2015 ) . We used a kmer-based approach to capture non-SNP variation ( Sheppard et al . , 2013 ) . Using the de novo assembled genome , all unique 31 base haplotypes were counted using dsk ( Rizk et al . , 2013 ) . If a kmer was found in the assembly , it was counted present for that genome , otherwise it was treated as absent . This produced a set of 10 , 744 , 013 variably present kmers , with the presence or absence of each determined per isolate . We identified a median of 2 , 801 , 000 kmers per isolate , including variably present kmers and kmers common to all genomes ( IQR 2 , 778 , 000–2 , 837 , 000 ) . We used the Genome-wide Efficient Mixed Model Association tool ( GEMMA ( Zhou and Stephens , 2012 ) ) to fit a univariate linear mixed model for association between a single phenotype ( pyomyositis vs asymptomatic nasal carriage ) . We calculated the relatedness matrix from kmers , and used GEMMA to estimate the proportion of variance in phenotypes explained by genotypic diversity in the sample set ( i . e . estimated heritability ) . Heritability estimates with and without a covariate ( e . g . the presence of high risk kmers ) are compared by testing for difference in proportions . We use the point estimate for heritability as the denominator to calculate the relative decrease proportion . We performed association testing using an R package bacterialGWAS ( https://github . com/jessiewu/bacterialGWAS ) , which implements a published method for locus testing in bacterial GWAS ( Earle et al . , 2016 ) . The association of each kmer on the phenotype was tested , controlling for the population structure and genetic background using a linear mixed model ( LMM ) implemented in GEMMA ( García et al . , 2013 ) . The parameters of the linear mixed model were estimated by maximum likelihood and likelihood ratio test was performed against the null hypothesis ( that each locus has no effect ) using the software GEMMA ( García et al . , 2013 ) . GEMMA was run using a minor allele frequency of 0 to include all SNPs . GEMMA was modified to output the ML log-likelihood under the null , and alternative and –log10 p values were calculated using R scripts in the bacterialGWAS package . Unadjusted odds ratios were reported because there was no residual heritability unexplained by the most significant kmers . To address the possibility of differing effect sizes between the two control cohorts , we have repeated the analysis after splitting the study into two groups – early ( 2008 and earlier , n = 276 , cases n = 54 , controls n = 222 ) and late ( 2009 and later , n = 242 , cases n = 47 , controls n = 195 ) . We then examined the maximum likelihood estimates produced by the LMM for kmers mapping to the PVL coding sequence in each region . The 95% CI of the estimate from each sub study were overlapping ( See Figure 2—figure supplement 4 ) . We tested for associations between lineage and phenotype using an R package bugwas ( available at https://github . com/sgearle/bugwas ) , which implements a published method for lineage testing in bacterial GWAS ( Earle et al . , 2016 ) . We tested lineages using principal components . These were computed based on biallelic SNPs using the R function prcomp . To test the null hypothesis of no background effect of each principal component , we used a Wald test , which we compared against a χ ( Chewapreecha et al . , 2014 ) distribution with one degree of freedom to obtain a p value . We used Bowtie ( Langmead and Salzberg , 2012 ) to align all 31 bp kmers from short-read sequencing to a draft reference ( the de novo assembly of a CC-121 pyomyosits strain PYO2134 ) . Areas of homology between the draft reference and well-annotated reference strains were identified by aligning sequences with Mauve ( Darling et al . , 2004 ) . For all 31 bp kmers with significant association with case-controls status , the likely origin of the kmer was determined by nucleotide sequence BLAST ( Altschul et al . , 1990 ) of the kmers against a database of all S . aureus sequences in Genbank . 31 bp kmers were counted for the 37 hybrid short-read and long-read assemblies using dsk ( Rizk et al . , 2013 ) . The presence or absence of all Illumina ( short-read ) kmers that mapped to the two PVL toxin-coding sequences and the upstream intergenic region plus the surrounding 1 kb were reassessed . For the 37 samples with hybrid assemblies , the presence/absence of these kmers was determined from the kmers counted from the hybrid assemblies . For all other samples , presence/absence was determined from the kmers counted from the short-read only assemblies . The new presence/absence patterns were tested for association with the phenotype controlling for population structure and genetic background using GEMMA ( Zhou and Stephens , 2012 ) , using the same relatedness matrix as the original short-read analysis . We used BLAST to check for the relative coverage of the PVL CDS ( From reference genome USA300_FPR3757 ( CP000255 . 1 ) positions 1546170–1548350 ) , as well as the entire sequence of 6 known PVL positive phages ( φ2958 ( NC_011344 . 1 ) , φPVL ( NC_002321 . 1 ) , φPVL108 ( NC_008689 . 1 ) , φSLT ( NC_002661 . 2 ) , φSa2MW ( NC_003923 . 1 ) , φSa2USA ( CP000255 . 1 ) ) ( Boakes et al . , 2011 ) , as well as the hypothesised prophage region from PYO2134 ( 1571177–1616957 ) , which we have calledφCC121 . For PVL genes , we determined relative coverage of the query sequence; over 98% coverage was used as threshold for gene presence . Multiple testing was accounted for by applying a Bonferroni correction ( Dunn , 1959 ) ; the individual locus effect of a variant ( kmer or PC ) was considered significant if its P value was smaller than α/np , where we took α = 0 . 05 to be the genome-wide false-positive rate and np to be the number of kmers or PCs with unique phylogenetic patterns , that is , unique partitions of individuals according to allele membership . We identified 236627 unique kmer patterns and 518 PCs , giving thresholds of 2 . 1 × 10−7 and 9 . 7 × 10−5 respectively . Sequence data have been submitted to Short Read Archive ( Bioproject ID PRJNA418899 ) . Clinical origins of sequenced strains are listed in supplementary information ( Supplementary File 4 ) .	Certain bacteria that normally live on the skin or in the nose without causing problems can sometimes lead to diseases elsewhere in the body . For example , the bacterium Staphylococcus aureus can cause blood infections or a severe and painful infection of the muscle called pyomyositis , which is very common in children who live in the tropics . Scientists believe that pyomyositis happens when S . aureus bacteria in the blood stream infect the muscles . Some strains of this bacteria are more likely to cause such infections , but why is unclear . One potential cause is a toxin produced by some S . aureus bacteria called Panton-Valentine leucocidin ( PVL ) . So far , studies looking at whether PVL-producing bacteria are more likely to cause pyomyositis have had conflicting results . Now , Young et al . show that the gene for PVL is always present in S . aureus strains that cause pyomyosistis in Cambodian children , but is rarely found in S . aureus taken from the noses of their healthy counterparts . In the experiments , bacteria were collected from 101 children with pyomyositis and from the noses of 417 healthy children at the Angkor Hospital for Children in Cambodia over a 5-year period . The DNA in these bacteria were compared using very sensitive genetic techniques . The comparisons showed having the gene for PVL increased the odds of having pyomyositis 130-fold , showing that this one toxin likely accounts for much of the risk of developing this disease . If more studies confirm the link between PVL and pyomyositis , developing vaccines that block the gene for PVL might be one way to protect children in the tropics from developing this infection . Treating children with pyomyositis with antibiotics that reduce the production of the PVL toxin may also be helpful .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease", "genetics", "and", "genomics" ]	2019	Panton–Valentine leucocidin is the key determinant of Staphylococcus aureus pyomyositis in a bacterial GWAS
The mechanisms that regulate growth and size of the regenerating limb in tetrapods such as the Mexican axolotl are unknown . Upon the completion of the developmental stages of regeneration , when the regenerative organ known as the blastema completes patterning and differentiation , the limb regenerate is proportionally small in size . It then undergoes a phase of regeneration that we have called the ‘tiny-limb’ stage , which is defined by rapid growth until the regenerate reaches the proportionally appropriate size . In the current study we have characterized this growth and have found that signaling from the limb nerves is required for its maintenance . Using the regenerative assay known as the accessory limb model ( ALM ) , we have found that growth and size of the limb positively correlates with nerve abundance . We have additionally developed a new regenerative assay called the neural modified-ALM ( NM-ALM ) , which decouples the source of the nerves from the regenerating host environment . Using the NM-ALM we discovered that non-neural extrinsic factors from differently sized host animals do not play a prominent role in determining the size of the regenerating limb . We have also discovered that the regulation of limb size is not autonomously regulated by the limb nerves . Together , these observations show that the limb nerves provide essential cues to regulate ontogenetic allometric growth and the final size of the regenerating limb . It is estimated that over 2 . 1 million Americans are living with limb loss or limb difference , which has profound effects on the function , health , and quality of life in these patients ( Ziegler-Graham et al . , 2008 ) . The long-term goal of regenerative medicine is to replace or repair damaged limbs by inducing endogenous regenerative responses in humans . Studies in tetrapods capable of regenerating complete and functional limb structures , such as the Mexican axolotl ( Ambystoma mexicanum ) , have been invaluable in terms of understanding the basic underlying biology of limb regeneration and the mechanisms that control this process ( McCusker et al . , 2015 ) . Much research in the axolotl system has focused on the essential aspects of the initial stages of regeneration; how the regeneration permissive environment is established , how mature limb cells become regeneration competent , and how the unique pattern of the regenerated limb structures is generated . However , to date , very little is known about the later stages of regeneration that are required for the regenerating structure to mature into a fully functional limb . One key occurrence at the later stages of limb regeneration is the growth of the limb regenerate to the size that is proportionally appropriate to the rest of the animal . When the limb is amputated , a transient regenerative organ known as the blastema develops at the site of injury and contains limb progenitor cells that will grow and eventually pattern and differentiate into the missing limb structure . After these developmental ( blastema ) stages of regeneration , the limb is small in size relative to the animal’s body and undergoes a period of rapid ontogenetic allometric growth until the size of the regenerate is proportional to the rest of the animal . Ontogenetic allometric growth refers to the rate of growth of a structure ( i . e . the limbs ) relative to the rest of the animal . Little is known about how this growth is regulated to impact the overall scaling of the regenerated limb . Moreover , axolotl are an indeterminately growing species , and continue to grow in size throughout their life cycle . Thus , the size of the limb at the time of amputation is different from that once the limb has completed regeneration ( Riquelme‐Guzmán et al . , 2021 ) . This simple observation indicates that rather than having a ‘set-point’ of size , growth must be dynamically regulated throughout the process of limb regeneration in the axolotl . Most of what is known about the factors that impact tetrapod limb size has been discovered by mutant analyses on determinately growing species , such as humans , mice , and chicken , which cease growing after they reach adulthood due to the closure of the epiphyseal growth plates . These studies reveal that mutations to genes that play roles in patterning , long bone growth , or overall cell physiology in the developing limb tissues can have large impacts on the final size of the limb . For example , the developing bat forelimb bud has increased expression of HoxD transcription factors compared to hindlimb bud and is required for the extreme positive allometric growth that occurs in this tissue as it grows into a wing ( Booker et al . , 2016 ) . At the later stages of tetrapod limb formation , allometric growth is primarily driven through the elongation of the long bones through the regulation of growth plate activity . As a result , alterations in expression of genes that impact paracrine signals which regulate the growth plates , such as fibroblast growth factor ( FGF ) , bone morphogenetic protein ( BMP ) , and Indian hedgehog ( IHH ) , can all positively and negatively regulate the final size of the limb ( Iwata et al . , 2001; Iwata et al . , 2000; Lee et al . , 2017; Segev et al . , 2000; Eswarakumar and Schlessinger , 2007; Toydemir et al . , 2006; Tseng et al . , 2010; Wen et al . , 2016; Duprez et al . , 1996; Aizawa et al . , 2012; Barna et al . , 2000; Bhattacharyya et al . , 2015; Evers et al . , 1996; Karst et al . , 1986; Klüppel et al . , 2005; Lallemand et al . , 2005; Zhang et al . , 2020; Caparrós-Martín et al . , 2013; Joeng and Long , 2009; Kim et al . , 2017; Long et al . , 2001; Mo et al . , 1997; Razzaque et al . , 2005; Ruiz-Perez et al . , 2007; Sohaskey et al . , 2008; St-Jacques et al . , 1999; Yoshida et al . , 2015; Zhang et al . , 2015; Bren-Mattison et al . , 2011; Minina et al . , 2002 ) . Thus , ontogenetic allometric growth in the developing limb bud can be impacted by intrinsic or extrinsic factors at both the developmental and maturation stages of limb formation . Given the similarities between the developing limb bud and the regenerating limb blastema , these mechanisms are likely to hold true during limb regeneration . Supporting this notion , studies using pharmacological inhibitors for FGF , BMP , or TGFβ signaling on the developing axolotl limb blastema all result in smaller limbs compared to controls by negatively impacting patterning , differentiation , or overall blastema physiology ( Lévesque et al . , 2007; Purushothaman et al . , 2019; Vincent et al . , 2020 ) . Conversely , treatment of the axolotl blastemas with exogenous retinoic acid ( RA ) , which negatively regulates HoxA13 ( a marker of distal limb ) ( Gardiner et al . , 1995 ) during blastema patterning , results in the elongation of skeletal elements in the regenerated limb ( Maden , 1983; Niazi et al . , 1985 ) . However , how these and other signals that regulate limb growth are so tightly coordinated during regeneration such that the size of the completed limb is proportional to the animal is not known . Identifying the source of this coordination is the focus of the current study . Multiple pieces of evidence point to the possible involvement of neural signaling in the regulation of allometric growth in limbs . In amphibians , neural signaling is an essential component of the developmental stages of limb regeneration in both the induction and maintenance of the blastema ( Singer , 1978 ) . Nerve signaling is essential for proliferation in the blastema , and loss of innervation at early stages of regeneration results in complete regenerative failure ( Singer , 1978 ) . Multiple protein-based nerve-dependent factors have been identified to play key roles in the maintenance of the limb blastema including anterior gradient ( AG ) , BMPs , FGFs , and neuregulin-1 ( Kumar et al . , 2007; Makanae et al . , 2014; Farkas et al . , 2016 ) . At the late-bud blastema stage , when the pattern of the regenerated limb has been established , the loss of nerve signaling results in the formation of complete , yet miniaturized , limb regenerates ( Mullen et al . , 1996 ) . Last , permanently miniaturized limbs , generated by the repeated removal of the limb bud , have a significant decrease in the abundance of innervation ( Bryant et al . , 2017 ) . Thus , the loss of neural signaling during the developmental stages of limb formation in the axolotl correlates with negative ontogenetic allometric growth and smaller limb size . Nerve abundance , or altered nerve signaling , also correlates with growth and limb size in humans . Damage to the limb nerves during the birth of a human infant results in impaired limb growth and size , which can be alleviated by surgical repair of the nerves ( Bain et al . , 2012 ) . Additionally , the presence of neurofibromas or hamartomas in human arms or hands , which increases the abundance of neural signals , results in the formation of proportionally large digits ( Frykman and Wood , 1978; Razzaghi and Anastakis , 2005; Tsuge and Ikuta , 1973 ) . These observations support the idea that neural signaling plays a role in the regulation of growth in the tetrapod limbs . In the current study , we directly test this possibility in the regenerating axolotl limb . Here , we characterize the growth in the regenerating limb and reveal that its growth is biphasic . By removing nerve signaling from regenerating limbs at different stages , we show that neural signaling is required for multiple cellular behaviors that contribute to growth , and that this requirement changes during the different phases of growth . Using the accessory limb model ( ALM ) assay , our data shows that nerve abundance positively correlates with the rate of growth and the ultimate size of the limb regenerates . To evaluate the potential role of non-neural signals in the regulation of growth , we have developed a new assay that we call the NM-ALM , which decouples the nerve source from the host animal . Using this assay , we found that non-neural extrinsic factors do not play a direct instructive role in the regulation of scaling in the regenerating limb tissue . We have additionally used the NM-ALM to evaluate whether neurons work autonomously to regulate growth in the regenerate and have discovered that they require a connection with their native environment to coordinate this process . These observations will be foundational to future work on the identification of the molecular mechanisms that regulate ontogenetic allomeric growth of the regenerating axolotl limb . Because there was so little known about how allometric growth occurs in the regenerating limb , we started by characterizing the growth of the regenerate until it had completed regeneration . We measured limb and body length to calculate both limb ‘proportionality’ ( ratio of limb length to body length ) ( Figure 1—figure supplement 1 ) and the growth rate in 10 cm sized animals over a period of 140 days , which is when the regenerated limb was no longer significantly different in size relative to the uninjured limbs on the same sized animal , and the growth rate of these limbs was also equivalent . Because axolotls are indeterminate growers , both their limb and body lengths continue to increase in size during the course of regeneration , thus the ratio of limb length to body length was used to normalize for the changes in animal size as time proceeded . We observed that the regenerating limb underwent at least three phases of growth before it reached the size of the unamputated limb , which occur during the blastema and the post-developmental stage of regeneration ( Figure 1A and B ) . The growth during the blastema stage starts immediately following amputation all the way through the digit stage of blastema development . Previous studies have shown that the expression of growth factors and signaling molecules associated with blastema development are lost by the digit staged regenerate , thus indicating the end of the blastema stage , and initiation of post-developmental regenerative processes ( day 38 ) ( Gerber et al . , 2018; Nacu et al . , 2016; Satoh et al . , 2008 ) . The digit staged regenerate is significantly smaller than the unamputated limb on size-matched animals ( Figure 1A and B ) . This small regenerate undergoes rapid ontogenetic allometric growth until it has reached the proportionally appropriate size , at which point the regenerate grows isometrically with the rest of the animal . We have named this post-blastema staged regenerate the ‘tiny limb’ . We observed that the allometric growth is biphasic in the tiny limb . The initial phase of growth is rapid , similar to the speed of growth during the blastema stage at 0 . 04 cm/day ( Figure 1B and Figure S2A ) . Approximately 4 weeks later ( in 10 cm sized animals ) , the rate of growth of the tiny limb slows significantly to 0 . 02 cm/day over the following 9 weeks and gradually decreases over time , until both the size and the growth rate of the regenerated limb are not significantly different than the unamputated limb on size-matched animals ( Figure 1B and Figure 1—figure supplement 2A ) . Based on this difference in the rate of allometric growth , we have separated the tiny limb stage of development into two phases of growth: the early tiny limb ( ETL ) stage and the late tiny limb ( LTL ) stage . Both the growth rate and the amount of time spent in each of the three stages ( blastema , ETL , and LTL ) are dependent on the size of the animal at the time of limb amputation . Comparison between 4 , 10 , and 20 cm animals ( snout to tail tip ) showed that as body length increases , the growth rate decreases ( Figure 1—figure supplement 2 ) , and the amount of time spent in the ETL and LTL phase increases ( Figure 1C ) . In fact , when comparing animals twofold different in size , the amount of time in the ETL phase increases by 30–40% ( Figure 1C ) and growth rate falls by 30–40% ( Figure 1—figure supplement 2 ) . This difference in growth rate ( measured in terms of regenerate elongation ) may be due to the difference in the amount of tissue that needs to be regenerated , since larger animals have more tissue to regenerate than the smaller animals . Additionally , smaller animals grow more rapidly than larger animals , which also might contribute to these differences ( Figure 1—figure supplement 2B ) . We next wanted to determine what cellular mechanisms were contributing to growth in the tiny limb . Multiple processes can contribute to tissue growth including increased cell number via regulation of cell proliferation and cell death , increased cell size , and increased extracellular matrix ( ECM ) deposition ( Conlon and Raff , 1999; Leevers et al . , 1996; Penzo-Méndez and Stanger , 2015; Stanger , 2008; Stocker and Hafen , 2000 ) . Thus , we quantified each of these processes during the different growth phases of regeneration , relative to uninjured limbs , to determine which could be contributing to growth of the tiny limb . Additionally , we speculated that the contribution of these cell processes could vary in the different tissue types in the regenerating limb . Rather than quantifying the above-described processes globally , we analyzed the epidermis , soft tissue ( including all tissues except for skeleton and epidermis ) , and skeletal tissue ( bone and cartilage ) separately . Regenerating limbs at the different stages of growth were sectioned transversely mid-zeugopod , and cell proliferation and cell death were each analyzed by either EdU incorporation or TUNEL staining , respectively . We observed significantly more cell proliferation and significantly less cell death in all tissues analyzed in the ETL and LTL staged regenerates compared to the unamputated and fully regenerated limbs ( Figure 2A and B ) . Interestingly , the fold increase or decrease in cell proliferation or death , respectively , differed depending on the tissue type . The largest increase in cell proliferation was observed in the epidermis ( 3 . 7-fold increase , Figure 2A ) , while soft tissue and skeletal tissue had slightly more modest increases ( 3 . 2- and 2 . 4-fold increases respectively , Figure 2A ) . The largest decrease in cell death was seen in the skeletal tissue ( 20 . 8-fold decrease , Figure 2B ) , while the soft tissues and epidermis exhibited more moderate decreases of 3 . 0- and 3 . 3-fold , respectively ( Figure 2B ) . Cell size and ECM area were analyzed using a combination of fluorescent and histological stains on sectioned limbs . Wheat germ agglutinin ( WGA ) was used to label the plasma membrane of the epidermal and muscle cells , and Alcian blue stained the ECM of the skeletal tissue ( Figure 2—figure supplement 1 ) ( more detail in Materials and methods ) . To standardize our quantification of the average cell size , we measured only the area of cells where the nucleus was observable ( Figure 2—figure supplement 1A ) . We observed that cell size was significantly smaller in the regenerating tissue than the uninjured tissue and increased as regeneration progressed ( Figure 2C ) . This was most profound in the muscle ( 4 . 5-fold smaller ) and least in the epidermis ( 1 . 4-fold smaller , Figure 2C ) . The ECM area was calculated indirectly by subtracting the total cellular area from the tissue area and dividing by the tissue area ( Figure 2—figure supplement 1B ) ( more detail in Materials and methods ) . However , we did not observe any significant differences in the extracellular compartment of limbs , indicating that ECM deposition does not play a significant role in growth of the tiny limb ( Figure 2D ) . We also observed stage-specific differences in the above-described growth characteristics . In almost all tissues , the largest differences were observed between the ETL stage and the unamputated limb , while the differences between the LTL and the unamputated limb were more modest . Apart from cell size in the muscle tissue , there were no significant differences in the growth characteristics when the unamputated and fully regenerated limb were compared . Together , this data indicates that a combination of increased cell proliferation , decreased cell death , and increased cell size contributes to the allometric growth of the tiny limb staged regenerate . While all tissue types showed the same trends in all the cell processes that we analyzed , our data suggests that different cell processes contribute more or less to growth in different tissue types . Future studies will be required to resolve these tissue-specific contributions to growth in more detail . Last , the tissue-specific contributions to growth also depend on the stage of growth of the regenerate and correlate with the overall growth rate of the regenerate at these stages . One interesting observation from the above-described characterization is that the abundance of cell proliferation , cell death , and cell size all show similar trends regardless of the tissue type assessed during each stage of growth in the regenerate . This suggests that there could be a singular signal that coordinates these processes such that the highest growth-promoting signal is occurring during the ETL stage when growth is most abundant and decreases as the growth rate slows during the LTL stage . Thus , we next sought to determine the source of the signal that regulates cell proliferation , death , and size during regeneration . Previous studies indicate that nerve signaling is required for growth in developing and regenerating limbs . The role and mechanism of neurotropic regulation during the early ( blastema ) stages of regeneration has been widely studied ( Farkas et al . , 2016; Farkas and Monaghan , 2017; Kumar et al . , 2010; Makanae et al . , 2014; Singer , 1978; Singer , 1952; Singer , 1946; Singer and Inoue , 1964 ) . It has been well established that nerve signaling is required for , and is a key driver of , blastemal cell proliferation ( Brockes , 1984; Brockes and Kintner , 1986; Lehrberg and Gardiner , 2015 ) . Furthermore , generation of permanently miniaturized limbs through repeated removal of limb buds results in ‘mini limbs’ that are hypo-innervated compared to controls ( Bryant et al . , 2017 ) . We therefore hypothesized that nerves could play a role in driving growth during the tiny limb stages of regeneration . To test this idea , we first characterized the abundance of nerves during these post-developmental stages of regeneration . We collected ETL and LTL staged regenerates , as well as unamputated and fully regenerated limbs for comparison , and sectioned them transversally through the zeugopod . The sections were stained with an anti-acetylated tubulin antibody ( Figure 3A ) , and we measured the abundance of innervation relative to the limb area ( Figure 3B ) . This quantification revealed significantly higher levels of relative innervation during the ETL and LTL stages compared to unamputated and fully regenerated limbs ( Figure 3B ) . Interestingly , the absolute abundance ( not normalized to tissue area ) of innervation is lowest at the ETL stage and increases during the later stages of regeneration . Thus , the decrease in relative innervation during the transition from the ETL to LTL stages is likely due to the substantial increase in limb area of the later staged regenerate . The relative abundance of innervation also correlates well with the growth rate in these tissues . We speculated that nerves could provide growth-promoting signals during regeneration , which decreases ( or is diluted ) as the limb grows in size and the relative abundance of innervation decreases , slowing the growth of the regenerate as it reaches its final size . To determine whether nerve signaling plays a functional role regulating ontogenetic allometric growth , we next tested whether nerve signaling is required to maintain growth in the tiny limb staged regenerate . Limbs were amputated and permitted to regenerate to the ETL stage , at which point nerve signaling was severed via denervation at the brachial plexus ( Figure 4A ) . To test for a possible dose-response or signaling threshold effect , we severed either one , two , or all three of the nerve bundles at the plexus . Mock denervation surgeries were performed as controls . The limbs were measured prior to denervation and 4 days post denervation , when they were collected for analysis . The growth rate , abundance of cell proliferation , abundance of cell death , and cell size were all analyzed ( Figure 4B–E ) . We observed that nerve signaling is required for growth of the tiny limb . Fully denervated ETLs had a 9-fold slower growth rate than ETLs with the nerve intact ( Figure 4B ) . Likewise , cell proliferation was negatively impacted by denervation in all tissue types analyzed ( 1 . 7- , 4 . 9- , and 4 . 8-fold decreases in the epidermis , soft tissue , and skeleton , respectively; Figure 4C ) . Cell death levels in all tissues were significantly increased following full denervation ( 21 . 6- , 1 . 6- , and 25 . 7-fold increases in epidermis , soft tissue , and skeleton , respectively; Figure 4D ) . Lastly , cell size appears to only be significantly affected in the muscle , where there is a near 2-fold decrease in average cell size in the denervated ETLs ( Figure 4E ) . It is possible that this change in muscle could be due to atrophy as a result of decreased muscle activity in the denervated limb . When LTLs were fully denervated , only the growth rate and abundance of cell proliferation were significantly decreased ( Figure 4—figure supplement 1 ) . Thus , the impact of neural signaling is different depending on whether the regenerating limb is in the early or late phase of growth . The partial denervations revealed either a dose-response or threshold response depending on the tissue and growth characteristic quantified . A dose response is reflected by a linear relationship between abundance of signal and the phenotype . A threshold response indicates a specific abundance of a factor is required for a phenotype , for example , a specific amount of nerves per limb volume is required for growth . We observed that cell proliferation in the soft tissue and skeleton decreased significantly , to full denervation levels , with partial denervations indicating that there is a high threshold of nerve abundance required for maintaining cell proliferation in these tissues ( Figure 4C ) . Conversely , cell death in the epidermis had a strong dose response , with significant incremental increases with increased denervation ( Figure 4D ) . These results indicate that each tissue responds differently to nerve abundance to maintain growth . This differential contribution of the different limb tissues could also explain why the overall growth rate of the regenerate is significantly impacted only when a full denervation is performed ( Figure 4B ) . Having established that nerves are required for growth during the late stages of limb regeneration , we next wanted to determine whether we could positively and negatively manipulate the size of the regenerate by altering the abundance of nerves the regenerating tissue is exposed to . To test this , we performed a grafting experiment between differently sized axolotl ( Figure 5A ) . The size of the nerve bundles originating from the spinal column increases as the animal grows , and quantification of the cross-sectional area of these bundles stained immunofluorescently in situ reveals that the size is almost 2-fold larger in 14 cm long animals compared to 7 cm animals ( Figure 5—figure supplement 1 ) . Thus , blastema grafts onto the limbs of large hosts will be exposed to larger nerve bundles than grafts on small host animals . To generate large and small animals we housed age-matched GFP+ ( donors ) and GFP- ( hosts ) axolotl at either 19°C or 4°C . Housing half the animals at 4°C stunted their growth while the 19°C animals grew at a faster rate . After approximately 2 months , the 19°C animals were 2-fold larger in body length and limb length than the 4°C animals ( Figure 5B ) . After the large and small siblings were both incubated at 19°C for 14 days , both forelimbs on small ( ~6 cm ) and large ( ~12 cm ) GFP+ axolotl were amputated and permitted to regenerate to the mid-bud blastema stage ( Figure 5A and B ) . The blastemas and approximately 2 mm of stump tissue were grafted onto regenerative permissive environments on small ( ~6 cm ) and large ( ~12 cm ) GFP-host animals ( Figure 5A ) . It has previously been found that cells from approximately 500 μm of stump tissue migrate and contribute to the regenerate ( Currie et al . , 2016 ) . Therefore , stump tissue was included with the blastemas to prevent the contribution of cells from the host environment to the regenerate . A regeneration permissive environment on the large animal hosts was generated by deviating the branchial nerve bundle to an anterior located wound site on the limb , using the standard accessory limb model ( ALM ) surgery ( Figure 5A; Endo et al . , 2004; McCusker and Gardiner , 2013 ) . Because the limb circumference of the small host animals was too small to receive the grafted blastema tissue from the large donors , we deviated the branchial nerve to a wound site on the flank of the host , where a larger skin wound could be made to fit the larger graft size ( Figure 5A ) . The length of the ectopic limbs was measured bi- or tri-weekly . They were considered fully regenerated when the rate of growth of the ectopic limb was no longer significantly different from the uninjured limbs on the host animals . The donor animal limbs that were amputated to harvest the blastema grafts were also continually measured during regeneration as an additional control to measure the final sizes of the regenerated limbs if left on their native environment . We hypothesized that if nerve abundance can regulate size of the limb regenerate , we would observe that the lengths of the grafted regenerates will correspond to the size of the host environment rather than the donor . Thus , blastemas from the small donors will produce large ectopic limbs when grafted to large hosts , and blastemas from large donors will produce small ectopic limbs when grafted to small hosts . Alternately , if nerve abundance does not influence regenerate size , then we would expect to see the blastemas from small donors on large hosts produce ectopic limbs smaller than the control grafts from large animals , and vice versa . We observed that ALM environments with different nerve abundances altered both the growth rate and overall size of the regenerated limb structure . The blastemas ( 15 days post amputation ) from small donors were initially 2-fold smaller than those from large donors when they were grafted onto the small and large hosts ( Figure 5C , left panels , 0 day post graft [DPG] ) . Interestingly , we observed that the grafted tissues were the same size by 18 DPG within each host type and remained that way until regeneration was completed ( Figure 5—figure supplement 2A ) . In comparison , the regenerated limbs that remained on the donor animals remained significantly different in size ( Figure 5D ) . Although there was no significant difference in the size of regenerates that formed within an individual host type regardless of the origin of the blastema , we did observe significant differences when we compared regenerate sizes between the large and small hosts . By approximately 38 DPG , there was a significant difference in size between the ectopic limbs on the large and small host animals , and this continued until 130 DPG , when the growth rate of the grafted limbs matched that of the host limbs ( Figure 5C , right panels; Figure 5—figure supplement 2A ) . This difference in size was due to a more rapid increase in the growth rate of the blastemas grafted to the large host animals ( Figure 5—figure supplement 2 ) . We hypothesize that the host environment on the large animals reaches a critical threshold of nerves to support growth at an earlier time and/or accumulates more nerve-dependent factors that promote growth in the regenerating tissues . Together these data indicate that the growth rate and sizing of the limb regenerate positively correlates with nerve abundance in the host environment . However , it does not rule out other potential influences from the host environments that may also contribute to growth . Thus , we next designed an assay that decouples nerve abundance from the size of the host to test whether non-neural signals contribute to the sizing of the limb regenerate . To evaluate the potential role of non-neural sources of growth regulation that may be present in the differently sized host animals , we developed a new regenerative assay called the NM-ALM that decouples the source of the nerves from the host environment ( Figure 6B ) . In the NM-ALM , the forelimb-specific dorsal root ganglia ( DRGs ) from GFP+ donor animals are harvested and grafted into lateral limb wounds on host animals ( one DRG per wound ) . The GFP+ DRGs are grafted below the mature skin next to the wound site , and the axon bundles are pulled into the middle of the wound site in a similar manner as the traditional ALM surgery . Mid-bud blastema staged regenerates from age-matched donors are then grafted onto the wound site ( Figure 6B ) . We measured the length of the ectopic limbs in the NM-ALM weekly or bi-weekly , and their growth rates ( cm/day ) were compared to the unamputated limbs on the donors . The grafted limbs were considered to have completed limb regeneration when their growth rates were not significantly different from the growth rates of the unamputated limbs on the control animals . We observed that the implanted DRG supported the continued development of the blastema into a completely patterned and differentiated limb ( Figure 6B’ ) . Thus , the implanted DRGs are able to provide the appropriate signals to support the regenerative process . To test whether non-neural extrinsic signals from the host environment provide cues that regulate the growth of the regenerating limb , we performed the NM-ALM on differently sized host animals ( Figure 7A ) . If non-neural extrinsic signals play a role in regulating allometric growth of the regenerate , then we expected to observe that the regenerates that grew in the NM-ALMs on the large ( 14 cm ) hosts would be larger in size than the ones that grew on the small ( 7 cm ) hosts ( Figure 7B ) . Alternately , if the limb nerves play the central regulatory role , we would not expect to see differences in the size of the regenerates on the different sized hosts . After 135 DPG , we observed that there continued to be no significant difference in growth rates and the length of the grafted regenerates on the different sized host animals ( Figure 5—figure supplement 2B and Figure 7C ) . This trend was observed in NM-ALMs that were performed with DRGs that were harvested from both large and small animals ( Figure 5—figure supplement 2B ) . In contrast , the control uninjured limbs on the small and large animals remained significantly different in size ( Figure 7D ) . Because the size of the limb regenerate was not impacted by the host environment when the abundance of nerves was held constant , we conclude that non-neural extrinsic factors do not play a direct role in regulating allometric growth of the regenerating limb structure . To test whether growth of the regenerate is autonomously regulated by the nerves , we leveraged our newly developed NM-ALM using DRGs that were harvested from different sized GFP+ donors ( 14 cm versus 7 cm ) ( Figure 8A ) . Our expectation was that if neural regulation occurs autonomously ( at the DRG level ) , then the growth rate and overall size of the regenerated limb that grew on the DRG from the large donor would be larger than that which grew on the one from the small donor ( Figure 8B ) . Conversely , if the DRGs require a connection with their native environment to regulate regenerate size , then we expected to see no difference in the regenerate growth rate and size regardless of the donor source of the DRG ( large or small ) ( Figure 8B ) . Because we had previously observed a dose response in limb growth in partially denervated regenerating limbs ( Figure 4 ) , we hypothesized that allometric growth was regulated by nerve abundance alone . We quantified the abundance of innervation from the GFP+ DRGs in the ectopic limbs that grew from the NM-ALMs and found significantly more GFP+ neurons in the limbs with the large versus small DRG implants ( Figure 8D ) . However , our data showed that there was no statistical difference in growth rates and the ectopic limb length between the regenerates from NM-ALMs with the DRGs from small or large animals ( Figure 5—figure supplement 2B and Figure 8C ) . This trend was observed on the NM-ALMs on both the large and small hosts ( Figure 5—figure supplement 2B ) . Thus , we concluded that nerve abundance alone does not regulate the final size of the regenerate and that factors from the DRGs’ endogenous environment are required for the neural regulation of ontogenetic allometric growth during the tiny limb stage of regeneration . It is possible that without these upstream signals in NM-ALM , the DRGs from large and small donor animals generate the same amount of growth-promoting factors . Alternately , because the DRGs only contain sensory neurons , it is possible that motor neurons are required to control growth . Future studies will focus on the cells and signals that play a role in this upstream regulation . Previous studies on developing amphibian limbs suggest that both intrinsic and extrinsic factors determine size during embryogenesis . One example is from the classic cross-grafting study performed by Ross Harrison using limb buds between two differently sized salamander species ( Harrison , 1924 ) . In this experiment , limb buds from Ambystoma tigrinum ( adult body length of 27 cm ) were grafted to Ambystoma punctatum ( adult body length of 16 cm , now known as Ambystoma maculatum ) hosts and vice versa ( Harrison , 1924 ) . It was observed that the grafted limb buds grew to sizes that failed to correspond with limbs from either donor or host . Rather , the limb buds from A . punctatum grew to a size that was smaller than both the host and donor limbs when grafted to A . tigrinum , while the A . tigrinum limb buds grew larger than the host and donor limbs when grafted to A . punctatum . Because the overall size of the grafted limb was influenced by both the host environment and the source of the graft , it was concluded that the growth of the limbs was regulated by both the intrinsic growth capacity of the limb bud cells and an extrinsic growth-promoting factor that is present in different abundances in the embryos of different host species ( Harrison , 1924 ) . It was later shown by Twitty and Schwind that when the host animals were fed to capacity , the growth rates of the grafted limb buds mimicked the donor animal’s growth rate , but final size of the limb was more closely matched to the host ( Twitty and Schwind , 1931 ) . This observation suggests that the rate of allometric growth appears to be intrinsically regulated , while the duration of this growth is impacted more by extrinsic factors . While our study also shows that extrinsic signals play a central role in the scaling of the regenerate , it is unknown whether the same nerve-dependent mechanisms are at play during limb bud development in the axolotl . During embryonic limb development , the apical ectodermal ridge ( AER ) signaling center is required for the production of paracrine factors such as FGFs and BMPs , which maintain an undifferentiated state and drive proliferation in the limb bud mesenchyme ( Purushothaman et al . , 2019; Ovchinnikov et al . , 2006 ) . The AER is established in an aneurogenic environment in multiple amphibian species including Xenopus laevis , Rana pipiens , and A . punctatum/maculatum ( Keenan and Beck , 2016; Kumar et al . , 2011; Nieuwkoop and Faber , 1958; Taylor , 1943; Yntema , 1959 ) , while regenerating amphibian limbs require signaling from the nerve to establish the analogous structure , the apical epithelial cap ( AEC ) ( Singer , 1978; Singer and Inoue , 1964 ) . However , axolotl develop limb buds as larvae , not as embryos , and it is not clear whether the axolotl limb bud is innervated or not . Therefore , it is possible that the regulation of growth is dependent on nerve signaling in both embryonic and regenerating limbs . It should be noted that the current study does not evaluate the influence of intrinsic factors that regulate the size of the regenerating limb . Because our study was performed on sibling A . mexicanum , there were minimal genetic differences between donor and host animals . Performing a similar cross grafting experiment , as was performed by Harrison , 1924 , between species of different sizes in the context of regeneration would yield valuable insight in this regard . Observations from multiple tetrapod species indicate that neural regulation of limb growth and size is a conserved mechanism . As we have explained previously , damage to the limb nerves in humans co-relates to decreased limb size , while overabundance of nerve signaling corresponds to limb or digit enlargement ( Bain et al . , 2012; Cerrato et al . , 2013; Labow et al . , 2016; Tsuge and Ikuta , 1973 ) . The current study shows that the limb nerves also influence limb size in regenerating amphibians as well . This poses the amphibian limb regenerate as an exciting new model to study the general mechanisms underlying neural regulation of limb allometry . Studies in developing embryos show that the alteration of factors that regulate patterning , elongation of the long bones , and cell physiology in the developing limb can all influence the final size of the limb . Interestingly , altering the expression of intrinsic factors , such as Hox transcription factors or paracrine factors such as FGFs , during the patterning or maturation stages of limb development , respectively , can both positively and negatively impact limb size ( Booker et al . , 2016; Cho et al . , 2008; Fromental-Ramain et al . , 1996; Hérault et al . , 1996 ) . In contrast , the alteration of factors that are required for the maintenance of cell physiology , such as DNA damage repair gene Trp63 , only negatively impact limb size ( Vernersson Lindahl et al . , 2013 ) . We interpret these observations to indicate that genes that regulate patterning or long bone growth are ‘instructive’ regarding the overall size of the limb , whereas genes that impact cell physiology play a more permissive role in this process . Whether and how the limb nerves regulate one or more of these different influencers of allometry during regeneration is an important outstanding question . However , it is interesting that the role of neural signaling appears to differ depending on the phase of growth . Limb denervation at the ETL phase negatively impacted proliferation , cell survival , and cell size , while the same manipulation at the LTL stage only impacted cell proliferation . It is unclear at this time whether the change in neural requirement is due to alterations in the generation of growth-promoting signals from the nerve , the responsiveness of the regenerating tissue to these signals , or both . While the molecular mechanism by which nerves regulate scaling has not been determined , the role of the limb nerves during the early stages of limb regeneration has been extensively studied and might be drawn upon to provide insight into their role in size regulation . Innervation of the wound epithelium during the early stages of limb regeneration is essential to establish the AEC . The AEC then produces growth factors essential to induce and maintain the cells in a dedifferentiated state and drive blastema cell proliferation ( McCusker et al . , 2015 ) . Satoh et al . , 2016 demonstrated that FGF8 and BMP7 are directly produced by the nerves in the DRGs and migrate to the wound epithelium . Furthermore , regenerative failure caused by the denervation of the regenerating limb can be rescued by cocktails of growth factor proteins including FGFs and BMPs ( Makanae et al . , 2014; Satoh et al . , 2016; Vieira et al . , 2019 ) , neuregulin-1 ( Farkas et al . , 2016 ) , or AG ( Kumar et al . , 2007 ) . Although the AEC is no longer present during the post-blastema stages of growth that we have focused on in the current study , nerves could both generate and induce the expression of similar growth-promoting factors in the late staged regenerating tissues . Alternately , or in conjunction with , the limb nerves may also regulate limb growth through their modulation of bioelectric signals which has been shown to regulate fin size in zebrafish ( Perathoner et al . , 2014; Daane et al . , 2018 ) . Regardless , the nature of the nerve-derived signals , and the downstream tissue and molecular targets of these signals , will be an essential focus of future research . The cessation of limb growth in determinately growing tetrapods , such as humans , is brought about by hormone-based changes which leads to alterations of chondrocyte behavior in the growth plate and completion of the terminal phase of differentiation in this tissue ( reviewed in Ağırdil , 2020 ) . In contrast , axolotl are an indeterminately growing species that maintain active growth plates in the long bones throughout life . Thus , the transition of positive to isometric ontogenetic allometric growth in the regenerated limb must occur without the closure of the axolotl growth plates and suggests an alternate mechanism of regulation . Our studies using the NM-ALM indicate that non-neural extrinsic factors , which would include systemic factors such as hormones , do not directly promote growth in the regenerating limb tissue , which further supports this idea ( Figure 7 ) . However , it is possible that systemic factors are permissive to regenerative growth . In determinately growing species , these factors play important roles in regulating the growth of long bones in preadult limbs ( Williams , 1981; Boersma and Wit , 1997; Penzo-Méndez and Stanger , 2015 ) . Our observation that neurons , when separated from their typical environment within the vertebrae , are unable to support the positive allometric growth of the regenerate may indicate that additional signals that are extrinsic to the regenerating limb tissue are required for the neural regulation of scaling ( Figure 8 ad Figure 5—figure supplement 2 ) . It is possible that either the vertebrae provide a permissive environment which allows the nerves to provide the appropriate cues to the regenerate , or there is some upstream signal that requires a direct connection of the limb nerves to the CNS to regulate growth . Alternately , because the NM-ALM assays that we performed only contain sensory neurons , it is possible that motor neurons can autonomously instruct growth in the regenerate . It will be important to resolve this issue in future studies . This study provides foundational knowledge on how proportion becomes reestablished in the regenerating limb in a continually growing system . Our data indicates that nerve signaling plays a central role in coordinating ontogenetic allometric growth , and future studies will focus on identifying the molecular mechanism ( s ) underlying this regulation . Furthermore , our data suggests that there is an upstream factor ( or factors ) required for the neural regulation of growth , potentially deriving from the endogenous environment of the limb nerves or the CNS . Lastly , as previously stated , our studies have not ruled out the likely intrinsic factors involved in size regulation . In total , there are still many unknowns , both up- and downstream of nerve signaling , that must be resolved to fully understand how allometric growth is regulated during axolotl limb regeneration . Furthermore , as regenerative medicine seeks to tap back into the developmental mechanism in order to regrow a fully functional limb , it will be important to study size regulation in multiple species to identify the shared mechanisms regulating this process . Ethical approval for this study was obtained from the Institutional Animal Care and Use Committee at the University of Massachusetts Boston ( Protocol # IACUC2015004 ) and all experimental undertakings were conducted in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . Axolotls ( A . mexicanum ) were spawned either at the University of Massachusetts Boston or at the Ambystoma Genetic Stock Center at the University of Kentucky . Experiments were performed on white-strain ( RRID: AGSC_101J ) , GFP-strain ( RRID: AGSC_110J ) , and RFP-strain ( RRID: AGSC_112J ) Mexican axolotls ( A . mexicanum ) . Animal sizes are measured snout to tail tip and described in the text for each experiment . They were housed in 40% Holtfreters on a 14/10 hr light/dark cycle and fed ad libitum . Animals were fed every day or three times a week depending on the size of the animal . Animals were anesthetized in 0 . 1% MS222 prior to surgery or imaging . Live images were obtained using a Zeiss Discovery V8 Stereomicroscope with an Axiocam 503 color camera and Zen software ( Zeiss , Oberkochen , Germany ) . To generate large and small animals , larval animals were either housed at 19°C or 4°C , which slows their growth rate . Animals were grown at these temperatures until their body lengths were approximately 2-fold different , at which point the smaller animals were moved to 19°C for 2 weeks prior to any surgical manipulation . When measuring limb length and body length to determine limb proportionality and growth during regeneration , limbs were measured from the trunk/limb interception to the elbow and from the elbow to the longest digit tip ( Figure 1—figure supplement 1 ) . Body length was measured from snout to tail tip . All measurements were taken in centimeter ( cm; Figure 1—figure supplement 1 ) . Measurements were recorded prior to experimentation and weekly following surgical manipulation . After 5 weeks measurements were bi-weekly , and after 10 weeks they were taken tri-weekly . Forelimb amputations are done mid-stylopod ( mid-humerus ) . If the bone protruded from the amputation plane after contraction of the skin and muscle , it was trimmed back to make a flat amputation plane . Limb regeneration stages were determined through an observation of patterning and growth rates . Limbs were considered in the ‘ETL’ stage when they reached the mid-digit stage of patterning ( Iten and Bryant , 1973 ) . Prior to this , they are considered in the ‘blastema’ stage of limb regeneration . The transition from ‘ETL’ to ‘LTL’ is determined by a statistically significant decrease in limb length growth rate ( cm/day ) . The regenerating limb is fully regenerated when the growth rate is no longer statistically significant from the unamputated control limbs . Denervation of limbs was done by making a posterior incision on the flank , at the base of the arm , and severing and removing a piece of the three nerve bundles that come from spinal nerves 3 , 4 , and 5 , proximal to the brachial plexus . A 2–3 mm piece of the nerve bundle was cut out in an effort to delay the regeneration of the nerves into the regenerates . Partial denervations were also performed by severing 1 , 2 , or all three of the limb nerve bundles . denervations were performed by severing spinal nerve 5 . 2/3 denervations were performed by severing spinal nerves 4 and 5 . Full denervations were performed by severing all three nerves . Mock denervations were also performed by creating the same incision on the posterior side of the limb and dissecting the nerve bundles as in a typical denervation but leaving the nerve bundles intact . Because limb denervation last approximately 7 days before nerves begin to re-innervate the limb , experimental limbs were analyzed 4 days post denervation . Limbs on large ( average 12 cm snout to tail tip ) and small ( average 6 cm snout to tail tip ) GFP+ donor axolotls were amputated mid-stylopod and allowed to regenerate until mid-bud stage . At that stage , they were amputated along with approximately 2 mm stump tissue , and grafted onto a regenerative permissive environment on large ( 12 cm snout to tail tip ) RFP+ host axolotls or small ( 6 cm snout to tail tip ) white host axolotls . Stump tissue was included in the graft to ensure the ectopic limb is composed primarily of donor animal cells , since stump cells contribute to the regenerate ( Currie et al . , 2016 ) . The regenerative permissive environment on the large hosts was created by removing an anterior patch of full thickness skin from the stylopod and deviating a limb nerve bundle to the wound site ( Endo et al . , 2004; McCusker and Gardiner , 2013 ) . The blastemas on the large donor animals were substantially larger than the limbs of the small host animals , making it impossible to perform this test on the small limbs . Thus , a regenerative permissive environment on the small host axolotl was generated by removing a patch of full thickness skin from flank of the animal posterior to the forelimb . The limb nerve bundle is dissected from the limb and deviated to the flank wound site on the small animal . After the blastemas are grafted on to the host wound sites , hosts are kept on ice and misted frequently for 1 hr , permitting attachment of the graft . We developed the NM-ALM assay to determine if non-neural extrinsic factors were contributing to size regulation . In the NM-ALM the nerve source and host environment are decoupled by grafting a limb DRG from a donor animal in lieu of the deviated nerve bundle from the host animal into the wound site , as in the standard ALM surgery . Anterior wounds are created on white-strain host animals . Limb DRGs were carefully extracted postmortem from GFP+ donor animals and implanted below the proximal skin surrounding the wound site ( Figure 6A ) . The limb axon bundle is positioned in the middle of the wound site ( Figure 6A ) . Mid-bud staged blastemas , along with approximately 2 mm of stump tissue , were amputated from white-strain donor animals and immediately grafted onto the DRG nerve bundle and wound site on the host animals . Animals were kept on ice and moist for 2 hr following grafting to ensure attachment of the graft . In the NM-ALM experiment reported here , we used large ( average 14 cm snout to tail tip ) and small ( average 7 cm snout to tail tip ) GFP+ and white-strain siblings , generated by crossing heterozygous GFP parents , as blastema donors , DRG donors , and NM-ALM hosts . DRGs from both large and small GFP+ donor animals were dissected postmortem and grafted into anterior limb wound sites on large and small white host animals ( Figure 6C ) . Mid-bud staged blastemas from small white-strain donors were then grafted onto the NM-ALM . Tissues were fixed overnight at 4°C in 4% formaldehyde ( RICCA Chemical Company , Arlington , TX ) , decalcified in 10% EDTA ( VWR , Radnor , PA ) for 5–14 days depending on tissue size , rehydrated in 30% sucrose for 2 days , and embedded in Tissue-Tek OCT Compound ( Sakura , Torrance , CA ) . The OCT blocks were then flash-frozen in liquid nitrogen and stored at –20°C . They were cryo-sectioned on a Leica CM 1950 ( Leica Biosystems , Buffalo Grove , IL ) through the mid-zeugopod at 7 μm thickness . Following staining , histological stained slides were mounted using Permount Mounting Medium ( Thermo Fisher Scientific , Waltham , MA ) . VECTASHIELD Antifade Mounting Medium ( Vector Laboratories , Burlingame , CA ) was used to mount coverslips on the fluorescently stained slides . Sections were imaged on the Zeiss Observer . Z1 at 20× magnification . Tile scans were taken of the entire tissue section and then stitched together using the ZenPro software ( Zeiss ) . For analysis of cell proliferation in the limbs , 100 ng of EdU ( Roche , Basel , Switzerland ) was injected into the intraperitoneal space on the flank of the animal . The limbs were then collected exactly 4 hr after injection and prepared for staining . Harvested tissues were process for cryo-sectioning as described above . The Roche Click-It EdU kit was used , following manufacturer’s protocol , and co-stained with DAPI ( 1:1000 dilution – Sigma-Aldrich , St Louis , MO ) . The Roche In Situ cell death detection kit ( Fluorescein ) was used to analyze apoptosis cell death using the manufacturer’s protocol . TUNEL-stained sections were also co-stained with DAPI to obtain percent cell death . Using the Open Source FIJI software , the number of either proliferating ( EdU+ ) or dying ( TUNEL+ ) cells and DAPI+ cells was counted and used to calculate the labeling indices for each . In each section , the tissues were visually separated based on morphology into three basic categories: epidermal , skeletal ( bone or cartilage ) , and soft tissue ( all other tissue ) , and the labeling indices were generated for each separately . Three technical ( three sections per limb ) and at least three biological replicates were performed for each sample . The quantification of average cell size and extracellular area were performed separately on the entire epidermal , muscle , and skeletal tissues in each tissue section . The epidermal and muscle tissues were identified based on morphology in WGA ( Thermo Fisher Scientific ) , Rhodamine phalloidin ( Thermo Fisher Scientific ) , and DAPI ( 1:1000 dilution – Sigma-Aldrich ) stained sections . Skeletal tissues ( bone or cartilage ) were identified in sections stained with Harris Hematoxylin ( Sigma-Aldrich ) , Eosin Y ( Thermo Fisher Scientific ) , and Alcian Blue ( Sigma-Aldrich ) . To measure the average cell size of epidermis and muscle in the fluorescent images , the area within WGA plasma membrane-stained cells , where the nucleus was observed ( DAPI signal ) , was quantified using the FIJI software and averaged ( Figure 2—figure supplement 1 ) . While muscle cells are elongated syncytial cells , our measurements only quantified a cross-sectional area . For the skeletal elements , the average cell area was quantified by measuring the area of Alcian blue negative areas in the element that contained a nucleus ( Figure 2—figure supplement 1 ) . To analyze ECM size for the epidermis and muscle , the area of all WGA internal cell spaces ( with and without DAPI ) were quantified . For skeletal tissue , the area of the Alcian blue negative cell spaces were quantified . The sum of these areas was calculated to obtain the ‘total cellular area’ . The area of the complete tissue ‘total tissue area’ was then quantified ( Figure 2—figure supplement 1B ) . Since the tissue sizes can vary , percent ECM was determined through the following equation:PercentECMDeposition=TotalTissueArea−TotalCellularAreaTotalTissueArea×100 The quantitative analysis of nerve abundance was performed on regenerating limb and flank sections using the Mouse Monoclonal Anti-Acetylated Tubulin antibody ( 1:200 dilution – Sigma-Aldrich ) , followed by the Goat-Anti-Mouse IgG Alexa Fluor 488 ( 1:200 dilution – Abcam , Cambridge , MA ) secondary antibody . They were co-stained with Rhodamine phalloidin and DAPI as a general tissue stain , and to provide positional context for the location of the axon bundles . Limb nerve abundance was quantified by determining the percentage of limb area that is positive for Anti-Acetylated Tubulin staining . To determine limb-bound axon bundle size , the sum of the area of axon bundles from DRGs 3 , 4 , and 5 were quantified as they emerged from the skeletal muscle surrounding the spine ( Figure 5—figure supplement 1 ) .	Humans’ ability to regrow lost or damaged body parts is relatively limited , but some animals , such as the axolotl ( a Mexican salamander ) , can regenerate complex body parts , like legs , many times over their lives . Studying regeneration in these animals could help researchers enhance humans’ abilities to heal . One way to do this is using the Accessory Limb Model ( ALM ) , where scientists wound an axolotl’s leg , and study the additional leg that grows from the wound . The first stage of limb regeneration creates a new leg that has the right structure and shape . The new leg is very small so the next phase involves growing the leg until its size matches the rest of the animal . This phase must be controlled so that the limb stops growing when it reaches the right size , but how this regulation works is unclear . Previous research suggests that the number of nerves in the new leg could be important . Wells et al . used a ALM to study how the size of regenerating limbs is controlled . They found that changing the number of nerves connected to the new leg altered its size , with more nerves leading to a larger leg . Next , Wells et al . created a system that used transplanted nerve bundles of different sizes to grow new legs in different sized axolotls . This showed that the size of the resulting leg is controlled by the number of nerves connecting it to the CNS . Wells et al . also showed that nerves can only control regeneration if they remain connected to the central nervous system . These results explain how size is controlled during limb regeneration in axolotls , highlighting the fact that regrowth is directly controlled by the number of nerves connected to a regenerating leg . Much more work is needed to reveal the details of this process and the signals nerves use to control growth . It will also be important to determine whether this control system is exclusive to axolotls , or whether other animals also use it .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine", "developmental", "biology" ]	2021	Neural control of growth and size in the axolotl limb regenerate
CRISPR adaptation immunizes bacteria and archaea against viruses . During adaptation , the Cas1-Cas2 complex integrates fragments of invader DNA as spacers in the CRISPR array . Recently , an additional protein Cas4 has been implicated in selection and processing of prespacer substrates for Cas1-Cas2 , although this mechanism remains unclear . We show that Cas4 interacts directly with Cas1-Cas2 forming a Cas4-Cas1-Cas2 complex that captures and processes prespacers prior to integration . Structural analysis of the Cas4-Cas1-Cas2 complex reveals two copies of Cas4 that closely interact with the two integrase active sites of Cas1 , suggesting a mechanism for substrate handoff following processing . We also find that the Cas4-Cas1-Cas2 complex processes single-stranded DNA provided in cis or in trans with a double-stranded DNA duplex . Cas4 cleaves precisely upstream of PAM sequences , ensuring the acquisition of functional spacers . Our results explain how Cas4 cleavage coordinates with Cas1-Cas2 integration and defines the exact cleavage sites and specificity of Cas4 . Bacteria and archaea use an adaptive immune system composed of clustered regularly interspaced short palindromic repeats ( CRISPR ) arrays and CRISPR-associated ( Cas ) proteins to defend against infection ( Barrangou et al . , 2007; Brouns et al . , 2008; Marraffini and Sontheimer , 2008 ) . Within this system , the CRISPR locus is programmed with ‘spacer’ sequences that are derived from foreign DNA and serve as a record of prior infection events ( Bolotin et al . , 2005; Mojica et al . , 2005; Pourcel et al . , 2005 ) . The host adapts to an infection event when Cas proteins insert short fragments from the invader DNA as new spacers between repeating sequence elements within the CRISPR locus ( reviewed in Jackson et al . , 2017 ) . The locus is transcribed and processed into short CRISPR RNAs ( crRNAs ) , which assemble with Cas proteins to form a RNA-guided surveillance complex ( reviewed in Hochstrasser and Doudna , 2015; Charpentier et al . , 2015 ) . Finally , the surveillance complex recognizes the target bearing complementarity to the crRNA sequence and a Cas nuclease cleaves or degrades the target during the interference stage ( reviewed in Marraffini , 2015 ) . Although the machinery and mechanisms involved in CRISPR interference are extremely diverse ( Koonin et al . , 2017 ) , the adaptation proteins Cas1 and Cas2 are conserved among most CRISPR systems , suggesting a common molecular mechanism for acquiring spacers . Cas1 and Cas2 form a heterohexameric complex that catalyzes spacer integration via two transesterification reactions mediated by nucleophilic attack of the 3ʹ-hydroxyl on each strand of a double-stranded prespacer substrate at the phosphodiester backbone within the CRISPR array . Integration occurs at the first repeat in the CRISPR array , with one attack occurring between the upstream leader sequence and the repeat and the other occurring on the opposite strand between the repeat and first spacer within the array ( Arslan et al . , 2014; Nuñez et al . , 2015a; Rollie et al . , 2015 ) . These reactions result in the insertion of the prespacer between two single-strand repeats , and this gapped intermediate is repaired by host factors ( Ivančić-Baće et al . , 2015 ) . In order to form a functional spacer , the adaptation complex must capture and process longer fragments of DNA from the invader containing a flanking sequence called a protospacer adjacent motif ( PAM ) ( Nuñez et al . , 2015b; Wang et al . , 2015; Xiao et al . , 2017 ) . The PAM is an essential motif during target recognition by the surveillance complex and must be present next to the target in order for interference to occur ( Deveau et al . , 2008; Redding et al . , 2015; Sashital et al . , 2012; Semenova et al . , 2011; Sternberg et al . , 2014 ) . However , the PAM is not part of the spacer and must be removed from the prespacer prior to integration through a processing step . In addition , integration must occur in the correct orientation to produce a crRNA that is complementary to the PAM-containing strand of the invader . In some systems , additional Cas proteins , such as Cas4 , are also required during adaptation . Cas4 is widespread in type I , II , and V systems ( Hudaiberdiev et al . , 2017 ) . In in vivo studies , deletion of cas4 reduced the adaptation efficiency ( Li et al . , 2014; Liu et al . , 2017 ) and resulted in the acquisition of non-functional spacers from regions that lacked a correct PAM ( Almendros et al . , 2019; Kieper et al . , 2018; Shiimori et al . , 2018; Zhang et al . , 2019 ) . Some systems have two cas4 genes that work together to define the PAM , length and orientation of spacers , suggesting that the two Cas4 proteins are involved in processing each end of the prespacer and that they may be present during integration ( Shiimori et al . , 2018 ) . Similarly , in vitro studies have suggested that Cas4 is involved in PAM-dependent prespacer processing ( Lee et al . , 2018; Rollie et al . , 2018 ) . Cas4 endonucleolytically cleaves PAM-containing 3ʹ-single-stranded overhangs that flank double-stranded prespacers ( Lee et al . , 2018 ) . Importantly , Cas4 cleavage activity is dependent on the presence of Cas1 and Cas2 , and Cas4 inhibits premature integration of unprocessed prespacers . These observations suggest that Cas4 associates with the Cas1-Cas2 complex , although direct biochemical and structural evidence for this Cas4-Cas1-Cas2 complex remains elusive ( Lee et al . , 2018; Plagens et al . , 2012 ) . Here we show that Cas4 forms a complex with Cas1-Cas2 in the presence of dsDNA . Using single-particle negative-stain electron microscopy ( EM ) , we determined the architecture of Bacillus halodurans type I-C Cas1-Cas2 and Cas4-Cas1-Cas2 complexes . Unlike the symmetrical Cas14-Cas22 structure , we observed a mixture of symmetrical ( Cas42-Cas14-Cas22 ) and asymmetrical ( Cas41-Cas14-Cas22 ) complexes , suggesting a structural mechanism for distinguishing between the PAM and non-PAM end of the prespacer following processing . The positioning of Cas4 places it in close proximity to the Cas1 active site , suggesting a mechanism for substrate handoff following prespacer processing . Surprisingly , the Cas4-Cas1-Cas2 complex processes single-strand DNA when an activator duplex DNA is provided in trans . Using this ssDNA cleavage assay , we show that the Cas4-Cas1-Cas2 complex is highly specific for PAM sequences and cleaves precisely upstream of the PAM . In a duplex substrate , the PAM must be positioned within a single-strand region for optimal cleavage , but Cas4 can cleave at various locations within this single-stranded overhang . Collectively , these findings provide the first structural information of the Cas4-Cas1-Cas2 adaptation complex and reveal the precision and specificity of prespacer processing prior to integration . We previously showed that B . halodurans type I-C Cas4 associates tightly with Cas1 but were unable to obtain the Cas4-Cas1-Cas2 complex due to instability of the Cas1-Cas2 complex in this system ( Lee et al . , 2018 ) . We hypothesized that a CRISPR DNA substrate may help stabilize the complex . To test this possibility , we designed a CRISPR hairpin ‘target’ substrate containing a 10 bp leader , the full 32 bp repeat , and a 5 bp spacer , mimicking the region of the CRISPR at which integration occurs ( Figure 1A ) . We incubated the individually purified Cas1 and Cas2 proteins with hairpin target DNA in equimolar amounts and removed unassociated DNA via ion-exchange chromatography followed by size-exclusion chromatography to remove free Cas proteins . Incubation of individual components with or without the hairpin target led to different elution volumes from a size-exclusion column ( Figure 1B ) . In the absence of the target , Cas1 and Cas2 proteins generated two separate peaks , which correspond to the peaks of each individual component . When Cas1 and Cas2 were incubated with the target DNA , the proteins eluted earlier as a single peak , while unassociated Cas2 eluted at the original elution volume ( Figure 1B–C ) . These data indicate that the Cas1-Cas2 complex from type I-C is stabilized in the presence of dsDNA . Next , we attempted to reconstitute the putative Cas4-Cas1-Cas2 complex in the presence of the CRISPR hairpin target ( Figure 1A ) . Following incubation of equimolar amounts of each component , Cas4 , Cas1 and Cas2 eluted in a single peak from the size-exclusion column ( Figure 1—figure supplement 1 ) , with an earlier elution volume than the Cas1-Cas2-target sample ( Figure 1B–C ) . In addition , we observed two peaks with the approximate elution volumes of free Cas1 and Cas2 ( Figure 1—figure supplement 1 ) . Both Cas1-Cas2 and Cas4-Cas1-Cas2 complexes contained the hairpin target DNA ( Figure 1D ) . Together , these data suggest that the proteins directly interact with the hairpin target , and the formation of the higher-order complex is also stabilized by the presence of dsDNA substrates . To characterize the molecular architecture of the complexes , we next performed single-particle electron microscopy ( EM ) of negatively stained Cas1-Cas2 or Cas4-Cas1-Cas2 complexes bound to the target . Raw micrographs and two-dimensional ( 2D ) class averages revealed particles with fairly homogenous size and symmetrical architecture consistent with the known structure of the Cas1-Cas2 complex ( Nuñez et al . , 2014; Nuñez et al . , 2015b; Wang et al . , 2015; Xiao et al . , 2017 ) ( Figure 2—figure supplement 1A–B ) . Some 2D class averages for the Cas4-Cas1-Cas2 complex contained clear additional density , suggesting the presence of ordered Cas4 within the complex ( Figure 1E–F ) . For the Cas1-Cas2 complex , we determined a 22 Å three-dimensional ( 3D ) reconstruction ( Figure 2A ) . The EM density revealed clear C2 symmetry , which was enforced during the final 3D refinement ( Figure 2—figure supplement 2A ) . Segmentation of the density revealed three clear domains , corresponding to two Cas1 dimers sandwiching a Cas2 dimer ( Figure 2A ) . We fit the B . halodurans Cas2 structure and a structural model of B . halodurans Cas1 ( see Materials and methods ) to the segmented densities . The structural models fit well , and the overall architecture of the complex was similar to that of the crystal structure of the type II-A Cas1-Cas2-prespacer complex from Enterococcus faecalis ( Figure 2—figure supplement 3A , B ) . In contrast , the E . coli type I-E Cas1-Cas2-prespacer complex ( Figure 2—figure supplement 3C ) did not fit well in the Cas1-Cas2 density , revealing that the architecture of type I-C Cas1-Cas2 may be more similar to type II-A than to another type I system ( Nuñez et al . , 2015b; Xiao et al . , 2017 ) . For the Cas4-Cas1-Cas2 complex , we determined a 20 Å 3D reconstruction of symmetrical particles enforcing C2 symmetry ( Figure 2—figure supplement 2B ) . The segmented density clearly reveals the base Cas1-Cas2 architecture , along with additional density corresponding to two molecules of Cas4 ( Figure 2B ) . During 3D classification of particles , we observed two classes containing approximately 50% of Cas4-Cas1-Cas2 particles that appeared to contain density for only a single Cas4 molecule ( Figure 2—figure supplement 2B ) . A subset of these particles was refined as a separate 21 Å 3D reconstruction without symmetry enforced , revealing an asymmetrical Cas4-Cas1-Cas2 complex with 1:4:2 stoichiometry ( Figure 2C ) . These particles may represent ternary complexes in a partially dissociated state , or incomplete formation of the 2:4:2 stoichiometry complex due to reduced affinity for Cas4 within the ternary complex . In both the symmetrical and asymmetrical Cas4-Cas1-Cas2 reconstructions , the Cas4 and Cas1 densities are contiguous but Cas4 appears to be distinct from the Cas2 density ( Figure 2B–C ) . This suggests that Cas4 interacts with Cas1 but not Cas2 within the ternary complex , similar to the tight interaction we have previously observed between Cas4 and Cas1 in the absence of Cas2 ( Lee et al . , 2018 ) . However , the interaction surface between Cas4-Cas1 appears to be different in the context of the binary and ternary complexes , suggesting that Cas4 and Cas1 may have two alternative modes of interaction . A high-resolution structure of the BhCas4 sequence is not available , and the closest homolog from Pyrobaculum calidifontis ( PDB 4R5Q ) contains a long N-terminal domain that is likely not present in the BhCas4 structure ( Figure 2—figure supplement 4A ) . Therefore , for modeling Cas4 into the Cas4-Cas1-Cas2 density , we used a predicted structural model ( see Materials and methods ) that reflects the putatively more compact BhCas4 structure ( Figure 2—figure supplement 4B ) . Fitting this structural model into the segmented Cas4 density revealed four possible orientations of the protein with respect to Cas1 ( Figure 2—figure supplement 4C–E ) . The segmented density for Cas4 is smaller than the structure and parts of the structure fit into the Cas1 density , suggesting that the segmentation between Cas1 and Cas4 densities was incomplete due to low resolution as has been observed previously ( Pintilie et al . , 2010 ) . All four orientations position the Cas4 active site with varying degrees of proximity ( 19–38 Å ) to the Cas1 active sites that bind the 3ʹ-OH end of the prespacer in structures of Cas1-Cas2-prespacer ( Figure 2—figure supplement 4D–E ) . These active sites act as integrases for the two steps of integration . The proximity between Cas4 and the Cas1 active site suggests that the product of Cas4 cleavage could transit to the integrase active sites following cleavage . The Cas1-Cas2 and Cas4-Cas1-Cas2 EM volumes did not contain any obvious cylindrical density corresponding to the hairpin target DNA , consistent with prior observations that DNA is not readily observable by negative-stain single-particle EM ( Hochstrasser et al . , 2014; Nogales and Scheres , 2015 ) . We also note that the Cas4 density lies along the same surface of Cas1 that is expected to interact with elements within the CRISPR array during half-site and full-site integration , based on crystal structures of these intermediates ( Figure 2—figure supplement 5 ) . This suggests that the CRISPR hairpin target is either not present in the complex , or that it is bound in an alternate location . The observation that DNA co-elutes with the complex from the size-exclusion column disfavors the former possibility ( Figure 1D ) . Notably , in the highest-resolution symmetrical Cas4-Cas1-Cas2 reconstruction , we observe additional density along the prespacer-binding surface of Cas2 , but not the repeat-binding surface ( Figure 2—figure supplement 5 middle ) . Although we cannot definitively define this density as DNA , this observation suggests the DNA may be bound on the prespacer side of the complex and that the CRISPR target in our complex is not bound along the same surface as in prior crystal structures . Together , these observations suggest that Cas4 binding may be mutually exclusive with CRISPR binding along the repeat-binding surface , and that Cas4 may dissociate prior to integration . Although we hypothesized that the CRISPR hairpin target would stabilize the Cas1-Cas2 complex by binding along the repeat-binding surface , our structural studies of complexes formed in the presence of the hairpin target suggest this is not the case . Notably , the prespacer-binding surface of Cas2 interacts with DNA through non-specific backbone contacts ( Nuñez et al . , 2015b; Wang et al . , 2015; Xiao et al . , 2017 ) , and no structure of Cas1-Cas2 bound to CRISPR DNA alone has been previously reported . This could suggest that any dsDNA , including the CRISPR target , may preferentially bind along the prespacer-binding surface and that other DNA substrates binding in that location could promote stable complex formation . We therefore attempted to form the Cas4-Cas1-Cas2 complex using a prespacer substrate containing a 24 bp duplex with 15-nt 3ʹ overhangs containing 5ʹ-GAA-3ʹ PAM sequences , which we previously showed is processed by Cas4 in the presence of Cas1 and Cas2 ( Lee et al . , 2018 ) . Using a modified protocol ( see Materials and methods ) , we were able to successfully purify a peak containing Cas4 , Cas1 , Cas2 , and prespacer DNA from a size-exclusion column ( Figure 1—figure supplement 2 ) . The peak eluted slightly later than the Cas4-Cas1-Cas2-target complex , likely due to the smaller size of the DNA substrate ( Figure 1—figure supplement 2A ) . As with the Cas4-Cas1-Cas2-target complex , all three proteins and DNA were present in the peak fractions ( Figure 1—figure supplement 2B–C ) . To determine whether the prespacer complex has a similar architecture to the complex bound to the CRISPR hairpin target , we performed negative stain EM analysis of Cas4-Cas1-Cas2-prespacer ( Figure 2—figure supplements 1 , 2 and 6 ) . The 2D class averages revealed fewer symmetrical classes ( Figure 2—figure supplement 1B ) , and 3D classification revealed a smaller proportion of particles containing two copies of Cas4 relative to particles containing one or no copies of Cas4 ( Figure 2—figure supplement 2C ) . These data suggest that Cas4 is less stably bound in the Cas4-Cas1-Cas2-prespacer complex than in a complex containing a longer dsDNA . Both EM densities were at a lower resolution than the complexes solved using the hairpin target ( symmetrical EM density 22 Å , asymmetrical EM density 24 Å ) . Although features are less well-defined for these lower resolution densities , the overall architecture is similar between the complexes containing the two different DNA substrates ( Figure 2—figure supplement 6 ) , indicating that complex formation occurs similarly regardless of the type of DNA substrate . The observation that the presence of dsDNA substrates stabilized the Cas4-Cas1-Cas2 complex prompted us to hypothesize that binding to the CRISPR DNA may enhance processing activity by stimulating complex formation . To test this hypothesis , we analyzed cleavage of the prespacer substrate containing a 24 bp duplex with 15-nt 3ʹ overhangs containing 5ʹ-GAA-3ʹ PAM sequences , in the absence or presence of CRISPR DNA ( Figure 3A ) . In addition , we tested cleavage of a single 39-nt strand of this substrate ( Figure 3A ) . Interestingly , for the ssDNA substrate , we observed cleavage product in the presence of the CRISPR DNA , while the DNA remained uncleaved in the absence of the CRISPR DNA or Cas4 ( Figure 3B ) . However , we observed similar amounts of cleavage for duplex prespacers with and without the CRISPR , suggesting that the presence of the CRISPR DNA only enhances cleavage of the ssDNA substrate ( Figure 3B ) . Importantly , we previously showed that cleavage of the duplex prespacer is dependent on having a PAM present in the single-stranded overhang , indicating that cleavage occurs in a PAM-dependent manner ( Lee et al . , 2018 ) . Consistently , both the ssDNA substrate and the duplex substrate produced the same length product , suggesting they were cleaved at the PAM site located within the 3ʹ-overhang of the duplex substrate . These data suggest that Cas4-Cas1-Cas2 can cleave PAM sites in ssDNA when a CRISPR DNA is provided in trans , and that the dsDNA region present within the duplex prespacer is sufficient to stimulate complex formation . We previously observed that Cas4 cleavage products can be integrated into mini-CRISPR arrays ( Lee et al . , 2018 ) ( Figure 3—figure supplement 1A ) . Notably , although the cleavage product for the duplex prespacer was integrated into the CRISPR ( Figure 3B ) , no integration product was visible for the single-stranded substrate following processing even at increased image contrast ( Figure 3—figure supplement 1B ) . In the absence of Cas4 , a small amount of unprocessed integration product was visible when the image contrast was increased for both the ssDNA and duplex prespacer . Together , these results suggest that although Cas1 can integrate ssDNA , optimal integration requires that the single-stranded region be attached to a duplex , which may facilitate handoff of the processed end from Cas4 to the Cas1 active site . Our results suggest that any dsDNA , not just the CRISPR DNA , may enhance ssDNA cleavage activity either when present in cis with the ssDNA region or when provided in trans . To test these possibilities , we performed a cleavage assay with two sets of substrates: a 25 bp duplex with one 25-nt 3ʹ overhang or 5ʹ-end-32P-labeled 25-nt ssDNA with an unlabeled blunt-end 25 bp duplex added in trans ( Figure 3C ) . As expected , the duplex substrate was cleaved within the single-stranded overhang region in the presence of Cas4 , similar to the duplex substrate tested in Figure 3B ( Figure 3D ) . For the ssDNA substrate , no cleavage was observed in the absence or at low concentrations of dsDNA , while cleavage product accumulated at higher dsDNA concentrations ( Figure 3D ) . A similar amount of cleavage product was observed at dsDNA concentrations ≥ 100 nM , which is higher than the expected concentration of the Cas4-Cas1-Cas2 complex ( 50 nM ) . These results are consistent with our model that dsDNA directly stabilizes the complex , and that excess dsDNA should not induce additional activation . Overall , our results show that type I-C Cas4-Cas1-Cas2 adaptation complex can be stabilized by dsDNA and that , within this complex , Cas4 is activated for ssDNA cleavage for substrate provided either in cis or in trans . We previously showed that Cas4 processes prespacers in a PAM-dependent manner ( Lee et al . , 2018 ) , but the exact cleavage sites and specificity of Cas4 remain unclear . The observation that Cas4-Cas1-Cas2 can process ssDNA allowed us to more precisely define the cleavage site . By using ssDNA substrates , we could more readily test sequences containing multiple PAM sites or a variety of sequences , and additionally include primer-binding sites for Sanger sequencing reactions . We first conducted prespacer processing assays using ssDNA substrates containing a 5ʹ-GAA-3ʹ PAM between T-rich sequences in the presence of activating dsDNA . Comparison with ddNTP Sanger sequencing reactions revealed that Cas4-Cas1-Cas2 precisely cleaved the ssDNA directly upstream of the PAM , while Cas1-Cas2 or Cas4 alone failed to cleave the substrates ( Figure 4A ) . This cleavage site is consistent with the expected processing site relative to the PAM required to form a functional spacer during spacer acquisition . We next tested whether the Cas4-Cas1-Cas2 could cleave at multiple PAM sites within a single substrate , and whether location of the PAM sites affects the processing activity . We designed ssDNA substrates containing three PAM sites at varying intervals . Substrates with 10 , 8 , 6 , 4 or 2-nt between three PAM sites generated three different sized cleavage products , resulting from cleavage directly upstream of each PAM site ( Figure 4B and Figure 4—figure supplement 1 ) . However , when we introduced three PAM sites consecutively on ssDNA substrates , we observed a predominant product at the first PAM position ( Figure 4C ) , indicating that processing was inhibited at the second and third PAM site . Together , these results suggest that the adaptation complex can process multiple PAM sites within a single substrate as long as the PAMs are not consecutive . To explore whether the PAM-flanking sequences affect cleavage and whether cleavage could be observed at non-PAM sequences , we used PAM-flanking sequences containing either A-T rich or ‘random’ sequences in which all four nucleotides were represented ( Figure 5 ) . Although we observed low levels of non-specific degradation in the presence of Cas4-Cas1-Cas2 , all substrates displayed a single predominant cleavage product directly upstream of the PAM site ( Figure 5A–D ) . These results indicate that Cas4 is highly PAM specific , as it did not cleave efficiently at other sites in the random sequences . We observed relatively low cleavage for substrates containing random sequences upstream of the PAM , suggesting that the PAM-flanking sequence may have an effect on cleavage efficiency . To test this possibility , we used ssDNA substrates containing degenerate PAM-flanking sequences for Cas4-Cas1-Cas2 cleavage ( Figure 5—figure supplement 1A–B ) . Using a PCR strategy that selectively amplified the uncleaved fraction ( Figure 5—figure supplement 1A , C ) , we analyzed the relative depletion of flanking sequence in the absence or presence of Cas4 by high-throughput sequencing . We detected no significant differences in relative sequence levels for the –Cas4 and +Cas4 samples ( Figure 5—figure supplement 1D–E ) , indicating that all PAM-flanking sequences were cleaved with the same efficiency . Together , our cleavage analyses strongly suggest that Cas4 cleavage specificity is dictated only by the presence of a PAM sequence within a single-stranded region of the substrate . Our data strongly suggests that Cas4 can cleave PAM sites within single-stranded DNA substrates , and that cleavage occurs at the same site when the ssDNA is provided in cis or in trans with the activating dsDNA ( Figure 3B ) . However , for a duplex substrate containing a ssDNA overhang , it is also possible that Cas1-Cas2-substrate binding positions the single-stranded overhangs into the Cas4 active site . In that case , Cas4 could cleave the DNA based on a ruler mechanism , with the cleavage site defined by the distance between the end of the duplex and the Cas4 active site . To test this possibility , we designed a panel of duplex substrates in which the PAM position was varied within the single-strand overhang ( Figure 6A ) . The four different substrates yielded products of different lengths , indicating that they were cleaved based on the position of the PAM within the single-stranded overhang ( Figure 6B ) . Interestingly , Cas4 processed these substrates to varying degrees ( Figure 6B–C ) . The substrate in which the PAM was located close to the duplex ( two nt between duplex and PAM site ) was cleaved inefficiently , while the other three substrates were cleaved with similar efficiency ( Figure 6C ) . These data , along with prior results showing that Cas4 cannot cleave a PAM contained within a dsDNA ( Lee et al . , 2018 ) , suggest that the PAM must be located at a longer distance from the duplex to be accommodated within the Cas4 active site . The negative-stain EM volumes for the Cas1-Cas2-target , asymmetrical Cas4-Cas1-Cas2-target , symmetrical Cas4-Cas1-Cas2-target , asymmetrical Cas4-Cas1-Cas2-prespacer and symmetrical Cas4-Cas1-Cas2-prespacer complexes have been deposited to EMDB under the accession numbers EMDB-20127 , EMDB-20128 , EMDB-20129 , EMDB-20130 and EMDB-20131 , respectively . Further information and requests for resources and reagents should be directed to and will be fulfilled by Dipali Sashital ( sashital@iastate . edu ) . Cas1 , Cas2 , and Cas4 were purified as previously described ( Lee et al . , 2018 ) . For complex formation using CRISPR hairpin DNA target , Cas1 , Cas2 , and hairpin DNA substrates or Cas4 , Cas1 , Cas2 , and hairpin DNA substrates were mixed in equal molar ratios ( 1:1:1 or 1:1:1:1 ) and dialyzed against 1 L of buffer A ( 20 mM HEPES ( pH 7 . 5 ) , 100 mM NaCl , 5% glycerol , 2 mM DTT , and 2 mM MnCl2 ) overnight at 4°C . The samples were loaded on a 5 mL HiTrap Q column ( GE Healthcare ) equilibrated with buffer A and the free Cas2 and free DNA were separated from the complex using a gradient of buffer B ( 20 mM HEPES ( pH7 . 5 ) , 1M NaCl , 5% glycerol , 2 mM DTT , 2 mM MnCl2 ) . Fractions containing all two or three proteins were pooled , concentrated , and further purified using a Superdex 75 10/30 GL column ( GE Healthcare ) in size exclusion buffer A ( 20 mM HEPES ( pH 7 . 5 ) , 100 mM KCl , 5% glycerol , 2 mM DTT , and 2 mM MnCl2 ) for Cas1-Cas2-target complex and size exclusion buffer B ( 20 mM HEPES ( pH 7 . 5 ) , 250 mM KCl , 5% glycerol , 2 mM DTT , and 2 mM MnCl2 ) for Cas4-Cas1-Cas2-target complex . For complex formation using prespacer DNA substrate , 9 . 8 µM Cas4 , 6 . 5 µM Cas1 , 3 . 3 µM Cas2 , and 5 µM prespacer were mixed ( final approximate ratio of 3:2:1:1 . 5 ) in a final volume of 500 µl in size exclusion buffer B and incubated for 30 min at 4°C . The Cas4-Cas1-Cas2-prespacer complex was purified using a Superdex 75 10/30 GL column ( GE Healthcare ) in size exclusion buffer B . All oligonucleotides were synthesized by Integrated DNA Technologies or Thermo Scientific . Sequences of DNA substrates are shown in Table 1 . All DNA substrates were purified on 10% urea-PAGE . Double-stranded DNA was hybridized by heating to 95°C for 5 min followed by slow cooling to room temperature in oligo annealing buffer ( 20 mM HEPES ( pH 7 . 5 ) , 25 mM KCl , 10 mM MgCl2 ) . Prespacers were labeled with [γ-32P]-ATP ( PerkinElmer ) and T4 polynucleotide kinase ( NEB ) for 5′-end labelling . Excess ATP was removed using Illustra Microspin G-25 columns ( GE Healthcare ) . To prepare grids for EM imaging , Cas1-Cas2-target , Cas4-Cas1-Cas2-target or Cas4-Cas1-Cas2-prespacer were diluted to ~100 nM , and 3 µL of sample was applied to a glow-discharged copper 400-mesh continuous carbon grid for one minute at room temperature . The excess sample was blotted with Whatman filter paper , followed by immediate application of 3 µL 2% ( w/v ) uranyl formate . The excess stain was blotted , followed by immediate application of 3 µL 2% uranyl formate . This step was repeated once more . The grids were allowed to dry for at least 5 min prior to imaging . Images were collected on a 200 keV JEOL 2100 transmission electron microscope equipped with a Gatan OneView camera at a nominal magnification of 60 , 000x and pixel size of 1 . 9 Å . The electron dose was between 30 and 40 electrons/Å2 . For each sample , images ( 200 for Cas1-Cas2-target and Cas4-Cas1-Cas2-target and 93 for Cas4-Cas1-Cas2-prespacer ) were collected manually at a defocus range of 1–2 . 5 µm . All image processing and analysis was performed in Scipion v . 1 . 2 ( de la Rosa-Trevín et al . , 2016 ) ( RRID:SCR_016738 ) ( available at http://scipion . i2pc . es/ ) . The contrast transfer function ( CTF ) for each micrograph was estimated using CTFFIND4 ( Rohou and Grigorieff , 2015 ) ( RRID:SCR_016732 ) . For each sample , ~200 particles were picked using Xmipp manual picking , followed by automated picking using the manually picked particles as a training set ( Abrishami et al . , 2013; Sorzano et al . , 2013; Vargas et al . , 2013 ) . In total , 95 , 669 Cas1-Cas2-target , 115 , 445 Cas4-Cas1-Cas2-target and 32 , 494 Cas4-Cas1-Cas2-prespacer particles were present in the initial data set . Particles were extracted using a 160 × 160 pixel box . To reduce computational requirements , the particles were down sampled by a factor of 2 to a final box size of 80 × 80 pixels ( ~152×152 Å ) . The particles were normalized and subjected to reference-free 2D classification using RELION 2 . 1 ( Scheres , 2012 ) ( RRID:SCR_016274 ) ( Figure 2—figure supplement 1B ) . The initial 100 class averages were inspected , and averages with clear structural features , the largest number of particles and size consistent with the molecular weight of the complex were selected for further analysis . These particles were subjected to a second round of 2D classification into 50 classes using RELION to further clean the particles . After selection of the best 2D classes , the Cas1-Cas2-target dataset contained 34 , 626 particles , the Cas4-Cas1-Cas2-target dataset contained 49 , 173 particles , and the Cas4-Cas1-Cas2-prespacer dataset contained 19 , 620 particles . Particles were next subjected to 3D classification in RELION using the X-ray crystal structure of E . coli Cas1-Cas2 ( PDB: 4P6I ( Nuñez et al . , 2014 ) ) low-pass-filtered to 90 Å as a starting model ( Figure 2—figure supplement 2 ) . The target-bound complex datasets were initially classified into six classes , while the prespacer-bound complex was classified into five classes due to the lower number of starting particles . For Cas1-Cas2-target , 11 , 636 particles from two similar 3D classes with the clearest density were combined ( Figure 2—figure supplement 2A ) . These particles were subjected to 3D refinement using RELION , and the refined volume was used to create a 3D mask . The refined particles were subjected to a second round of classification into three classes using the 3D mask as a reference mask ( Figure 2—figure supplement 2A ) . A class containing 5279 particles with the clearest density was selected . This class contained clear C2 symmetry . These particles were subjected to 3D refinement while enforcing C2 symmetry and using the 3D mask yielding the final reconstruction . For Cas4-Cas1-Cas2-target , 37 , 051 particles from four out of six initial 3D classes that appeared to contain Cas1-Cas2 with additional density were selected for further refinement ( Figure 2—figure supplement 2B ) . These particles were subjected to 3D refinement and then further classified into four 3D classes . The resulting classes had clearly defined extra density in comparison to the Cas1-Cas2 3D reconstruction . For two of these classes , the density displayed clear C2 symmetry ( Figure 2—figure supplement 2B ) , with extra density extending from each Cas1 lobe , while for the other two classes , the extra density was only observed extending from one Cas1 lobe . Particles ( 18 , 290 ) from the two symmetrical 3D classes were combined and subjected to 3D refinement while enforcing C2 symmetry . Particles ( 9 , 160 ) from one of the asymmetrical 3D classes were subjected to 3D refinement with C1 symmetry . For Cas4-Cas1-Cas2-prespacer , 11 , 001 particles from two out of five initial 3D classes that appeared to contain density in addition to the Cas1-Cas2 core were selected for further refinement ( Figure 2—figure supplement 2C ) . These particles were subjected to 3D refinement and then further classified into three 3D classes . The resulting classes had varying degrees of extra density . Class 1 ( 3 , 682 particles ) contained no apparent extra density in comparison to Cas1-Cas2-target , while the other two classes resembled the asymmetrical ( 4668 particles ) and symmetrical ( 2651 particles ) Cas4-Cas1-Cas2-target densities . Particles from the symmetrical 3D class were subjected to 3D refinement while enforcing C2 symmetry . Particles from the asymmetrical 3D classes were subjected to 3D refinement with C1 symmetry . For all Cas4-Cas1-Cas2 samples , the refined 3D reconstructions were used to create 3D masks , and each set of particles was subjected to a final round of refinement using the 3D mask as reference mask . The resolutions of the final 3D reconstruction were 22 . 1 Å , 19 . 7 Å , 21 . 4 Å , 21 . 6 Å and 24 . 4 Å for Cas1-Cas2-target , symmetrical Cas4-Cas1-Cas2-target , asymmetrical Cas4-Cas1-Cas2-target , symmetrical Cas4-Cas1-Cas2-prespacer and asymmetrical Cas4-Cas1-Cas2-prespacer , respectively , based on Fourier Shell Correlation ( FSC ) at a cutoff of 0 . 5 ( Figure 2—figure supplement 1C–D ) . The 0 . 5 FSC criterion was used to ensure that resolution was not overestimated . The Euler angle plots for the final 3D reconstructions revealed some preferred orientations but indicated a wide angular distribution in the data ( Figure 2—figure supplement 1E ) . Volumes were segmented using Segger ( Pintilie et al . , 2010 ) in UCSF Chimera ( Pettersen et al . , 2004 ) . Structural models for the B . halodurans Cas1 and Cas4 sequences were predicted using the Phyre2 webserver ( http://www . sbg . bio . ic . ac . uk/phyre2/html/page . cgi ? id=index ) ( Kelley et al . , 2015 ) . The top results provided structural models based on the closest homologs of BhCas1 and BhCas4 available in the protein databank . For Cas1 , the closest homolog is Cas1 from Archaeoglobus fulgidus ( PDB 4N06 , 28% identical , 65% similar ) ( Kim et al . , 2013 ) . For Cas4 , the closest available homolog structure is from Pyrobaculum calidifontis ( PDB 4R5Q , 15% identical , 44% similar ) ( Lemak et al . , 2014 ) . The crystal structure of B . halodurans Cas2 ( PDB 4ES3 ) was used for fitting to density assigned to Cas2 ( Nam et al . , 2012 ) . Fitting of individual copies of Cas1 dimers , Cas2 dimer or Cas4 monomers into assigned densities was performed using the ‘Fit in Segments’ tool in Segger within UCSF Chimera ( RRID:SCR_004097 ) ( Pettersen et al . , 2004; Pintilie et al . , 2010 ) . For Cas4 , the top four fits are shown for segmented volumes of the symmetrical and asymmetrical Cas4-Cas1-Cas2-target complexes in Figure 2—figure supplement 4D–E . The cross-correlation score provided by Segger is reported in the figure . The distance between the alpha carbon atom of the Cas1 H234 and Cas4 K110 active site residues was measured using the ‘Distances’ tool in UCSF Chimera . For analyzing fit of Cas1-Cas2 crystal structures in the Cas1-Cas2-target density , the protein subunits of the X-ray crystal structure of E . faecalis Cas1-Cas2 bound to prespacer ( PDB: 5XVN [Xiao et al . , 2017] ) or E . coli Cas1-Cas2 bound to prespacer ( PDB 5DS4 [Nuñez et al . , 2015b] ) were docked into the final 3D reconstruction using the Fit to Segments tool in Segger ( Figure 2—figure supplement 3 ) . Prespacer processing assays were performed using 25 nM of 5′-radiolabeled substrate with 500 nM Cas4 , 200 nM Cas1 , 200 nM Cas2 and 1 μM mini-CRISPR DNA ( as indicated ) or indicated amount of activating dsDNA in integration buffer ( 20 mM HEPES ( pH 7 . 5 ) , 100 mM KCl , 5% glycerol , 2 mM DTT , and 2 mM MnCl2 ) . The proteins and activating dsDNA were incubated on ice for 10 min prior to addition of the 5ʹ-labeled substrate . All reactions were performed at 65°C for 20 min . Reactions were quenched with 2X RNA loading dye ( NEB ) and heated at 95°C for 5 min followed by cooling on ice for 3 min . Samples were run on 12% urea-PAGE . The gels were dried and imaged using phosphor screens on a Typhoon imager . Sequenase Version 2 . 0 DNA sequencing kit ( Applied Biosystems ) was used for Sanger sequencing lanes . Samples were prepared by hybridizing template with 5ʹ-radiolabeled primer at 65°C for 2 min and slowly cooling to RT for 30 min . Samples were incubated with the chain terminators ( ddGTP , ddCTP , ddATP or ddTTP ) at 37°C for 15 min and quenched with 2X RNA dye . The cleaved products were prepared in the presence of 1 . 2 μM Cas4 , 600 nM Cas1 , 600 nM Cas2 , and 1 μM of activator 48 bp dsDNA and quenched with 2X RNA loading dye . All samples were heated at 95°C for 5 min followed by cooling on ice for 3 min . Samples were run on 0 . 4 mm 8% urea-PAGE . The gels were dried and imaged using a Typhoon imager . Cas4 cleavage assays for the panel of duplex substrates in which the PAM position was varied within the single-strand overhang were performed as described above , with the following alterations . For preparing duplex substrates , 25 nM 5ʹ radiolabeled top strand in which the PAM position was varied was annealed to 50 nM unlabeled bottom strand in buffer ( 20 mM HEPES ( pH 7 . 5 ) , 100 mM KCl , 5% glycerol , 2 mM DTT , 2 mM MnCl2 , ) . The hybridization reactions were incubated at 95°C for 3 min followed by slow cooling to room temperature . A final concentration of 5 nM radiolabeled substrate was used in the reaction . Processing reactions were performed in triplicate , and the intensity of bands was measured by densitometry using ImageJ ( Schneider et al . , 2012 ) . The fraction cleaved was calculated by dividing the product band by the sum of both bands . The values from three replicates were averaged , and error is reported as standard deviation between the replicates . The ssDNA substrates containing between 1–4 degenerate nucleotides upstream or downstream or the PAM were purified with 10% urea-PAGE . For cleavage , 20 nM of each substrate was incubated with 600 nM Cas1 , 600 nM Cas2 and 1 μM activator dsDNA in the presence or absence of 1 . 2 μM Cas4 . Samples without Cas4 were used as negative control . Three separate samples were prepared for each condition and treated as separate replicates . After ssDNA cleavage , the uncleaved products were hybridized with a primer that was complementary to the 3ʹ end and extended using Klenow Fragment ( NEB ) . Samples were extracted with phenol-chloroform-isoamyl alcohol and precipitated with ethanol . The samples were amplified by PCR using Platinum SuperFi DNA polymerase ( Thermo ) with primers containing Nextera adapters followed by a second round of PCR with primers containing i5 and i7 barcodes . Amplification products were analyzed on 2% SYBR Safe stained agarose gels and quantified using densitometry . Samples were mixed in equal quantities and were run on 2% agarose gel . The band was excised and DNA was purified using QIAquick Gel Extraction kit ( Qiagen ) . Samples were submitted for Illumina MiSeq sequencing to the Iowa State University DNA facility . The de-multiplexed datasets were analyzed separately to determine the relative read counts for each possible sequence in the degenerate regions . The degenerate regions of the sequences were cut from the reads , and the number of counts for each unique sequence was determined using bash commands . The reads were normalized by dividing the number of reads for each sequence by the total number of reads for the dataset . The normalized reads from three replicates were averaged for each substrate for reactions performed in the absence or presence of Cas4 , and error was calculated as standard deviation between the replicates .	Many people have now heard of CRISPR , or CRISPR-Cas9 , as a gene editing technology . Yet CRISPR evolved in bacteria to protect them against viral infections . While parts of the CRISPR system are now being widely used , the research community still does not know everything about how the system operates in its natural setting . In bacteria , CRISPR protects against infection by making lasting records of viruses a cell has encountered . It cuts short sections from the viral DNA and keeps them as a way to fight the virus if it ever returns . The key proteins in collecting and storing the virus DNA are called Cas1 , Cas2 and Cas4 . Previous work suggests that Cas4 is important for cutting suitable lengths of DNA for storage . Yet , how Cas4 , Cas1 and Cas2 work together to select , cut and store DNA is not well studied . Lee et al . have now used electron microscopy to examine how Cas1 , Cas2 and Cas4 cooperate in the CRISPR system . The proteins studied came from bacteria called Bacillus halodurans . The structure revealed direct links between the Cas1 and Cas4 proteins that likely help to ensure these proteins are coordinated correctly to cut and store the DNA sections . Specifically , it showed that two Cas4 proteins interact with the two key active sites of Cas1 . The findings also highlight that Cas4 cuts DNA at specific locations to make sure the resulting DNA sections are suitable for CRISPR protection . The close association between Cas1 and Cas4 could be a critical aspect of the reliability of the CRISPR system in protecting bacteria from viruses . There are more bacteria on Earth than any other living thing . Understanding their biology has wide ranging environmental , health and bioengineering applications . In addition , learning more about the CRISPR system could further expand its potential to drive revolutionary biotechnology tools derived from these bacterial immune systems .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2019	The Cas4-Cas1-Cas2 complex mediates precise prespacer processing during CRISPR adaptation
Human standing balance relies on self-motion estimates that are used by the nervous system to detect unexpected movements and enable corrective responses and adaptations in control . These estimates must accommodate for inherent delays in sensory and motor pathways . Here , we used a robotic system to simulate human standing about the ankles in the anteroposterior direction and impose sensorimotor delays into the control of balance . Imposed delays destabilized standing , but through training , participants adapted and re-learned to balance with the delays . Before training , imposed delays attenuated vestibular contributions to balance and triggered perceptions of unexpected standing motion , suggesting increased uncertainty in the internal self-motion estimates . After training , vestibular contributions partially returned to baseline levels and larger delays were needed to evoke perceptions of unexpected standing motion . Through learning , the nervous system accommodates balance sensorimotor delays by causally linking whole-body sensory feedback ( initially interpreted as imposed motion ) to self-generated balance motor commands . The nervous system learns and maintains motor skills by forming probabilistic estimates of self-motion . The resulting inferred relationships between sensory and motor signals form a representation of the world and self that allows the brain to identify unexpected behavior and adapt motor control ( Friston , 2010; Krakauer and Mazzoni , 2011; Wolpert et al . , 2011 ) . Due to neural conduction delays , these estimates of self-motion rely on the expected timing between motor commands and resulting sensory feedback . As such , errors associated with self-generated movement increase with larger feedback delays ( Gifford and Lyman , 1967; Miall et al . , 1985; Smith et al . , 1960 ) . Through repeated exposure to an imposed delay , the brain can learn to expect the delayed feedback associated with self-motion , leading to improvements in movement control with delays up to 430 ms ( Cunningham et al . , 2001; Miall and Jackson , 2006 ) . When balancing upright , sensory feedback associated with lower-limb motor commands is delayed by up to ~100–160 ms ( Forbes et al . , 2018; Kuo , 2005; van der Kooij et al . , 1999 ) . As a consequence of these relatively long delays , computational feedback models of upright standing predict that balance controllers cannot adjust their sensorimotor gains and stabilize balance in the anteroposterior ( AP ) direction with imposed delays larger than ~300–340 ms ( Milton and Insperger , 2019; van der Kooij and Peterka , 2011 ) . These predictions contrast the reported upper-limb sensorimotor adaptation to imposed delays ( Cunningham et al . , 2001; Miall and Jackson , 2006 ) . The present study aims to directly quantify the destabilizing effects of imposed delays between ankle torque and whole-body motion during standing balance , and to determine the underlying mechanisms responsible for any subsequent adaptation and learning . Imposed delays inserted within the balance control task are expected to increase postural oscillations and , past a critical delay , lead to falls ( Bingham et al . , 2011; Milton and Insperger , 2019; van der Kooij and Peterka , 2011 ) . The sensorimotor mechanisms underlying these predicted effects , however , are unknown . Of particular interest is the vestibular control of balance due to its task-dependent modulations ( Fitzpatrick and McCloskey , 1994; Forbes et al . , 2016; Luu et al . , 2012; Mian and Day , 2014 ) , which rely on predictable associations between self-generated motor and resulting sensory signals . For example , participants exposed to novel vestibular feedback of balance motion initially exhibit increased postural oscillations but decreased muscle responses to a vestibular error signal ( Héroux et al . , 2015 ) . With practice , participants improve their balance and vestibular-evoked muscle responses return to baseline amplitudes , suggesting that the brain updated its vestibular estimates of self-motion . Based on these observations , we hypothesized that increasing balance delays would initially increase whole-body motion and attenuate vestibular-evoked responses but these effects would diminish following a learning period . Another critical feature of probabilistic associations between motor and sensory cues is our ability to perceptually distinguish between self-generated or externally imposed motions . Imposed sensorimotor delays during self-generated movements evoke a sensation interpreted to arise from external causes rather than oneself ( Blakemore et al . , 1999; Farrer et al . , 2008; Wen , 2019 ) . Repeated exposures to delayed self-generated touch can re-align the perceived timing of the contact with the imposed delay ( Kilteni et al . , 2019; Stetson et al . , 2006 ) . Therefore , we further hypothesized that balance behavior under imposed delays would be inferred as externally imposed motion but this likelihood would decrease through repeated exposure to the delay . To test these hypotheses , we performed three experiments where participants balanced in a robotic simulator ( Figure 1 ) in the AP direction with imposed delays ranging from 20 to 500 ms . These delays were in addition to the physiological delays ( ~100–160 ms ) inherent to standing balance . In Experiment 1 , we characterized standing behavior across this range of delays . Generally , whole-body sway variability increased with larger imposed delays and participants repeatedly fell into virtual limits of the balance simulation ( i . e . , 6° anterior and 3° posterior ) for added delays larger than ~200–300 ms . In Experiment 2 , participants trained to balance upright with a 400 ms added delay ( testing beyond the critical delay previously proposed ) for 100 min over five consecutive days . We probed the vestibular-evoked muscle responses and the perception of body motion before , after and 3 months following training to assess how the brain adapted to and processed the delayed sensory feedback . Initially , participants exhibited increased postural oscillations while the vestibular contributions to balance decreased and their perception of unexpected balance motion increased . After training , participants’ balance behavior improved , their vestibular-evoked responses increased , and larger imposed delays were needed to elicit perceptions of unexpected balance motion . To further evaluate the effect of imposed delays on the vestibular control and perception of balance , we exposed participants to transient delays ( Experiment 3 ) . Within a few seconds of transitioning to a 200 ms delay , whole-body sway variability increased , vestibular responses attenuated ( ~70–90% decline ) and participants perceived unexpected balancing motion . Collectively , our findings demonstrate how novel sensorimotor delays disrupt standing balance and suggest that the nervous system can learn to maintain standing balance with imposed delays by associating delayed whole-body motion with self-generated balancing motor commands . Thirteen healthy participants were instructed to stand quietly on a robotic balance simulator for 60 s trials ( Materials and methods ) while experiencing fixed imposed delays ( 20 , 100 , 200 , 300 , 400 , 500 ms ) between the torques generated at their feet and resulting whole-body motion . These delays were in addition to the ~100–160 ms sensorimotor feedback delays inherent to standing balance . Whole-body sway was recorded throughout these trials to quantify the effect of the additional delay on standing balance . The robot was programmed to rotate the whole body in the AP direction about the participant’s ankles . Angular position limits of 6° anterior and 3° posterior from vertical were imposed into the simulation to represent the physical limits of sway during standing balance , whereby the robot constrained the angular rotation when these virtual limits were exceeded ( see Materials and methods ) . While balancing on the robotic simulator at the 20 ms delay ( baseline condition ) , all participants maintained standing balance with small postural oscillations around their preferred upright posture ( sway velocity variance: 0 . 07 ± 0 . 07 [°/s]2 [mean ± standard deviation] ) . Whole-body oscillations increased with the imposed delays , leading to marked difficulties in maintaining a stable posture when a 400 ms delay was imposed . Representative data ( Figure 2A ) illustrate a participant exceeding the virtual balance limits ( i . e . , whole-body position traces exceeding dashed lines ) 20 times within a 60 s period . This observation was confirmed in the group data . No participant exceeded these virtual balance limits at the 20 ms condition ( only one participant reached the limit during the 100 ms condition ) whereas every participant exceeded the virtual limits at least once within the 60 s balance period when delays were ≥200 ms . There was a main effect of delay on sway velocity variance ( extracted over 2 s windows of continuous balance; Materials and methods ) , such that sway velocity variance was smallest for the 20 ms condition ( 0 . 07 ± 0 . 07 [°/s]2 ) and increased with the magnitude of the imposed delay and reached a maximum at 400 ms ( 21 . 08 ± 15 . 41 [°/s]2; p<0 . 001; Table 1 and Figure 2B ) . We also quantified the percent time participants balanced within the virtual limits . There was a main effect of delay on the percent time within the balance limits ( p<0 . 001; Table 1 and Figure 2B ) , which decreased from 100% ± 0% during the 20 ms condition to 54% ± 9% during the 500 ms condition . Decomposition of the main effects revealed that participants exhibited greater sway velocity variance and lower percentage of trial duration within the virtual limits compared to the 20 ms condition when imposed delays were ≥200 ms ( all p-values <0 . 05 ) . In a second set of experiments , we tested whether humans can adapt and learn to stand with imposed sensorimotor delays . Participants ( n = 16 ) performed a training protocol over five consecutive days ( two 10 min trials per day ) where they balanced on the robot with a 400 ms delay . To explore the neural processes involved in balancing with novel sensorimotor delays , we characterized the participants’ vestibular control of balance ( vestibular testing , see below ) or their perceptual detection of unexpected motion ( perceptual testing , see below ) before and after training . Twelve participants also returned ~3 months later to examine whether any learning was retained . Within the first minute of training with the 400 ms delay , no participant could remain upright: on average , they reached the forward or backward virtual balancing limits 18 ± 5 times ( see representative participant in Figure 3A ) and could only remain within the balancing limits for 64% ± 9 % of the time ( or 38 . 5 s ) . This unstable balancing behavior was characterized by large whole-body sway velocity variance ( 12 . 62 ± 9 . 03 [°/s]2 ) . During training , participants progressively reduced the variance of their sway velocity and increased the percentage of time they balanced within the virtual limits . The first minute of each day ( i . e . , start of every 20 min interval ) was characterized by an increase of sway velocity variance and a decrease of percentage of time within the limits relative to the last min of the previous day ( see filled circles , Figure 3B ) . By the end of training ( 100 min ) , participants exhibited an ~80% decrease in sway velocity variance and a 51% increase in percent time within the virtual limits . First-order exponential fits estimated the changes in sway velocity variance and percent time within the limits . The time constant for the decrease in sway velocity variance ( i . e . , 63 . 2% attenuation ) was 27 . 9 min ( corresponding to a value of 6 . 44 [°/s]2 ) , and the time constant for the increase in percent time within limits ( i . e . , 63 . 2% increase ) was 32 . 5 min ( corresponding to a value of 85% within the balance limits ) . By the last 60 s of training , participants could balance the robot within the simulation limits on average for 97% ± 3% of the time ( or 59 . 2 s ) , with four participants capable of balancing for the final 60 s interval without reaching a limit . However , the smallest sway velocity variance observed with a 400 ms imposed delay remained ~38× greater than the baseline condition ( 400 ms at 93rd min vs . 20 ms variance: 1 . 91 ± 1 . 12 [°/s]2 vs . 0 . 05 ± 0 . 05 [°/s]2; t ( 15 ) = 6 . 74: p<0 . 001 ) . When participants ( n = 12 ) returned for retention testing ~3 months later , these balance improvements were partially maintained . Sway velocity variance in the first minute of retention testing was ~60 . 8% lower than the sway velocity variance from the first minute of training ( 4 . 95 ± 2 . 32 [°/s]2 vs . 12 . 62 ± 9 . 03 [°/s]2; independent samples t-test: t ( 26 ) = –2 . 86 , p<0 . 01 ) . Sway velocity variance at the first minute of retention testing , however , remained greater than the last minute of training ( 4 . 95 ± 2 . 32 [°/s]2 vs . 2 . 55 ± 1 . 76 [°/s]2; independent samples t-test: t ( 26 ) = 3 . 11 , p<0 . 01 ) . Similarly , the first minute of retention was associated with a greater percentage of time within the balancing limits compared to the first minute of training ( 88% ± 9% vs . 64% ± 9%; independent samples t-test: t ( 26 ) = 6 . 67 , p<0 . 001 ) , but less than the last minute of training ( 88% ± 9% vs . 97% ± 3%; independent samples t-test: t ( 26 ) = –3 . 68 , p<0 . 01 ) . When using only data from participants who performed the retention session ( n = 12; paired t-tests with df = 11 ) , sway velocity variance and percent time within the balance limits revealed identical results ( all p-values < 0 . 01 ) . Overall , these results indicate that while standing with an imposed 400 ms delay is initially difficult ( if not impossible ) , participants learn to balance with the delay with sufficient training ( i . e . , >30 min ) and this ability is partially retained 3 months later . During vestibular testing , we probed the vestibular contribution to soleus muscle activity by exposing participants ( n = 8 ) to a non-painful electrical vestibular stimulus ( EVS ) while they balanced on the robot at different delays ( 20–500 ms; Materials and methods ) before , after , and 3 months following training . Vestibular-evoked muscle responses are known to attenuate when actual sensory feedback does not align with expected estimates from balancing motor commands ( Héroux et al . , 2015; Luu et al . , 2012 ) . Therefore , we hypothesized that increasing the delay between ankle torques and body motion would progressively diminish the vestibular response . We further hypothesized that learning to control balance with imposed delays would allow the brain to update its sensorimotor estimates of balance motion and consequently increase the vestibular-evoked muscle responses . Frequency domain measures ( coherence and gain; see Materials and Methods ) were evaluated qualitatively using the pooled participant estimates because with delays ≥ 200 ms , single-participant coherence only exceeded significance at sporadic frequencies and significant coherence is needed to obtain a reliable gain estimate . Our time-domain measure ( cross-covariance; see Materials and methods ) , which estimates the net vestibular contribution to muscle activity at all stimulated frequencies , was extracted on a participant-by-participant basis and used for statistical analysis . Participants exhibited the largest vestibular-evoked muscle responses ( coherence , gain , and cross-covariance ) for the 20 and 100 ms delay conditions , where significant coherence was observed at frequencies between 0 and 25 Hz and cross-covariance responses were characterized by short ( ~60 ms ) and medium ( ~100 ms ) latency peaks exceeding the 95% confidence interval ( see Figure 4A ) . Prior to learning , pooled coherence and gain decreased with imposed delays ≥ 200 ms , and coherence fell below the significance threshold at most frequencies for delays ≥ 300 ms . Similarly , cross-covariance amplitudes decreased with increasing delay ( ≥200 ms ) , with only five out of eight participants showing significant biphasic muscle responses ( cross-covariance ) for the 400 ms delay condition ( and six out of eight participants at 500 ms ) . Across training conditions ( pre , post , retention ) , increasing the delay reduced the cross-covariance peak-to-peak amplitudes ( main effect of delay , p<0 . 001; Figure 4A and B , Table 1 ) . Following training , pooled vestibular-evoked muscle responses ( coherence , gain , cross-covariance ) partially recovered in both the post-learning and retention phases . Every participant exhibited biphasic muscle responses that exceeded significance thresholds for every delay after training . A significant interaction between delay and learning was observed for the cross-covariance ( p<0 . 001 , Table 1 ) , suggesting that the recovery of vestibular responses was dependent on the delay magnitude . Planned comparisons ( Wilcoxon sign-rank test , Bonferroni corrected ) revealed that cross-covariance response amplitudes were larger during post-learning relative to pre-learning for delays ≥ 200 ms ( all p-values <0 . 05 ) and were larger during retention relative to pre-learning for 300 and 400 ms ( p<0 . 05 ) . Similar to Experiment 1 , sway velocity variance generally increased with increasing delays ( Figure 4C , Table 1 , Table 2; p<0 . 001 ) , while learning decreased sway velocity variance at almost all imposed delays ( Figure 4C , Table 1; p<0 . 001 ) , resulting in a significant delay × learning interaction ( Figure 4C , Table 1; p<0 . 001 ) . Planned comparisons ( paired t-tests , Bonferroni corrected ) revealed that sway velocity variance decreased during post-learning relative to pre-learning for delays between 20 and 400 ms ( all p-values <0 . 05 ) and decreased during retention relative to pre-learning for delays between 100 and 400 ms ( all p-values <0 . 05 ) . Because training was only performed with the 400 ms delay , these training-related changes in vestibular responses ( cross-covariance ) and sway behavior across delays indicate that learning generalized to different sensorimotor delays . During perceptual testing , we assessed whether participants ( n = 18 ) perceived unexpected balance motion when transient delays ( 20–350 ms applied for 8 s periods; see Materials and methods ) were imposed while balancing on the robot . Behavioral studies in humans suggest that delayed self-motion is perceived as unexpected ( or externally imposed ) because it does not align with prior expectations of the intended movement ( Blakemore et al . , 1999; Farrer et al . , 2008; Wen , 2019 ) . Therefore , we hypothesized that increasing the delay in the control of standing balance would evoke motion that is increasingly perceived as unexpected . We further hypothesized that learning to control balance with imposed delays would allow the brain to update its estimates of balance motion and consequently greater delays would be needed to elicit a perception of unexpected balance movements . When delays were transiently imposed during balance , whole-body sway became more variable and , as delays increased , participants perceived unexpected balance motion more often . Data from a representative participant ( see Figure 1E ) show missed detections of the 100 and 150 ms imposed delays , and this participant had a resulting 70% correct detection threshold occurring at a delay of 136 ms . Across participants , the probability of perceiving unexpected postural motion increased with delays: from 4% detection for the 50 ms delay up to 100% for the 350 ms delay ( Table 3 ) . We found no significant difference in the 70% correct detection thresholds for participants who did not participate in training ( i . e . , the no-learning group ) and those who did during the pre-learning phase ( 156 ± 33 ms vs . 147 ± 21 ms; independent samples t-test: t ( 16 ) = 0 . 706 , p=0 . 49 ) . On average , when unexpected balance motion was correctly detected , participants pressed the button at least ~2 s after the delay was imposed across all delays and learning conditions ( Table 3 ) . After participants ( n = 8 ) completed training with the 400 ms delay , psychometric functions shifted to the right , resulting in increased thresholds for detecting unexpected balance movements ( Figure 5A and B , Table 1; p=0 . 008 , Bonferroni corrected ) . Thresholds were larger during post-learning relative to pre-learning ( Figure 5B; 192 ± 40 ms vs . 147 ± 21 ms; p=0 . 032 , Bonferroni corrected ) and were larger for retention relative to pre-learning ( 209 ± 82 ms vs . 147 ± 21 ms; p=0 . 014 , Bonferroni corrected ) . For the whole-body oscillations ( sway velocity variance extracted over the 8 s periods when delays were imposed , see Table 1 ) , we again confirmed significant main effects of delay ( p<0 . 001 ) and learning ( p<0 . 001 ) as well as a delay × learning interaction ( p=0 . 023 ) . Planned comparisons ( paired t-tests , Bonferroni corrected ) revealed that sway velocity variance decreased during post-learning relative to pre-learning for delays ranging from 150 to 250 ms ( all p-values <0 . 05 ) and decreased during retention relative to pre-learning for the 150 , 200 , 300 , and 350 delays ( all p-values <0 . 05 ) . Our results from the vestibular and perceptual testing of Experiment 2 indicate that reduced vestibular responses ( Figure 4 ) and a higher probability of perceiving unexpected balance behavior ( Figure 5 ) were accompanied by increased sway velocity variance . These two experiments , however , involved different methodologies ( trials with constant delays vs . transient changes in delay ) , making it difficult to compare sway behavior between data sets or determine if vestibular modulations coincided with perceptual changes . Therefore , in Experiment 3 we tracked the time course of vestibular responses and the occurrence of perceptual detections together with whole-body oscillations when delays were transiently imposed . Here , participants ( n = 7 ) were exposed to a transient delay of 200 ms because Experiment 2 revealed that this delay increases sway variability , attenuates vestibular responses , and elicits frequent perceptions of unexpected standing motion . To quantify balance variability throughout the transitions , sway velocity variance was estimated across a 2 s sliding window on a point-by-point basis ( Figure 6A ) . Furthermore , to link changes in vestibular responses and perception , we compared vestibular response attenuation – estimated with a time-frequency analysis of coherence and gain ( see Materials and methods ) – at the time of perceptual detection of unexpected balance motion . Out of the 588 total transitions , 489 were perceived as eliciting unexpected balance motion and used for all analyses . During periods of standing preceding the imposed delays , participants swayed with low-velocity variance ( Figure 6A ) and with consistent vestibular control of balance as shown by coherence and gain estimates between the vestibular stimulus and soleus EMG activity ( Figure 6B and C ) . Transitions between baseline and delayed balance control increased whole-body sway velocity variance throughout most of the delay period ( see Figure 6A ) . Over the same period , coherence and gain between EVS and EMG decreased . To characterize the time course of the decrease in this vestibular contribution to balance , we fit an exponential decay function to the mean coherence ( i . e . , coherence averaged over 0 . 5–25 Hz at each time point ) from each participant . The 63 . 2% attenuation ( i . e . , the time constant ) for coherence occurred at 1 . 5 ± 0 . 6 s following delay onset while the 95% attenuation ( i . e . , 3× time constant ) occurred at 4 . 4 ± 2 . 6 s following delay onset . For gain , we fit an exponential decay only to the mean gain estimated from the pooled data because for some participants coherence decreased below significance at all frequencies for some periods of the delay exposure ( see Figure 6C , left-lower panel ) . The 63 . 2% attenuation from this mean gain estimate occurred at 2 . 3 s while the 95% attenuation occurred at 6 . 8 s . Perceptual detection times occurred over a similar time period , with the group averaged ( 489 detections ) detection occurring at 3 . 4 ± 1 . 8 s after delay onset ( see magenta line in Figure 6 , right panel ) and the 95th percentile for detection time occurring at 6 . 9 ± 0 . 8 s . Comparing the perceptual detection timing with the exponential decrease in coherence estimated from each participant , we found that the average time to detect the imposed delay ( 3 . 4 ± 1 . 8 s ) corresponded to a 90% ± 7% average reduction in mean coherence . For the mean gain estimate ( from all participants ) , this average perceptual time aligned with a 73% reduction in gain . The peak velocity variance leading up to a perceptual detection was also extracted for each perceived transition and resulted in an average peak velocity variance of 8 . 42 ± 8 . 62 [°/s]2 . Overall , these results indicate that the perception of unexpected motion and increased sway variability arising from an imposed delay are accompanied by an ~70–90% attenuation of vestibular contributions to balance . When delays were first inserted in the control of standing balance , sway velocity variance increased with larger delays to the point that upright balance could not be maintained for at least 60 s at imposed delays ≥ 200 ms ( i . e . , total sensorimotor delay ≥300–360 ms ) . These results support predictions from computational models that upright standing is destabilized with added delays and that balance is impossible passed a critical delay of ~300–340 ms ( Milton and Insperger , 2019; van der Kooij and Peterka , 2011 ) . This critical delay was previously estimated by varying the parameters of a proportional-derivative ( PD ) feedback control model of standing balance , and suggests that even with training , the balance controller is incapable of adjusting the gain of its feedback to maintain upright stance with delays beyond this upper limit . Our results , in contrast , clearly demonstrate that participants improved their balance behavior ( i . e . , reduced sway velocity variance ) when given the opportunity to train with a 400 ms delay ( ~63% by 28–33 min ) , and that this improvement was retained , at least partially , 3 months later . Therefore , the nervous system can learn to control standing balance ( although with more variability ) with a net delay of up to 500–560 ms . The notable learning observed after training may have been partly due to participants being passively supported past the virtual balance limits because it prevented certain nonlinear behaviors that disrupt continuous balance control such as taking steps or falls . This remarkable ability for humans to adapt and maintain upright stance with delays raises questions regarding the principles underlying the neural control of balance . Compared to feedback controllers ( i . e . , PD and proportional-integral-derivative ) , which are not optimal in the presence of delays , optimal controllers can model the control of human standing ( Kiemel et al . , 2002; Kuo , 1995; Kuo , 2005; van der Kooij et al . , 1999; van der Kooij et al . , 2001 ) and theoretically stabilize human standing with large ( >500 ms ) delays ( Kuo , 1995 ) . This ability , however , rapidly declines with increasing center of mass accelerations ( Kuo , 1995 ) , including those driven by external disturbances ( Zhou and Wang , 2014 ) . Although feedback and optimal controllers assume that the nervous system linearly and continuously modulates the balancing torques to stand ( Fitzpatrick et al . , 1996; Masani et al . , 2006; van der Kooij and de Vlugt , 2007; Vette et al . , 2007 ) , intermittent corrective balance actions ( Asai et al . , 2009; Bottaro et al . , 2005; Gawthrop et al . , 2011; Loram et al . , 2011; Loram et al . , 2005 ) may represent a solution when time delays rule out continuous control ( Gawthrop and Wang , 2006; Arsan et al . , 1999 ) . Intermittent muscle activations are also sufficient to stabilize the upright body during a robotic standing balance task similar to the one used in the present study ( Huryn et al . , 2014 ) . The nervous system may use a combination of the controllers during standing ( Elias et al . , 2014; Insperger et al . , 2015 ) , but our study did not explicitly test for evidence of these different controllers or their ability to stabilize upright stance with large delays . Increasing the imposed delay between balancing motor commands and the whole-body motion associated with those actions progressively attenuated vestibular-evoked muscle responses and led to more frequent perceptual detections of unexpected movement . The attenuation of vestibular-evoked responses to increased sensorimotor delays could be explained through processes of sensory reweighting ( Cenciarini and Peterka , 2006; Peterka , 2002 ) , where the decreasing reliability in sensory cues when balancing with additional delays decreases feedback gains . Indeed , feedback control models of standing predict that sensorimotor gains should decrease with increasing sensorimotor delays ( Bingham et al . , 2011; Le Mouel and Brette , 2019; van der Kooij and Peterka , 2011 ) . However , our results show the partial return of vestibular response amplitude and shifted perceptual responses after training at a delay of 400 ms , indicating the involvement of alternative processes of sensorimotor recalibration . Both the vestibular and perceptual measures are influenced by whether the brain interprets sensory feedback as self-generated ( expected ) or externally imposed ( unexpected ) , and their adaptation indicates that the nervous system learned to expect the delayed balancing-related feedback associated with self-generated balancing motor commands . Similar recalibration of the vestibular control of balance has been observed when humans stand with manipulated vestibular feedback gains ( Héroux et al . , 2015 ) . When first standing with altered vestibular feedback ( via a head-coupled EVS ) , variability in balance behavior increased and the amplitude of vestibular-evoked muscle responses ( probed with an independent EVS signal ) decreased . After a short ( 240 s ) calibration period with the novel head-coupled vestibular stimulus , postural sway and vestibular responses returned to baseline levels ( Héroux et al . , 2015 ) , suggesting that the brain learned to expect the modified vestibular feedback . Notably , this could not be explained by sensory reweighting mechanisms since adaptation did not occur when applying matching levels of EVS that were uncoupled from head motion . Perceptually , there is similar evidence of recalibration to new sensorimotor delays . Following repeated exposure to imposed sensorimotor delays , the perceived timing between motor actions and sensory feedback can be shifted according to the magnitude of the delay ( Stetson et al . , 2006 ) and the perceived intensity ( i . e . , force , tickle sensation ) of delayed touch returns to baseline ( i . e . , no delay ) levels ( Kilteni et al . , 2019 ) . Our results suggest that these forms of recalibration are possible for the vestibular control of standing balance and the perception of standing balance . Because we naturally experience changing neural delays during growth and aging ( Eyre et al . , 1991 ) , it is crucial that the nervous system adapts and recalibrates to unexpected temporal relationships between sensory and motor signals to maintain stable balance control and perception of standing movement . What neural substrates could be responsible for the brain learning to expect standing balance feedback that is initially unexpected ? When faced with a new sensorimotor relationship , deep cerebellar neurons ( rostral fastigial nucleus ) initially increase their activity to vestibular signals because the motor commands result in unexpected sensory feedback ( Brooks et al . , 2015; Brooks and Cullen , 2013 ) . Learning this new sensorimotor relationship results in a gradual increase in the probability that this unexpected feedback arises from desired motor commands , leading to a gradual return of the normal firing patterns of the deep cerebellar neurons ( Brooks et al . , 2015 ) . These neuronal recordings further indicate that the nervous system scales responses ( i . e . , not a switch-like mechanism ) to sensory inputs based on the relative probability that sensory feedback is caused by motor actions . Our vestibular experiment results reflect a similar mechanism – vestibular response amplitudes gradually decreased with the magnitude of the delay – extending this framework to a standing balance control context . Additionally , training led to modified vestibular and perceptual responses as well as decreased sway velocity variance that transferred to different imposed delays , indicating a generalized effect of learning . This suggests that the nervous system did not specifically recalibrate sensorimotor cues to the 400 ms delay but estimated the source of the unexpected balance motion ( i . e . , distorted motor command – whole-body motion relationship ) and broadly updated its control ( Berniker and Kording , 2008; Braun et al . , 2010; Krakauer and Mazzoni , 2011 ) to accommodate for imposed delays . Our third experiment showed that within a few seconds of balancing with a 200 ms delay , an ~70–90% attenuation of vestibular responses accompanied increased sway velocity variance and the perception of unexpected standing motion . Part of the variability in whole-body sway caused by sensory manipulations is considered to represent errors in balance estimates ( Kiemel et al . , 2002 ) . Because imposed delays should evoke unexpected feedback errors ( Blakemore et al . , 1999; Farrer et al . , 2008; Haering and Kiesel , 2015; Wen , 2019 ) , the changes in vestibular-evoked muscle responses and perception with sway velocity variance may reflect that the two behavioral responses are linked to balance errors . However , it is not explicitly clear from our data what specific component of the standing behavior can be attributed to discrepancies between actual and expected feedback . Therefore , changes in standing behavior following adaptation to a delay could also be attributed to other factors induced by the imposed delays , such as a change in control policy or increased volitional control to balance ( Elias et al . , 2014; Ozdemir et al . , 2018; Peterson and Ferris , 2019 ) . When accounting for this limitation , it is also plausible that the observed changes in vestibular and perceptual responses were partially driven by whole-body motion ( i . e . , magnitude and variability ) and not solely prediction errors . For instance , vestibular responses are known to be modulated by standing kinematics ( Day et al . , 1997; Rasman et al . , 2018; Son et al . , 2008 ) and perceiving balance motion is related to whether the experienced motion exceeds sensory detection thresholds ( Fitzpatrick and McCloskey , 1994 ) . The important questions regarding whether movement variability is attributed to control errors and their resulting influence on balance control , perception , and adaptation are inherently challenging for standing balance because the task is continuous and does not have an effective end-point target ( thus no computable end-point error ) . Future studies using carefully designed sensorimotor manipulations of balance control ( Rasman et al . , 2018 ) with differing effects ( perhaps directional ) on unexpected errors and whole-body sway behavior may resolve these critical issues . During aging and certain diseases ( e . g . , diabetic neuropathy or multiple sclerosis ) , the sensory and/or motor conduction times increase and may affect an individual’s ability to maintain their body center of mass position within their base of support . According to our results , the standing balance system can be trained to accommodate a larger range of sensorimotor delays than previously predicted , which may be valuable for clinical populations who are thought to experience instability and an increased risk of falls as a result of increased neural delays . Adaptation to increased neural delays with aging , for example , is thought to occur through decreased sensorimotor gains and increased ankle stiffness ( Le Mouel and Brette , 2019 ) . Because our results show that training with an additional 400 ms delay can restore vestibular contributions of balance across a broad range of imposed delays ( and are retained over a 3 month period ) , it may be possible to explore targeted rehabilitation of aging-related balance impairments by training with delays to restore and sustain sensory contributions to standing balance . We manipulated the delay between ankle-produced torques ( measured from the force plate ) and the resulting whole-body motion ( angular rotation about the ankle joints ) . This manipulation altered the timing between the net output of self-generated balance motor commands ( i . e . , ankle torques ) and resulting sensory cues ( visual , vestibular , and somatosensory ) encoding whole-body and ankle motion . However , the timing between motor commands and part of the somatosensory signals from muscles ( muscle spindles and Golgi tendon organs ) and/or skin ( cutaneous receptors under the feet ) that are sensitive to muscle force ( and related ankle torque ) or movements and pressure distribution under the feet were unaltered by the imposed delays to whole-body motion . This may have led to potential conflicts in the sensory coding of balance motion and may have influenced the ability to control and learn to stand with imposed delays . As methodologies to probe and manipulate the sensorimotor dynamics of standing improve , future experiments can be envisioned to replicate and modify specific aspects ( i . e . , specific sensory afferents ) of the physiological code underlying standing balance . Such endeavors are needed to unravel the sensorimotor principles governing balance control . We observed that increasing the imposed sensorimotor delays in standing balance results in unstable standing balance , attenuated vestibular control of balance , and perceptual detections of unexpected motion . The nervous system , however , can adapt to novel sensorimotor delays and learn to maintain upright standing despite their initially disruptive influence . This learning is accompanied by vestibular contributions to balance partially returning and standing motion more likely to be perceived as self-generated . Thus , our results suggest that the nervous system can learn to control standing balance with added sensorimotor delays by causally relating delayed whole-body sensory feedback ( initially deemed unexpected ) with self-generated balancing motor commands . A total of 46 healthy adult participants ( 32 males , age: 24 . 0 ± 3 . 9 years [mean ± SD]; range 19–34 years ) with no known history of neurological deficits participated in this study . The experimental protocol was verbally explained before the experiment and written informed consent was obtained . The experiments were approved by the University of British Columbia Human Research Ethics Committee and conformed to the Declaration of Helsinki , with the exception of registration to a database . Three experiments were conducted to investigate how imposed sensorimotor delays affect standing balance . We first performed an experiment evaluating standing balance behavior when participants balanced upright with different imposed sensorimotor delays . We then conducted a second experiment to determine ( 1 ) whether participants could learn to maintain standing balance with a 400 ms delay and ( 2 ) whether learning to control balance would lead to changes in the vestibular control of balance ( Experiment 2 – vestibular testing ) and perception of postural behavior ( Experiment 2 – perceptual testing ) . To probe the vestibular control of balance , we delivered EVS while participants balanced upright with different delays . To assess perception , we applied transient delays during ongoing standing balance and asked participants to report when they consciously perceived unexpected postural motion . Finally , we performed a third experiment to track the time course of modulations in vestibular contributions to balance caused by imposed delays and how it follows changes in sway behavior variability and perception of unexpected balance motion . For all experiments , participants stood on a custom-designed robotic balance simulator programmed with the mechanics of an inverted pendulum to replicate the load of the body during standing ( Figure 1A ) . Specifically , the simulator used a continuous transfer function that was converted to a discrete-time equivalent for real-time implementation using the zero-order hold methodIθ¨−mmgLθ=TθT=1Is2−0 . 971mgL as described by Luu et al . , 2011 , where θ is the angular position of the body’s center of mass relative to the ankle joint from vertical and is positive for a plantar-flexed ankle position , T is the ankle torque applied to the body , mm is the participant’s effective moving mass , L is the distance from the body’s center of mass to the ankle joint , g is gravitational acceleration ( 9 . 81 m/s2 ) , and I is mass moment of inertia of the body measured about the ankles ( mmL2 ) . The body weight above the ankles was simulated by removing the approximate weight of the feet from the participant’s total body weight so that the effective mass was calculated as 0 . 971m , where m is the participant’s total mass . The balance simulator was controlled by a real-time system ( PXI-8119; National Instruments , TX , USA ) running at 2000 Hz and consisted of an ankle-tilt platform and rigid backboard independently controlled by two rotary motors ( resolution of 0 . 00034°; SCMCS-2ZN3A-YA21 , Yaskawa , Japan ) . The backboard was lined with a layer of medium-density foam and memory foam . Participants were secured to the backboard through seat belts at the waist and shoulders , and the backboard orientation was adjusted relative to the frame to account for the participant’s natural standing posture . Participants stood on a force plate ( OR6-7-1000; AMTI , MA , USA ) secured to the ankle-tilt platform . In all experiments , the ankle-tilt platform was held horizontal ( earth-fixed reference ) while the backboard moved the upright body in the AP direction in response to ankle plantar- and dorsiflexion torques , thus replicating whole-body AP movements associated with standing ( Luu et al . , 2011 ) . The delay between a position command and the measured position of the motor was estimated to be 20 ms ( Shepherd , 2014 ) . A visual projection screen was located to the left of the robotic device on which a 3D scene of a city courtyard with a water fountain was presented ( Vizard 2013 software; WorldViz , CA , USA ) . Rendering and projection of the visual scene took approximately 70 ms; therefore , a linear least-squares predictor algorithm was used to synchronize the visual motion ( i . e . , predict visual motion occurring 50 ms later ) together with the motors at a delay of 20 ms ( Shepherd , 2014 ) . The linear prediction model used six data points , and the coefficients were selected by fitting data of participant sway to the corresponding data shifted by the appropriate delay using a linear least-squares method . Participants wore active 3D glasses ( DLP Link 3D Glasses; BenQ , Taipei , Taiwan ) , modified to block out peripheral vision and limit the participant’s field of view to approximately ±45° horizontally and ±30° vertically . Participants also wore earplugs and noise canceling headphones ( Bose Soundlink Around , Bose Corporation , MA , USA ) with audio of a water fountain to minimize acoustic cues of motion produced by the motors as well as other extraneous sounds . To represent the physical limits of sway during standing balance , the backboard rotated in the AP direction about the participant’s ankles with virtual angular position limits of 6° anterior and 3° posterior ( Luu et al . , 2011; Shepherd , 2014 ) . When the backboard position exceeded these position limits , the program gradually increased the simulated stiffness such that the participants could not rotate further in that direction regardless of the ankle torques they produced . This was implemented by linearly increasing a passive supportive torque to a threshold equivalent to the participant’s body load over a rotation range of 1° beyond the balance limits ( i . e . , passively maintaining the body at that angle ) . Any active torque applied by participants in the opposite direction would enable them to get out of the limits . Finally , to avoid a hard stop at these secondary limits ( i . e . , 7° anterior and 4° posterior ) , the supportive torque was decreased according to an additional damping term that was chosen to ensure a smooth attenuation of motion . The balance simulation was modified to add a delay between the measured ankle torque ( i . e . , motor command ) and whole-body sway ( i . e . , sensory feedback ) . Delays were programmed by buffering ankle torque recordings such that the signals driving motor position commands , and therefore whole-body sway , could be delivered based on the torque participants performed up to 500 ms in the past . The natural sensorimotor delays within the standing balance controller are estimated to be ~100–160 ms ( Forbes et al . , 2018; Kuo , 2005; van der Kooij et al . , 1999 ) . Here , as our delays are in the robotic system , they need to be added to the internal delays to estimate the overall standing balance delays . Throughout this study , we refer to the delays added to the robotic simulator ( 20–500 ms ) , but the total sensorimotor delays for the standing balance task are ~100–160 ms larger . All participants were naïve to the delay protocols and were simply told that ‘at times during the balance simulation , control may change such that your body movement may seem unexpected or abnormal , and standing balance may become more difficult . ’ In all experiments , participants were instructed to stand upright normally at their preferred standing angle ( typically ~1–2° anterior ) . In trials with delays ≥ 200 ms , participants had difficulty maintaining a stable upright posture and would often reach the virtual balancing limits ( i . e . , 6° anterior or 3° posterior ) . When this happened , participants were immediately instructed by the experimenter to get out of the limits and continue to balance upright . Despite these efforts , trials with larger imposed delays were often characterized by brief periods ( ~2–5 s ) of active balancing before crossing the virtual balancing limits ( i . e . , angular whole-body position exceeding the angular position limits ) . After a trial was completed , the robot was returned to a neutral position ( 0° ) at a fixed velocity ( 0 . 5°/s ) in preparation for the next trial . Signals used to control the robotic simulation in real time were processed via a multifunction reconfigurable module ( PXI-7853R; National Instruments , Austin , TX , USA ) . The backboard motor encoder provided backboard angular position data regarding the participant’s whole-body position during the task . The delay ( ms ) used during the simulation was recorded for all experiments . Force plate signals ( forces and torques ) were amplified ×4000 ( MSA-6; AMTI , Watertown , MA , USA ) prior to being digitized . Vestibular stimuli ( see Experiment 2 – vestibular testing and Experiment 3 ) and button switch signals ( see Experiment 2 – perceptual testing and Experiment 3 ) were also recorded . For Experiment 2 – vestibular testing and Experiment 3 , surface electromyography ( EMG ) was collected from the right soleus to measure the vestibular-evoked responses . We measured activity from this muscle for two reasons: ( 1 ) the head was turned left during all experiments and this orientation aligns the vestibular-evoked error with soleus muscle’s line of action ( see Vestibular stimulation ) and ( 2 ) the soleus was shown to have the most consistent activity during pilot testing at the manipulated delay conditions , which is a prerequisite to estimate a vestibular-evoked response in lower-limb muscles ( Dakin et al . , 2007; Forbes et al . , 2013 ) . The skin over the muscle was cleaned with an alcohol swab and abraded with gel ( Nu-Prep , Weaver and Company , Aurora , CO , USA ) . Self-adhesive Ag-AgCl surface electrodes ( Blue Sensor M , Ambu A/S , Ballerup , Denmark ) were positioned over the belly of the muscle in a bipolar configuration , with an inter-electrode distance of 2 cm center-to-center . For each participant , we noted the electrode placement on the soleus ( by measuring the distance from the electrodes to the heel ) during the first experimental session and used the same placement for consistency across experimental sessions ( pre-learning , post-learning , retention ) . To reduce electrical noise from the motors , two ground electrodes were used: a nickel-plated disc electrode coated with electrode gel ( Spectra 360 Parker Laboratories , NJ , USA ) secured to the right medial malleolus , and a Velcro strap electrode secured around the right lower leg overlaying the tibial tuberosity and head of the fibula . Surface EMG signals were amplified ( ×5000 , Neurolog , Digitimer Ltd . , Hertfordshire , UK ) and band-pass filtered ( 10–1000 Hz ) prior to digitization . Force plate , vestibular stimuli , EMG , and button switch signals were recorded via a data acquisition board ( PXI-6289; National Instruments ) . All signals were digitized at 2000 Hz . For all experiments , a balance session was first completed to familiarize the participant with the control of the robot . Instructions were given on the nature of movement control; that is , similar to standing , applying torque to the support surface ( force plate ) will control the motion of the upright body ( backboard ) . In a forward leaning position , plantar-flexor torque is required to stabilize the body and an increase in plantar-flexor torque greater than the gravitational torque will cause the body to accelerate backward . Similarly , a dorsiflexor torque is required when standing in a backward leaning position and an increase in dorsiflexor torque will accelerate the body forward . Participants were instructed to sway back and forth and allow the robot to reach its limits ( 6° anterior , 3° posterior ) , which occurs if the magnitude of the generated ankle torque is not large enough to resist the toppling torque of gravity . After becoming familiar with the control of the robot , participants were then asked to stand quietly and maintain an upright posture ( normal standing ) . Participants were also instructed to focus on the sensation of balance control and told that this setting was the ‘normal’ condition , which was of particular importance for perception experiments ( Experiment 2 – perceptual testing and Experiment 3 ) . Participants performed this familiarization period until they were accustomed to the sensation of standing balance on the robot . This ensured that participants could confidently discern between normal and novel ( i . e . , unexpected ) standing balance behavior caused by manipulating delays . They completed the entire familiarization session within 5–7 min . In Experiment 2 – vestibular testing and Experiment 3 , we used transmastoid EVS to probe vestibular-evoked muscle responses . Vestibular stimulation was delivered in a binaural bipolar configuration to modulate the firing rates of all vestibular afferents ( Goldberg et al . , 1984; Kim and Curthoys , 2004; Kwan et al . , 2019 ) and provide an isolated vestibular error signal that evoked a virtual sensation of head motion about an axis-oriented posteriorly and superiorly by ~17–19° above Reid’s plane ( Fitzpatrick and Day , 2004; Peters et al . , 2015 ) . The head was pitched up such that Reid’s plane was oriented ~17–19° up from horizontal . In this head position , EVS evokes a net signal of angular head rotation orthogonal to gravity ( Chen et al . , 2020; Fitzpatrick and Day , 2004; Schneider et al . , 2002 ) and , through integration with an internal estimate of gravity , an inferred interaural linear acceleration signal ( Khosravi-Hashemi et al . , 2019 ) . While standing , this imposed vestibular error evokes stereotypical compensatory muscle and whole-body responses ( Dakin et al . , 2007; Day et al . , 1997; Fitzpatrick and Day , 2004; Forbes et al . , 2014 ) . The head was also turned ~90° to the left , which aligns the vestibular-evoked error signal with the AP direction of balance and line of action for the soleus , maximizing the muscle response ( Dakin et al . , 2007; Forbes et al . , 2016 ) . We chose stochastic vestibular stimuli ( a white noise signal low-pass filtered to contain a set bandwidth of frequencies ) rather than square-wave stimuli as it improves signal-to-noise ratio and reduces testing time ( Dakin et al . , 2007; Reynolds , 2011 ) . EVS signals with a 0–25 Hz frequency bandwidth were generated offline using custom-designed computer code ( LabVIEW 2013 , National Instruments ) . The stimuli were delivered to participants through carbon rubber electrodes ( 9 cm2 ) coated with Spectra 360 electrode gel and secured over the mastoid processes in a binaural bipolar configuration . The stimuli were sent as analog signals via a data acquisition board ( PXI-6289 , National Instruments ) to an isolated constant current stimulator ( DS5 , Digitimer , Hertfordshire , England ) . Throughout the trials , head position was monitored and maintained using a head-mounted laser and verbal feedback , respectively . Computational models indicate that adding sensorimotor delays will destabilize balance control ( Insperger et al . , 2015; Milton and Insperger , 2019; van der Kooij et al . , 1999 ) and that upright standing cannot be maintained past a critical delay of ~340 ms ( Milton and Insperger , 2019; van der Kooij and Peterka , 2011 ) . To determine how imposed balance delays destabilize and limit balance control , 13 participants balanced the robotic system for 60 s with different imposed delays ( 20 , 100 , 200 , 300 , 400 , and 500 ms ) . Trial order was the same for each participant , with delays presented in ascending order to avoid crossover effects from larger delays to smaller delays . Participants were not given any information regarding the different delay conditions or strategies on how to best control the robot . Based on observations from Experiment 1 , we designed a training protocol ( Experiment 2 ) to determine if participants could learn to stand when the robotic balance control simulation operated with 400 ms delay ( i . e . , a total feedback delay of 500–560 ms ) . This delay was chosen because it is larger than the ~300–340 ms critical feedback delay of standing balance proposed previously ( Milton and Insperger , 2019; van der Kooij and Peterka , 2011 ) . To investigate the underlying physiological mechanisms of this learning process , we conducted Experiment 2 to evaluate how imposed delays influenced the vestibular-evoked balance responses and perception of standing motion , respectively , before and after training as well as 3 months later . Sixteen participants ( vestibular testing , n = 8; perceptual testing , n = 8 ) performed the training protocol over five consecutive days ( see Figure 1C ) . On the first day , prior to any training , participants completed pre-learning session ( vestibular or perceptual testing; see below ) . Each training session then started with a 60 s standing balance trial with the 20 ms delay . This was followed by two 10 min training trials at the 400 ms delay . To minimize fatigue , participants rested for 2–3 min between these two trials . Participants were not given any specific instruction on how to improve their balance . On each day , the training trials were followed by a final 60 s trial at the 20 ms delay . In total , participants performed 100 min of training over 5 days at the 400 ms delay condition . After finishing the training session on the fifth day , participants completed the post-learning session ( vestibular or perceptual testing ) . Finally , 12 of the 16 participants ( seven vestibular , five perceptual ) returned ~3 months later ( range: 81–110 days ) to perform a retention session . The retention testing session began with a short familiarization period ( <60 s ) of balancing the robot with the 20 ms delay . Participants then completed vestibular or perceptual testing followed by two 3 min standing balance trials: one with the 20 ms delay and one with the 400 ms delay ( order randomized between participants ) . Eight of the 16 training participants in Experiment 2 completed vestibular testing ( Figure 1C ) . Here , we probed the vestibular-evoked responses in the soleus muscle while participants maintained standing balance at six imposed delays ( see below ) . Vestibular testing was conducted for pre-learning , post-learning , and retention sessions . Seven participants returned ~3 months after the post-learning session to complete the retention session . Because vestibular-evoked responses are attenuated when actual sensory feedback does not align with estimates from balancing motor commands ( Héroux et al . , 2015; Luu et al . , 2012 ) , we hypothesized that increasing the delay between ankle torques and body motion would progressively attenuate the vestibular response . We further hypothesized that learning to control balance with imposed delays would allow the brain to update its sensorimotor estimates of balance motion and consequently increase the vestibular-evoked muscle responses . For all vestibular response testing , participants stood on the robotic balance simulator with different delays while being exposed to EVS ( Figure 1D ) . Six delay conditions were tested: 20 , 100 , 200 , 300 , 400 , and 500 ms . Each trial began with a short period of data collection ( ~3–5 s ) while the participants stood quietly on top of the robotic balance simulator , after which the electrical stimulus was delivered . For each delay condition , EVS was delivered in four 20 s trials , resulting in a total of 80 s of data ( 24 total trials ) . We performed short trials to minimize potential adaptation during a trial . Four different EVS signals ( 0–25 Hz bandwidth ) were generated offline with root-mean-square ( RMS ) ranging from 1 . 47 to 1 . 61 mA ( due to stochastic variation in signal generation ) and each EVS stimulus was delivered once per delay condition ( 24 trials ) . We presented the trials in four subgroups , each containing the six delay conditions ordered randomly . A total of 18 participants , 8 of which participated in the training protocol of Experiment 2 , completed perceptual testing ( Figure 1C ) . Here , we examined if participants perceived unexpected postural behavior while being exposed to intervals of imposed sensorimotor delays ( see below ) . All participants completed the pre-learning session , eight then completed the training protocol and were tested in the post-learning session and five of those eight returned ~3 months later for the retention session . If adding balance delays reduces the probability that sensory feedback is associated with motor commands , we hypothesized that participants would perceive self-generated balance oscillations as unexpected motion under delayed balance conditions . If training to stand with the imposed delay allows the brain to update its sensorimotor estimates of balance motion , we hypothesized that after training participants would be less prone to detecting unexpected standing motion when balancing with experimental delays ( i . e . , psychometric functions would shift rightward ) . During the perception trials , the delay imposed by the robotic simulator transitioned while participants were actively balancing ( see Figure 1E ) . Participants were instructed to indicate by pressing and holding a hand-held button switch when they perceived unexpected balance movements and release the button when balance felt normal . Delays were manipulated using a variation of the method of constant stimuli . We used seven different delays in the trials . For the 10 participants who did not participate in the training protocol ( no-learning group ) , the delays were 20 ( catch trial ) , 50 , 100 , 150 , 200 , 250 , and 300 ms . We limited the delay to a maximum of 300 ms because preliminary testing showed that participants always perceived unexpected balance at the 300 ms delay . Participants that partook in the training protocol were presented with 50 , 100 , 150 , 200 , 250 , 300 , and 350 ms delays . We expected that after training some 300 ms delay periods would not elicit perceptions of unexpected motion ( see Results ) , and therefore increased the experimental delay to a maximum of 350 ms . Participants balanced themselves in the robotic simulator for ~260 s in 10 separate trials . While participants actively balanced , we randomly inserted delays in the balance control for 8 s . During each trial , the robot transitioned instantaneously from the baseline 20 ms balance delay to one of the experimental delays ( 20–350 ms ) before returning to 20 ms . The inter-transition interval varied randomly between 7 and 10 s . In this manner , 14 delay periods ( two of each delay level ) were presented in a random order for each trial , resulting in a total of 20 delay periods for each experimental delay ( yielding a 5% resolution ) . The catch delay period ( 20 ms ) from the no-learning group did not reveal any different performance from the existing inter-delay intervals and both the no-learning and pre-learning participants had 100% success at detecting the 300 ms delay . Therefore , we present the psychometric functions ( see Psychometric functions ) of these two groups together . Our results from the vestibular and perceptual testing in Experiment 2 showed that with added delays , ( 1 ) vestibular-evoked muscle responses are attenuated and participants perceive unexpected behavior , and ( 2 ) both vestibular and perceptual response modulations occur together with increases in sway variability ( i . e . , sway velocity variance ) ( see Results ) . Vestibular stimulation trials in Experiment 2 – vestibular testing had single , fixed delays ( no transition between delays within the trial ) ; consequently , we could not assess the time course of vestibular response attenuation . In Experiment 3 , we tracked the time course of the vestibular , balance , and perceptual responses to imposed delays by repeatedly exposing participants ( n = 7 ) to transient periods ( 8 s ) of a 200 ms delay period while delivering EVS and instructing them to report perceptions of unexpected balance . Specifically , we determined ( 1 ) vestibular response attenuation in relation to when participants perceive unexpected standing behavior , and ( 2 ) the sway velocity variance associated with modulations in the vestibular and perceptual responses . Participants balanced on the robotic balance system for six separate trials lasting ~260 s each and were instructed to indicate perceived changes in balance control via button press . To probe the vestibular-evoked muscle responses , right soleus EMG was recorded and stochastic EVS was delivered ( RMS: 1 . 38 mA ) throughout each trial . The trials were similar to those of Experiment 2 – perceptual testing , except that only the 200 ms experimental delay was presented . We chose 200 ms based on our results from Experiments 1 and 2 ( see Results ) . This specific delay increases sway velocity variance , attenuates vestibular-evoked balance responses , and elicits frequent perceptual detections ( ~80% ) of unexpected standing motion . During each trial , the robot transitioned from the baseline balance condition ( 20 ms delay ) to a 200 ms delay before returning to baseline . Transition to the delayed period occurred randomly over an inter-transition interval of 8–9 s and each delay period lasted exactly 8 s . Each trial consisted of 14 delay periods and a total of six trials were performed , providing 84 delay periods . All non-statistical processing and analyses were performed using custom-designed routines in MATLAB ( 2018b version , MathWorks , Natick , MA , USA ) and LabVIEW software ( LabVIEW 2013 , National Instruments ) . To quantify how balance behavior was affected by imposed delays in Experiments 1 and 2 , we estimated the variance of whole-body sway velocity within each trial . Sway velocity variance was estimated only from data in which whole-body angular position was within the virtual balance limits ( 6° anterior and 3° posterior ) . Specifically , data were extracted in non-overlapping 2 s windows , when there was at least one period of two continuous seconds of balance within the virtual limits . Data windows were only extracted in multiples of 2 s . For instance , if there was an 11 s segment of continuous balance , we only extracted five 2 s windows ( i . e . , first 10 s of the segment ) . Sway velocity variance was estimated over these 2 s windows ( see Figure 2A ) because ( 1 ) participants could only balance within the balance limits for periods of ~2–5 s during some trials with delays ≥ 200 ms and ( 2 ) the transient delays in perceptual testing lasted only 8 s . Although too short to evaluate low-frequency postural sway position , this analysis window was considered appropriate to evaluate the variance of sway velocity because the velocity signal primarily consists of frequencies > 0 . 5 Hz ( van der Kooij et al . , 2011 ) . On a participant-by-participant basis , we then averaged the sway velocity variance from the 2 s windows to provide an estimate of sway velocity variance for each participant in each balance condition ( e . g . , Experiment 1 , 200 ms delay ) . For Experiments 1 and 2 ( training trials ) , we also computed the percentage of time ( over 60 s intervals ) that whole-body sway position was within the virtual balance limits . For the training trials in Experiment 2 , sway velocity variance was estimated from non-overlapping 2 s windows taken across 1 min intervals ( thus a maximum of 30 available windows per minute ) throughout the training and retention phases , as well as during standing balance prior to training . These sway velocity variances were then averaged for every minute across all participants , providing a minute-by-minute representation of sway velocity variance in the training trials . For perceptual testing , where delays transitioned during perception trials , we evaluated the variance of sway velocity from 2 s windows over the period during which a delay was imposed . Since each delay was presented 20 times , this provided 160 s of data for each delay . For Experiment 3 , we identified peak sway velocity variance leading up to a perceptual detection . We therefore used a 2 s sliding window to extract the time course of sway velocity variance . Time-varying variance was calculated using the movvar MATLAB function , which calculated variance over 2 s segments while repeatedly moving over the data on a point-by-point basis . On a participant-by-participant basis , all delay periods ( 84 total ) were classified as detected ( button pressed during delay ) or missed ( button not pressed during delay ) . Using only perceptually detected trials ( 499 out of 588 ) , we then extracted peak sway velocity variance over a period starting at the onset of the delay and ending at the perceptual detection from each perceived trial . Transitions where a button press occurred before delay onset were removed from the analysis ( group data: 10 out of 499 transitions ) , providing 489 transitions in total . To estimate the presence and magnitude of vestibular-evoked muscle responses in Experiment 2 - vestibular testing , we used a Fourier analysis to estimate the relationship between vestibular stimuli and muscle activity in the frequency and time domains . Specifically , we computed the coherence , gain , and cross-covariance functions between the electrical stimulus and soleus EMG for each participant ( Dakin et al . , 2007 ) . Because vestibular stimulation trials were performed on three separate days ( pre-learning , post-learning , retention sessions ) , we scaled soleus EMG to each session’s baseline EMG measure . Specifically , for each testing session ( e . g . , pre-learning ) , we estimated the mean soleus EMG amplitude from a baseline quiet standing trial . We calculated the mean EMG amplitude recorded while participants maintained standing balance within ±0 . 25° around their preferred standing posture . We then scaled soleus EMG recorded during the vestibular stimulation trials by dividing it by the calculated mean EMG amplitude . In this manner , each participant’s EMG during vestibular trials from a given session was scaled to the baseline EMG measured during that session . EMG data were high-pass filtered ( 30 Hz , zero lag , sixth-order Butterworth ) and full-wave rectified . Data were cut into segments of 2048 data points ( ~1 s ) providing a frequency resolution of ~0 . 98 Hz . For each participant and condition ( e . g . , pre-learning 100 ms delay ) , data were concatenated over the four balance trials providing 76 segments equating to 77 . 82 s of data . Over each segment ( no overlap between segments ) , the autospectra for EVS and soleus EMG as well as the cross-spectra between EVS and EMG were calculated . The spectra were then averaged in the frequency domain to estimate coherence , gain , and cross-covariance . Coherence is a measure of the linear relationship between two signals in the frequency domain and is given by Cf= \|Psrf\|2/Pss ( f ) Prr ( f ) where Psrf is the stimulus-response cross spectrum , Pss ( f ) is the autospectrum of the EVS stimulus , and Prr ( f ) is the autospectrum of the rectified EMG . At each frequency point , coherence ranges from 0 ( no linear relationship ) to 1 ( noise-free linear relationship ) . We interpreted coherence at each frequency point as significant if it exceeded the 95% confidence limit derived from the number of disjoint segments ( Halliday et al . , 1995 ) . Gain was computed by dividing the EVS-EMG cross spectrum by the EVS autospectrum and represents the ratio of the output signal to the input signal . Gain must be assessed alongside coherence because its estimate is unreliable at frequency points where coherence is below the significance threshold . In the time domain , we estimated the non-normalized cross-covariance , which provides a time-domain measure of EVS-EMG association . We estimated cross-covariance for individual participants and group pooled data by taking the inverse Fourier transform of the EVS-EMG cross spectra ( Halliday et al . , 1995 ) . Cross-covariance ( sometimes referred to as cumulant density ) estimates are used in neurophysiological and motor control studies to characterize the correlation between two signals ( Brown et al . , 1999; Brown et al . , 2001; Halliday et al . , 1995; Halliday et al . , 2006; Hansen et al . , 2005; Nielsen et al . , 2005; Tijssen et al . , 2000 ) . These measures have been widely used to estimate time domain vestibulo-motor responses to EVS during postural control activities ( Dakin et al . , 2010; Dakin et al . , 2007; Mackenzie and Reynolds , 2018; Mian and Day , 2009; Mian and Day , 2014; Reynolds , 2011 ) . During standing balance , lower-limb EVS-EMG cross-covariance responses exhibit a biphasic pattern with opposite peaks , defined as short- ( 50–70 ms ) and medium-latency ( 100–120 ms ) responses that match the responses elicited by square wave stimuli ( Dakin et al . , 2010; Dakin et al . , 2007 ) . For statistical analysis in Experiment 2- vestibular testing , the peak-to-peak amplitude of the cross-covariance estimates was extracted from each participant’s response and used as a measure of response magnitude . When either of the short- or medium-latency cross-covariance peaks did not surpass the 95% confidence interval , the value of that peak was set to zero ( Dakin et al . , 2007; Forbes et al . , 2016 ) . Therefore , if both short- and medium-latency peaks did not exceed the confidence intervals , the cross-covariance vestibular response amplitude was zero and considered absent . For Experiment 3 , we tracked time-varying changes in the vestibular-evoked muscle responses during transitions to 200 ms delayed balance control by estimating the time-varying coherence and gain between EVS and rectified EMG activity . Segments of 24 s of data , including 8 s prior to and after the imposed delay , from each trial were used for analysis . Because estimates of vestibulomuscular coherence and gain required multiple repetitions of the imposed delay , we did not evaluate vestibular responses on a trial-by-trial basis . Time–frequency vestibulomuscular coherence and gain were calculated by convolving the input EVS and output EMG with a set of complex Morlet wavelets ( Blouin et al . , 2011; Zhan et al . , 2006 ) , defined as complex sine waves tapered by a Gaussian ( Cohen , 2014; Cohen , 2019 ) . The peak frequencies of the wavelets were linearly spaced from 0 . 5 to 25 Hz in 40 steps , and the number of cycles used for each wavelet was logarithmically spaced from 3 to 12 with increasing wavelet peak frequency . This corresponds to a full-width at half maximum ( FWHM ) ranging from 168 ms to 2249 ms and a spectral FWHM range of 0 . 39–5 . 25 Hz ( Cohen , 2019 ) . For each participant , the wavelet analysis was performed on a data set consisting of all segments of data ( each 24 s in length ) and averaged across the number of transitions . To compare vestibular responses with the reported perception of unexpected balance , we only included transitions that were perceived by the participants ( group data: 489 out of 584 ) . A pooled wavelet analysis was also performed on a concatenated data set of all 489 participant data segments . Time–frequency contour plots of coherence and gain are presented with the first and last 2 s of data removed due to the distortion by window edge effects when applying wavelet analysis . For illustrative purposes , and because gain is only reliable when coherence is significant , we plotted non-significant coherence points and their gain counterparts as zero in the time-frequency contour plots ( Figure 6B ) . To compare the overall strength of the EVS-EMG relationship during the period of baseline balance control vs . delayed balance control , we computed mean coherence and mean gain ( Luu et al . , 2012 ) . Mean coherence and mean gain were calculated by averaging values across the 0 . 5–25 Hz bandwidth at each time point . For the mean gain estimate , only gain values for which coherence was significant at each corresponding time and frequency points were used ( which was primarily across the 0–10 Hz bandwidth ) . To characterize the time course of vestibular response attenuation induced by the imposed delay , we fit exponential decay functions to each participant’s average coherence and during the 8 s period of 200 ms delay . For gain , we fit an exponential only to the group mean gain because for some participants coherence decreased below significance at all frequencies for some periods of the delay exposure ( see Figure 6C , left-lower panel ) . For perception trials , psychometric functions were generated using the participant button switch responses during the experimental delay periods . The following criteria were used to classify a participant’s response: ( 1 ) detected , identified if the button was pressed at any time during an 8 s delay period; and ( 2 ) missed , identified if the button was not pressed during an 8 s delay period . If the button was pressed prior to the onset of the delay and held until the 8 s delay period started , the trial was removed from analysis ( see Table 3 ) . Mean detection rates for each delay were computed by dividing the number of detected balance control transitions ( delay trial presented with button switch on ) by the total number of used delay trials ( see Table 3 ) . Detection time was computed as the time between the onset of the delay and the button press . Psychometric functions were generated by relating the participant’s proportion of detected responses to the magnitude of the delay ( ms ) to estimate a delay threshold level of detected unexpected balance behavior induced by the simulation delays . Using custom software ( LabVIEW 2013 , National Instruments; Rasman et al . , 2021; available at https://doi . org/10 . 5683/SP2/IKX9ML ) , a sigmoidal cumulative normal function was fit to each participant’s response data across the different delays using a robust Bayesian curve fitting procedure ( Peters et al . , 2016; Peters et al . , 2015 ) . Briefly , we parameterized each participant’s psychometric function using the following mixture model:pdataμ , σ , δ=δ2+ ( 1-δ ) ∫-∞x12πσ2 e- ( u-μ ) 22σ2 where µ is the position of the sigmoidal curve along the x-axis , σ is the slope of the sigmoidal curve , and δ is the realistic probability of a lapse in performance ( e . g . , loss of concentration , accidental button pushes , etc . ) on some small proportion of trials ( Goldreich et al . , 2009; Kontsevich and Tyler , 1999 ) . We set the range of possible µ values from 0 to 0 . 5 ( seconds ) in steps of 0 . 005 , σ ( standard deviation ) values from 0 . 01 to 5 in steps of 0 . 01 , and δ ( lapse rate ) values from 0 . 01 to 0 . 05 ( % ) in steps of 0 . 01 . We began with a uniform prior distribution over this parameter space . We marginalized over the δ , a nuisance parameter in this case , to obtain a joint posterior probability distribution over µ and σ . The best-fitting curve was taken as the mode of this joint posterior probability distribution . Participants’ perceptual threshold was calculated by averaging the interpolated 70% correct threshold value across the joint posterior probability distribution . For Experiment 3 , we used only a 200 ms experimental delay because we were primarily interested in comparing perceptual detection times with vestibular response attenuation and relating those latencies to standing behavior ( sway velocity variance ) . Because only one stimulus level ( 200 ms ) was presented , we were unable to use psychometric functions to estimate a perceptual threshold and instead we simply report the proportion of 200 ms transitions that were detected . All statistical analyses were performed using SPSS22 software ( IBM ) and the significance level was set at 0 . 05 . Group data in text , tables , and figures are presented as mean ± standard deviations unless otherwise specified .	When standing , neurons in the brain send signals to skeletal muscles so we can adjust our movements to stay upright based on the requirements from the surrounding environment . The long nerves needed to connect our brain , muscles and sensors lead to considerable time delays ( up to 160 milliseconds ) between sensing the environment and the generation of balance-correcting motor signals . Such delays must be accounted for by the brain so it can adjust how it regulates balance and compensates for unexpected movements . Aging and neurological disorders can lead to lengthened neural delays , which may result in poorer balance . Computer modeling suggests that we cannot maintain upright balance if delays are longer than 300-340 milliseconds . Directly assessing the destabilizing effects of increased delays in human volunteers can reveal how capable the brain is at adapting to this neurological change . Using a custom-designed robotic balance simulator , Rasman et al . tested whether healthy volunteers could learn to balance with delays longer than the predicted 300-340 millisecond limit . In a series of experiments , 46 healthy participants stood on the balance simulator which recreates the physical sensations and neural signals for balancing upright based on a computer-driven virtual reality . This unique device enabled Rasman et al . to artificially impose delays by increasing the time between the generation of motor signals and resulting whole-body motion . The experiments showed that lengthening the delay between motor signals and whole-body motion destabilized upright standing , decreased sensory contributions to balance and led to perceptions of unexpected movements . Over five days of training on the robotic balance simulator , participants regained their ability to balance , which was accompanied by recovered sensory contributions and perceptions of expected standing , despite the imposed delays . When a subset of participants was tested three months later , they were still able to compensate for the increased delay . The experiments show that the human brain can learn to overcome delays up to 560 milliseconds in the control of balance . This discovery may have important implications for people who develop balance problems because of older age or neurologic diseases like multiple sclerosis . It is possible that robot-assisted training therapies , like the one in this study , could help people overcome their balance impairments .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2021	Learning to stand with unexpected sensorimotor delays
Vertebrate centrioles normally propagate through duplication , but in the absence of preexisting centrioles , de novo synthesis can occur . Consistently , centriole formation is thought to strictly rely on self-assembly , involving self-oligomerization of the centriolar protein SAS-6 . Here , through reconstitution of de novo synthesis in human cells , we surprisingly found that normal looking centrioles capable of duplication and ciliation can arise in the absence of SAS-6 self-oligomerization . Moreover , whereas canonically duplicated centrioles always form correctly , de novo centrioles are prone to structural errors , even in the presence of SAS-6 self-oligomerization . These results indicate that centriole biogenesis does not strictly depend on SAS-6 self-assembly , and may require preexisting centrioles to ensure structural accuracy , fundamentally deviating from the current paradigm . Centrioles are microtubule-based , ninefold symmetrical structures essential for centrosome and cilia formation . In cycling cells , centrioles are maintained in fixed numbers , and formed through canonical duplication depending on pre-existing ( or mother ) centrioles . In the absence of pre-existing centrioles , however , de novo synthesis can occur ( Khodjakov et al . , 2002 ) . The number of centrioles formed through the de novo pathway is highly variable ( Khodjakov et al . , 2002; La Terra et al . , 2005 ) , providing an explanation for why canonical duplication dominates in dividing cells . In contrast to cycling cells , in post-mitotic cells such as multi-ciliated epithelia , the genes required for centriole assembly are highly up-regulated ( Hoh et al . , 2012 ) to produce large , variable numbers of centrioles prior to ciliogenesis , a process thought to primarily depend on de novo assembly ( Dirksen , 1991 ) . Interestingly , a recent study showed that the production of high quantities of centrioles in mouse multi-ciliated epithelia is in fact driven by the pre-existing centriole rather than through de novo assembly ( Al Jord et al . , 2014 ) , suggesting that the presence of pre-existing centrioles may have additional roles other than the number control for centriole biogenesis . Centriole biogenesis , canonical or de novo , starts with cartwheel assembly , a geometric scaffold that defines the shape and structural integrity of centrioles ( Anderson and Brenner , 1971 ) . The backbone of the cartwheel is characterized by a central hub from which nine spokes emanate ( Anderson and Brenner , 1971 ) and is primarily made of the centriolar protein SAS-6 ( Kitagawa et al . , 2011; van Breugel et al . , 2011 ) . SAS-6 exists as dimers , which can self-oligomerize in vitro via an N-terminal head domain , forming a ring resembling the central hub , and C-terminal tails pointing outwards as spokes ( Kitagawa et al . , 2011; van Breugel et al . , 2011; van Breugel et al . , 2014 ) , albeit not always ninefold symmetric in vitro ( Cottee et al . , 2011 ) . Nevertheless , these elegant discoveries raise an exciting proposal that the self-assembly property associated with the N terminus of SAS-6 drives cartwheel and centriole formation . In contrast to the SAS-6 self-assembly model , a template-based assembly model , dependent on the interaction of the C-terminal tail of SAS-6 with the lumen of mother centrioles , has recently been proposed to initiate canonical duplication ( Fong et al . , 2014 ) . During S phase , SAS-6 molecules are first recruited to the proximal lumen of the mother centriole prior to centriole duplication , adopting a cartwheel-like organization through interactions with the luminal wall , rather than via their self-oligomerization activity . This leads to a proposal that mother centrioles may function as the template to shape SAS-6 assembly , thereby preserving the geometric shape of the centriole that otherwise cannot be ensured by SAS-6 self-assembly alone . Notably , the template-based model appears incompatible with de novo centriole synthesis in which no pre-existing centrioles are required . However , as the nature of de novo synthesis , for example , whether it is indeed based on SAS-6 self-assembly , has not been determined , it is premature to accept or reject any of these ideas . To characterize de novo formation of centrioles or centrosomes , we used CRISPR/Cas9 gene targeting to sequentially inactivate p53 and SAS-6 genes in retinal pigment epithelial cells ( RPE1 ) , generating stable , acentriolar cell lines ( SAS-6-/-; p53-/- ) in which de novo centrosome formation can be subsequently reconstituted ( Izquierdo et al . , 2014 ) ( see ‘Materials and methods’ ) . Pure acentriolar cell lines were established through clonal propagation from single cells , a process taking 4–5 weeks , before these cells were used for experiments . In two independent SAS-6-/-; p53-/- cell lines we generated ( clone #1 and #2 ) , frameshift mutations present in the beginning of the coding region were found in each SAS-6 allele ( Figure 1—figure supplement 1A ) , predicting to produce severely truncated products consisting of 19 or 23 amino acids in clone #1 or #2 , respectively . Consistently , Western blot analyses using antibodies against either the N- or C-terminal region of SAS-6 failed to detect any SAS-6 signal in these cells ( Figure 1B; Figure 1—figure supplement 1C ) . Similar frameshift mutations were also seen in TP53 alleles ( Figure 1—figure supplement 1B ) , leading to loss of p53 function ( Izquierdo et al . , 2014 ) . Importantly , both SAS-6 knockout cell lines completely lack centrioles or centrosomes as expected ( Figure 1A for clone #1; Figure 1—figure supplement 2A for clone #2 ) , but can continue to proliferate in the absence of p53 ( Figure 1C for clone #1; Figure 1—figure supplement 2B for clone #2 ) ( Bazzi and Anderson , 2014; Izquierdo et al . , 2014; Lambrus et al . , 2015; Wong et al . , 2015 ) , although their M phase is significantly lengthened ( Figure 1D ) . Intriguingly , when exogenous wild-type , full length SAS-6 ( SAS-6FL ) was inducibly expressed in SAS-6-/- cells ( see ‘Materials and methods’ for details ) , either in clone #1 or #2 , variable numbers of centrosomes formed robustly in the absence of pre-existing centrosomes ( Figure 1E , G for clone #1; Figure 1—figure supplement 2C , D for clone #2 ) , a result consistent with previous reports ( Lambrus et al . , 2015; Wong et al . , 2015 ) . As clone #1 and clone #2 cell lines behave similarly , we used clone #1 to establish a stable , cell-based system in which the role of SAS-6 in de novo centrosome synthesis can be analyzed ( see below ) . 10 . 7554/eLife . 10586 . 003Figure 1 . De novo centrosome formation in the absence of SAS-6 self-oligomerization . ( A ) Wild-type ( WT ) or acentrosomal ( p53-/-; SAS-6 -/-; clone #1 ) RPE1 cells were stained with anti-centrin ( green ) and γ-tubulin ( red ) . DNA ( DAPI , blue ) . ( B ) Western blot analysis of WT or SAS-6-/- cells with indicated antibodies , including both N- and C-terminal SAS-6 antibodies ( SAS-6N; SAS-6C ) . ( C ) WT or acentrosomal cells in mitosis were stained with anti-centrin ( green ) , α-tubulin ( red ) , and pericentrin ( blue ) . ( D ) The duration of mitosis in WT or SAS-6-/- cells was measured through time-lapse imaging of live cells . ( E ) A schematic diagram showing various SAS-6 mutants tagged with Flag and HA ( FH ) . Clone #1 p53-/- SAS-6-/- cells were infected with lentiviruses carrying various of SAS-6 constructs , and the ability of each construct in rescuing centrosome formation was indicated and quantified in infected cells expressing detectable HA-tagged SAS-6 . n , number of infected cells examined . ( F ) Isogenic SAS-6-/-; p53-/- cells carrying indicated SAS-6 constructs ( see Methods ) were induced for SAS-6 expression for 3 days and then analyzed by western blot with Flag antibodies or C-terminal SAS-6 antibodies ( SAS-6C ) . For analyses with N-terminal SAS-6 antibodies ( SAS-6N ) , please see Figure 1—figure supplement 1D . Note that in Flag staining , DM6 leaked to the lane of DM5 . The expression level of each SAS-6 construct , indicated as fold changes ( Folds ) relative to the endogenous SAS-6 or SAS-6FL , is shown . ( G ) Isogenic SAS-6-/-; p53-/- cells carrying indicated SAS-6 constructs were treated as ( F ) and analyzed by immunofluorescence microscopy using indicated antibodies ( Scale bar: 20 μm in A and G , 10 μm in C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 00310 . 7554/eLife . 10586 . 004Figure 1—figure supplement 1 . Characterization of SAS6-/-; p53-/- cell lines ( clone #1 and #2 ) . ( A , B ) Sequence analyses of SAS-6 and p53 alleles in clone #1 and clone #2 as indicated . The sequence of wild-type and edited alleles are in black and blue , respectively . The lesion ( deletion or insertion ) is shown in red . ( C ) Western blot analysis of WT , SAS6-/-; p53-/- clone #1 , and SAS6-/-; p53-/- clone #2 cells with indicated antibodies . SAS-6C and SAS-6N are antibodies recognizing the N-terminus and C-terminus of SAS-6 , respectively . ( D ) Isogenic SAS6-/-; p53-/- cells carrying indicated SAS-6 constructs ( see ‘Methods’ ) were induced to express indicated SAS-6 mutants for 3 days as in Figure 1F , and then processed for Western blot with N-terminal SAS-6 antibodies ( SAS-6N ) , Flag antibodies , and tubulin antibodies as the loading control . The expression level of each SAS-6 construct , indicated as fold changes ( Folds ) relative to the endogenous SAS-6 or SAS-6FL , is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 00410 . 7554/eLife . 10586 . 005Figure 1—figure supplement 2 . Induction of de novo centrosome formation in clone #2 SAS6-/-; p53-/- cells in the absence of SAS-6 self-oligomerization . ( A , B ) Wild-type ( WT ) or clone #2 acentrosomal ( p53-/-; SAS-6 -/- ) RPE1 cells were stained with indicated antibodies during interphase ( A ) and mitosis ( B ) . DNA ( DAPI , blue ) . ( C , D ) Clone #2 acentrosomal cells were infected with lentiviruses , and induced to express indicated SAS-6 constructs for 16h ( C ) or 3 days ( D ) . The ability of each SAS-6 mutant in rescuing de novo centriole ( C ) or centrosome ( D ) formation was examined with indicated antibodies . HA staining marks SAS-6 . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 005 To determine which domains of SAS-6 is required and sufficient for de novo centrosome formation , full length SAS-6 ( FL ) or various SAS-6 deletion mutants ( DMs ) were made to allow controlled expression under the doxycycline inducible promoter ( Figure 1E ) . Isogenic , acentriolar SAS-6-/-; p53-/- cell lines stably carrying specific SAS-6 expression constructs were isolated in the absence of doxycycline ( see Methods ) . Upon induction by doxycycline treatments ( Figure 1F; Figure 1—figure supplement 1D ) , the ability of each SAS-6 mutant in driving de novo centrosome formation was examined ( Figure 1G ) . Note that similar results were also seen in reconstitution experiments done in non-isogenic condition ( Figure 1E; Figure 1—figure supplement 2 ) . All SAS-6 fragments lacking the C-terminal domain ( DM1 , DM3 , and DM5 ) failed to induce any centrosome formation ( Figure 1E , G ) , although expressed at similar or higher levels ( Figure 1F; Figure 1—figure supplement 1D ) . The C-terminal tail alone ( DM6 ) was also insufficient . However , when the full-length coiled-coil domain ( DM2 ) or a small portion of it ( DM4 ) was present together with the C-terminal tail , it effectively drove de novo centrosome formation in all cells ( Figure 1E , G; 100 ) . Consistently , SAS-6 harboring the F131E mutation ( F131E ) that has been previously shown to disrupt the oligomerization property of SAS-6 ( Kitagawa et al . , 2011; van Breugel et al . , 2011 ) also efficiently drove de novo centrosome formation ( Figure 1E–G , 100% ) , with an expression level slightly higher than the endogenous level ( Figure 1F; Figure 1—figure supplement 1D ) . A recent report showed that the fly SAS6 mutant equivalent to human SAS-6F131E could support the formation of some centriole-like structures in vivo ( Cottee et al . , 2015 ) . Our result thus suggests that in contrary to the SAS-6 self-assembly model , the main activity of SAS-6 in promoting centriole assembly is possessed by its C-terminal portion , rather than the N-terminal domain mediating SAS-6 self-oligomerization . To determine whether centrioles form the core of these de novo centrosomes , SAS-6-/- cells arrested in S phase were induced to express various types of SAS-6 mutants , and examined for de novo centriole assembly ( see ‘Materials and methods’ ) . Consistently , FL , F131E , DM2 , and DM4 , all of which contain intact C-termini , efficiently induced the formation of de novo centrioles in all cells ( 100% ) , as revealed by the co-localization of centrin , SAS-6 , STIL , CEP135 , and CPAP ( Figure 2; Figure 2—figure supplement 1 ) . Moreover , some of these de novo centrioles , even those formed with DM4 that lacks the entire N-terminal half of SAS-6 , could later mature properly by acquiring distal appendages . Importantly , upon G1 arrest , some of these mature centrioles were able to support ciliogenesis ( Figure 2B ) , suggesting that functionally normal centrioles can arise from de novo assembly through SAS-6 mutants that are incapable of self-oligomerization . 10 . 7554/eLife . 10586 . 006Figure 2 . De novo centrioles formed in the absence of SAS-6 self-oligomerization can ciliate and duplicate . ( A ) Isogenic SAS-6-/-; p53-/- cells carrying indicated SAS-6 constructs ( see ‘Methods’ ) were arrested in S phase and induced to express indicated SAS-6 mutants for 16 hr and then immunostained with indicated antibodies to visualize de novo centrioles ( centrin , green; SAS-6/HA , red ) . DNA ( DAPI , blue ) . ( B ) Isogenic SAS-6-/-; p53-/- cells carrying indicated SAS-6 constructs ( see ‘Methods’ ) were induced to express indicated SAS-6 mutants for 3 days , arrested in G1 for additional 36 hr , and then processed for immunofluorescence microscopy to visualize cilia ( GT335 , green ) and distal appendage ( CEP164 , red ) . DNA ( DAPI , blue ) . ( C ) Isogenic SAS-6-/-; p53-/- cells carrying indicated SAS-6 constructs ( see ‘Methods’ ) were induced to express indicated SAS-6 mutants for 3 days , and then processed for BrdU pulse-chase before fixation for immunofluorescence . Centriole duplication was revealed with indicated antibodies ( centrin , green; daughter centriole marker , STIL , red; PCM marker , γ-tubulin , blue ) . S-phase cells labeled with BrdU were shown ( blue ) . ( D ) Quantification of the centriole duplication efficiency from ( C ) . S-phase cells were identified ( BrdU+ ) and their centrioles were analyzed by immunofluorescence with centrin , STIL , and γ-tubulin antibodies . Error bars represent standard error of the mean ( SEM ) ; n > 150 , N = 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 00610 . 7554/eLife . 10586 . 007Figure 2—figure supplement 1 . Characterization of de novo centrioles formed in the absence of SAS-6 self-oligomerization . p53-/-; SAS-6 -/- cells ( clone #1 ) exogenously expressing full-length SAS-6 ( FL ) or various SAS-6 mutants as indicated were analyzed by immunofluorescence with indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 007 We next examine whether de novo centrioles formed with F131E or DM4 can later support canonical duplication in S phase after they have been converted to centrosomes ( Wang et al . , 2011 ) . Strikingly , both F131E- and DM4-derived centrioles could duplicate , as revealed by centrin and STIL staining , with an efficiency grossly similar to or only slightly less than that of centrioles derived from wild-type SAS-6 ( Figure 2C , D ) . This result , however , appears inconsistent with previous reports , where neither F131E nor DM4 could fully support centriole duplication ( Fong et al . , 2014; Kitagawa et al . , 2011; van Breugel et al . , 2011 ) , although both were able to initiate SAS-6/cartwheel assembly in the presence of pre-existing centrioles ( Fong et al . , 2014 ) . We do not understand the reason behind this discrepancy , but the result that F131E or DM4 can support the duplication of F131E- or DM4-derived centrioles , respectively ( Figure 2C , D ) , but not pre-existing , wild-type SAS-6-derived centrioles ( Fong et al . , 2014; Kitagawa et al . , 2011; van Breugel et al . , 2011 ) raises an idea that perhaps pre-existing centrioles have an active , dominant role in guiding canonical duplication . As SAS-6 self-oligomerization is also thought to define the ninefold symmetric shape of centrioles , it is possible that de novo centrioles derived from F131E or DM4 can function normally as centrosomes or basal bodies , but their shape or structural integrity may be defective . To determine the structural integrity of de novo centrioles , serial sectioning transmission electron microscopy ( TEM ) was used to analyze SAS-6-/- cells inducibly expressing FL , F131E or DM4 . To our surprise , normal looking centrioles , which are characterized as 200 nm x 500 nm in size , made of nine microtubule triplets , and equipped with distal and sub-distal appendages , could be easily found in cells expressing either F131E or DM4 ( Figure 3A–C ) . In particular , EM tomography reconstruction of a centriole from DM4 expressing cells revealed a perfect ninefold symmetric shape ( Figure 3B; Video 1 ) . Moreover , EM analyses also confirmed that at least some DM4-derived centrioles could ciliate ( Figure 3C ) and duplicate ( Figure 3D ) , firmly demonstrating that neither structural assembly nor shape determination of centrioles strictly depends on the self-oligomerization activity of SAS-6 . Rather , the C-terminal half of SAS-6 is both required and sufficient for centriole biogenesis . 10 . 7554/eLife . 10586 . 008Figure 3 . SAS-6 self-assembly is not essential for the ninefold symmetry of centrioles . Isogenic SAS-6-/-; p53-/- cells carrying indicated SAS-6 constructs ( see ‘Methods’ ) were induced to express indicated SAS-6 mutants for 3 days , and then processed for serial sectioning electron microscopy . ( A , B ) Mature canonical centrioles in WT RPE1 cells , or de novo centrioles formed in SAS6-/- cells expressing full-length ( FL ) or mutant SAS-6 as indicated are shown in longitudinal ( A ) or cross sectional ( B ) views . Note that these centrioles were mature and have acquired appendages . ( C ) Ciliated centrioles from indicated cell types were serially sectioned and examined by EM . ( D ) Duplicated , engaged centriole pairs from indicated cell types were serially sectioned and examined by EM . Scale bar: 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 00810 . 7554/eLife . 10586 . 009Video 1 . EM tomography of a normal centriole from SAS-6DM4 expressing cells ( related to Figure 3 ) . Note that a centriole on the right in cross-sectional views contains nine MT triplets . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 009 While SAS-6 self-assembly is not essential for centriole biogenesis , structurally normal centrioles can indeed form in the absence of pre-existing centrioles , questioning the necessity of having mother centrioles in canonical duplication . We thus asked if the defect-free rate of de novo centriole production is comparable to that of canonical duplication . EM analyses revealed that canonically duplicated centrioles are essentially free of error in their size , structural integrity , and geometric shape ( Figure 4E , 100% , n = 70 ) . In striking contrast , we consistently observed a small but significant portion of de novo centrioles that are abnormal , including centrioles missing variable numbers of MT triplets , or varying in size or shape ( Figure 4A–D; Video 2 ) . Importantly , similar types of abnormalities were seen regardless of whether centrioles were derived from SAS-6FL , SAS-6F131E , or SAS-6DM4 . Nevertheless , these abnormally looking centrioles appeared to have normal activities , suggesting that they are not simply unstable structures in the process of non-specific disintegration . For example , we observed SAS-6FL-derived centrioles that had only six MT triplets , but able to duplicate , producing daughter centrioles also made of an incomplete set of MT triplets ( Figure 4A-#1 ) . SAS-6FL-derived centrioles that appeared larger than normal centrioles ( Figure 4A-#2 ) , but able to mature and acquire appendages were also found . In addition , we observed SAS-6FL-derived centrioles made of disorganized MT triplets ( Figure 4A-#3 ) , or centrioles comprising of 9 MT triplets but existing as a distorted open cylinder ( Figure 4A-#6 ) . Consequently , the width ( outer diameter ) of SAS-6FL-derived centrioles is more variable than that of canonically duplicated centrioles ( Figure 4D ) . De novo centrioles made of disorganized , or abnormal numbers of , MT triplets have also been seen to form in cells depleted of the endogenous centrioles by laser ablation ( La Terra et al . , 2005 ) . 10 . 7554/eLife . 10586 . 010Figure 4 . Centrioles formed through de novo assembly are error-prone . ( A–C ) Isogenic SAS-6-/-; p53-/- cells carrying indicated SAS-6 constructs ( see ‘Methods’ ) were induced to express indicated SAS-6 mutants for 3 days , and then processed for serial sectioning electron microscopy . ( A1 ) A mature , SAS-6FL-derived centriole missed three MT triplets but was able to duplicate , producing daughter centrioles also made of an incomplete set of MT triplets ( see stars in sections I and II ) . It appears that two daughter centrioles were formed at the same time . ( A2 ) A mature , SAS-6FL-derived centriole 33% wider than normal centrioles was able to acquire appendages . ( A3 ) A SAS-6FL-derived centriole made of disorganized MTs . ( A4&5 ) Cross-sectional views of SAS-6FL-derived , abnormal centrioles missing MT triplets . ( A6 ) A SAS-6FL-derived centriole carrying 9 MT triplets ( arrow heads ) but existing as a distorted open cylinder ( arrow ) . ( B ) Abnormal centrioles derived from SAS-6DM4 were shown in cross-sectional or longitudinal views . ( C ) Abnormal centrioles derived from SAS-6F131E were shown in cross-sectional views . Note that a larger centriole made of 11 MT triplets was shown ( C1 ) . ( D ) The outer diameter of centrioles was quantified for both canonical centrioles ( in normal RPE1 cells ) , and SAS-6FL-derived de novo centrioles . ( E ) Quantifications showing the error rates of de novo and canonical centrioles . Sample size ( n ) is indicated . Note that arrows in all images indicate the missing of MT triplets . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 01010 . 7554/eLife . 10586 . 011Video 2 . EM tomography of an abnormal centriole from SAS-6DM4 expressing cells ( related to Figure 4 ) . Note that a centriole on the right in cross-sectional views contains eight MT triplets . DOI: http://dx . doi . org/10 . 7554/eLife . 10586 . 011 Similarly , abnormal centrioles were observed in the population of F131E- or DM4- derived centrioles , including smaller/incomplete centrioles ( Figure 4B , C ) , or bigger centrioles that were made of more than nine MT triplets ( Figure 4C-#1 ) . Notably , DM4-mediated , but not F131E-mediated , de novo synthesis clearly has an error rate higher than those mediated by SAS-6FL ( Figure 4E ) , suggesting that the N-terminal half of SAS-6 is involved in some additional processes that ensure the accuracy or quality control of centriole assembly . Taken together , these results suggest that in the absence of pre-existing centrioles , de novo centriole synthesis can occur , largely independent of SAS-6 self-oligomerization , but the process is inherently error-prone , and may require the presence of pre-existing centrioles to achieve high accuracy . We have established a cell-based system for reconstitution of de novo centriole assembly in human cells . Using this system , we found that self-oligomerization of SAS-6 , an activity proposed to drive cartwheel/centriole formation , is in fact not essential for the structural assembly or shape determination of centrioles . Instead , de novo centriole formation can be sufficiently driven by the C-terminal half of SAS-6 lacking the self-oligomerization activity , fundamentally deviating from the current paradigm for centriole biogenesis . Moreover , our results show that in human cells , de novo centriole assembly is an error-prone process that generates abnormal centrioles in a significantly higher frequency comparing to canonical duplication . The error rate of de novo assembly , which was estimated from later stages of centrioles , could potentially be underestimated , as it is plausible that the de novo centrioles formed initially are actually much more error-prone , and that over time ninefold symmetric centrioles persist because they are more stable or more efficient to duplicate . In any case , our results suggest that if de novo assembly must be used for centriole production , the structural accuracy may need to be insured or reinforced by other mechanisms . The expression level of the SAS-6FL or SAS-6F131E in our cell-based assay is higher than the endogenous level of SAS-6 responsible for canonical duplication ( Figure 1F; Figure 1—figure supplement 1D ) , a condition that may potentially affect the accuracy of de novo centriole assembly . However , as de novo centriole assembly is normally blocked under physiological conditions ( for number control ) , it is unclear what the ‘right’ level of SAS-6 should be , or if the level really matters that much , when the number control is not a relevant issue for de novo assembly . Consistently , it has been reported that de novo centrioles made of disorganized , or abnormal numbers of MT blades are formed with the endogenous level of SAS-6 ( La Terra et al . , 2005 ) . Conversely , during multiciliogenesis where SAS-6 level is highly up-regulated ( Hoh et al . , 2012 ) , almost every new centriole forms correctly . These results suggest that when cells carry pre-existing centrioles , they could perhaps tolerate a wider range of SAS-6 level for proper structural assembly of centrioles . Numerical control of centriole biogenesis , however , is a different issue , as it is highly sensitive to the sas-6 level ( Strnad et al . , 2007 ) , even in the presence of pre-existing centrioles . Our results thus lead to an interesting question: Is de novo synthesis a reliable , physiologically relevant pathway for centriole production in animals ? Large quantities of centrioles naturally arising from the acentriolar pathway has been reported in planarians ( Azimzadeh et al . , 2012 ) , but whether the same process occurs in vertebrates under normal physiological conditions is unclear . It will be very interesting to see if planarians carefully specify the structural integrity of de novo centrioles , and if so , how the accuracy is achieved in the absence of pre-existing centrioles . In vertebrate somatic cycling cells , the birth of a new centriole is normally given by the pre-existing centriole ( canonical duplication ) , a process known to regulate the number at which newborn centrioles can form . It has led to a belief that de novo assembly would take over the canonical duplication , when large , variable numbers of centrioles are needed , such as in the post-mitotic , multi-ciliated epithelium . Strikingly , a recent report showed that in multi-ciliated epithelia , de novo assembly is not responsible for the production of hundreds of centrioles ( Al Jord et al . , 2014 ) ; instead , all new centrioles that later form cilia are initiated , directly or indirectly , from the pre-existing centriole , supporting the idea that the presence of pre-existing centrioles has additional roles other than the number control for centriole reproduction . Our results that canonically duplicated centrioles are predominantly free of error , while de novo centriole assembly is error-prone , suggest that perhaps pre-existing centrioles can specify some steps of the assembly process that define the structural integrity of the centriole ( e . g . cartwheel assembly ) , an idea consistent with the template-based model of centriole formation proposed recently ( Fong et al . , 2014 ) . Notably , mouse embryos show exceptions that break all the known rules . The mouse embryogenesis starts with a fertilized egg lacking apparent centrioles/centrosomes , and finishes with an intact animal in which nearly all cells carry centrioles that are both numerically and structurally normal ( Abumuslimov et al . , 1994; Gueth-Hallonet et al . , 1993; Howe and FitzHarris , 2013; Palacios et al . , 1993 ) . How acentrosomal cells in mouse embryos escape from p53 dependent elimination acting against acentrosomal mitosis ( Bazzi and Anderson , 2014; Lambrus et al . , 2015; Wong et al . , 2015 ) , and how they later form and maintain proper centrioles in the absence of pre-existing centrioles are completely not understood . With our surprising observation that SAS-6 self-oligomerization is not essential for centriole assembly , perhaps other unanticipated mechanisms are specifically involved in centriole biogenesis in early mouse embryos . Human telomerase-immortalized retinal pigment epithelial cells ( hTERT-RPE1 or RPE1 ) were cultured in DME/F-12 ( 1:1 ) medium supplemented with 10% FBS and 1% penicillin-streptomycin . The expression constructs of the various SAS-6 DMs under the control of tetracycline-inducible promoter were made using the lentiviral vector pLVX-Tight-Puro vector ( Clonetech ) as described previously ( Fong et al . , 2014 ) . Antibodies used in this study include anti-α-tubulin ( Sigma-Aldrich ) , anti-HA . 11 ( Covance ) , anti-polyglutamylated tubulin ( GT-335; Adipogen ) , anti-pericentrin and CPAP ( Proteintech ) , anti-centrin2 ( Millipore ) , anti-BrdU ( AbD Serotec ) , CEP164 ( Novus Biologicals ) , anti-STIL ( Bethyl laboratories , Inc ) , anti–p53 and hSAS-6 ( Santa Cruz Biotechnology; sc-6243 , sc-376836 and sc-81431 ) , and anti-CEP135 ( abcam , ab75005 ) . The antibody for N-terminal SAS-6 ( sc-376836 ) is a monoclonal antibody raised against amino acids 1–300 , with the actual epitope being mapped within amino acids 173–300 ( Figure 1—figure supplement 1D ) . The antibody for C-terminal SAS-6 ( sc-81431 ) is a monoclonal antibody raised against amino acids 404–657 , with the epitope being mapped within amino acids 519–657 ( Figure 1F ) . RNA-guided targeting of genes in human cells was achieved through coexpression of the Cas9 protein with gRNAs using reagents prepared by the Church group ( Mali et al . , 2013 ) , which are available from the Addgene ( http://www . addgene . org/crispr/church/ ) . The targeting sequence for TP53 and SAS-6 is 5’-GGCAGCTACGGTTTCCGTC-3’ and 5’-GTGAAATGCAAAGACTGTG-3’ , respectively , which were cloned into the gRNA cloning vector ( Addgene plasmid #41824 ) via the Gibson assembly method ( New England Biolabs , Ipswich , MA ) as described previously ( Mali et al . , 2013 ) . To obtain stable acentriolar cells lacking SAS-6 , the TP53 gene in RPE1 cells was targeted by the CRISPR method prior to inactivation of SAS-6 . Six days after SAS-6 inactivation , we observed that about 10–15% of cells were devoid of centrioles or centrosomes . Pure acentriolar cell lines were subsequently established through clonal propagation from single cells , a process taking additional 4–5 weeks ( before these cells were used for experiments ) , generating a number of independent SAS-6-/-; p53-/- cell lines all behaving similarly ( with clone #1 and #2 being characterized here ) . SAS-6-/-; p53-/- cells actively proliferate or divide , but take longer periods of time to go through mitosis ( Figure 1D ) . For genotyping , the following PCR primers were used: 5’-ATCGGAATTCGGCCAAGTCTCTTACGCCTT-3’ and 5’- CTAGTCTAGAATGTGAGCCGGCTTCCTAAC-3’ for SASS6 alleles , and 5’- ACGCGGATCCACCCATCTACAGTCCCCCTTG-3’ and 5’-CTAGTCTAGAGCATCCCCAGGAGAGATGCT-3’ for TP53 alleles . PCR products were cloned and sequenced . To examine the role of SAS-6 in de novo centriole formation , SAS-6-/-; p53-/- cell lines generated above were infected with lentiviruses carrying various of SAS-6 constructs , and induced to express wild-type or mutant SAS-6 with 50 ng/ml Doxycycline for 16 hr . To examine the function of de novo centrioles to form centrosomes , to duplicate , or ciliate , infected cells were incubated with doxycycline for 3 days , followed by serum starvation if ciliogenesis was to be examined . Isogenic , acentriolar cell lines stably carrying specific SAS-6 expression constructs ( SAS-6-expression-ready cells ) were isolated and propagated from single cells in the absence of doxycycline , which allow us to directly induce de novo centriole/centrosome formation with doxycycline addition . Our reconstitution of de novo centriole/centrosome formation was successfully done in acentriolar cells infected with viruses and then treated with doxycycline ( Figure 1E; Figure 1—figure supplement 2C , D ) , or in isogenic , SAS-6-expression-ready cells treated with doxycycline ( Figures 1G , 2 ) . Cells were fixed with methanol at −20°C for 5 min . Slides were blocked with 3% bovine serum albumin ( w/v ) with 0 . 1% Triton X-100 in PBS before incubating with the indicated primary antibodies . Secondary antibodies were from molecular probes and were diluted 1:500 . DNA was visualized using 4′ , 6-diamidino-2-phenylindole ( DAPI; Molecular Probes ) . Fluorescent images were acquired on an upright microscope ( Axio imager; Carl Zeiss ) equipped with 100× oil objectives , NA of 1 . 46 , and a camera ( ORCA ER; Hamamatsu Photonics ) . For time-lapse experiments , hTERT-RPE1 cell ( wild-type , P53-/- or p53-/-; SAS-6-/- ) were imaged using a Zeiss Axiovert microscope configures with a 10X objective , motorized temperature-controlled stage , environmental chamber , and CO2 enrichment system ( Zeiss , Germany ) . Images were acquired and processed by axiovision software ( Zeiss , Germany ) . Cells grown on coverslips made of Aclar film ( Electron Microscopy Sciences ) were fixed in 4% paraformaldehyde and 2 . 5% glutaraldehyde with 0 . 1% tannic acid in 0 . 1 M sodium cacodylate buffer at room temperature for 30 min , postfixed in 1% OsO4 in sodium cacodylate buffer for 30 min on ice , dehydrated in graded series of ethanol , infiltrated with EPON812 resin ( Electron Microcopy Sciences ) , and then embedded in the resin . Serial sections ( ∼90-nm thickness ) were cut on a microtome ( Ultracut UC6; Leica ) and stained with 1% uranyl acetate as well as 1% lead citrate . Samples were examined on JEOL transmission electron microscope . Tomography data was collected using a JEOL 1400 Plus TEM operated at 120 kV . Double tilt series were recorded from –60° to 60° with 1° increments using SerialEM software ( Mastronarde , 2005 ) . Tomograms were reconstructed using IMOD programs ( Kremer et al . , 1996 ) .	Cells pass on their characteristics or “traits” to new generations in the form of DNA molecules . DNA provides the instructions to make proteins , which may then assemble into larger structures without using any external templates in a process called self-assembly . However , when a cell divides , DNA is not the only element that is passed on to the daughter cells; many large protein structures that have assembled in mother cells are also divided between the daughter cells . The daughter cells may then produce extra copies of these protein structures , but it is not known whether the pre-existing structures are involved in this process . Centrioles are complex structures made of proteins and play a crucial role in cell division . One of the main components of centrioles is a protein called SAS-6 . Recent studies have shown that SAS-6 molecules can bind to each other to form “oligomers” . This process , which is called self-oligomerization , has been proposed to drive the formation of centrioles . Now , Wang et al . examine whether centrioles can form properly in cells when no other centrioles are present . The experiments show that centrioles can indeed form , but they are prone to structural errors . In contrast , centrioles that form in the presence of older centrioles are essentially free of errors . The experiments used human eye cells that were missing the gene that encodes SAS-6 . These cells could not make centrioles , but when SAS-6 was re-introduced into these cells , new centrioles formed . Unexpectedly , re-introducing a mutant form of SAS-6 that cannot form oligomers into the cells still allowed new centrioles to form , which shows that self-oligomerization of SAS-6 is not essential for the assembly of centrioles . Together , Wang et al . ’s findings challenge the idea that SAS-6 self-oligomerization is involved in the formation of centrioles , and suggest that preexisting centrioles may help to minimize errors in the formation of new centrioles .	[ "Abstract", "Introduction", "Results", "Discussion", "Material", "and", "methods" ]	[ "cell", "biology" ]	2015	De novo centriole formation in human cells is error-prone and does not require SAS-6 self-assembly
Anaerobic thermophiles inhabit relic environments that resemble the early Earth . However , the lineage of these modern organisms co-evolved with our planet . Hence , these organisms carry both ancestral and acquired genes and serve as models to reconstruct early metabolism . Based on comparative genomic and proteomic analyses , we identified two distinct groups of genes in Thermovibrio ammonificans: the first codes for enzymes that do not require oxygen and use substrates of geothermal origin; the second appears to be a more recent acquisition , and may reflect adaptations to cope with the rise of oxygen on Earth . We propose that the ancestor of the Aquificae was originally a hydrogen oxidizing , sulfur reducing bacterium that used a hybrid pathway for CO2 fixation . With the gradual rise of oxygen in the atmosphere , more efficient terminal electron acceptors became available and this lineage acquired genes that increased its metabolic flexibility while retaining ancestral metabolic traits . Deep-branching , anaerobic , thermophilic Bacteria and Archaea inhabit relic environments that resemble the early Earth ( Baross and Hoffman , 1985; Martin et al . , 2008 ) . Thermophily ( Di Giulio , 2003 , 2000 ) , anaerobic metabolism ( Baross and Hoffman , 1985; Martin et al . , 2008; Schopf , 1983 ) and reliance on substrates of geothermal origin are among the proposed ancestral traits of these microorganisms ( Baross and Hoffman , 1985; Di Giulio , 2003 , 2000; Lane et al . , 2010; Martin et al . , 2008; Russell and Martin , 2004; Schopf , 1983 ) . At the same time , their lineages have co-evolved with Earth and their genomes also carry more recently acquired traits . Therefore , these microorganisms can be used as models to reconstruct the evolution of metabolism . Thermovibrio ammonificans is part of the phylum Aquificae , a deep-branching group of thermophilic bacteria found in geothermal environments ( Lebedinsky et al . , 2007; Sievert and Vetriani , 2012 ) . Based on phylogenetic analyses of the 16S rRNA gene as well as whole genomes , Aquificae are believed to be the earliest bacterial lineages having emerged on Earth along with the phyla Thermotogae and Thermodesulfobacteria ( Di Giulio , 2003; Pitulle et al . , 1994; Battistuzzi et al . , 2004 ) ( Figure 1 , Figure 1—figure supplement 1 ) . All the cultured members of this phylum are chemolithoautotrophs that use hydrogen as an energy source , have optimum growth temperatures between 65˚C and 95°C and rely on the reductive tricarboxylic acid cycle ( rTCA ) to convert carbon dioxide into biomass ( Hügler et al . , 2007; Hügler and Sievert , 2011; Sievert and Vetriani , 2012 ) ( Table 1 ) , making them ideal candidates to investigate the evolution of early metabolism . 10 . 7554/eLife . 18990 . 003Figure 1 . Phylogenetic tree of 16S rRNA sequences computed from the current version of the aligned 16S rRNA database obtained from the arb-SILVA project ( http://www . arb-silva . de/ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 00310 . 7554/eLife . 18990 . 004Figure 1—figure supplement 1 . Maximum likelihood tree showing the 16S rRNA relationship of the Aquificae phylum . Thermotoga spp . were used as outgroup . Bootstrap values based on 1000 replication are shown at branch node . Bar , 5% estimated substitution . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 00410 . 7554/eLife . 18990 . 005Table 1 . Characteristic of representative members of Aquificae phylum . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 005FamilyOrganismGrowth tempEnergy sourceCarbon sourceTerminal electron acceptorIsolated fromGenome sequence accession numberReferences AquificaceaeAquifex aeolicus VF595°CH2CO2O2Underwater volcanic vents , Aeolic Islands Sicily , ItalyNC_000918 . 1 ( Deckert et al . , 1998; Huber et al . , 1992 ) Hydrogenivirga sp . 128–5 R1-1175°CS2O32- , S0CO2NO3- , O2Lau Basin hydrothermal vent area , Pacific OceanNZ_ABHJ00000000 ( Nakagawa et al . , 2004; Reysenbach et al . , 2009 ) Hydrogenobacter thermophilus TK-675°CH2CO2O2Hot springs in Izu and Kyushu , JapanNC_017161 . 1 ( Arai et al . , 2010; Kawasumi et al . , 1984 ) Hydrogenobaculum sp . Y04AAS1----Marine hydrothermal area , Vulcano Island , ItalyNC_011126 . 1 ( Reysenbach et al . , 2009; Stohr et al . , 2001 ) Thermocrinis ruber DSM 1217385°CH2 , S2O32- , S0CO2O2Octopus Spring , Yellowstone National Park , Wyoming , USAPRJNA75073 ( Huber et al . , 1998 ) DesulfurobacteriaceaeBalnearium lithotrophicum 17S70–75˚CH2CO2S0Deep-sea hydrothermal vent chimney , Suiyo Seamount , JapanNot sequenced ( Takai et al . , 2003b ) Desulfurobacterium thermolithotrophum DSM 1169970°CH2CO2S0 , S2O32- , SO2-Deep-sea hydrothermal chimney , mid-Atlantic ridge , Atlantic OceanNC_015185 . 1 ( Göker et al . , 2011; L'Haridon et al . , 1998 ) Phorcysia thermohydrogeniphila HB-875°CH2CO2S0 , NO3-Tube of Alvinella pompejana tubeworms , deep-sea hydrothermal vents 13 ˚N , East Pacific Rise , Pacific OceanNot sequenced ( Pérez-Rodríguez et al . , 2012 ) Thermovibrio ammonificans HB-175°CH2CO2S0 , NO3-Deep sea hydrothermal vents 9˚N , East Pacific Rise , Pacific OceanNC_014926 . 1 ( Giovannelli et al . , 2012; Vetriani et al . , 2004 ) HydrogenothermaceaeHydrogenothermus marinus VM165°CH2CO2O2 ( 1–2% ) Deep sea hydrothermal vents 9˚N , East Pacific Rise , Pacific OceanNot sequenced ( Stohr et al . , 2001 ) Persephonella marina EX-H173°CS0CO2O2 , NO3-Deep sea hydrothermal vents 9˚N , East Pacific Rise , Pacific OceanNC_012440 . 1 ( Götz et al . , 2002; Reysenbach et al . , 2009 ) Sulfurihydrogenibium sp . YO3AOP1170°CS2O32- , S0CO2O2Calcite Hot Springs , Yellowstone National Park , USANC_010730 . 1 ( Reysenbach et al . , 2009; Takai et al . , 2003a ) Venenivibrio stagnispumantis70°CH2CO2O2Terrestrial hot spring Champagne Pool , Waiotapu , New ZealandNot sequenced ( Hetzer et al . , 2008 ) Incertae sedisThermosulfidibacter takai ABI70S6T70°CH2CO2S0deep-sea hydrothermal field at Southern Okinawa Trough , JapanNot sequenced ( Nunoura et al . , 2008 ) – Strain not formally described whose genome sequence is available . For these strains , the physiological information reported in Table S1 have been collected from MIG associated with the sequencing or from the closest validly published species . The ability to synthesize new biomass from inorganic precursors , i . e . autotrophic carbon fixation , is a critical step in the global carbon cycle and is considered a key invention during the early evolution of metabolism ( Braakman and Smith , 2012; Fuchs , 2011; Hügler and Sievert , 2011 ) . Among the six known pathways of carbon fixation ( reviewed in Fuchs , 2011; Hügler and Sievert , 2011 ) , the rTCA cycle ( present in the Aquificae , among others ) and the reductive acetyl-CoA pathway ( present in acetogenic bacteria and methanogens ) , represent good candidates for the ancestral carbon fixation pathway ( Martin and Russell , 2003; Russell and Hall , 2006 ) . The emergence of a reductive acetyl-CoA pathway has been associated with the FeS-rich minerals at alkaline hydrothermal vents ( Martin and Russell , 2007 , 2003; Wächtershäuser , 1988a ) and the presence of homologous core enzymes in both Bacteria and Archaea potentially support its ancestral nature ( Fuchs , 2011; Russell and Martin , 2004 ) . The rTCA cycle is considered the modern version of a proposed prebiotic autocatalytic cycle fueled by the formation of the highly insoluble mineral pyrite in sulfur-rich hydrothermal environments ( Wächtershäuser , 1988b , 1990 ) . The rTCA cycle is widespread among anaerobic and microaerophilic bacteria including all Aquificae , chemolithoautotrophic Epsilonproteobacteria , Chlorobi , and Nitrospirae ( Berg , 2011; Fuchs , 2011; Hügler and Sievert , 2011 ) in addition to few other bacterial strains . A recent phylometabolic reconstruction hypothesized that all extant pathways can be derived from an ancestral carbon fixation network consisting of a hybrid rTCA cycle / reductive acetyl-CoA pathway ( Braakman and Smith , 2012 ) . Here , we used a comparative genomic approach coupled to proteomic analyses to reconstruct the central metabolism of Thermovibrio ammonificans ( Giovannelli et al . , 2012; Vetriani et al . , 2004 ) , and provide evidence of ancestral and acquired metabolic traits in the genome of this bacterium . We suggest that the last common ancestor within the Aquificae possessed the reductive acetyl-CoA pathway in addition to the previously described rTCA cycle . Furthermore , we show that these two pathways may still coexist within the Desulfurobacteraceae lineage . The simultaneous presence of both pathways of carbon fixation may represent a modern analog of the early carbon fixation phenotype , and suggests that the redundancy of central metabolic pathways could have been common in ancestral microorganisms . The genome of T . ammonificans HB-1 consists of one chromosome and one plasmid of 1 , 682 , 965 bp and 76 , 561 bp , respectively ( Figure 2 ) . The genome encodes for 1890 genes , 1831 of which are protein coding genes and 75 . 16% of the protein coding genes could be assigned a putative function ( Giovannelli et al . , 2012 ) . The chromosome contains three ribosomal RNA operons , the first two with a 5 S-23S-16S alignment ( coordinates 149 , 693–152 , 859 bp and 1 , 168 , 776–1 , 172 , 141 bp antisense strand , respectively ) and the other with a 16 S-23S-5S ( coordinates 1 , 604 , 527–1 , 609 , 585 bp ) . The second copy of the operon is flanked by the largest CRISPR found in the genome ( Figure 2 , circle 6 ) . Several other repeats were identified in the chromosome ( Figure 2 , circle 5 ) . The plasmid contained mainly ORFs with unknown hypothetical functions ( Figure 2 , circle 9 ) with the exception of a putative RNA polymerase sigma factor of probable archaeal origin , a DNA Topoisomerase I ( 43 . 2% similarity to Hydrogenivirga sp . 128–5 R1-6 ) , a type II secretion protein E ( identified also in the preliminary proteome , Figure 3 , Figure 3—source data 1 ) and an ArsR-like regulatory protein , both of which had homologs among the Firmicutes . All other plasmid genes were ORFans with no significant similarities in the database and contained highly repeated DNA ( Figure 2 , circle 5 ) . 10 . 7554/eLife . 18990 . 006Figure 2 . Thermovibrio ammonificans HB-1 genomic map highlighting main genomic features . The circular chromosome and plasmid are drawn together . Several other repeats were identified in the chromosome ( circle 5 ) . From the outer circle: 1 – Dimension of the genome in base pairs ( chromosome in violet , plasmid in pink ) ; 2 and 3 – sense and antisense coding sequences colored according to their COGs classification ( see legend in the figure ) ; 4 – similarity of each coding sequence with sequences in the non-redundant database; 5 – position of tandem repeat in the genome . A total of 46 tRNA were identified , comprising all of the basic 20 amino acid plus selenocysteine; 6 – position of CRISPRs; 7 – GC mol% content , the red line is the GC mean of the genome ( 52 . 1 mol% ) ; 8 – GC skew calculated as G+C/A+T; 9 – localization of coding genes arbitrarily colored according to the metabolic pathways reconstructed ( see Figure 1 ) . Colors are consistent trough the entire paper . Total dimension and accession number for the chromosome and plasmid are given . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 00610 . 7554/eLife . 18990 . 007Figure 3 . Central metabolism of T . ammonificans HB-1 . Enzyme names are reported together with the gene locus number ( Theam_number ) . Primary compounds involved in reactions were also reported , however visualized reactions are not complete . Pathways were arbitrarily color coded according to their reconstructed function and are consistent throughout the paper . Solid shapes represent genes for which the enzyme was found in the proteome , while outlined shapes were only identified in the genome . Circled numbers 1 to 6 represent reaction numbers of the putative reductive acetyl-CoA pathway as described in the text . Abbreviations: NITRATE AMMONIFICATION: NapCMADGH – Periplasmic nitrate reductase complex; NirA – Putative nitrite reductase; AmtB – ammonia transporter; GlnA – L-Glutamine synthetase; GltA – Glutamate synthetase . HYDROGEN OXIDATION: HynABC – Ni-Fe Membrane bound hydrogenase; FrhACB – Cytosolic Ni-Fe hydrogenase/putative coenzyme F420 hydrogenase; HupD – Cyrosolic Ni-Fe hydrogenase maturation protease; HypA-F – Hydrogenases expression/synthesis accessory proteins; EchABCEF – Ech membrane bound hydrogenase complex; HydAB – Cytosolic Ni-Fe hydrogenases potentially involved in ferredoxin reduction; Hdr ? – Missing heterodisulfide reductase CoB-CoM; FdhABC – Formate dehydrogenase . ENERGY PRODUCTION: NADH dehydrogenase and ATP synthetase are reported without the names of the single units; Nhe – Sodium/hydrogen symporter . SULFUR REDUCTION: Sqr – Putative sulfate quinone reductase involved in sulfur respiration; Nsr ? – FAD/NAD nucleotide-disulphide oxidoreductase; Aps – Sulfate adenylyl transferase and kinase involved in assimilation of sulfur . FLAGELLAR COMPLEX: for simplicity single unit names are not reported . REDUCTIVE ACETYL-CoA: Fhs – Proposed reverse formyl-THF deformylase; PurHT – Formyltransferases ( phosphoribosylaminoimidazolecarboxamide and phosphoribosylglycinamide ) FolD – Methenyl-THF cyclohydrolase and dehydrogenase; MetVF – Methylene-THF reductase; MetR – Putative methyl-transferase; CodH – Carbon monoxyde dehydrogenase; AcsA – Acetyl-CoA ligase/synthase; AccABCD – Acetyl-CoA carboxylase . REDUCTIVE CITRIC ACID CYCLE: AclAB – ATP-citrate lyase; Mdh – Malate dehydrogenase; FumAB – Fumarate hydratase; FrdAB – Fumarate reductase; SucCD – Succinyl-CoA synthetase; OorABCD – 2-Oxoglutarate synthase; Idh2 – Isocitrate dehydrogenase/2-oxoglutarate carboxylase; AcnB – Aconitate hydratase; PorABDG – Pyruvate synthase; PycAB - Pyruvate carboxylase; PpsA – Phosphoenolpyruvate synthase water dikinase; PyK – Pyruvate:water dikinase . GLUCONEOGENESIS: Eno – Enolase; Pgm – Phosphoglycerate mutase; Pgk – Phosphoglycerate kinase; Gapor – Glyceraldehyde-3-phosphate dehydrogenase; Gapdh – Glyceraldehyde 3-phosphate dehydrogenase; Fba – Predicted fructose-bisphosphate aldolase; Tim – Triosephosphate isomerase; Fbp – Fructose-1 , 6-bisphosphatase I; Pgi – Phosphoglucose isomerase; Pgm – Phosphoglucomutase . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 00710 . 7554/eLife . 18990 . 008Figure 3—source data 1 . List of the proteins identified in the proteome of T . ammonificans grown under nitrate reducing conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 00810 . 7554/eLife . 18990 . 009Figure 3—figure supplement 1 . Reductive TCA cycle variants found in extant bacterial lineages . ( A ) two-step variant of the rTCA cycle , also known as symmetric variants , were the carboxylation of 2-oxoglutarate is performed in two steps by the enzymes 2-oxoglutarate carboxylase and oxalosuccinate reductase , and the cleavage of citrate is performed in two step by the citryl-CoA synthetase and citryl-CoA lyase enzymes; ( B ) one-step variant of the rTCA cycle , also known as asymmetric variants , were the carboxylation of 2-oxoglutarate is performed in a single step by the enzymes isocitrate dehydrogenase , and the cleavage of citrate to acetyl-CoA is performed in a single reaction by the ATP citrate lyase enzyme . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 00910 . 7554/eLife . 18990 . 010Figure 3—figure supplement 2 . The activated methyl cycle of T . ammonificans reconstructed from the genome . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 010 By combining genome-scale analyses with known physiological traits , we were able to reconstruct the metabolic network of T . ammonificans ( Figure 3 ) . As part of the central metabolism of T . ammonificans , we identified the major pathways for carbon fixation , hydrogen oxidation and nitrate reduction to ammonia , and confirmed them by identifying the corresponding expressed enzymes in the proteome ( Figure 3 , Figure 3—source data 1 ) . Further , we compared the central metabolism of T . ammonificans with closely related species within the Aquificae , and with ecologically similar Epsilonproteobacteria . Comparative genomic analyses with closely related strains provide information about the evolutionary history of the group , while comparison with phylogenetically distant , but ecologically similar species may reveal common adaptive strategies to similar habitats . 780 proteins , comprising approximately 43% of all predicted ORFs in the database , were identified in cell samples of T . ammonificans grown under nitrate reducing conditions ( Figure 3 , Figure 3—source data 1 ) , indicating a very high congruence between database entries and sample peptides . Among the most abundant proteins were – besides those involved in general housekeeping functions like the translation elongation factor Tu , ribosomal proteins and chaperones – many rTCA cycle enzymes . The four key enzymes of the rTCA cycle ( see next paragraph for details ) together yielded a relative abundance of 6% of all identified proteins in the samples . Most of the enzymes putatively involved in the reductive acetyl-CoA pathway ( Table 2 ) , including the carbon-monoxide dehydrogenase ( CodH ) , were also detected , although at a lower abundance of 1 . 25% of all proteins . 10 . 7554/eLife . 18990 . 011Table 2 . Enzymes involved in the putative reductive acetyl-CoA pathway in T . ammonificans HB-1 , closest relative homolog and homologs within the Aquificae phylum . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 011Desulfurobac teriaceaeHydrogenothe rmaceaeAquificaceaeReactionPutative enzymePutative gene locusClosest relativePutative origin‡D . thermolitho trophumDesulfuro bacterium sp . TC5-1P . marinaSulfurihydro genibium sp . YO3AOP1H . themophilusHydrogeno baculum sp . Y04AAS1HHydrogeni virga sp . 128–5 R1-6T . ruberA . aeolicus 1Formate dehydrogenaseTheam_1020Nitratiruptor sp . SB155-2 sim . 55% YP_001357016Methanogensf . e . § - 30% YP_00428203#-*56% YP_002730364f . e . § - 32% YP_001930236f . e . § - 33% YP_003433330--f . e . § - 27% YP_003474076- 25-formyltetra hydrofolate cyclo-ligaseTheam_1206Hydrogenivirga sp . 128–5 R1-1 sim . 41% WP_008285842Deltaproteo bacteria / Gram +41% YP_00428133944% WP_022846876 . 1f . e . § - 36% YP_002729868f . e . § - 38% YP_001931834f . e . § - 38% YP_003431710f . e . § - 30% YP_00212153941% WP_00828584240% YP_00347298143% 1SOU_Aformyltetra hydrofolate deformylase / hydrolaseTheam_0826Hydrogenivirga sp . 128–5 R1-1 sim . 70% WP_008287030Bacteria82% YP_00428107777% WP_022846479----70% WP_00828703065% NP_214247-phosphori bosylglycina mide formyl transferase 2Theam_1211P . marina sim . 76% YP_002731257Methanogens / Deltaproteo bacteria78% YP_00428133564% WP_02284751276% YP_00273125773% YP_001931730f . e . § - 28% YP_00343307267% YP_00749947974% WP_008288476f . e . § - 25% YP_003473050f . e . § - 30% NP_213168phosphori bosylaminoi midazole carboxamide formyltransferase /IMP cyclohydrolaseTheam_0328P . marina sim . 59% YP_002731268Aquificae / Bacteria86% YP_00428168178% WP_02284735959% YP_00273126856% YP_00193123953% YP_00343327050% YP_00749947357% WP_00828836052% YP_00347433756% NP_214344 3methylenetetra hydrofolate dehydrogenase / methenyltetra hydrofolate cyclohydrolaseTheam_1261A . aeolicus sim . 60% NP_214304Aquificae / Gram +87% YP_00428114774% WP_02284621755% YP_00273147658% YP_00193124059% YP_00551131858% YP_001931240-58% YP_00347310460% NP_214304 4methylenetetra hydrofolate reductaseTheam_0919Methanosrcina barkeri sim . 48% YP_305816Methanogens87% YP_00428114774% WP_02284621755% YP_00273147658% YP_00193124059% YP_00551131858% YP_002121558f . e . § - 24% WP_00828678358% YP_00347310460% NP_214304 55-methyltetra hydrofolate corrinoid/iron sulfurprotein methyltransferaseTheam_1184Clostridium glycolicum sim . 42% WP_018589788Gram +82% YP_00428148472% WP_022846603f . e . § - 31% YP_002729885f . e . § - 38% YP_001930381f . e . § - 37% YP_003432344f . e . § - 34% YP_002122031f . e . § - 38% WP_008287326f . e . § - 36% YP_003473438f . e . § - 34% NP_213375 6†Carbon-monoxide dehydrogenaseTheam_1337Candidatus Methano peredens sp . BLZ1 sim . 56% KPQ43483Archaea / Gram +87% WP_013638427 38% WP_01363803076% WP_02284714341% YP_002729904------* – Reactions are numbered according to Figure 1 . For reaction two we reported the possible enzymes that could substitute the missing 10-fomyl-THF synthetase ( Fhs ) . The enzymes are listed in order of decreasing likelihood of their involvement in the reaction based on putative substrate affinity . † – Reaction six is catalyzed by the Acs/CodH complex , reported here separately . ‡ – Putative origin of the gene was calculated as consensus taxonomic assignment of the first 100 bastp hits against the nr database . Double assignment implies equal number of assignments to the two taxonomic groups . § – f . e . = functional equivalent annotated in the genome with similarity below 40% with T . ammonificans equivalent gene . Not considered a true homolog in the present study . # – Pairwise similarity to T . ammonificans translated gene and accession number of the homologs . ** – Missing homolog or functional annotated equivalent in the genome . The genome and proteome of T . ammonificans ( Figure 3 ) supports previous evidence , based on detection of key genes and measurements of enzyme activities , that carbon fixation occurs via the rTCA cycle ( Hügler et al . , 2007 ) . The rTCA is widespread among anaerobic and microaerophilic bacteria including all Aquificae , chemolithoautotrophic Epsilonproteobacteria , Chlorobi , Nitrospina aaaand Nitrospirae , in addition to few other bacterial strains ( e . g . Desulfobacter hydrogenophilus , Magnetococcus marinus MC-1 and the endosymbiont of the vent tubeworm , Riftia pachyptila; Hügler and Sievert , 2011 ) . We identified all the genes responsible for the functioning of the rTCA in the genome and proteome of T . ammonificans ( Figure 3 ) , including the key enzymes fumarate reductase ( frdAB , gene loci Theam_1270–1275 ) , ATP-citrate ( pro-S ) -lyase ( aclAB , 1021–1022 ) , 2-oxoglutarate synthase ( oorABCD , 1410–1413 ) and isocitrate dehydrogenase ( idh2 , 1023 ) . T . ammonificans synthesizes 2-oxoglutarate from succinyl-CoA rather than via citrate , a feature in common with methanogenic Archaea and Desulfurococcales , and present in numerous strict anaerobes that use the rTCA cycle ( Fuchs , 2011 ) . Comparative analyses showed that several genes associated with the rTCA cycle and the gluconeogenesis pathway were highly conserved in all Aquificae , with the exception of the genes coding for the ATP-citrate lyase and the isocitrate-dehydrogenase , which are absent in the Aquificaceae ( see details below ) ( Hügler et al . , 2007 ) . In general , T . ammonificans genes had a high percent identity to the genes of D . thermolithotrophum , its closest relative whose genome sequence is available . Two separate variants of the rTCA cycle are known ( Braakman and Smith , 2012; Hügler et al . , 2007 ) . In the two-step variant – also known as symmetric variant , with respect to the enzymatic reaction catalyzed in the two arcs of the rTCA cycle – citrate cleavage is accomplished by the combined action of the enzymes citryl-CoA synthetase ( CCS ) and citryl-CoA lyase ( CCL ) and the carboxylation of 2-oxoglutarate is catalyzed by 2-oxoglutarate carboxylase and oxalosuccinate reductase ( Aoshima , 2007; Aoshima and Igarashi , 2008; Hügler and Sievert , 2011 ) . This variant of the rTCA cycle was previously proposed as ancestral ( Braakman and Smith , 2012; Hügler et al . , 2007 ) . By contrast , in the one-step variant of the rTCA cycle , also known as asymmetric variant and more widely distributed , the cleavage of citrate and the carboxylation of 2-oxoglutarate are performed in a one-step fashion ( Braakman and Smith , 2012; Hügler et al . , 2007 ) ( Figure 3 , Figure 3—figure supplement 1 ) . This variant is found in T . ammonificans . The genome of T . ammonificans also codes for an incomplete reductive acetyl-CoA pathway ( Figure 3 and Table 2 ) . The key enzyme of this carbon fixation pathway , which is found in acetogens , methanogens , sulfate-reducers and anaerobic ammonium oxidizers ( anammox; Berg et al . , 2010 ) , is the bifunctional heterotetramer enzyme carbon monoxide dehydrogenase/acetyl-CoA synthase ( CODH/ACS ) . The reductive acetyl-CoA pathway is believed to be the most ancient carbon fixation pathway on Earth ( Fuchs , 2011 ) . In T . ammonificans , the genes coding for the CO-dehydrogenase ( codh catalytic subunit Theam_1337; Table 2 and Figure 3 ) is present and also expressed . However , rather than coding for the classical bifunctional type II CodH , T . ammonificans codes for the type V , which is present in numerous deep-branching organisms ( Figure 4 ) and whose function is unknown ( Techtmann et al . , 2012 ) . Sequence analysis of the T . ammonificans CodH protein sequence revealed the presence of conserved residues necessary for the interaction with Acs and for the coordination of the NiFeS cluster . We hypothesize that the type V CodH , which , based on structural and size considerations may represent the most ancestral version of this class of enzymes ( Frank Robb , personal communication ) , may catalyze the reduction of CO2 to CO . The codh gene of T . ammonificans shared high similarity with that of D . thermolithotrophum ( 87 similarity ) , while the similarity with the next closest homologous genes dropped below 54% . The genome of D . thermolithotrophum also codes for a classical bifunctional Type II CodH ( Dester_0417 and Dester_0418; Figure 4 ) in addition to the type V CodH . The two genomes also share a CodH iron-sulfur accessory protein ( Theam_0999 and Dester_0418 , 81% similarity ) . Additional laboratory analyses will be required to confirm the role of the type V CodH in carbon fixation . 10 . 7554/eLife . 18990 . 012Figure 4 . Neighbor-joining tree showing the position of the carbon monoxide dehydrogenase , CodH , of T . ammonificans ( in bold ) . Bootstrap values based on 1000 replications are shown at branch nodes . Only bootstrap values above 50% are reported . Bar , 10% estimated substitutions . The metabolism of each organism is reported on the side . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 012 T . ammonificans uses the Embden-Meyerhof-Parnas pathway to synthesize pentose and hexose monosaccharides . We identified in the genome the key genes of this pathway , fructose-1 , 6-biphosphatase ( fbp , Theam_1323; identified also in the proteome , Figure 3 ) and all other genes necessary for the functioning of the pathway ( Figure 3 ) . While the fbp gene is present in Desulfurobacterium thermolithotrophum , and in the genome of Sulfurohydrogenibium sp . YO3AOP1 , Persephonella marina and Hydrogenivirga sp . 128–5 R1-6 , it is absent from the genome of A . aeolicus ( Deckert et al . , 1998 ) and H . thermophilus , suggesting that in those species an alternative unidentified pathway may be active . Hydrogen is the only electron donor used by T . ammonificans and its importance in the metabolism of this bacterium is reflected by the diversity of hydrogenases found in the genome , whose encoding genes are organized in two large clusters ( Figure 2 circle 9 ) . The first cluster is composed of eight genes encoding for the complete Group 4 Ech hydrogenases ( echABCDEFG , Theam_0476–0482 ) and the alpha , beta , gamma and delta subunits of the Group 3b multimeric cytoplasmic Hyd sulfhydrogenase ( hydABCD , Theam_0484–0487 ) . The second hydrogenase cluster in the genome codes for the maturation protein HypF ( hypF , Theam_1116 ) , the three subunit of the Group 3a cytoplasmic F420-reducing hydrogenase Frh ( frhACB , Theam_1117–1119 ) and the Group 1 hydrogenases Hyn ( hynABC , Theam_1121–1123 ) with associated maturation factors/chaperones Hyp ( Theam_1124–1128 ) . We also found other hydrogenase maturation factors ( Theam_0922–0924 ) and the formate dehydrogenase Fdh ( fdhABC , Theam_1020 and Theam_0966–0967 respectively ) scattered in the genome . We reconstructed the phylogeny of T . ammonificans [NiFe]-hydrogenases ( Figure 5 ) , and we assigned them to known classes of hydrogenases ( Groups 1 through 4 ) according to their phylogenetic position in the tree and the presence of known conserved amino acid motifs typical of each group ( Vignais and Billoud , 2007 ) . Group 1 [NiFe]-hydrogenases are normally coupled to anaerobic respiration through electron transfer to the membrane quinone and menaquinone pool . Comparative analyses of the hydrogenases revealed that the Group 1 cluster in T . ammonificans ( Hyn ) is more similar to the hydrogenases found in Deltaproteobacteria than to homologs in other Aquificae ( Figure 5 ) . This finding raises interesting questions about the origin of the hydrogen oxidation Group 1 enzymes in T . ammonificans and suggests that an event of lateral gene transfer occurred between T . ammonificans and the Deltaproteobacteria . Further , the periplasmic nitrate reductase ( napA ) and the nap operon structure in T . ammonificans also appears to be closely related to that of Deltaproteobacteria , suggesting that a horizontal gene transfer event involved the entire hydrogen oxidation ( Hyn ) /nitrate reduction respiratory pathway ( see discussion ) . 10 . 7554/eLife . 18990 . 013Figure 5 . Neighbor-Joining tree showing the position and classification of the [NiFe]-hydrogenases found in the genome of T . ammonificans . Bootstrap values based on 1000 replication are shown at branch node . Bar , 20% estimated substitution . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 013 Contrarily to members of the Aquificaceae and Hydrogenothermaceae , T . ammonificans and D . thermolithotrophum genomes code for both Group 3 cytoplasmic hydrogenases , Group 3a F420-reducing hydrogenases ( frhACB ) and Group 3b multimeric cytoplasmic sulfhydrogenases ( hydABCD; Figure 5; Jeon et al . , 2015 ) . Comparative analyses of these two clusters revealed that other members of the Desulfurobacteriaceae family were the closest relatives , while the next closest enzymes were those found in hyperthermophilic Euryarchaeota ( i . e . Pyrococcus abyssi ) and in methanogens ( Figure 5 ) . Overall , these results imply a polyphyletic origin for the group 3 cytoplasmic hydrogenases . Comparative analyses revealed similarities with hyperthermophilic Euryarchaeota and methanogens ( Figure 5 ) , suggesting a possible common ancestry for the group 3 cytoplasmic hydrogenases of Archaea and Desulfurobacteriaceae . Together with Group 3b sulfhydrogenases and Group 4 Ech , F420-reducing hydrogenases are in fact typically found in methanogens in which the F420-cofactor is used as an electron carrier in the reduction of CO2 or other C1 compounds to methane ( Vignais and Billoud , 2007 ) . Cultures of T . ammonificans grown with nitrate as the terminal electron acceptor did not present the typical autofluorescence of methanogens due to F420-cofactor fluorescence when observed under UV in epifluorescence microscopy and the genome lacks the gene involved in the biosynthetic pathway of the F420-coenzyme . Due to these observations , we suggest that these hydrogenases may work in conjunction with ferredoxin in T . ammonificans . Group 3b Hyd sulfhydrogenases appear to be of mixed origins despite being present in a single operon . Three of the subunits share similarities with the Thermococcales ( Euryarchaeota; hydBCD ) while the catalytic subunit ( HydA ) appears to be unique to the Desulfurobacteriacea , not having any other significant homolog in the database ( Figure 6 ) . Comparative analyses indicated that Ech shared similarity with Epsilonproteobacteria Group 4 hydrogenases ( Figure 5 ) . Interestingly , Group 2 hydrogenases are found in microaerobic members of the Aquificaceae and Hydrogenothermaceae and in microaerobic Epsiloproteobacteria but are not found in anaerobes , suggesting a possible link with oxygen adaptation . In contrast , Group 3a is found in strict anaerobes , including T . ammonificans and methanogens ( Figure 5 ) . 10 . 7554/eLife . 18990 . 014Figure 6 . Blastp best hit ( cut off at 40% similarity ) for the T . ammonificans CDS: ( a ) Desulfurobacterium thermolithotrophum is included in the database; ( b ) D . thermolithotrophum and Desulfurobacterium sp . TC5-1 are excluded from the database ( best hits are outside of the Desulfurobacteriaceae family ) . The interactive versions of the Krona plots are available for download at DOI: 10 . 6084/m9 . figshare . 3178528 . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 014 Finally , Ech hydrogenases are the only Group 4 hydrogenases found in T . ammonificans . Both the reductive TCA cycle and the reductive acetyl-CoA pathway involve reaction steps that are driven by reduced ferredoxin ( Fuchs , 2011 ) . In the case of T . ammonificans , the enzymes formate dehydrogenase , CO-dehydrogenase , pyruvate synthase and 2-oxoglutarate synthase require reduced ferredoxin . We hypothesize that T . ammonificans can accomplish ferredoxin reduction in two different ways: the first involves cytosolic hydrogenases ( Group 3 , Frh and Hyd; Figure 3 ) commonly found in methanogens , where they reduce ferredoxin via electron bifurcation ( Fuchs , 2011 ) ; the second involves membrane bound hydrogenases ( Group 4 , Ech; Figure 3 ) . With the exception of the F420-reducing ( Frh ) hydrogenases , representatives of the remaining three groups were identified in the proteome of T . ammonificans ( Figure 3 ) . T . ammonificans HB-1 can use nitrate as a terminal electron acceptor and reduces it to ammonium ( Vetriani et al . , 2004 ) . Periplasmic nitrate reductases ( Nap ) have been studied extensively in some Proteobacteria , and are comprised of the catalytic enzymes – typically a heterodimer ( NapAB ) – associated with various other subunits involved in channeling electrons to the NapA reactive center ( Moreno-Vivián et al . , 1999; Vetriani et al . , 2014 ) . In T . ammonificans , the genes encoding for the Nap complex are organized in a single operon , napCMADGH . Comparative genomic analyses with genomes from Aquificae and Proteobacteria indicate that the nap operon structure of T . ammonificans is unique to this organism , and shares similarity with Desulfovibrio desulfuricans and other Deltaproteobacteria . All other Aquificae whose genomes are available have a Nap operon similar to that of Epsilon- and Gammaproteobacteria ( Vetriani et al . , 2014 ) . Moreover , unlike other members of the Aquificae , the nitrate reductase of T . ammonificans ( encoded by gene Theam_0423 , expressed in the proteome; Figure 3 ) appears to be of the monomeric type ( sensu; Jepson et al . , 2006 ) , and the NapB subunit is missing ( Figure 3 ) . We also identified a putative nitrite reductase-encoding gene possibly involved in the reduction of nitrite to ammonium ( NirA; Theam_1000 , Figure 3 ) , which was also expressed in the proteome ( Figure 3 ) . However , the exact mechanism through which T . ammonificans reduces nitrite to ammonium remains to be experimentally elucidated . While T . ammonificans is able to reduce elemental sulfur to hydrogen sulfide ( Vetriani et al . , 2004 ) , the sulfur reduction pathway remains unclear . We identified in the genome a putative polysulfide reductase of the sulfide-quinone reductase family ( sqr , Theam_0841 , Figure 3 ) . We propose that this membrane-bound complex can utilize polysulfide formed from the reaction of S0 with sulfide ( naturally enriched in hydrothermal systems ) via quinone oxidation ( Figure 3 ) . The putative sqr gene of T . ammonificans shares similarities with other Aquificae and Epsilonproteobacteria ( average 56% similarity to both groups ) . In particular , the sqr gene appears to be conserved also in Caminibacter mediatlanticus ( Figure 7; Giovannelli et al . , 2011 ) , whose sulfur reduction pathway is yet to be elucidated . Another possibility is that T . ammonificans uses a NAD or FAD-dependent reductase ( nsr , Theam_1321 ) to reduce sulfur , a mechanism previously described in the sulfur-reducing archaeon Pyrococcus furiosus ( Schut et al . , 2013 , 2007 ) . However , the putative polysulfide reductase of T . ammonificans has only about 33% identity to homologous enzymes of sulfur reducing bacteria and archaea identified in the database . 10 . 7554/eLife . 18990 . 015Figure 7 . Hives plot presenting the comparative genomic analysis of the T . ammonificans ( Tam ) genome with the closest relative available genomes of D . thermolithotrophum ( Des ) and the ecologically similar Epsilonproteobacterium C . mediatlanticus ( Cam ) . The three axes represent the organism’s linearized genomes with the origin of replication at the center of the figure . Genome size was normalized for visualization purpose . ( A ) Localization of homologous genes and syntenic regions among the three genomes . Lines colored according to the general pathway code adopted in Figure 3 connect homologous coding sequences . Lines opacity is proportional to similarity between sequences ( darker = higher similarity ) , line width is proportional to the extent of the syntenic area . ( B ) Hives panel of the homologous genes divided by pathways . Collinear lastZ alignments and interactive dot-plot replicating the analysis are accessible at the GenomeEvolution website with the following permanent addresses: T . ammonificans and D . thermolithotrophum ( http://genomevolution . org/r/9vb3 ) ; T . ammonificans and C . mediatlanticus ( http://genomevolution . org/r/9vb4 ) ; D . thermolithotrophum and C . mediatlanticus ( http://genomevolution . org/r/9vb8 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 015 T . ammonificans is a strict anaerobe . However , its genome codes for genes involved in the detoxification of oxygen radicals , including a catalase/peroxidase ( Theam_0186 ) , whose activity was previously detected ( Vetriani et al . , 2004 ) , a putative superoxide reductase ( Theam_0447 ) , a cytochrome-C peroxidase ( Theam_1156 ) and a cytochrome bd complex ( Theam_0494–0496 , Figure 3 ) . The latter three enzymes were identified in the proteome ( Figure 3 ) . The cythochrome bd complex has been shown to contribute to oxygen tolerance in other anaerobic bacteria ( Das et al . , 2005 ) , and to contribute to the detoxification of nitrous oxide radicals ( Mason et al . , 2009 ) . The genome of T . ammonificans contains two gene clusters coding for the NADH dehydrogenase ( Theam_0731–0745 and Theam_1488–1498 ) . The genes appear to have undergone an inversion during the duplication process and the second copy is missing four genes in the middle of the operon ( nuoEFG and a transcriptional regulator ) . The two clusters share on average 60% gene similarity . Only one copy of the gene for the synthesis of ATP synthetase complex was present ( Theam_1605–1606 and Theam_1656–1662 ) . Three different hydrogen-sodium symporters ( nhe , Theam_0443 , 0413 and 0647 ) were present with low pairwise similarity . NADH dehydrogenase second cluster and ATP synthetase complex appear to be conserved in the close relative D . thermolithotrophum and in the Epsilonproteobacterium , C . mediatlanticus TB-2 ( Figure 3 ) . T . ammonificans posses one to two terminal flagella ( Vetriani et al . , 2004 ) . The genes involved in flagella formation and assembly are prevalently organized in a single large cluster in the genome ( from Theam_1440 to Theam_1468 , Figure 7 ) , putatively organized in three distinct operons . The hook-filament junction proteins genes flgK and flgL ( Theam_1384 and Theam_1385 ) , the filament and filament cap proteins fliC and fliD ( Theam_1160 and Theam_1162 ) together with few flagellar maturation factors and motor switch genes ( motA and motB , Theam_1087 and Theam_1085–1086 , respectively ) are instead scattered in the genome . The entire cluster of genes shares similarities with D . thermolitotrophum , although the similarity for some of the proteins is as low as 50% . Despite this , comparative analyses revealed a similar organization in D . thermolitotrophum where the cluster appears inverted in the same relative genomic position and constituted one of the main region of synteny ( Figure 7b ) . By contrast , the flagellar genes in the Epsilonproteobacterium , Caminibacter mediatlanticus TB-2 , are scattered throughout the genome ( Figure 7b ) . Numerous genes for adhesion and pili were present , generally organized in small clusters ( Figure 7b ) . Four different putative methyl-accepting chemotaxis sensory transducer proteins were found in the genome ( mcp , Theam_0157 , 0165 , 0845 and 1027 ) . One of those , Theam_0845 , is surrounded by receptors cheW and cheA , and by the response regulator cheY . The entire group is organized in a single operon , cheYVBWAZ conserved in D . thermolithotrophum and other Aquificales with the exception of cheW , which shares higher similarities with the Epsilonproteobacterium , C . mediatlanticus TB-2 ( Figure 6 ) . We failed to find in the genome of T . ammonificans a homolog of the primary receptor encoding gene cheR ( Wadhams and Armitage , 2004 ) . We identified in the genome numerous membrane transporters for molybdenum ( modCBA , Theam_0787–0789 ) , iron ( fhuDBC , Theam_0192–0194 ) , zinc ( znuAB , Theam_0238 and 0689 ) and cobalt/nickel ( cbiOQMN , Theam_1522 , 1523 , 1525 and 1526 ) all involved in the uptake of important micro-elements for enzyme and cofactor synthesis . We also identified the complete type II secretion pathway , which is conserved in gram-negative bacteria and responsible for protein and toxin translocation to the extracellular milieu . Proteins secreted by the type II pathway include proteases , cellulases , pectinases , phospholipases , lipases , toxins and in some cases type III and IV pili ( Douzi et al . , 2012 ) . In some bacteria , the expression of secretion type II pathway genes are regulated either by quorum-sensing or by the environmental factors at the site of colonization ( Douzi et al . , 2012; Sandkvist , 2001 ) . Biofilm formation in bacteria is often linked to quorum-sensing mechanisms , regulating the settlement of planktonic cells in relationship to environmental sensing , substrate suitability and cell densities . It is likely that T . ammonificans maintains a mostly attached lifestyle in hydrothermal vent environments due to the high turbulence associated with fluid flux and mixing with seawater . We speculate that the type II pathway genes , together with chemotaxis and flagellar/adhesion genes may play a key role in the formation of biofilm . Extracellular secreted proteins may in fact be used to ‘condition’ the colonized substrate and to interact/remodel existing and newly formed biofilms . The activated methyl cycle is a conserved pathway in which S-methyl groups from L-methionine are converted into a biochemically reactive form through insertion into an adenosyl group ( Vendeville et al . , 2005 ) . S-adenosyl-L-methionine provides activated methyl groups for use in the methylation of proteins , RNA , DNA and certain metabolites central to the core cell machinery . The activated methyl cycle exists in two recognized forms , involving respectively luxS or sahH genes ( Vendeville et al . , 2005 ) . Genome-based reconstruction of the activated methyl cycle indicated that , in T . ammonificans , involves the sahH gene , while a luxS homolog is absent ( Figure 3—figure supplement 2 ) . Methionine methyl groups are adenosylated by a methionine adenosyltransferase ( samS , Theam_1075 ) . The methyl group is thus activated and readily available for the S-adenosylmethionine-dependent methyltransferase ( samT ) for methylation of substrate . In T . ammonificans , samT is a pseudogene , as it contains two stop codons ( Theam_0630 ) . Homologs of samT are absent in the genome of the other Aquificae and the pathway appears to be incomplete . Only in P . marina and D . themolithotrophum the functionally equivalent DNA-cytosine methyltransferase ( dcm ) is present . The presence of a gap in such a central pathway in the Aquificae and the retrieval of a samT-like pseudogene in T . ammonificans raises interesting question on the functioning of the activated methyl cycle in those organisms . An interesting working hypothesis is that enzyme thermal stability at the physiological temperature of these bacteria might have driven the loss of samT and the replacement by a not yet identified alternative enzyme in most Aquificae . The reductive acetyl-CoA pathway is present in acetogenic bacteria , methanogens , sulfate-reducers and anammox bacteria ( Berg , 2011; Fuchs , 2011; Hügler and Sievert , 2011 ) . It is believed to be the oldest carbon fixation pathway on Earth , due to the centrality of acetyl-CoA in metabolism , the low energy requirement ( ~1 ATP required vs 2–3 ATP for the rTCA cycle ) and the limited necessity of de novo protein assembly ( Berg et al . , 2010; Fuchs , 2011 ) . We hypothesize that the reductive acetyl-CoA pathway is used as an additional or alternative pathway to fix CO2 in T . ammonificans . The finding of a potentially active reductive acetyl-CoA pathway ( reactions 1 to 6; Figure 3 ) in T . ammonificans and other members of the Desulfurobacteraceae is intriguing . The pathway is missing reaction 2 , the formyl-THF synthesis for which no obvious 10-formyl-THF synthetase was found ( Figure 3 ) . However , the 10-formyl-THF synthetase of the reductive acetyl-CoA pathway may have been replaced by alternative enzymes , such as 5-formyl-THF cycloligase ( Theam_1206 ) or 10-formyl-THF deformylase ( Theam_0826 ) working in reverse ( Table 2 , Figure 3 ) . Promiscuous enzymes that function in more than one pathway have been long hypothesized and recently described in Escherichia coli ( Jensen , 1976; Kim et al . , 2010 ) . Assuming low substrate specificity , other candidate enzymes with similar substrate requirements are the phosphoribosylglycinamide formyltransferase ( Theam_1211 ) or phosphoribosylaminoimidazolecarboxamide formyltransferase/IMP cyclohydrolase ( Theam_0328 ) , which could catalyze the synthesis of formyl-THF , albeit with sub-optimal kinetics . While all four enzymes were detected in the proteomic analysis ( Figure 3 ) , we think that the possible involvement of a 5-formyl-THF cycloligase in the synthesis of formyl-THF in T . ammonificans is particularly appealing , as this would provide an evolutionary link between the N5-formate uptake of methanogens and N10-formate uptake in acetogens ( Braakman and Smith , 2012 ) . Despite the absence of the 10-formyl-THF synthetase , the genome of T . ammonificans encodes and expresses a type V CodH ( Figure 3 ) . We investigated the available genomes of other members of the Aquificae for the presence of genes homologous to the putative reductive acetyl-CoA pathway of T . ammonificans ( Table 2 ) . The similarity of the reductive acetyl-CoA pathway-related genes of T . ammonificans to those from organisms outside the Desulfurobacteriaceae is low . The type V codH gene identified in T . ammonificans is found exclusively in members of the family Desulfurobacteriaceae , with exception of a low similarity homolog present in Persephonella marina ( 41% amino acid identity ) ( Table 2 ) . Furthermore , most of the genes involved in reaction 3 , 4 and 5 ( Figure 3 ) have homologs in other Aquificae or are substituted by functional equivalents with low amino acid identity ( <40% ) . To understand the relationship of the T . ammonificans CodH with the homologous enzymes of methanogens , homoacetogens and sulfate-reducers , we reconstructed the phylogenetic history of these enzymes with a particular focus on type V CodH ( Figure 4 ) . Sequences from the Desulfurobacteriaceae appear to be related to those present in the Archaea , Ferroglobus placidus and Archaeoglobus spp . , and similar to Methanosarcina spp . and homoacetogenic bacteria . The CodH found in CO-utilizing bacteria is only distantly related to the one found in T . ammonificans ( Techtmann et al . , 2012 ) , while the CodH sequence of Persephonella marina appears to be one of a kind , with the enzyme of Ferroglobus placidus as its closest relative ( 44% similarity ) . Taken together , these findings suggest that , within the Aquificae , the type V CodH is exclusive to the Desulfurobacteriaceae . The retrieval of the CodH enzyme from the proteome of T . ammonificans ( Figure 3 , Figure 3—source data 1 ) , along with the catalytic subunit of the formate dehydrogenase ( FdhA ) , suggests that the reductive acetyl-CoA pathway could be operational in T . ammonificans in addition to the rTCA cycle . The simultaneous presence of two distinct carbon fixation pathways has been confirmed so far only in the endosymbiont of the deep-sea vent tubeworm Riftia pachyptila ( Markert et al . , 2007 ) . Recent findings suggest that the Riftia symbiont can switch from the Calvin-Benson-Bassham ( CBB ) to the rTCA cycles in response to low-energy supply ( e . g . , low concentrations of hydrogen sulfide in the vent fluids , being the rTCA cycle more energetically favorable than the CBB cycle; Markert et al . , 2007 ) . Genomic evidence of the simultaneous presence of different carbon fixation pathways came also from genome analysis of Ammonifex degensii and Ferroglobus placidus ( Berg , 2011; Hügler and Sievert , 2011 ) , where the reductive acetyl-CoA pathway seems to be coupled to an incomplete CBB cycle . The use of different carbon fixation strategies could be advantageous under energy limiting conditions ( e . g . , deep biosphere ) , could optimize overall carbon fixation or provide different precursors for biosynthetic pathways , and could be more widespread than originally thought . Further laboratory analyses will identify the exact role of the type V CodH in T . ammonificans and the potential contribution of the reductive acetyl-CoA to overall carbon fixation . The genome of T . ammonificans HB-1 displays a large degree of mosaicism ( Figure 6 ) . When comparing the coding sequences ( CDS ) of T . ammonificans with available genomes , 34% of the CDS shared higher similarity to genes outside the Aquificae , suggesting that these genes were acquired from more distantly related taxa . We carried out comparative genomic analyses among T . ammonificans and all available Aquificae genomes ( Table 3 ) . Desulfurobacterium thermolithotrophum DSM 11699 ( its closest relative whose genome was sequenced ) and Caminibacter mediatlanticus TB-2 ( an Epsilonproteobacterium which , albeit phylogenetically distant , shares the same physiology and occupies a similar ecological niche as T . ammonificans , only at lower temperature; Figure 7 ) were further used as a direct comparison . We identified areas of synteny between T . ammonificans and the closely related D . thermolithotrophum in the region surrounding the genes encoding for key enzymes responsible for the citrate cleavage in the rTCA cycle ( Figure 7b and Figure 8 ) . This region is conserved also within the Hydrogenothermaceae ( in particular P . marina ) and members of the Epsilonproteobacteria . 10 . 7554/eLife . 18990 . 016Figure 8 . Syntheny diagram presenting the genome organization around the rTCA key enzyme ATP citrate lyase . In T . ammonificans the two subunits of the ATP citrate lyase enzyme are organized in a single operon together with the isocitrate dehydrogenase and the aconitate dehydratase . The numbers inside each gene represent the locus number for the organism . Shaded color connects synthenic regions . The website address below each organism name is a permanent link to the pairwise analysis performed on the Genome Evolution server ( http://genomevolution . org/ ) against T . ammonificans . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 01610 . 7554/eLife . 18990 . 017Table 3 . List of the genomes belonging to the Aquificae phylum used for comparative genomic analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 017OrganismGenome acc . numberGenome lengthGC contentNum . genes Aquifex aeolicus VF5NC_0009181 , 590 , 79143%1782 Desulfurobacterium sp . TC5-1NZ_ATXC010000011 , 653 , 62540%1680 Desulfurobacterium thermolithotrophum DSM 11699NC_0151851 , 541 , 96835%1561 Hydrogenivirga sp . 128–5 R1-1NZ_ABHJ010005513 , 038 , 24044%3756 Hydrogenobacter thermophilus TK-6NC_0137991 , 743 , 13544%1897 Hydrogenobaculum sp . Y04AAS1NC_0111261 , 559 , 51435%1631 Persephonella marina EX-H1NC_0124401 , 983 , 96637%2067 Persephonella sp . IF05-L8NZ_JNLJ010000011 , 828 , 85835%1920 Sulfurihydrogenibium azorense Az-Fu1NC_0124381 , 640 , 87733%1722 Sulfurihydrogenibium sp . YO3AOP1NC_0107301 , 838 , 44232%1832 Sulfurihydrogenibium subterraneum DSM 15120NZ_JHUV010000011 , 610 , 18132%1701 Sulfurihydrogenibium yellowstonense SS-5NZ_ABZS010002281 , 534 , 47133%1637 Thermocrinis albus DSM 14484NC_0138941 , 500 , 57747%1631 Thermocrinis ruber DSM 23557NZ_CP0070281 , 521 , 03745%1625 Thermocrinis sp . GBSNZ_JNIE010000011 , 315 , 62541%1417 Thermosulfidibacter takaii ABI70S6NZ_AP0130351 , 816 , 67043%1844 Thermovibrio ammonificans HB-1NC_0149261 , 759 , 52652%1820 The conserved regions between the genomes of T . ammonificans and D . thermolithotrophum included the two hydrogenase clusters , the flagellum and the NADP dehydrogenase complex . The order and position of the hydrogenase and NADP dehydrogenase clusters were highly conserved also in C . mediatlanticus , while flagellar genes were scattered throughout the genome ( Figure 7 ) . These results suggest that the genes encoding hydrogen utilization and functions related to energy conversion are among the oldest genomic regions present in those organisms , and overall are conserved across phyla . When we analyzed the central metabolism of T . ammonificans using comparative genomic approaches , the presence of two distinct groups of genes became evident: the first group was related to early-branching bacterial or archaeal lineages and coded for enzymes involved in several central metabolic pathways , while the second group of genes appeared to have been acquired later . Among the first group of genes , we identified: ( I ) the cytoplasmic [Ni-Fe]-hydrogenases Group 3 that were related to the enzymes found in methanogens and thermophilic Euryarchaeota , in which they are involved in ferredoxin reduction ( Figure 3 and Figure 8 ) ; ( II ) sulfur-reduction related genes ( Figure 3 ) ; and ( III ) the genes coding for the enzymes of the reductive acetyl-CoA pathway ( Figure 3 ) . These pathways are either present in early branching Archaea , or are directly involved in metabolic reactions that do not require oxygen ( or oxygen by-products ) and use substrates of geothermal origin that are likely to have been abundant on Early Earth . Moreover , most key reactions in these pathways are catalyzed by enzymes that are extremely sensitive to oxygen . Our findings are consistent with a recent reconstruction of the ancestral genome of the last universal common ancestor ( LUCA ) , which suggest that LUCA was a hydrogen-dependent autotroph capable of S utilization that could fix CO2 via the reductive acetyl-CoA pathway ( Weiss et al . , 2016 ) . Taken together , these observations suggest that part of the central metabolism of T . ammonificans is ancestral and emerged prior to oxygenic photosynthesis . By contrast , some genes and associated metabolic pathways appear to have been acquired by the T . ammonificans lineage at a later time . One example is the ability to conserve energy by nitrate reduction . The hypothesis that nitrate respiration is an acquired trait in T . ammonificans is supported by the structure of the nap operon . The monomeric napA gene of T . ammonificans is homologous to that of Desulfovibrio desulfuricans and other Deltaproteobacteria ( Figure 6; online interactive version ) and apparently has been acquired laterally . Furthermore , it shares high similarity with assimilatory nitrate reductases ( Nas ) and it could be an evolutionary intermediate form between assimilatory Nas and the dimeric respiratory Nap . In line with this observation , phylogenetic analyses suggest that the membrane bound [Ni-Fe]-hydrogenases of Group 1 and 4 are of delta/epsilonproteobacterial origin ( Figure 5 ) . Based on the finding that Group 1 Hyn hydrogenases were expressed when T . ammonificans was grown under nitrate-reducing conditions ( Figure 3 ) , we suggest that these enzymes are linked through the membrane quinone pool to nitrate reduction and may have been acquired simultaneously with the nap genes . A second example is represented by the oxygen radical detoxification enzymes encoded by the genome of T . ammonificans . We propose that such genes are not part of the core , or ancestral genome of T . ammonificans , but that they evolved as a response to exposure to toxic oxygen radicals . In the modern ocean , catalase and peroxidase may provide protection to oxygen-sensitive enzymes during transient exposure to oxygen associated with entrainment of deep seawater in hydrothermal fluids , or when the organism is displaced into the water column . This second scenario may have important implication for the dispersal of vent microorganisms to other vent sites . We hypothesize that T . ammonificans is able to deal with oxygen exposures using two different mechanisms: the first response mechanism would be activated following exposure to oxygen at physiological temperatures ( 60–80°C ) . In that case , the oxygen-stress related genes are induced , protecting the cell machinery against damages . The second mechanism would be a response to prolonged exposure to oxygen at temperatures below the physiological threshold . Such prolonged exposure would trigger a metabolic shut-down , enabling survival of the organism in cold seawater . The latter hypothesis is supported by experiments where batch cultures of T . ammonificans were exposed for a prolonged time at 4°C , which revealed the survival of the bacterium despite being exposed to oxygen ( data not shown ) . In conclusion , the nitrate reduction pathway , as well as oxygen stress-related genes , may represent adaptations acquired by the T . ammonificans lineage to cope with the rise of oxygen on Earth . Six different pathways of carbon fixation are known to date ( Fuchs , 2011 ) . In the modern biosphere , the Calvin-Benson-Bassham cycle is the dominant mechanism of CO2 fixation . Yet , other anaerobic pathways have been investigated and are proposed to represent the ancestral autotrophic carbon fixation pathway . In particular , the reductive acetyl-CoA pathway is thought to be among the earliest carbon fixation pathways that have emerged on Earth ( Fuchs , 2011 ) . This hypothesis is supported by numerous observations: ( I ) the presence of this pathway in the early branching methanogenic archaea; ( II ) its low energy requirements and low need of de novo protein synthesis; ( III ) its capacity to incorporate different one-carbon compounds and carbon monoxide of geothermal origin , and ( IV ) the extreme oxygen sensitivity of its key enzymes which have common roots in Bacteria and Archaea ( reviewed in Berg et al . , 2010; Fuchs , 2011; Hügler and Sievert , 2011 ) . Despite differences in their structure , all extant carbon fixation pathways can be theoretically derived from a putative rTCA cycle/reductive acetyl-CoA hybrid pathway , as proposed by a recent phylometabolic reconstruction of the evolution of carbon fixation ( Braakman and Smith , 2012 ) . According to this study , both Aquificae and acetogenic bacteria represent the closest living example of the archetypal network , and both diverged from the pre-LUCA pathway under the selective pressure of energy-efficiency ( acetogens ) and oxygen sensitivity ( Aquificae ) . Further , a recent reconstruction of the genome of LUCA based on gene phyletic pattern reconstruction is consistent with some of these findings and suggests that LUCA’s genomic makeup point to autotrophic acetogenic and methanogenic roots and to the ancestry of the reductive acetyl-CoA pathway ( Weiss et al . , 2016 ) . The presence of a CodH type V enzyme in the genome of T . ammonificans and other members of the Desulfurobacteriaceae , and the presence of a type II CodH in Desulfurobacterium thermolithotrophum suggest that the reductive acetyl-CoA pathway could be operational in extant members of the Desulfurobacteriaceae . Furthermore , this finding supports the scenario proposed by Braakman and Smith that a complete and operational reductive acetyl-CoA pathway was present in the ancestor of the phylum Aquificae ( Braakman and Smith , 2012 ) . Comparative analyses revealed that the CodH enzyme is conserved only within the Desulfurobacteriaceae , consistent with the strict anaerobic nature of the members of this family and the extreme oxygen sensitivity of CodH . The rise of oxygen has been interpreted as one of the factors responsible for the diversification of carbon fixation from the ancestral pathway , and could explain the subsequent loss of the CodH in the generally facultative microaerobic Hydrogenothermaceae and Aquificaceae within the Aquificae ( Table 2 ) . Comparative genomic analyses of rTCA cycle genes allow the description of a possible evolutionary scenario for the rTCA cycle in Aquificae . Members of the Aquificaceae ( e . g . , Aquifex aeolicus; Figure 9 and Figure 3—figure supplement 1 ) have the two-step version of the rTCA cycle ( Figure 3—figure supplement 1A ) , where citrate cleavage is accomplished by the combined action of the enzymes citryl-CoA synthetase and citryl-CoA lyase ( encoded by the ccl gene ) , and the carboxylation of 2-oxoglutarate is catalyzed by the two enzymes 2-oxoglutarate carboxylase and oxalosuccinate reductase ( Aoshima et al . , 2004; Braakman and Smith , 2012 ) . Since citryl-CoA synthetase and 2-oxoglutarate carboxylase likely evolved by duplication of the genes for the succinyl-CoA synthetase and pyruvate carboxylase ( Aoshima , 2007; Braakman and Smith , 2012 ) , a complete rTCA cycle could have evolved in the Aquificaceae from an ancestral , incomplete version of the cycle . In contrast , the two other groups of Aquificae , the Hydrogenothermaceae and the Desulfurobacteriaceae ( Figure 1—figure supplement 1 ) , use the one-step – and more recent – version of the rTCA cycle ( Figure 10 , Figure 9 and Figure 3—figure supplement 1B ) , involving ATP citrate lyase ( ACL , encoded by the acl gene ) and isocitrate dehydrogenase , that carry out citrate cleavage and 2-oxoglutarate carboxylation in two single enzyme reactions , respectively ( Braakman and Smith , 2012 ) . 10 . 7554/eLife . 18990 . 018Figure 9 . Phylogenetic tree of ATP citrate lyase amino acid sequences . The tree was constructed with Bayesian inference and maximum likelihood methods and presents the phylogenetic relationship among the concatenated subunit of the ATP citrate lyase in different organisms known to use the rTCA cycle . The numbers near the nodes represent the bayesian posterior probability ( left number ) and the maximum-likelihood confidence values based on 1000 bootstrap replications ( right number ) . Bar , 30% estimated substitutions . Accession numbers are reported in Figure 9—source data 1 . Squares – symmetric rTCA cycle variant characterized by the presence of two-step enzyme reactions for the cleavage of citrate and the carboxylation of 2-oxoglutarate . Circles – asymmetric rTCA cycle variant characterized by one-step reaction enzymes catalyzing the above reactions . Triangles – presence of a citryl-CoA lyase in the genome . See Supplementary Materials for details . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 01810 . 7554/eLife . 18990 . 019Figure 9—source data 1 . Accession number of the ATP citrate lyase sequences used for the reconstruction of the phylogenetic history of the enzyme presented in Figure 2 in the main text . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 01910 . 7554/eLife . 18990 . 020Figure 10 . Proposed evolution of the carbon fixation pathway in the Aquificae phylum . Proposed evolution of the carbon fixation pathway in the Aquificae phylum and reconstruction of the ancestral carbon fixation for the last common ancestor ( LCA ) based on the results of integrated phylogenetic , comparative genomic and phylometabolic analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 18990 . 020 The enzyme responsible for citrate cleavage , the ATP citrate lyase , likely evolved through gene fusion of the genes of CCS and CCL ( Aoshima et al . , 2004 ) . Phylogenetic analyses of ACL suggests that this gene fusion event did not happen within the Aquificae , as Nitrospira and Chlorobia have an evolutionary older version of ACL than Hydrogenothermaceae and Desulfurobacteriaceae ( Figure 9; Hügler et al . , 2007 ) . Hence , it is likely that these two groups acquired ACL through HGT ( Figure 9 and Figure 8; Hügler et al . , 2007 ) . Furthermore , our comparative analysis showed that: ( I ) in the Hydrogenothermaceae and in the Desulfurobacteriaceae , the enzymes of the first half of the rTCA cycle , as well as aconitase ( acnA ) , share a common ancestor with the Aquificaceae and can be considered part of the core genome of the phylum Aquificae , while the remaining enzymes are either the result of gene duplication or have been acquired by horizontal gene transfer ( Hügler et al . , 2007 ) ; ( II ) the two-step citrate cleavage is exclusive to the Aquificacea ( Figure 9 ) ; ( III ) the ccl gene is still present in the Hydrogenothermacea ( in addition to acl ) , and one sequenced strain of the Hydrogenothermaceae ( S . azorense ) still uses the two-step carboxylation of 2-oxoglutarate , suggesting that the two-step version of the rTCA cycle was present in the ancestor of both the Aquificaceae and the Hydrogenothermaceae ( Figure 9 and Figure 10 ) . However , neither of the genes for the two-step citrate cleavage or two-step 2-oxoglutrate carboxylation is present in the Desulfurobacteriaceae ( Figure 9 ) . Finally , synteny analyses show that the four enzymes necessary to complement the one-step rTCA cycle variant in the Desulfurobacteriaceae from a theoretical ancestral linear rTCA are organized in a single operon ( Figure 8 ) . Taking into consideration evidence from comparative genomics and phylogenetic analyses , the most parsimonious interpretation of the data suggests that the last common ancestor of the Aquificae had an incomplete form of the rTCA that did not proceed past the synthesis of 2-oxoglutarate , and that later on the cycle was closed following two independent evolutionary trajectories ( Figure 10 ) : ( I ) Gene duplication in the lineage that lead to the Aquificaceae and Hydrogenothermaceae; and ( II ) gene acquisition by horizontal gene transfer in the Desulfurobacteriaceae and Hydrogenothermaceae ( the latter replaced the two-step version of the rTCA cycle with the one-step version ) ( Figure 9 and Figure 10 ) . The reasons behind the presence of two distinct rTCA cycle variants within the extant Aquificae are not known . Braackman and Smith hypothesized that the one-step reactions might have evolved as a way to improve the thermodynamic efficiency of the rTCA cycle ( Braakman and Smith , 2012 ) . We hypothesize that temperature may also have played a role in preserving the ancient , more symmetric , two-step citrate cleavage rTCA cycle variant in Aquificaceae ( Figure 10 ) . Members of this group have optimum growth temperatures ( 75–95°C ) higher than those of the two other groups ( 60–75°C; Table 1 ) , and the ‘ancient’ enzymes might be more stable at these high temperatures . In contrast , the ‘newer’ enzymes that catalyze the one-step citrate cleavage might have evolved to function optimally at lower temperatures . Different members of the Aquificae use either one of the two versions of the rTCA cycle , and all the extant members of this phylum possess the genes encoding for the enzymes of the reductive acetyl-CoA pathway , with the exception of codH ( encoding for the CO-dehydrogenase ) , which is only found in the Desulfurobacteriaceae . Therefore , we hypothesize that the last common ancestor of the Aquificae possessed the complete reductive acetyl-CoA pathway ( Figure 10 ) . While members of the Desulfurobacteriaceae kept the CodH due to their obligate anaerobic lifestyle , microaerophilic Aquificae ( Hydrogenothermaceae and Aquificaceae ) lost this extremely oxygen-sensitive enzyme ( Figures 4 and 10 ) . The presence of the gene encoding CodH in P . marina ( Hydrogenothermaceae ) is the only exception , and suggests that codH was lost independently in the two lineages ( Figure 10 ) . Alltogether , our results suggest that the last common ancestor of the Aquificae combined the reductive acetyl-CoA pathway with an incomplete form of the rTCA that did not proceed past the synthesis of 2-oxoglutarate ( Figure 10 ) . A similar incomplete version of the rTCA pathway , consisting only of the reactions from acetyl-CoA to 2-oxoglutarate , is present in extant methanogens ( Berg , 2011 ) . Thus , our phylometabolic reconstruction of the ancestral state of carbon fixation in the Aquificae ( Figure 10 ) is conceptually consistent with chemoautotrophic processes of extant Bacteria and Archaea , and may represent the earliest carbon fixation pathway . We propose that the ancestor of Thermovibrio ammonificans was originally a hydrogen oxidizing , sulfur reducing bacterium that used a hybrid carbon fixation pathway for CO2 fixation . The simultaneous presence of the rTCA cycle and of the reductive acetyl-CoA pathway of carbon fixation in T . ammonificans may represent a modern analog of the early carbon fixation phenotype , and suggests that the redundancy of central metabolic pathways was common in ancestral microorganisms . With the gradual rise of oxygen in the atmosphere , more efficient terminal electron acceptors became available and this lineage acquired genes that increased its metabolic flexibility - e . g . , the capacity to respire nitrate - along with enzymes involved in the detoxification of oxygen radicals . However , this lineage also retained its core , or ancestral , metabolic traits . Given the early branching nature of the phylum Aquificae and the ability of T . ammonificans and the Desulfurobacteriaceae to thrive in hydrothermal environments relying on energy sources of geothermal origins ( namely carbon dioxide , hydrogen and elemental sulfur ) , we argue that these microorganisms represent excellent models to investigate how metabolism co-evolved with Earth’s changing environmental conditions . Manual curation of the genome was performed using blastn and blastp ( McGinnis and Madden , 2004 ) against the non-redundant database ( Pruitt et al . , 2007 ) , the conserved domain database ( Marchler-Bauer et al . , 2005 ) , the Kyoto Encyclopedia of Genes and Genomes ( Kanehisa and Goto , 2000 ) and the PFAM database ( Sonnhammer et al . , 1998 ) . Coding sequence similarities were compared using translated protein sequence . Metabolic pathways were reconstructed on the basis of available genomic , physiologic and biochemical information and using Kyoto Encyclopedia of Genes and Genomes ( Kanehisa and Goto , 2000 ) and SEED ( Overbeek et al . , 2014 ) pathways as template . Genome maps were drawn using Circos ( Krzywinski et al . , 2009 ) , parsing blast results with ad hoc bash scripting . Comparative analyses between the genomes of T . ammonificans , those of representative members of the Aquificae ( reported in Table 3 ) , Desulfurobacterium thermolithotrophum ( Göker et al . , 2011 ) and Caminibacter mediatlanticus ( Giovannelli et al . , 2011 ) were performed using the GenomeEvolution pipeline CoGe ( Lyons et al . , 2008 ) . Aquificae genomes were selected among all available genome within this phylum to maximize diversity while minimizing redundancy . All the available genomes of validly published Aquificae species were selected for analysis . To this set we added the reference genomes for the genera Hydrogenivirga and Hydrogenobaculum , as no genome sequence is available for validly published species of these genera . We also selected two additional genomes belonging to the genera Desulfurobacterium ( the closest relative to the genus Thermovibrio ) and Persephonella , respectively . Excluded genomes include either alternative assemblies of selected genomes or closely related genomes with a gapped genome similarity above 90% . LastZ pairwise alignments of selected genomes were used to draw three-way plots using the Hive Plot software ( Krzywinski et al . , 2012 ) . Gene context and operonic structures were reconstructed using BioCyc ( Karp et al . , 2005 ) and FgenesB . Operons were manually screened and their structure selected based on gene context and available literature on the specific gene . When it was not possible to discriminate between the two alternative operonic predictions , both were reported in the text . Similarities between T . ammonificans genes and other prokaryotic genomes were found performing blastp analyses against the nr database . The top three blast results were retained and further analyzed , ranking the genes for their best hits . The procedure was repeated removing from the database the genome of D . thermolithotrophum and Desulfurobacterium sp . TC5-1 , thus searching for the best hit outside of the Desulfurobacteriaceae family . The results were analyzed and interactive Krona plots ( Ondov et al . , 2011 ) drawn linking the T . ammonificans genes with its closest match in the database . The interactive plots are accessible at DOI: 10 . 6084/m9 . figshare . 3178528 . Images were drawn or edited using the open source vector drawing program Inkscape ( http://inkscape . org/ ) . Phylogenetic analyses were performed using the approaches defined below for each tree . The phylogenetic tree presented in Figure 1 was computed from the current 16S rRNA database alignment available from the ARB-SILVA project ( http://www . arb-silva . de ) . The maximum likelihood three was computed from the ARB-SILVA alignment using PHYLML ( Guindon and Gascuel , 2003 ) and the GTR model . The phylogenetic tree in Figure 1—figure supplement 1 was constructed by aligning complete or near complete 16S rRNA sequences obtained from NCBI and representing the phylum Aquificae . The 16S rRNA sequences were aligned with ClustalO ( Thompson et al . , 1997 ) and Gblocks ( Castresana , 2000 ) and the alignment was manually refined using SEAVIEW ( Galtier et al . , 1996 ) . The maximum likelihood phylogeny was inferred from the alignment of 1455 sites using PHYML ( Guindon and Gascuel , 2003 ) , the GRT model and 1000 bootstrap replications . The CODH tree presented in Figure 4 was computed using a selected set of amino acid sequences and the neighbor-joining method . Alignments were obtained using Muscle ( Edgar , 2004 ) and Gblocks ( Castresana , 2000 ) , manually refined using SEAVIEW , and phylogenetic distances calculated using 255 sites and the Observed Divergence matrix . The neighbor-joining method was used to evaluate tree topologies using Phylo_win ( Galtier et al . , 1996 ) and their robustness was tested by bootstrap analysis with 1000 resamplings . The tree for the catalytic subunit of the [NiFe]-hydrogenases presented in Figure 5 was computed using a selected set of amino acid sequences and the same approach described above for the CODH tree and was based on 539 sites . The accession number for the sequences used in trees presented in Figure 1—figure supplement 1 , Figures 4 and 5 are reported within each tree . The ATP citrate lyase phylogenetic tree was reconstructed with Bayesian inference and maximum likelihood methods . Both subunit of the ATP citrate lyase ( AclB and AclA ) were aligned individually to retrieved homologs using Muscle ( Edgar , 2004 ) and SEAVIEW ( Galtier et al . , 1996 ) . The alignments were concatenated and refined using Gblock ( Castresana , 2000 ) . A hypothetical ancestral ATP citrate lyase enzyme was manually constructed by concatenating the citryl-CoA synthetase and citryl-CoA lyase of H . thermophilus and A . aeolicus , respectively , and used as the outgroup ( Hügler et al . , 2007 ) . The maximum likelihood phylogeny was inferred from the alignment using PHYML with the Wag substitution model ( Whelan and Goldman , 2001 ) and 1000 bootstrap resamplings . The substitution model was selected based on AIC values using ProTest3 ( Darriba et al . , 2011 ) . Bayesian phylogeny was inferred using MrBayes ( Ronquist and Huelsenbeck , 2003 ) performing 500 , 000 generations with two parallel searches with the Wag amino acid matrix model ( selected using forward selection among all possible substitution models ) and a burn-in of 125 , 000 generations . Both tree were computed on 1035 identified sites . Accession numbers for the tree presented are reported in Figure 9—source data 1 . A combination of phylogenetic analysis and comparative genomic analyses were used to identify lateral gene transfer events . Phylometabolic analysis ( Braakman and Smith , 2012 ) was used to investigate the carbon fixation pathway in T . ammonificans and its evolutionary relationship to the carbon fixation pathways present in other members of the Aquificae phylum . In phylometabolic analyses , the metabolic pathways of the organism under investigation are compared to those found in related organisms both within and across neighboring clades . By focusing on the pathways , the comparison may reveal variations in multi-enzyme functional units , providing context for the completion of the pathway within the networks of individual organism , while also allowing for the identification of ancestral states and horizontal gene transfer events . The resulting phylometabolic tree includes multiple complete pathways to common essential metabolites , and suggests which evolutionary substitutions are allowed ( at either organism or ecosystem levels ) among these pathways ( see [Braakman and Smith , 2012] and reference therein for a more extensive description of the principles underlying this approach ) . We reconstructed the carbon fixation pathways in representative genomes of the Aquificae , and compared them . Information regarding the carbon fixation metabolic network was implemented using phylogenetic and comparative genomic information to help reconstruct possible ancestral states of the carbon fixation network based on maximum parsimony principles . T . ammonificans cell pellets were washed in TE buffer ( 10 mM Tris-HCl pH 7 . 5 , 10 mM EDTA pH 8 . 0 , containing Roche cOmplete protease inhibitor ) and soluble proteins were extracted as described by ( Heinz et al . , 2012 ) . Briefly , cells were disrupted by sonication ( 2 × 25 s ) , cell debris was pelleted and protein concentrations in the supernatant were determined according to ( Bradford , 1976 ) . 20 µg of protein extract were loaded onto a precast 10% polyacrylamide mini gel in technical triplicates for 1D PAGE ( 150 V , 45 min ) . After staining with Coomassie Brilliant Blue , protein-containing gel lanes were excised and cut into 10 equal subsamples each , which were destained ( 200 mM NH4HCO3 , 30% acetonitrile ) and digested with trypsin solution ( 1 µg/ml , Promega , Madison WI , USA ) at 37°C over night , before peptides were eluted from the gel pieces in an ultrasonic bath ( 15 min ) . As described by ( Xing et al . , 2015 ) , peptide mixes were subjected to reversed phase C18 column chromatography on a nano-ACQUITY-UPLC ( Waters Corporation , Milford , MA , USA ) . Mass spectrometry ( MS ) and MS/MS data were recorded with an online-coupled LTQ-Orbitrap mass spectrometer ( Thermo Fisher Scientific Inc . , Waltham , MA , USA ) . MS data were searched against the forward-decoy T . ammonificans protein database using Sequest ( Thermo Fisher Scientific , San Jose , CA , USA; version 27 , revision 11 ) and identifications were filtered and validated in Scaffold ( http://www . proteomesoftware . com ) , as described previously ( Heinz et al . , 2012 ) . Data for all three technical replicates were merged and exclusive unique peptide count values of all proteins were exported for calculation of normalized spectral abundance factors ( NSAF ) , which are given in Figure 3—source data 1 in % , i . e . , as relative abundance of each protein in % of all identified proteins .	Life may have arisen on our planet as far back as four billion years ago . Unlike today , the Earth’s atmosphere at the time had no oxygen and an abundance of volcanic emissions including hydrogen , carbon dioxide and sulfur gases . These dramatic differences have led scientists to wonder: how did the ancient microorganisms that inhabited our early planet make a living ? And how has microbial life co-evolved with the Earth ? One way to answer these questions is to study bacteria that live today in environments that resemble the early Earth . Deep-sea hydrothermal vents are regions of the deep ocean where active volcanic processes recreate primordial conditions . These habitats support microorganisms that are highly adapted to live off hydrogen , carbon dioxide and sulfur gases , and studying these modern-day microorganisms could give insights into the earliest life on Earth . Thermovibrio ammonificans is a bacterium that was obtained from an underwater volcanic system in the East Pacific . Giovannelli et al . have now asked if T . ammonificans might have inherited some of its genetic traits from a long-gone ancestor that also thrived off volcanic gases . The genetic makeup of this microorganism was examined for genes that would help it thrive at a deep-sea hydrothermal vent . Next , Giovannelli et al . compared these genes to related copies in other species of bacteria to reconstruct how the metabolism of T . ammonificans might have changed over time . This approach identified a group of likely ancient genesthat allow a microorganism to use chemicals like hydrogen , carbon dioxide and sulfur to fuel its growth and metabolism . These findings support the hypothesis that an ancestor of T . ammonificans could live off volcanic gases and that the core set of genes involved in those activities had been passed on , through the generations , to this modern-day microorganism . Giovannelli et al . also identified a second group of genes in T . ammonificans that indicate that this bacterium also co-evolved with Earth’s changing conditions , in particular the rise in the concentration of oxygen . The findings of Giovannelli et al . provide insight into how the metabolism of microbes has co-evolved with the Earth’s changing conditions , and will allow others to formulate new hypotheses that can be tested in laboratory experiments .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "microbiology", "and", "infectious", "disease" ]	2017	Insight into the evolution of microbial metabolism from the deep-branching bacterium, Thermovibrio ammonificans
Naive B cells co-express two BCR isotypes , IgM and IgD , with identical antigen-binding domains but distinct constant regions . IgM but not IgD is downregulated on autoreactive B cells . Because these isotypes are presumed to be redundant , it is unknown how this could impose tolerance . We introduced the Nur77-eGFP reporter of BCR signaling into mice that express each BCR isotype alone . Despite signaling strongly in vitro , IgD is less sensitive than IgM to endogenous antigen in vivo and developmental fate decisions are skewed accordingly . IgD-only Lyn−/− B cells cannot generate autoantibodies and short-lived plasma cells ( SLPCs ) in vivo , a fate thought to be driven by intense BCR signaling induced by endogenous antigens . Similarly , IgD-only B cells generate normal germinal center , but impaired IgG1+ SLPC responses to T-dependent immunization . We propose a role for IgD in maintaining the quiescence of autoreactive B cells and restricting their differentiation into autoantibody secreting cells . The pre-immune mature naïve B cell compartment must balance the need for a diverse antibody repertoire with the risk of autoantibody-mediated disease . Early in development , B cells randomly rearrange their immunoglobulin genes through VDJ recombination and encounter a series of tolerance checkpoints that serve to remove autoreactive B cell receptors ( BCRs ) from the repertoire . Strongly autoreactive immature B cells rearrange additional light chains to ‘edit’ their autoreactivity , and they ultimately undergo deletion if this is unsuccessful ( Shlomchik , 2008 ) . Yet , despite stringent counter-selection of autoreactivity , the mature follicular ( Fo ) B cell compartment retains cells reactive towards endogenous antigens ( Wardemann et al . , 2003; Zikherman et al . , 2012 ) . How these cells are restrained from mounting autoimmune responses is not fully understood . We previously described a BAC Tg reporter mouse ( Nur77-eGFP ) in which GFP expression is under the control of the regulatory region of Nr4a1 , an immediate early gene rapidly induced by antigen receptor signaling ( Mittelstadt and DeFranco , 1993; Winoto and Littman , 2002 ) . We showed that Nur77-eGFP expression in naïve B cells is proportional to strength of antigenic stimulation and consequently to autoreactivity ( Zikherman et al . , 2012 ) . The most prominent characteristic of GFPhi reporter B cells is decreased surface IgM BCR relative to GFPlo cells . Goodnow and colleagues first suggested that IgM downregulation may mark autoreactive B cells in the normal mature repertoire shortly after they reported selective downregulation of IgM in the IgHEL ( hen egg lysozyme ) BCR Tg/soluble HEL Tg model system of B cell autoreactivity ( Goodnow et al . , 1988; Goodnow et al . , 1989 ) . Indeed , multiple studies in mice and humans have identified naturally occurring IgD+IgMlo cells as autoreactive and ‘anergic’ or functionally unresponsive ( Duty et al . , 2009; Kirchenbaum et al . , 2014; Quách et al . , 2011; Zikherman et al . , 2012 ) . These results corroborate observations with several BCR transgenic model systems ( Cambier et al . , 2007 ) . However , whether or how IgM downregulation might constrain autoreactivity remains unclear because naturally-occurring autoreactive B cells maintain high expression of the IgD BCR isotype ( Zikherman et al . , 2012 ) . IgM and IgD are splice isoforms of a common precursor heavy chain mRNA ( Moore et al . , 1981 ) . While they differ in their Fc domains , both BCR isotypes contain the same antigen-binding domain , as well as identical 3-amino acid cytoplasmic tails , and they pair with Igα/β in order to initiate the canonical BCR signaling cascade ( Blum et al . , 1993; Radaev et al . , 2010 ) . However , the expression pattern of the isotypes differs; IgM expression begins as soon as heavy and light chains recombine early in B cell development , and persists until class switch recombination occurs following B cell activation ( Chen and Cerutti , 2010 ) . IgD is uniquely co-expressed with IgM during a narrow developmental window on late transitional and mature naïve Fo B cells as a result of alternate splicing regulated by the zinc-finger protein ZFP318 ( Enders et al . , 2014; Pioli et al . , 2014 ) . This suggests that IgD may play a critical role specifically in mature naïve B cells . Initial characterization of IgM- and IgD-deficient mice revealed only mild phenotypes and substantial redundancy; each isotype could mediate B cell development , initiate antibody responses to T-dependent and -independent immunization , and induce normal levels of steady-state serum IgG ( Lutz et al . , 1998; Nitschke et al . , 1993; Roes and Rajewsky , 1993 ) . This is consistent with prior studies in BCR transgenic model systems demonstrating that each isotype alone can mediate B cell development , deletion , and activation ( Brink et al . , 1992 ) . IgM and IgD differ structurally; IgD has a much longer , flexible hinge region linking its Fab and Fc regions than IgM does ( Chen and Cerutti , 2010 ) . Recently , Ubelhart , Jumaa , and colleagues demonstrated that a short and inflexible hinge confers upon IgM the unique capacity to signal in response to monovalent antigens , while both isotypes can respond to multivalent antigens ( Übelhart et al . , 2015 ) . However , the Goodnow lab subsequently showed that IgHEL Tg splenic B cells expressing either the IgD or IgM BCR exclusively could mobilize calcium in response to soluble HEL antigen in vitro , and each isotype can mediate a common gene expression program characteristic of anergy in vivo ( Sabouri et al . , 2016 ) . Therefore , it remains unclear whether IgM and IgD BCRs expressed in an unrestricted B cell repertoire differentially sense bona fide endogenous antigens in vivo , particularly as the identity and nature of such antigens is largely unknown and not restricted to soluble monovalent antigens . For example , 5–10% of circulating naïve B cells in healthy humans express the well-characterized unmutated and autoreactive human heavy chain V-segment IGHV4-34 , and these cells exhibit anergy , selective IgM downregulation , and recognize cell-surface antigens on erythrocytes and B cells in vivo ( Quách et al . , 2011; Reed et al . , 2016 ) . We and others previously hypothesized that downregulation of IgM on Fo B cells might serve to restrain their response to endogenous antigen . Conversely , high expression of IgD on these cells might play a ‘tolerogenic’ role ( Zikherman et al . , 2012 ) . To determine how IgM and IgD differentially regulate B cell responses to endogenous antigens , we generated Nur77-eGFP reporter mice deficient for either IgM or IgD ( Lutz et al . , 1998; Nitschke et al . , 1993 ) . Using this reporter of antigen-dependent signaling , we show that IgD is less sensitive than IgM to bona fide endogenous antigens in vivo despite higher surface expression and robust responsiveness to receptor ligation in vitro . Indeed , marginal zone ( MZ ) and B1a cells that express IgD but lack IgM induce less Nur77-eGFP than WT , and this is not attributable to repertoire differences . To further support these observations , we examine a series of cell fate decisions for which in vivo BCR signaling requirements have been previously defined . We show that IgD drives a pattern of development consistent with reduced endogenous antigen recognition , favoring MZ B cell fate and disfavoring B1a cell fate . Similarly , IgD alone is less efficient than IgM at driving SLPC expansion in response to endogenous antigens , a process that requires robust BCR signaling . However , IgD is sufficient for germinal center ( GC ) B cell differentiation . Our data suggest that reduced endogenous antigen sensing ( i . e . signal transduction in response to antigen binding ) by the IgD BCR isotype shunts autoreactive IgDhi IgMlo Fo B cells away from differentiation into SLPCs . We propose that predominant IgD expression maintains the quiescence of autoreactive B cells in response to chronic endogenous antigen stimulation , and limits autoantibody secretion in the context of rapid immune responses . We previously showed that antigen recognition was both necessary and sufficient for Nur77-eGFP reporter expression by B cells in vivo ( Figure 1A ) ( Zikherman et al . , 2012 ) . Thus , reporter expression reflects endogenous antigen recognition in vivo . Conversely , under steady-state conditions in vivo , the Nur77-eGFP reporter does not reflect signaling through other receptors expressed in B cells; indeed , loss of either CD40 , or TLR3 , 7 , and 9 signaling has no effect on reporter expression in B cells in vivo ( Figure 1—figure supplement 1A ) . Moreover , neither BAFF , nor IL-4 , nor CXCR4-dependent signaling can regulate reporter expression in vitro ( Figure 1—figure supplement 1B–C ) ( Zikherman et al . , 2012 ) . To further confirm that Nur77 expression in naïve B cells under steady state conditions is not regulated by microbial stimulation of pattern recognition receptors ( e . g . TLRs 1 , 2 , 4 , 6 , and the TLR-4-like molecule RP105 ) , we studied mice raised under either germ-free conditions or conventional specific-pathogen-free conditions . We observed no induction of endogenous Nur77 protein in splenic B cells or endogenous Nr4a1 transcript in splenocytes in the presence of commensal flora ( Figure 1—figure supplement 1D , E ) . Moreover , MyD88-deficient and MyD88-sufficient splenocytes and peritoneal B1a cells express comparable amounts of endogenous Nr4a1 transcript and protein respectively under steady state conditions ( Figure 1—figure supplement 1F , G ) . Taken together , these data demonstrate that Nur77-eGFP expression in B cells under steady-state conditions in vivo is a specific readout of antigen-dependent signaling through the BCR ( Table 1 ) . We therefore sought to take advantage of the Nur77-eGFP reporter in order to probe the responsiveness of a diverse BCR repertoire of mature B cells expressing either IgM or IgD alone to the vast range of endogenous antigens they may encounter in vivo . Surface IgM , but not IgD , expression is inversely correlated with endogenous antigen recognition and GFP expression across a diverse repertoire of mature naïve follicular ( Fo ) B cells , such that B cells reactive to endogenous antigens express high levels of IgD and low levels of IgM on their surface ( Figure 1B ) ( Zikherman et al . , 2012 ) . The variation in surface IgM expression across the B cell repertoire is profound , spanning a 100-fold range . However , because IgD is expressed at high and invariant levels in all WT naïve Fo B cells , total surface BCR , unlike IgM , has little dynamic range across the WT repertoire ( Figure 1—figure supplement 2A ) . To explore how IgM and IgD differentially regulate B cells reactive to endogenous antigens , we generated Nur77-eGFP mice deficient for either IgM or IgD ( Lutz et al . , 1998; Nitschke et al . , 1993 ) . In IgM−/− mice , all B cells express IgD alone prior to class switch recombination ( CSR ) , and in IgD−/− mice , all B cells express IgM alone prior to CSR . We observed that IgM is still downregulated on IgD−/− GFPhi Fo B cells , but the dynamic range of IgM expression is highly restricted relative to the broad range observed on WT cells ( Figure 1B ) . We previously showed that IgM down-modulation on GFPhi reporter B cells largely accounts for impaired signaling through IgM ( Figure 1C ) ( Zikherman et al . , 2012 ) . However , the narrow dynamic range of IgM on IgD-deficient cells renders reporter B cells unable to fine-tune signaling through the IgM BCR across the GFP repertoire ( Figure 1C ) . Furthermore , IgD-mediated signaling is not altered across the GFP repertoire in the presence or absence of IgM ( Figure 1C ) . As a result , reduced responsiveness of GFPhi B cells is most profound in B cells that express the IgM isotype in the presence of IgD . This suggests that IgD expression may be necessary to establish and/or maintain a broad and functionally dynamic range of surface IgM expression across the B cell repertoire . This is consistent with a previously proposed survival function for the IgD BCR ( Roes and Rajewsky , 1993; Sabouri et al . , 2016 ) . Deletion of either the IgM or the IgD BCR isotype results in compensatory upregulation of the remaining isotype . This may be due in part to a change in competition for pairing with Igα/β heterodimers , which is essential for trafficking of IgM to the cell surface ( Hombach et al . , 1990; Sabouri et al . , 2016 ) . IgM−/− B cells express about twice as much surface BCR as IgD−/− B cells , as measured by surface anti-light chain staining ( Figure 1D and Figure 1—figure supplement 2B–E ) . Nevertheless , distribution of Nur77-eGFP expression across the naïve B cell repertoire is nearly identical in IgM−/− and IgD−/− Fo B cells ( Figure 1E ) . Further , at every level of Nur77-eGFP , cells that express IgD alone require more surface BCR to drive an equivalent amount of GFP compared to cells that express IgM alone ( Figure 1F ) . This suggests that on a per-receptor basis , IgD BCR may be less efficient at inducing Nur77-eGFP in response to endogenous antigens . Importantly , this is not attributable to differential dependence upon IgM or IgD expression for signaling via CXCR4 or downstream of canonical TLR ligands ( Figure 1—figure supplement 1C , Figure 1—figure supplement 3A ) . Because IgD drives less Nur77-eGFP per receptor than IgM in vivo , we wanted to determine whether there were defects in signal transduction downstream of IgD . To mimic antigen binding to the membrane-distal end of the BCR and allow for direct comparison between the isotypes , we stimulated IgD-only ( IgM−/− ) and IgM-only ( IgD−/− ) cells with anti-Igκ and compared downstream signaling . BCR ligation in IgD-only B cells induced equivalent amounts of Erk phosphorylation relative to IgM-only B cells ( Figure 2A ) , and there were no gross differences in pErk kinetics ( Figure 2B ) . Additionally , IgD-only cells induced more robust intracellular calcium increase and S6 phosphorylation than IgM-only cells ( Figure 2C and Figure 2—figure supplement 1A ) . This was not due to differential effects of the Fc portion of the stimulatory antibody on IgM-only and IgD-only B cells as anti-Igκ-F ( ab’ ) 2 and anti-Igκ stimulation produced identical signaling ( Figure 2—figure supplement 1B ) . Due in part to increased surface receptor expression , IgD-only B cells induce comparable Erk phosphorylation and enhanced calcium mobilization relative to IgM-only B cells stimulated with anti-Igκ . As a result , we observe significantly more CD86 and Nur77-eGFP induction in IgD-only B cells after 18 hr of anti-Igκ stimulation in vitro ( Figure 2D–F ) . However , IgD-only cells induced less Nur77-eGFP and MHC-II upregulation when cultured in the absence of exogenous stimulus ( Figure 2G and H ) . This discrepancy suggests that despite increased receptor expression and efficient coupling to downstream signaling machinery , residual endogenous antigens occupying the IgD BCR are less efficient at inducing signaling ex vivo . Consistent with this hypothesis , basal calcium analyzed immediately ex vivo is depressed in cells that express only IgD ( Figure 2I ) . These data suggest that while signal transduction downstream of the IgD BCR is robust in vitro , IgD responds less efficiently than IgM to the relevant antigens it encounters in vivo . While IgD-only and IgM-only Fo B cells express comparable levels of Nur77-eGFP ( Figure 1E ) , we suspected that competitive pressures for survival might constrain the acceptable range of BCR signaling among mature Fo B cells . We further speculated that compensatory mechanisms such as altered surface receptor and altered BCR repertoire might fine tune how much signaling Fo B cells experience in vivo in order to lie within this range . In contrast to Fo B cells , IgD-only marginal zone ( MZ ) and B1a cells expressed significantly less Nur77-eGFP than WT cells despite higher levels of surface BCR ( Figure 3A , Figure 1—figure supplement 2B–E ) . To determine whether this difference was due to isotype or altered BCR repertoire , we probed GFP expression in B1a cells specific for phosphatidylcholine ( PtC ) , an endogenous antigen exposed on the surface of dying cells that is thought to select developing B cells into the B1a compartment ( Baumgarth , 2011 ) . We found a large difference in GFP even among B1a cells with a common specificity , and this occurs in spite of high surface PtC binding in IgD-only B1a cells ( Figure 3B and C ) . We took an independent and complementary approach to address the same question in MZ B cells; we crossed IgM−/− and IgD−/− reporter mice to a genetic background that drives over-expression of the B cell survival factor BAFF ( Gavin et al . , 2005 ) . BAFF overexpression ‘unrestricts’ the BCR repertoire by removing competition for survival , thereby reducing both positive and negative selection pressures and permitting cells with either insufficient or excessive BCR signaling to persist in the repertoire , resulting in a massively expanded MZ compartment ( Stadanlick and Cancro , 2008 ) . Importantly , BAFF does not directly regulate reporter GFP expression , and therefore GFP reflects endogenous antigen stimulation on this genetic background ( Zikherman et al . , 2012 ) . In BAFF Tg mice , IgD-only MZ cells expressed much less GFP than those expressing IgM alone ( Figure 3D and E ) . This was not explained by differences in surface receptor levels between these populations as they are comparable on this genetic background ( Figure 3—figure supplement 1A ) . Indeed , IgD-only cells with low GFP and IgM-only cells with high GFP appear to be preferentially rescued in mice with excess BAFF , suggesting that distinct BCR repertoires , far from accounting for GFP differences between IgM-only and IgD-only MZ B cells , actually obscure these differences . Conversely , either ‘fixing’ or unrestricting the BCR repertoire unmasks impaired GFP upregulation by IgD relative to IgM . As a result of allelic exclusion of the heavy chain locus during B cell development , IgM+/− heterozygous mice develop two genetically distinct sets of B cells in which half express only IgD and the other half express both IgM and IgD . We exploited this property to confirm that the Nur77-eGFP difference in the MZ and B1a compartments is cell intrinsic and not a consequence of the presence or absence of serum IgM ( Figure 3—figure supplement 1B ) . Since IgM−/− and IgD−/− mice were generated on the Balb/c and 129 genetic backgrounds respectively , their germline VDJ loci are different from that of C57BL/6 mice to which they have been back-crossed . Throughout our study , we have validated our findings in either Balb/c-B6 F1 mice or IgHa/b heterozygous mice , which have two sets B cells with identical IgM and IgD isotype expression but distinct germline VDJ loci . We took this approach to confirm that differences in MZ GFP were due to BCR isotype and not VDJ locus ( Figure 3—figure supplement 1C ) . Generation of the B1a compartment requires endogenous antigen recognition and strong BCR signaling , while the opposite is true for MZ B cells ( Figure 4—figure supplement 1A ) ( Cariappa and Pillai , 2002; Casola et al . , 2004 ) . Indeed , B1a cells are modestly reduced in IgM-deficient mice , while MZ B cells are modestly increased ( Figure 4—figure supplement 1B–C ) . Because serum IgM deficiency leads to increased B1a numbers , we assessed B1a and MZ B cell development in a competitive setting to isolate the cell-intrinsic effects of IgM and IgD BCRs ( Boes et al . , 1998 ) . Analogous to the IgM+/− mice described above , IgM+/− IgD−/+ heterozygous mice develop two genetically distinct sets of B cells in which half express only IgD and the other half express only IgM ( Figure 4A ) . Although IgD-only B cells can partially populate the B1a compartment in the absence of competition ( Lutz et al . , 1998; Übelhart et al . , 2015 ) , IgD-only cells are virtually excluded from this compartment in a competitive setting ( Figure 4B ) . Indeed , this parallels the loss of PtC-binding IgD-only peritoneal B cells observed in competition with wild type cells ( Übelhart et al . , 2015 ) . In contrast , IgD-only B cells preferentially populate the MZ B cell compartment in competition with IgM-only B cells ( Figure 4B ) . These data are consistent with a signal strength model of B1a/MZ B cell development ( Figure 4—figure supplement 1A ) , and this suggests that in vivo signaling , antigen recognition , or both are reduced in IgD-only B cells . These differences in signaling could in turn modulate the generation and/or survival of IgD-only B1a and MZ B cells . We sought to determine whether skewed development of B cell lineages was driven primarily by properties of IgD , IgM , or both by comparing development of B cell populations in IgM+/− IgD−/+ , IgM+/− , and IgD+/− mice ( Figure 4C–4E ) . We conclude that skewing of MZ and B1a fates is primarily attributable to lack of IgM expression on IgD-only cells rather than lack of IgD expression on IgM-only cells because we observe little skewing in IgD+/− mice ( Figure 4E ) . Further , the competitive advantage of IgD-only cells in the marginal zone niche is reduced by BAFF over-expression , consistent with loosening of competitive selection pressures ( Figure 4F ) . IgM-only B cells exhibit a disadvantage in the mature Fo compartment relative to IgD-only B cells ( Figure 4C ) . We find that absence of IgD rather than excess IgM expression accounts for some of this disadvantage as WT and IgD-only cells compete equally well in this compartment ( Figure 4D ) . This may reflect impaired positive selection or survival in absence of IgD . Indeed , similar to our observations , a disadvantage for Fo B cells lacking IgD was observed in IgD+/− mice previously ( Roes and Rajewsky , 1993 ) . As with differences in Nur77-eGFP expression , we confirmed that the developmental fates of IgM-only and IgD-only cells were due to BCR isotype and not VDJ locus ( Figure 4—figure supplement 1D ) . Mice deficient for either IgM or IgD can mount T-independent and T-dependent immune responses to model antigens ( Lutz et al . , 1998; Nitschke et al . , 1993; Roes and Rajewsky , 1993 ) . However , endogenous antigens may have unique properties that are absent in model antigen systems . To test whether each BCR isotype was competent to mediate autoimmune responses to endogenous antigens in lupus-prone mice , we generated IgM+/− and IgD+/− mice on the Lyn−/− background ( Chan et al . , 1997 ) . The Src family kinase Lyn is essential to mediate ITIM-dependent inhibitory signals in B cells and myeloid cells ( Scapini et al . , 2009; Xu et al . , 2005 ) . Lyn−/− mice consequently develop a spontaneous lupus-like disease characterized by anti-DNA antibodies and nephritis on the C57BL/6 genetic background ( Chan et al . , 1997; Hibbs et al . , 1995; Nishizumi et al . , 1995 ) . It has been shown that both B-cell-specific MyD88 expression and T cells are essential for IgG2a/c anti-dsDNA autoantibody production in Lyn−/− mice , and conditional deletion of Lyn in B cells is sufficient for autoimmunity ( Hua et al . , 2014; Lamagna et al . , 2014 ) . Secreted natural IgM is thought to play important homeostatic functions that repress autoimmunity , particularly clearance of dead cell debris ( Boes et al . , 2000; Manson et al . , 2005 ) . By performing our analysis in mice with wild type and IgM-only or IgD-only B cells in a common milieu ( IgM+/− Lyn−/− and IgD+/− Lyn−/− mice ) , we could control for this and isolate the B cell-intrinsic effects of the IgM and IgD BCRs . Importantly , precursor Fo B cells are present at roughly equal ratios from WT and IgM-only or IgD-only loci in these mice , and this is unaffected by VDJ background ( Figure 5—figure supplement 1A–C ) . Moreover , Lyn−/− B cells expressing either IgM or IgD alone exhibit enhanced BCR signaling in vitro relative to Lyn+/+ B cells , suggesting that neither isotype is uniquely coupled to ITIM-containing receptors ( Figure 5—figure supplement 2A–B ) . Polyclonal B cell activation in Lyn−/− mice precedes and is genetically separable from autoantibody production; it is driven by B cell-intrinsic loss of Lyn , is thought to result from enhanced BCR signaling , and is independent of MyD88 expression ( Hua et al . , 2014; Lamagna et al . , 2014 ) . Indeed , Lyn−/− reporter B cells have elevated Nur77-eGFP expression consistent with enhanced BCR signal transduction in vivo ( Figure 5—figure supplement 3A ) . We had expected that activation of IgD-only Lyn−/− B cells might be impaired due to reduced sensing of endogenous antigens , but neither CD86 nor CD69 upregulation were significantly different on WT and IgD-only B cells in the absence of Lyn ( Figure 5A ) . This implies that both BCRs can provide sufficient signals to drive polyclonal activation by endogenous antigens . Although GC fate is disfavored relative to SLPC fate in Lyn−/− mice , GCs do arise spontaneously with time ( Hibbs et al . , 1995; Hua et al . , 2014 ) . We found that both IgD-only and IgM-only cells in Lyn-deficient mice made comparable contributions to the GC compartment in competition with wild type B cells , independent of VDJ background ( Figure 5B and Figure 5—figure supplement 3B–C ) . We propose that endogenous antigen sensing by IgD , while normally dampened , is sufficient to drive both polyclonal activation and GC differentiation on the Lyn−/− background . This further implies that antigen capture by IgD and presentation on MHC-II is robust enough to recruit adequate T cell help to support these cell fates . IgM- and IgD-deficient B cells express BCR heavy chain allotype [a] ( IgHa ) , and their secreted antibodies can be differentiated from WT B6 antibodies ( IgHb ) even after isotype switching ( e . g . IgG2a[a] vs . IgG2a[b] – also referred to as IgG2c ) . Allotype-specific antibodies have been previously used in the context of lupus mouse models to track autoantibodies generated by B cells of distinct genetic origin ( Mills et al . , 2015; Pisitkun et al . , 2006 ) . We took an analogous approach to assess IgG2a/c anti-dsDNA autoantibodies emanating from each genetic locus in IgM+/− , IgD+/− , and IgHa/b control mice deficient for Lyn . In order to compensate for reduced penetrance of autoantibody production in IgM+/− Lyn−/− mice , we collected serum from this genotype at time points beyond 24 weeks to increase the number of samples in which tolerance had been broken . While B cells expressing IgM ( of either IgH locus ) generated IgG2a/c anti-dsDNA , IgD-only B cells were relatively protected from generating anti-dsDNA antibodies , even at late time points ( Figure 5C–G , Figure 5—figure supplement 3D–E ) . Importantly , this is attributable to BCR isotype and not VDJ locus ( Figure 5—figure supplement 3D–E ) . We noted that anti-dsDNA antibody titers fluctuated substantially over time in Lyn−/− mice , in some cases dropping markedly over a four-week period ( Figure 5C and D ) . Anti-dsDNA IgG autoantibodies in patients with systemic lupus erythematosus ( SLE ) similarly fluctuate over time – in contrast to other anti-nuclear specificities - and correlate with disease flares ( Liu et al . , 2011 ) . Furthermore , unmutated germline BCRs from naïve B cells are recruited directly into the circulating plasmablast compartment during SLE flares ( Tipton et al . , 2015 ) . Taken together , this suggests that anti-dsDNA antibodies in mice and humans are secreted at least in part by SLPCs emerging from an extra-follicular immune response , rather than long-lived plasma cells ( LLPCs ) that originate in GCs . Our data suggests that , in the absence of Lyn , the IgD BCR is sufficient to mediate B cell activation and GC entry , but the IgM BCR is essential to produce dsDNA-specific SLPCs . In addition to stochastic generation of autoantibodies over time , Lyn−/− mice exhibit a massive expansion of the splenic IgM plasma cell compartment that corresponds to a roughly 10-fold increase in serum IgM ( Hibbs et al . , 1995 ) . While natural IgM is thought to be secreted by B1a-derived plasma cells ( Baumgarth , 2011 ) , elevated serum IgM in Lyn−/− mice arises from the Fo B2 compartment since Lyn−/− mice lack MZ B cells , and bone marrow chimeras lacking the B1a compartment reconstitute the hyper-IgM secretion phenotype ( Luo et al . , 2014 ) . Moreover , while BCR signaling is prominently enhanced in Lyn−/− B2 B cells , it is markedly dampened in Lyn−/− B1a cells ( Figure 6—figure supplement 1A ) ( Skrzypczynska et al . , 2016 ) . IgM plasma cell expansion is temporally and genetically separable from autoimmunity in Lyn−/− mice; this phenotype develops in mice as young as 7–10 weeks of age with 100% penetrance , is B-cell intrinsic , and is not dependent upon MyD88 expression ( Hua et al . , 2014; Infantino et al . , 2014; Lamagna et al . , 2014 ) . A pathway involving Btk , Ets1 , and Blimp-1 is thought to drive this phenotype . Ets1 inhibits plasma cell differentiation in part by antagonizing the key transcriptional regulator of plasma cell fate , Blimp-1 ( John et al . , 2008 ) . Mice deficient for Ets1 display IgM plasma cell expansion , Ets1 levels are reduced in Lyn−/− mice , and restoration of Ets1 expression normalizes IgM plasma cell numbers ( Luo et al . , 2014 ) . Furthermore , reducing levels of Btk , a critical kinase that mediates BCR signaling , is sufficient to suppress IgM plasma cell expansion in Lyn−/− mice , and Ets1 levels are concomitantly restored to normal levels in Btklo Lyn−/− mice ( Gutierrez et al . , 2010; Luo et al . , 2014; Mayeux et al . , 2015 ) . Therefore , IgM plasma cell expansion in Lyn-/- mice is driven by exaggerated Btk-dependent BCR stimulation of B2 B cells by endogenous antigens ( Figure 6—figure supplement 2A ) . Indeed , enhanced BCR signaling has been shown to favor SLPC fate over GC differentiation in Lyn-sufficient mice as well ( Chan et al . , 2009; Nutt et al . , 2015; Paus et al . , 2006 ) . We therefore wondered whether impaired endogenous antigen sensing by IgD influenced Ets1 expression and unswitched PC expansion in Lyn−/− mice . As previously shown , Lyn−/− B cells downregulate Ets1 protein expression ( Figure 6A ) ( Luo et al . , 2014 ) . However , Lyn-deficient B cells expressing only IgD did not do so , consistent with a reduced ability of IgD to transmit antigen-dependent signals in vivo ( Figure 6A ) . As previously reported , we found that the frequency of plasma cells in the spleens of Lyn−/− mice was increased fourfold relative to Lyn+/+ mice , but the bone marrow plasma cell compartment size was relatively unaffected ( Figure 6B–C ) ( Infantino et al . , 2014 ) . The increase in splenic plasma cells was attributable entirely to unswitched ( IgM+ ) plasma cells ( Figure 6C ) . However , unswitched plasma cell expansion was completely absent in IgM−/− Lyn−/− spleens ( Figure 6C ) . This was reflected in steady-state serum antibody levels; while Lyn−/− mice have elevated serum IgM relative to WT mice , IgM−/− Lyn−/− mice did not have elevated serum IgD relative to IgM−/− mice ( Figure 6D–E ) . To confirm that resistance to unswitched plasma cell expansion was a cell-intrinsic feature of IgD-only Lyn−/− B cells , we assessed this phenotype in IgM+/− Lyn−/− mice . Similar to the non-competitive setting , IgD-only Lyn−/− cells did not contribute to an expansion in the unswitched plasma cell compartment ( Figure 6F-G ) . This effect was due to isotype and not VDJ allele usage because both IgHa and IgHb B cells in IgHa/b Lyn-/- mice generated unswitched plasma cells in a competitive setting ( Figure 5—figure supplement 3C ) . IgD-only Lyn−/− Fo B cells are completely prevented from generating an expanded unswitched SLPC compartment in response to chronic endogenous antigen stimulation , but they are competent to enter the GC . To determine how this defect relates to responses towards exogenous model antigens , we sought to determine whether IgD-only Lyn+/+ B cells exhibit skewing of cell fate in the context of a T-dependent response to a model antigen . To generate a robust polyclonal B cell response in which T cell help should not be limiting , we used the classic immunogen sheep RBCs . During this immune response , Fo B cells can either enter the germinal center or rapidly differentiate into IgG1+SLPCs ( Chan et al . , 2009; Paus et al . , 2006 ) . We found that wild type , IgM−/− , and IgD−/− mice all produced extremely robust GC responses 5 days after SRBC injection , but B cells expressing IgD alone were unable to produce IgG1+ SLPCs at this time point ( Figure 7A–D ) . Since serum IgM is absent in IgM−/− mice and is known to enhance PC generation ( Boes et al . , 1998 ) , we wanted to assess the cell-intrinsic effect of IgD and IgM BCRs on the immune response to SRBCs . To do so , we immunized IgM+/− mice in which half of Fo B cells express IgD alone and half express both IgM and IgD . We could track the relative contribution of these two cell populations to the GC response because the bulk of GC B cells have not yet isotype-switched and express surface IgM or IgD . Both cell types made a robust contribution to the germinal center response ( Figure 7E ) . Next we tracked the generation of IgG1+ PCs by detecting allotype [a] or allotype [b] IgG1 . In contrast to the GC response , IgD-only B cells were significantly disfavored in this compartment , although the defect was less severe than that observed in IgM−/− mice , which lack serum IgM ( Figure 7F ) . To control for differences in the VDJ locus , we performed this experiment using IgHa/b control mice and observed no significant differences between allotypes , implying that BCR isotypes rather than VDJ locus accounted for our observations ( Figure 7—figure supplement 1A–B ) . In contrast to impaired IgG1 SLPC responses by IgD-only B cells , we observed that IgD-only B cells were able to generate unswitched plasma cells in response to SRBC immunization in both non-competitive and competitive settings ( Figure 7—figure supplement 1C–D ) . Unswitched PC numbers correlate well with MZ B cell numbers in these mice ( Figure 4—figure supplements 1B and Figure 4D ) . Moreover , MZ B cells efficiently generate short-lived unswitched plasma cells even in response to TD-antigens ( Phan et al . , 2005; Song and Cerny , 2003 ) . We therefore suspect that IgG1 PCs emanate from the Fo B cell compartment , while the unswitched PC response may originate in the MZ B cell compartment . Normal SLPC generation by IgM-deficient MZ B cells in response to immunization would be consistent with previously reported normal T-independent II responses in IgM-deficient mice ( Lutz et al . , 1998 ) . However , we cannot exclude a contribution of follicular-derived PCs that have failed to class switch to IgG1 . We sought to confirm that the IgG1+ PC defect was generalizable to other T-dependent immunogens . The B cell response to NP hapten is stereotyped and makes predominant use of the VH186 . 2 heavy chain ( Loh et al . , 1983 ) . We therefore immunized both IgM+/− and IgHa/b mice with NP hapten conjugated to rabbit serum albumin ( NP-RSA ) in order to control for differences in VDJ locus . Although allotype [a] makes a less robust response to NP than allotype [b] , we normalized responses of IgD-only cells to wild type allotype [a] cells and found that GC responses at early time points were intact , but IgG1+ PCs and NP-specific IgG1 titers were significantly reduced 7–8 days after immunization ( Figure 7G–I ) . These data suggest that IgD-only follicular B cells are competent to enter the GC , but they have defects in adopting the SLPC fate ( or class-switching ) in response to T-dependent immunization . We and others previously hypothesized that a major tolerance strategy employed by autoreactive Fo B cells is selective downregulation of IgM . However , since all Fo B cells express a high and invariant amount of surface IgD , it was unclear whether and how loss of IgM expression could affect B cell responses to antigenic stimulation . It was recently reported that IgM , but not IgD , is uniquely responsive to monomeric antigens due to a short and inflexible linker region coupling the Fab and Fc portions of IgM ( Übelhart et al . , 2015 ) . This could account for selective quiescence of mildly autoreactive B cells to monovalent endogenous antigens . However , the identity and structural characteristics of endogenous antigens are not well understood , and include cell-surface antigens on the surface of red cells and B cells ( Quách et al . , 2011; Reed et al . , 2016 ) . Moreover , recent work from Goodnow and colleagues showed that HEL-specific IgD BCR , when expressed on primary mouse splenocytes , can indeed mobilize calcium in vitro and can mediate gene expression changes in vivo in response to monovalent cognate antigen ( Sabouri et al . , 2016 ) . How do our observations help move beyond and reconcile the seemingly contradictory published work on antigen sensing by IgM and IgD ? In vitro signaling studies of HEL-specific BCRs by the Jumaa and Goodnow labs differ in several important ways; the former work assesses calcium signaling in pre-B cells that lack Slp-65 expression , while the latter studies splenic B cells with a considerably more mature phenotype . In addition , differences in reagents or strength of stimulus may also contribute to opposing conclusions . Our data suggests that the reduced sensitivity of IgD to bona fide endogenous antigen may lie somewhere between these two extremes; we find that IgD can sense endogenous antigens , just less efficiently than IgM . Consistent with this conclusion , both HEL-specific IgM and IgD BCRs can mediate B cell anergy and drive a common gene expression program in response to chronic high affinity soluble antigen exposure in vivo ( Brink et al . , 1992; Sabouri et al . , 2016 ) . Here we show that B cells expressing IgD BCR alone inefficiently sense endogenous antigens , and exhibit impaired in vivo cell fate decisions that are dependent upon BCR signal strength . This is epitomized by dramatic skewing of IgD-only B cells away from the B1a and towards the MZ fate in a competitive setting . Importantly , signal transduction mediated by the IgD BCR in response to anti-Igκ stimulation in vitro is intact for all assayed parameters , suggesting that antigen sensing and signal initiation rather than signal transduction downstream of IgD is impaired . We propose that this impaired sensing is conferred by structural properties of IgD , either through direct binding of antigen and signal transduction through the BCR or through differential pairing with co-receptors that directly modulate the BCR signaling pathway . Characterization of bona fide endogenous antigens that drive Nur77-eGFP expression in the polyclonal repertoire is necessary in order to further dissect the biophysical basis for reduced antigen sensing by IgD in vivo . These antigens may not be exclusively restricted to those that are germline-encoded or generated by host enzymes; rather they may include non-inflammatory antigens taken up through the gastrointestinal tract and recognized by specific BCRs . At least some relevant endogenous antigens may be membrane-bound ( Quách et al . , 2011; Reed et al . , 2016 ) , and recent studies have illustrated the importance of membrane spreading , contraction , and stiffness discrimination in BCR signaling ( Fleire et al . , 2006; Shaheen et al . , 2017 ) . We propose that relevant endogenous antigens are presented in contexts that do not efficiently trigger signaling through the flexible structure of IgD . This might explain the discrepancy between weak responsiveness of IgD to endogenous antigens in vivo and strong signaling in response to crosslinking antibodies in vitro . An alternative mechanism is that differential clustering of co-receptors with IgD and IgM might influence how BCR signals are integrated in vivo . It has long been proposed that there are not only quantitative , but also qualitative differences between IgM and IgD signaling . Early studies of IgM and IgD signaling in cell lines suggested that kinetics but not quality of signaling triggered by each isotype differed ( Kim and Reth , 1995 ) . However , we have been unable to identify a difference in kinetics of BCR signaling in primary IgM- or IgD-deficient B cells in vitro . More recently , studies using TIRF , dSTORM , and PLA ( proximity ligation assay ) showed that IgD is more densely clustered on the cell surface than IgM , and these distinct IgM- and IgD-containing ‘islands’ are differentially associated with co-receptors such as CD19 in resting and activated B cells ( Kläsener et al . , 2014; Maity et al . , 2015; Mattila et al . , 2013 ) . Since function of the ITIM-containing inhibitory co-receptor CD22 depends on its ability to efficiently access the BCR , differences in cluster density of IgM and IgD might render them differentially sensitive to such inhibitory tone ( Gasparrini et al . , 2016 ) . It is possible that differential association with such co-receptors contributes to differences in the function of the IgM and IgD BCRs in vivo by modulating BCR signal strength , or by selectively perturbing specific downstream signaling events . In addition to B cell co-receptors that have long been known to directly modulate canonical BCR signaling , a growing list of immunoreceptors expressed on B cells have more recently been shown to require expression of the BCR for optimal function; Becker et al . have shown that expression of the IgD BCR is critical for CXCR4-dependent signaling ( Becker et al . , 2017 ) . This may influence the biology of B cells expressing only IgM; indeed , CXCR4-deficient B cells exhibit reduced plasma cell migration from spleen to bone marrow ( but intact splenic SLPC responses ) ( Nie et al . , 2004 ) . Importantly , defective CXCR4 signaling in IgM-only B cells could not account for reduced Nur77-eGFP expression in IgD-only B cells because reporter expression is insensitive to CXCR4 signaling ( Figure 1—figure supplement 1C ) . Nor would it account for skewed cell fate decisions by IgD-only B cells in competition with wild type B cells , both of which retain intact CXCR4 signaling . B cell responses to TLR4 ligands also require expression of the BCR , but importantly , canonical TLR ligands do not exhibit selective dependence on either the IgM or the IgD isotypes ( Figure 1—figure supplement 3A ) . Finally , recent work establishes a role for the Fc-receptor for IgM in modulating B cell responses , and could therefore play a role downstream of serum IgM that is absent in IgM−/− mice ( Nguyen et al . , 2017a; Nguyen et al . , 2017b ) . We control for this effect by confirming that all phenotypes in this study are cell-intrinsic and independent of secreted IgM . Accumulating evidence suggests that strong BCR signals favor SLPC over GC fate ( Chan et al . , 2009; Nutt et al . , 2015; Paus et al . , 2006 ) . This is thought to be transcriptionally mediated at least in part by loss of Ets1 expression and induction of Irf4 in a BCR signal-strength dependent manner ( Nutt et al . , 2015 ) . Here we show that IgM and IgD are each sufficient to drive GC responses , but IgD is less efficient at promoting SLPC fate in response to endogenous antigens , implying that IgDhi IgMlo B cells may similarly be shunted away from SLPC responses . This discrepancy is unlikely to be attributable to differences in antigen capture and presentation by the IgM and IgD BCRs , as GC entry is highly T cell-help dependent . Instead , we provide evidence to suggest that reduced antigen-dependent BCR signals are transduced in vivo by IgD . We show that , in the absence of Lyn , Ets1 downregulation and unswitched PC expansion require the IgM BCR . IgM−/−Lyn−/− mice phenocopy BtkloLyn−/− mice; both strains are protected from Ets1-downregulation , PC expansion , and anti-dsDNA autoantibody production in vivo ( Mayeux et al . , 2015; Whyburn et al . , 2003 ) . This suggests that robust Btk-dependent Ets-1 downregulation in response to endogenous antigens relies upon efficient antigen sensing by the IgM BCR and is important to drive unswitched PC expansion in the absence of Lyn . It will be important to explore whether a similar mechanism accounts for impaired IgG1+ SLPC responses by Lyn-sufficient B cells with low or absent IgM expression . We propose that the purpose ( and consequence ) of IgM downregulation on autoreactive B cells is to limit their direct differentiation into SLPCs in response to chronic endogenous antigen stimulation , and prevent secretion of auto-antibodies in the context of an acute humoral immune response ( Figure 7—figure supplement 2A ) . However , IgD is sufficient to drive germinal center differentiation without the contribution of IgM . This has been well-demonstrated both in the present study as well as prior work ( Lutz et al . , 1998 ) . Indeed , not only are autoreactive IgDhi IgMlo B cells competent to enter the germinal center , they appear to do so with greater efficiency , perhaps due in part to improved survival ( Sabouri et al . , 2016; Sabouri et al . , 2014 ) . It is worth emphasizing that selection pressures and tolerance mechanisms operating within the germinal center must independently ensure that somatic hypermutation both abolishes germ-line-encoded autoreactivity and prevents de novo acquisition of autoreactivity . Goodnow and colleagues recently showed that autoreactive BCRs can indeed be ‘redeemed’ by somatic hypermutation , providing proof-of-principle that entry of autoreactive B cells into the GC does not necessarily pose a risk to the organism ( Sabouri et al . , 2014 ) . Why , though , are autoreactive B cells preserved in the periphery and not deleted ? Why is IgD necessary at all ? One suggestion is that such BCRs are retained in the pre-immune B cell compartment to fill holes in the repertoire , and that IgD is essential for their survival; indeed , it has been shown that loss of IgD expression , with or without compensatory upregulation of IgM , leads to loss of mature naïve B cells in competition ( Roes and Rajewsky , 1993; Sabouri et al . , 2016 ) . We similarly find that IgD-only B cells have a profound competitive advantage in the follicular B cell compartment relative to IgM-only ( but not WT ) B cells , implying an obligate survival function for the IgD isotype BCR . This is consistent with a well-appreciated pro-survival function of the BCR as demonstrated by Rajewsky and colleagues ( Kraus et al . , 2004; Lam et al . , 1997 ) . Since IgD expression is needed to keep IgMlo B cells alive , this may explain why IgD expression facilitates a broad dynamic range of IgM expression across the B cell repertoire ( Figure 1B , C ) , allowing B cells to fine-tune the intensity of their SLPC responses according to their autoreactivity . IgD expression on IgMlo cells could serve to mediate antigen capture and participation in T-dependent immune responses , while simultaneously limiting SLPC responses as we show here . We demonstrate here for the first time that IgD BCRs sense endogenous antigen more weakly than IgM BCRs in vivo , are inefficient at driving SLPC responses to endogenous antigens , and thereby may function to divert autoreactive follicular B cells from direct PC differentiation . We propose that this property of the IgD BCR limits generation of auto-reactive antibodies in the context of immediate humoral immune responses . Taken together with recent work from the Jumaa and Goodnow groups , we propose a unified model of how IgM downregulation in the face of high IgD expression regulates the fate of naïve autoreactive B cells in vivo while retaining their contribution to the mature BCR repertoire . Nur77-eGFP mice and Lyn−/− mice have been previously described ( Chan et al . , 1997; Zikherman et al . , 2012 ) . BAFF Tg mice were originally generated in the Nemazee lab , and express BAFF under the control of the myeloid promoter human CD68 ( Gavin et al . , 2005 ) ( source: MMRRC UC Davis ) . IgD-/- and IgM−/− mice were previously described , and the former were generously shared by Dr . Hassan Jumaa ( Lutz et al . , 1998; Nitschke et al . , 1993 ) . TLR7−/− , CD40L−/− , Unc93b13d/3d , Nr4a1−/− , and MB1-Cre MyD88fl/fl mice were previously described ( Hobeika et al . , 2006; Hou et al . , 2008; Lee et al . , 1995; Lund et al . , 2004; Renshaw et al . , 1994; Tabeta et al . , 2006 ) . All strains were backcrossed to the C57BL/6 genetic background for at least six generations . Mice were used at 5–12 weeks of age for all functional and biochemical experiments unless otherwise noted . Germ-free ( GF ) and specific pathogen free ( SPF ) C57BL/6 mice used for direct comparison were purchased from Taconic and Jackson Laboratory and kept in microisolators under GF or SPF ( ISOcage P - Bioexclusion System , Tecniplast ) conditions . All GF mice were housed in closed caging systems and provided with standard irradiated chow diet , acidified water and housed under a 12 hr light cycle; 7-week-old males mice were used . GF and control mice were generously provided by Drs . Sergio Baranzini and Anne-Katrin Proebstel at UCSF . All other mice used in our studies were housed in a specific pathogen free facility at UCSF according to the University Animal Care Committee and National Institutes of Health ( NIH ) guidelines . Fluorescently-conjugated or biotin-conjugated antibodies to B220 , CD5 , CD19 , CD21 , CD23 , CD69 , CD86 , CD93 ( AA4 . 1 ) , CD95 ( Fas ) , CD138 , CXCR4 , IgA , IgM , IgM[a] , IgM[b] , IgD , IgD[a] , IgG1[a] , IgG1[b] , Igκ , Igλ , GL-7 , MHC Class II , and fluorescently-conjugated streptavidin were from Biolegend , eBiosciences , BD Biosciences , Tonbo , or Life Technologies . NP-PE was from Biosearch Technologies . 100 nm Rhodamine PtC liposomes were from FormuMax . Antibodies for intra-cellular staining , pErk Ab ( clone 194g2 ) and pS6 Ab ( 2F9 ) , were from Cell Signaling Technologies , and Nur77 Ab ( clone 12 . 14 ) conjugated to PE was from eBioscience . Goat anti-mouse IgM F ( ab’ ) 2 was from Jackson Immunoresearch . Goat anti-mouse Igκ and goat F ( ab’ ) 2 anti-mouse Igκ were from Southern Biotech . Anti-IgD was from MD Biosciences . CXCR4 ligand ( CXCL12/hSDF-1α ) was from Peprotech . LPS ( Cat . L8274 ) was from Sigma . CpG DNA ( ODN 1826 Biotin; Cat . tlrl-1826b ) and Pam3CSK4 ( Cat . tlrl-pms ) were from InvivoGen . Cells were stained with indicated antibodies and analyzed on a Fortessa ( Becton Dickson ) as previously described ( Hermiston et al . , 2005 ) . Data analysis was performed using FlowJo ( v9 . 7 . 6 ) software ( Treestar Incorporated , Ashland , OR ) . Statistical analysis and graphs were generated using Prism v6 ( GraphPad Software , Inc ) . Figures were prepared using Illustrator CS6 v16 . 0 . 0 . Median Igk levels ( Figure 1F ) were calculated for 200 equally sized ‘bins’ spanning the Nur77-eGFP spectrum using Canopy v1 . 4 . 1 ( Enthought ) ; source code is provided , and parameters can be modified to condense any 2D FACS plot for comparisons of multiple samples . We define biological replicates as independent analyses of cells isolated from different mice of the same genotype , and we define technical replicates as analyses of cells isolated from the same mouse and used in the same experiment . When technical replicates were used , we averaged them to calculate a value for the biological replicate . All reported values and statistics correspond to biological replicates only , and all ‘n’ values reported reflect the number of biological replicates . Wherever MFIs are directly compared , we collected all samples in a single experiment to avoid potential error that could arise from fluctuations in our flow cytometers . In some instances , MFI ratios were compiled from different experiments for statistical purposes , but qualitative findings were consistent experiment to experiment . As our panels reproducibly generated flow plots with well-defined populations , population sizes were calculated from data compiled from different experiments . When comparing two groups , we employed Welch’s t test , which is more stringent than Student’s t test and does not assume that the groups have the same standard deviation . We used paired difference t tests when studying parameters in allotype-heterozygous mice that are sensitive to cell-extrinsic factors ( e . g . dose of immunogen or severity of autoimmune disease ) . When comparing three genotypes ( e . g . receptor levels on WT vs . IgM−/−vs . IgD−/− B cells ) , we used one-way ANOVA and Tukey’s multiple comparison test to calculate p values . Ex vivo or following stimulation , cells were fixed in 2% paraformaldehyde , permeabilized with 100% methanol , and stained with anti-Nur77 , anti-pErk , or anti-pS6 followed by lineage markers and secondary antibodies if needed . Cells were fixed in 2% paraformaldehyde and permeabilized in BD Perm/Wash . Stains for antibody isotypes and allotypes were prepared in BD Perm/Wash . Cells were loaded with 5 μg/mL Indo-1 AM ( Life Technologies ) and stained with lineage markers . Cells were rested at 37 C for 2 min , and Indo-1 fluorescence was measured immediately prior to stimulation to calculate basal calcium , and for at least two minutes following addition of stimulus . Splenocytes or lymphocytes were harvested into single cell suspension , subjected to red cell lysis using ACK buffer , and plated at a concentration of 7 . 5 × 105 cells/200 μL in round bottom 96 well plates in complete RPMI media with stimuli for 18 hours prior to analysis . In vitro cultured cells were stained using fixable near IR live/dead stain ( Life technologies ) per manufacturer’s instructions . Serum antibody titers for total IgM , total IgD , anti-dsDNA IgG2a , and NP-specific IgG1 were measured by ELISA . For total IgM and total IgD , 96-well plates ( Costar ) were coated with 1 μg/mL anti-IgM F ( ab’ ) 2 ( Jackson ) or 1 μg/mL anti-IgD ( BD 553438 ) , respectively . Sera were diluted serially , and total IgM and total IgD were detected with anti-IgM-biotin ( eBioscience ) and SA-HRP ( Southern Biotech ) or anti-IgD-HRP ( American Research Products ) . dsDNA plates were generated by serially coating plates with 100 μg/mL poly-L-lysine ( SigmaAldrich ) and 0 . 2 U/mL poly dA-dT ( SigmaAldrich ) in 0 . 1 M Tris-HCL pH 7 . 6 . Sera were diluted serially on dsDNA plates , and autoantibodies were detected with IgG2[a]-biotin , IgG2a[b]-biotin , and SA-HRP . NP plates were generated by coating 96-well plates with 1 μg/mL NP-BSA ( Biosearch , conjugation ratio 23 ) in PBS . Sera from NP-RSA immunized and unimmunized mice were diluted serially , and NP-specific IgG1 was detected with IgG1[a]-biotin , IgG1[b]-biotin , and SA-HRP . All ELISA plates were developed with TMB ( Sigma ) and stopped with 2N sulfuric acid . Absorbance was measured at 450 nm . For total IgM , total IgD , and dsDNA IgG2a , OD values or OD ratios were calculated and displayed . For NP-specific IgG1 , titers were calculated at OD = 0 . 2 and normalized such that the average IgG1[a]/IgG1[b] ratio in immunized Balb/c-C57BL/6 F1 mice equals 1 . 0 . Splenic B cells were purified by negative selection with a MACS kit according to manufacturer’s protocol ( Miltenyi Biotech , cat# 130-090-862 ) . Purified B cells were lysed in 2N 4X SDS and boiled at 95 C for 5 min . 500 , 000 cells per lane were loaded onto NuPAGE 4–12% Bis-Tris gels ( Invitrogen NP0335 ) , run for 50 min at 200V , and transferred onto PVDF membranes using an XCell II Blot Module ( Invitrogen ) . Membranes were blocked with 3% BSA . Proteins were detected with rabbit monoclonal anti-Ets1 ( Epitomics , custom lot provided by Lee Ann Garrett-Sinha ) , mouse anti-mouse GAPDH ( Santa Cruz Biotechnology 32233 ) , and HRP-conjugated secondary antibodies goat anti-rabbit IgG ( H + L ) and goat anti-mouse IgG ( H + L ) ( SouthernBiotech ) . Membranes were developed with Western Lightning Plus ECL ( Perkin Elmer 0RT2651 and 0RT2751 ) and imaged using a ChemiDoc Touch Imaging System ( Bio-Rad ) . Quantifications were performed using Image Lab v . 5 . 2 . 14 ( Bio-Rad ) . Mice were immunized i . p . with 200 uL of 10% sheep red blood cells ( Rockland R406-0050 ) diluted in PBS . Mice were sacrificed 5 days after immunization , and serum and spleens were harvested . Mice were immunized i . p . with 100 ug of NP-conjugated rabbit serum album ( NP-RSA , Biosearch N-5054–10 , conjugation ratio 10 ) prepared in PBS with Alhydrogel 1% adjuvant ( Accurate Chemical and Scientific Corp . 21645-51-2 ) . NP-RSA immunized mice were sacrificed 7–8 days following immunization , and serum and spleens were harvested . To ensure that our calculated values accurately reflect the magnitude and variability of immune responses induced by immunization , we analyzed all mice that had larger plasma cell and germinal center compartments than unimmunized mice . One IgM+/− mouse immunized with SRBCs ( Figure 7E–F ) was excluded because it did not display an expanded plasma cell compartment; no other mice were excluded from analysis . Splenocytes were lysed in TRIzol ( Invitrogen ) and stored at −80 C . cDNA was prepared using a SuperScriptIII kit ( Invitrogen ) . Quantitative PCR reactions were run on a QuantStudio 12K Flex thermal cycler ( Applied Biosystems ) with FastStart Universal SYBR Green Master Mix ( Roche ) . Nr4a1 ( forward: gcctagcactgccaaattg; reverse: ggaaccagagagcaagtcat ) and GAPDH ( forward: aggtcggtgtgaacggatttg; reverse: tgtagaccatgtagttgaggtca ) primers were used at 250 nM each .	To defend an organism against invaders such as viruses and bacteria , cells of the immune system need to recognize and respond to foreign microbes . However , these immune cells must also avoid attacking ‘self’ – for example , the healthy tissues of the body – as this could lead to autoimmune disease . B cells are a type of immune cell that is essential in order to produce antibodies , protective proteins that can identify and defend against a broad range of germs: in addition , certain antibodies can also recognize ‘self’ . When a B cell first develops , it places its antibody on its surface and uses this protein as a receptor ( termed ‘B cell receptor’ ) to sense its surroundings . Prior to mounting an immune response , B cells carry two closely related versions of the B cell receptor on their surface: IgM and IgD . Both IgM and IgD perform many of the same roles and can largely substitute for one another . However , B cells that recognize ‘self’ decrease their levels of IgM but keep high amounts of IgD on their surface . It is unclear why this is the case , but one possibility is that IgM and IgD may see ‘self’ differently . To investigate this , Noviski et al . used mice with B cells that only carry either IgM or IgD , and tracked how these cells reacted to molecules from ‘self’ and foreign origins . IgM-only B cells reacted more strongly to ‘self’ molecules than IgD-only cells , which suggests that IgM is more sensitive to ‘self’ than IgD . In fact , in mice which are at risk for an autoimmune disease similar to lupus , deleting the IgM receptor prevented antibodies against ‘self’ from being produced . Therefore , reducing the amount of IgM receptors may be a way to keep B cells that are reactive against ‘self’ from inappropriately attacking the host . Meanwhile , the IgD receptor still allows B cells to mount protective antibody responses against foreign microbes . Future studies are necessary to determine whether these differences between IgM and IgD also exist in humans . If this is the case , blocking the IgM receptor could become a new kind of treatment for certain autoimmune diseases .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2018	IgM and IgD B cell receptors differentially respond to endogenous antigens and control B cell fate
Diploid transgenic organisms are either hemi- or homozygous . Genetic assays are , therefore , required to identify the genotype . Our AGameOfClones vector concept uses two clearly distinguishable transformation markers embedded in interweaved , but incompatible Lox site pairs . Cre-mediated recombination leads to hemizygous individuals that carry only one marker . In the following generation , heterozygous descendants are identified by the presence of both markers and produce homozygous progeny that are selected by the lack of one marker . We prove our concept in Tribolium castaneum by systematically creating multiple functional homozygous transgenic lines suitable for long-term fluorescence live imaging . Our approach saves resources and simplifies transgenic organism handling . Since the concept relies on the universal Cre-Lox system , it is expected to work in all diploid model organisms , for example , insects , zebrafish , rodents and plants . With appropriate adaptions , it can be used in knock-out assays to preselect homozygous individuals and thus minimize the number of wasted animals . Life sciences , especially cell and developmental biology , rely on model organisms . The most frequently used vertebrates are mouse and zebrafish . Amongst insects , the fruit fly Drosophila melanogaster and the red flour beetle Tribolium castaneum are the two prevailing species . An important standard technique is transgenesis , that is , the insertion of recombinant DNA into the genome of the model organism ( Gama Sosa et al . , 2010 ) . Since model organisms are typically diploid , the genotype has to be considered , which leads to a certain experimental complexity . Usual mating schemes result in ( i ) non-transgenic wild-type progeny , ( ii ) hemizygous transgenic progeny , that is , only the maternal or only the paternal chromosome carries the transgene , and ( iii ) homozygous transgenic progeny , that is , both the maternal and paternal chromosomes carry the transgene . In rare cases , the phenotype reveals the genotype , but usually , either two of the three or even all three outcomes cannot be distinguished . Transformation markers can be used to separate wild-type from transgenic , but not hemi- from homozygous individuals . Thus , additional experiments are necessary to determine the genotype , for example genetic assays , which are invasive and require manpower as well as consumables . In our AGameOfClones ( AGOC ) vector concept , all genotypes are directly identifiable by specifically designed distinct phenotypes , which permits the systematic creation of homozygous transgenic lines . Our approach relies on two clearly distinguishable transformation markers embedded in interweaved , but incompatible Lox site pairs . Cre-mediated recombination results in hemizygous individuals that retain only one of the two markers and are thus phenotypically distinguishable from each other and the wild type . In the next generation , descendants that express both markers are identified as heterozygous for the transgene . Finally , a cross of two heterozygotes results in homozygous progeny that are selected by the lack of one marker . The proof-of-principle of the AGOC vector concept relied on the red flour beetle Tribolium castaneum , an emerging insect model organism ( Klingler , 2004; Brown et al . , 2009 ) , in conjunction with the piggyBac transposon system ( Lorenzen et al . , 2003; Berghammer et al . , 2009 ) , which allows semi-random genomic insertion . We developed the transformation-ready pAGOC vector ( Figure 1—figure supplement 1 ) that contains mOrange-based ( Shaner et al . , 2008 ) and mCherry-based ( Shaner et al . , 2004 ) eye-specific ( Berghammer et al . , 1999 ) transformation markers ( mO and mC , respectively ) . Both fluorescent proteins are spectrally separable by appropriate excitation bands and emission filters ( Shaner et al . , 2005 ) . Each marker is flanked upstream by a LoxP site ( Hamilton and Abremski , 1984 ) and downstream by a LoxN ( Livet et al . , 2007 ) site , resulting in interweaved Lox site pairs ( Figure 1 ) . Due to variations in the spacer sequences , LoxP and LoxN sites are incompatible with each other . We injected this vector together with a piggyBac transposase-expressing helper vector ( that is , pATub’piggyBac ) into pre-blastoderm embryos to achieve germline transformation . All survivors , that is , F1 potential mosaics , were mated with wild types and in six of these crosses , at least one F2 ( mO-mC ) founder female was found among the progeny . For each cross , one founder female was mated with a wild-type male and the progeny were scored to confirm that only a single insertion had occurred ( Supplementary file 1 ) . Transgenic descendants were collected to establish six proof-of-principle cultures , which carry the same transgene , but in different genomic locations . These F3 ( mO-mC ) pre-recombination hemizygous sublines were called AGOC #1 to #6 . To roughly estimate homozygous viability , two F3 ( mO-mC ) pre-recombination hemizygous siblings were mated and the progeny were scored ( Supplementary file 2 ) . Additionally , the insertion locations of the transgenes were determined in four of the six AGOC sublines ( Supplementary file 3 ) . Up to this step , our scheme did not differ from most standard procedures to establish transgenic lines . The mating procedure for the systematic creation of homozygous transgenic lines ( Figure 2 , an comprehensive scheme is provided in Figure 2—figure supplement 1 ) spanned four generations and involved a transgenic helper line , ICE{HSP68’NLS-Cre} #1 . This line expresses a nuclear-localized Cre recombinase ( Peitz et al . , 2002 ) under control of the heat shock protein 68b promoter ( Schinko et al . , 2012 ) and carries a mCerulean-based ( Markwardt et al . , 2011 ) eye-specific transformation marker ( mCe ) . The procedure was performed with all six AGOC sublines and phenotypically documented for #5 and #6 ( Figure 3 ) . Throughout all generations , the subline-specific scores matched the expectations . No significant differences between the respective arithmetic means and the theoretical Mendelian ratios were found . Importantly , all expected phenotypes , and thus all expected genotypes , were found in all generations and consequently , F7 ( mO/mO ) as well as ( mC/mC ) homozygotes were obtained for all six AGOC sublines . A description of the generations and their characteristics is found in Supplementary file 4 . Two controls were performed with the AGOC #5 and #6 sublines to confirm proper function of the AGOC vector concept: The F3 to F7 crossing procedure was successfully conducted with ( i ) swapped genders ( Figure 3—figure supplement 2A , Supplementary file 5 ) and ( ii ) an alternative helper subline ( Figure 3—figure supplement 2B , Supplementary file 5 ) , ICE{HSP68’NLS-Cre} #2 , which carries the same Cre recombinase-expressing cassette as the #1 subline , but at a different genomic location . Based on pAGOC , we developed the intermediate pAGOC{#P’#O ( LA ) -mEmerald} vector ( Figure 1—figure supplement 2 ) , which contains an open-reading frame for mEmerald-labeled ( Shaner et al . , 2005 ) Lifeact , a small and universal peptide tag derived from Saccharomyces cerevisiae that binds to filamentous actin ( Riedl et al . , 2008 ) . With this vector , three transformation-ready derivates were created that allow expression of mEmerald-labeled Lifeact under control of either the tubulin alpha 1-like protein ( Siebert et al . , 2008 ) , the zerknüllt 1 ( van der Zee et al . , 2005; ; Sharma et al . , 2013; Panfilio et al . , 2013; Hilbrant et al . , 2016 ) or the actin-related protein 5 promoter . Additionally , we developed two more derivates that contain expression cassettes for either mEmerald-labeled beta-galactoside alpha-2 , 6-sialyltransferase 1 or histone H2B under control of the tubulin alpha 1-like protein promoter . With these vectors , we created five functional lines with one , three , two , three and four sublines , respectively ( that is , 13 in total ) , which are primarily designed for fluorescence live imaging . Twelve of these sublines went through the procedure successfully , only the AGOC{ATub’H2B-mEmerald} #4 subline turned out to be heterozygous/homozygous sterile ( Supplementary file 6 for the mEmerald-labeled Lifeact-expressing sublines and Supplementary file 7 for the mEmerald-labeled beta-galactoside alpha-2 , 6-sialyltransferase 1- and histone H2B-expressing sublines ) . We performed long-term fluorescence live imaging of the embryonic development ( Strobl and Stelzer , 2016 ) with three of the functional ( mC/mC ) homozygous sublines . We used a digital scanned laser light-sheet-based fluorescence microscope ( Keller et al . , 2008; Keller and Stelzer , 2010 ) in conjunction with previously published sample preparation protocols for Tribolium ( Strobl and Stelzer , 2014; Strobl et al . , 2015; Strobl et al . , 2017a ) . The AGOC{Zen1’#O ( LA ) -mEmerald} #2 subline allows the characterization of actin dynamics within certain extra-embryonic membrane progenitor cells during gastrulation , visualizing the actomyosin cable that closes the serosa window ( Figure 4A and Video 1 ) . The AGOC{ARP5’#O ( LA ) -mEmerald} #1 subline provides strong fluorescence signal in the brain and ventral nerve cord and moderate signal throughout the remaining embryonic tissue ( Figure 4B and Video 2 ) . In contrast , the AGOC{ARP5’#O ( LA ) -mEmerald} #2 subline provides uniform fluorescence intensity during gastrulation , germband elongation , germband retraction and dorsal closure ( Figure 4C and Video 3 ) . We explained the abstract genetic background of the AGOC vector concept and confirmed its straightforward applicability with Tribolium . The unique feature of our approach is that temporary ambiguities are avoided in any generation , since all genotypes are directly identified by specifically designed distinct phenotypes . Hence , AGOC-based workflows can be used to systematically create progeny with relevant genotypes , as exemplified in this study for the creation of homozygous lines . Consequently , our concept provides many advantages that apply not only to Tribolium but also to many other model organisms: ( i ) Our approach saves manpower . For example , genotyping 30 to 40 Tribolium adults with genetic assays takes about one afternoon ( Strobl et al . , 2017b ) , while processing the same number of individuals with a stereo microscope takes less than ten minutes . ( ii ) The concept does not require any further consumables . ( iii ) When genetic assays are used , the ‘slowest’ member of a group defines the earliest convenient time point for synchronized genotyping , while our concept also supports unsynchronized genotyping of single organisms . ( iv ) Our approach is non-invasive and thus favorable when invasive procedures are incompatible with the experimental workflow . It can be performed even when sufficient amounts of genomic DNA cannot be obtained without severely injuring or even sacrificing the individual . ( v ) The concept simplifies transgenic organism handling since genotypes are determined directly . Quick and reliable quantification , selection , mating and/or grouping of individuals can be performed during nearly all developmental stages . ( vi ) Our approach is less error-prone than genetic assays . In more than 300 independent instances , the progeny scores confirmed the phenotypically determined parental genotypes . ( vii ) Although homozygous transgenic lines can be systematically created with slightly less waiting time by using balancer chromosomes , a convenient number of balancer lines is only available for Drosophila ( Ashburner , 1989 ) . Furthermore , in the balancer-based approach , the insertion location has to be known , while our approach performs properly in random and semi-random insertion assays . ( viii ) Many special cases of transgenesis ( four cases that occurred during our study are described within the Materials and methods section ) can be explicitly identified and/or attended to . ( ix ) Specifically designed distinct phenotypes foster automation . For example , several approaches for the computer-controlled allocation of zebrafish embryos to 96-well plates have been suggested ( Graf et al . , 2011; Mandrell et al . , 2012 ) . Automation devices , equipped with a phenotype-adapted detection unit , in our case fluorescence , can be used to sort organisms with different genotypes according to their markers . The functionality of the AGOC vector concept was confirmed with Tribolium , but due to the universality of the Cre-Lox system , it should work in all diploid model organisms . These include various insects , zebrafish , rodents , and even plants . For many insect species , modifying the basic architecture of the vector is not necessary . It has been shown that both the piggyBac transposon system and the 3×P3 promoter function properly in Drosophila melanogaster ( Sarkar et al . , 2006 ) , its close relative Drosophila suzukii ( Schetelig and Handler , 2013 ) and many other dipterans ( Hediger et al . , 2001; Warren et al . , 2010; Caroti et al . , 2015 ) , including epidemiologically relevant mosquito species such as Aedes aegypti ( Kokoza et al . , 2001 ) and Aedes albopictus ( Labbé et al . , 2010 ) . This also applies to multiple lepidopterans such as Bicyclus anynana ( Marcus et al . , 2004 ) and Bombyx mori ( Thomas et al . , 2002 ) , other coleopterans ( Kuwayama et al . , 2006 ) as well as some hymenopterans , for example the honeybee Apis mellifera ( Schulte et al . , 2014 ) . For several other dipteran species , such as the African malaria mosquito Anopheles gambiae ( Grossman et al . , 2001 ) as well as several tephritid ( Handler and Harrell , 2001; Schetelig et al . , 2009; Raphael et al . , 2011 ) and calliphorid ( Heinrich et al . , 2002; Allen et al . , 2004 ) species , the marker cassettes have to be modified , for example by replacing the artificial 3×P3 promoter with the Drosophila polyubiquitin promoter . If fluorescence-based markers interfere with the experimental workflow , pigmentation-based markers can be used . Eye pigmentation markers are available for Drosophila melanogaster ( Adams and Sekelsky , 2002 ) and Tribolium castaneum ( Lorenzen et al . , 2002 ) , but require the appropriate background strain . Another convenient and apparently universal option for insects are arylalkylamine-N-acetyl transferase-based markers , which lighten pigmentation throughout the cuticle and thus can be detected without microscopes ( Osanai-Futahashi et al . , 2012 ) . Although the piggyBac transposon system works properly in zebrafish ( Lobo et al . , 2006 ) and the 3×P3 promoter is believed to work in a broad variety of animal species ( Berghammer et al . , 1999 ) , it may be convenient to transit to the well-established Tol2 transposon system ( Kawakami et al . , 2000 ) and to replace the 3×P3 promoter in the marker cassettes with endogenous alternatives , for example the eye-specific cryaa or the muscle-specific 503unc promoter ( Berger and Currie , 2013 ) . For mouse , epidermal ( Ikawa et al . , 1995; Zhu et al . , 2005 ) and eye-specific ( Cornett et al . , 2011 ) fluorescence-based as well as fur color-based ( Zheng et al . , 1999 ) markers have been established . In this study , we used the AGOC vector concept in conjunction with transposon-mediated transgenesis to systematically create functional homozygous Tribolium lines that are primarily designed for fluorescence live imaging of embryonic development . However , our approach can also be used in insertional mutagenesis knock-out assays , independent of whether large-scale transposon-mediated remobilization with subsequent screening is performed ( Trauner et al . , 2009 ) , or genes or genetic elements are specifically rendered inoperative , for example , by using genome engineering techniques such as CRISPR/Cas9 , where AGOC-based transgenes can be integrated into targeted genomic locations via either homology-based repair or non-homologous end-joining ( Gilles and Averof , 2014; Gilles et al . , 2015 ) . Studies that utilize genetically manipulated organisms require researchers to rear their lines for many years . A certain numbers of individuals with known genotypes are required during this period , either to maintain the lines or to use them in experiments . This results in a total demand of hundreds to thousands of organisms . The AGOC vector concept supports well-designed experimental strategies in the following scenarios: ( i ) Transgenic lines that are , for example , specifically designed for fluorescence live imaging are easily maintained as homozygotes since continuous genotyping and/or curation are not necessary . However , the initial workload following transgenesis that is required to create homozygous lines can be very high and thus a limiting factor . As shown in this study , these efforts are significantly reduced by using the AGOC vector concept . ( ii ) In knock-out assays of genes that result in homozygous lethality , the respective lines are maintained as hemizygotes , which are usually viable and phenotypically inconspicuous . When two hemizygous organisms are mated , only one quarter of their progeny are homozygous for the knock-out , while half are hemizygous and one quarter resemble the wild type . Certain experimental approaches , for example , fluorescence live imaging or transcriptome/proteome analyses , require the researcher to commence with all descendants and to select the homozygous knock-out individuals as soon as discrimination is possible , that is , when the phenotype manifests or when biological material for genetic assays can be obtained . By using our approach with appropriate markers , a preselection can be performed . This narrows the efforts down to relevant individuals and appropriate controls , for example , down to about one quarter of the currently required number . The AGOC vector concept , broadly adapted to established and emerging model organisms , contributes significantly to ethically motivated endeavors to minimize the number of wasted animals . For this study , a T . castaneum ( NCBITaxon:7070 ) double mutant background strain was created , which carries the pearl ( Grubbs et al . , 2015 ) and light ocular diaphragm ( Mocelin and Stuart , 1996 ) mutations that result in completely unpigmented eyes . This strain was called Plain-White-As-Snow ( PWAS ) and used as a donor for genomic DNA and messenger RNA as well as for the creation of transgenic lines . Cultures were kept in groups of 150–500 individuals on growth medium ( full grain wheat flour ( 113061006 , Demeter ) supplemented with 5% ( wt/wt ) inactive dry yeast ( 62–106 , Flystuff ) in 1 l glass bottles in a 12:00 hr light/12:00 hr darkness cycle at 25°C and 70% relative humidity ( DR-36VL , Percival Scientific ) . Approximately 20 PWAS adults ( 40 mg ) were starved for 24 hr before genomic DNA was extracted with the Blood and Tissue Kit ( 69504 , Qiagen ) according to the manufacturer’s instructions . From approximately 10 adults ( 20 mg ) , messenger RNA was extracted with TRIzol Reagent ( 15596026 , Thermo Fisher Scientific ) and complementary DNA was transcribed with the Superscript III reverse transcriptase ( 12080093 , Thermo Fisher Scientific ) using random hexamer primers . For germline transformation , the piggyBac transposon system ( Handler , 2002 ) was chosen , which is highly active in Tribolium . This study utilized a set of vectors based on in silico design and de novo synthesis . This section explains the architecture of the three most important vectors in general , while the detailed molecular biological procedure is explained within the following sections . All intermediate and transformation-ready vectors used in this study are based on pAVOIAF{#1–#2–#3–#4} ( Figure 1—figure supplement 3 ) . Between the unique AatII and PciI sites , this vector carries a transposon cassette which consists of ( i ) the piggyBac 3’ terminal repeat , ( ii ) a four-slot ( #1 to #4 ) cloning site and ( iii ) the 5’ piggyBac terminal repeat . The repeats have the minimal length ( 235 and 310 bp , respectively ) necessary for efficient transposition ( Li et al . , 2005 ) . The four-slot cloning site consists of four restriction enzyme site pairs , XmaI/SpeI for #1 , HindIII/XbaI for #2 , XhoI/NheI for #3 and AflII/AvrI for #4 . The pairs are separated by 18 bp Phe-Arg-Glu-Asp-Asp-Tyr ( FREDDY ) spacers . For convenience , PmeI sites were placed at upstream and downstream of the four-slot cloning site as well as between the restriction enzyme site pairs . In #3 and #4 of pAVOIAF{#1–#2–#3–#4} , mOrange- and mCherry-based eye-specific transformation markers ( mO and mC , respectively ) were inserted ( in reverse orientation ) that consist of ( i ) the artificial 3×P3 promoter ( Berghammer et al . , 1999 ) , ( ii ) the codon-optimized open-reading frames for the respective fluorescent protein , that is , mOrange2 ( Shaner et al . , 2008 ) or mCherry ( Shaner et al . , 2004 ) and ( iii ) the SV40 poly ( A ) ( van den Hoff et al . , 1993 ) . Both markers are flanked by incompatible Lox sites ( the spacer is underlined , deviations are marked bold ) : upstream by a LoxP ( 5’-ATAACTTCGTATAGCATACATTATACGAAGTTAT-3’ ) and downstream by a LoxN ( 5’-ATAACTTCGTATAGTATACCTTATACGAAGTTAT-3’ ) site . For convenience , the resulting vector was termed pAGOC ( Figure 1—figure supplement 1 ) . In #2 of pAGOC , a modular fluorescent protein expression cassette was inserted , which consists of ( i ) a two-slot cloning site composed of a promoter ( #P ) and an open-reading frame ( #O ) slot , ( ii ) a 9 bp Ala-Ala-Ala linker , ( iii ) the codon-optimized mEmerald open-reading frame ( Tsien , 1998 ) and ( iv ) an elongated variant of the SV40 poly ( A ) ( van den Hoff et al . , 1993 ) . The #P slot can be accessed by the AscI/FseI site pair , or alternatively scarlessly by the double BtgZI site pair . The #O slot carries the open-reading frame for the Saccharomyces cerevisiae Lifeact peptide tag ( Riedl et al . , 2008 ) per default and can be accessed by the FseI/NotI site pair . The mEmerald open-reading frame can be accessed by the NotI/SbfI site pair . For convenience , the resulting vector was termed pAGOC{#P’#O ( LA ) -mEmerald} ( Figure 1—figure supplement 2 ) . Any exogenous or endogenous promoter can be inserted in the #P slot to generate the spatiotemporal activity pattern of choice . For the #O slot , the Lifeact open-reading frame can be replaced with any open-reading frame to change the subcellular localization of the fluorescent protein . In the default configuration , the Lifeact peptide tag will guide mEmerald to the actin cytoskeleton . In this study , 25 vectors ( Figure 1—figure supplement 4 and Supplementary file 8 ) plus the commercial subcloning vector pGEM-T Easy ( A1360 , Promega ) were used . Three were ordered as gene synthesis plasmids , two were previously published and obtained from the respective laboratories or from Addgene , six are library vectors , while the remaining 13 vectors are derivates . For all PCRs , Phusion High Fidelity DNA polymerase ( M0530L , New England BioLabs ) was used , and T4 DNA ligase ( M0202L , New England BioLabs or provided with the pGEM-T Easy vector ) for all ligations . Cloning primers are listed in Supplementary file 9 . The respective promoter and open-reading frame sequences were amplified from genomic or complementary DNA by using the appropriate extraction PCR primer pairs ( C1 for tubulin alpha 1-like protein ( ATub’ ) , C2 for zerknüllt 1 ( Zen1’ ) , C3 for actin-related protein 5 ( ARP5’ ) and C4 for heat shock protein 68b ( HSP68’ ) as well as C5 for beta-galactoside alpha-2 , 6-sialyltransferase 1 transcription variant X1 ( ’SiaTr ) and C5 for histone H2B ( ’H2B ) ) . Amplification was followed by A-tailing using the Recombinant Taq DNA polymerase ( 10342020 , Thermo Fisher Scientific ) and ligation into pGEM-T Easy . The resulting vectors were termed pTC-ATub’-GEM-T Easy , pTC-Zen1’-GEM-T Easy , pTC-ARP5’-GEM-T Easy , pTC-HSP68’-GEM-T Easy , pTC-’SiaTr-GEM-T Easy and pTC-’H2B-GEM-T Easy . To create the hybrid promoter/open-reading frame library vectors , the sequences were amplified from the library vectors or pTriEx-HTNC ( Peitz et al . , 2002 ) with the respective fusion PCR primer pairs ( either C7 for HSP68’ / ’NLS-Cre , or C8 for ATub’ / ’H2B ) and fused in a secondary PCR reaction using both PCR products as a template and the promoter forward primer ( C7-1 or C8-1 , respectively ) and the open-reading frame reverse primer ( C7-4 or C8-4 , respectively ) . The primer pairs introduce upstream an AscI and downstream a NotI site or upstream a NheI and downstream a XhoI site , respectively . The fusion PCR products were inserted into pGEM-T Easy as described above . The resulting vectors were termed pTC-ATub’H2B-GEM-T Easy and pTC-HSP68’NLS-Cre-GEM-T Easy . A hybrid sequence , consisting of ( i ) the transposon cassette as well as ( ii ) mO and mC and their flanking Lox sites in #3 and #4 as described above , was de novo synthetized and inserted into the unique NdeI and PstI sites of pUC57-Kan ( GeneBank accession number JF826242 . 2 ) . The resulting vector was termed pUC57[AGOC] . The insert was PCR amplified with primer pair C9 , which introduced upstream an AatII and downstream a PciI site . The PCR product and pUC57[AGOC] were digested accordingly , and the insert was reintegrated into the vector , removing 629 functionless bp . The resulting vector was termed pAGOC and used ( i ) as a transformation-ready vector for germline transformation , and ( ii ) as an intermediate vector for further cloning operations . A hybrid sequence , consisting of ( i ) a HindIII site , ( ii ) the modular fluorescent protein expression cassette as described above and ( iii ) a XbaI site , was de novo synthetized and inserted into the unique SfiI site of pMK-RQ ( Thermo Fisher Scientific ) . The resulting vector was termed pGS[#P’#O ( LA ) -mEmerald] . The insert was excised from the backbone with HindIII/XbaI and inserted into #3 of the pAGOC vector . The resulting vector was termed pAGOC{#P’#O ( LA ) -mEmerald} and used as an intermediate vector for further cloning operations . The respective promoter sequences were amplified from the library vectors with the respective primer pairs ( C10 for ATub’ , C11 for Zen1’ and C12 for ARP5’ ) , which introduced upstream an AscI and downstream a BsmBI ( ATub’ ) or BsaI ( Zen1’ and ARP5’ ) site . The PCR products were digested accordingly , and the pAGOC{#P’#O ( LA ) -mEmerald} vector was digested with BtgZI , which led to compatible overhangs and allowed scarless insertion of the promoter sequences into #P . The resulting vectors were termed pAGOC{ATub’#O ( LA ) -mEmerald} , pAGOC{Zen1’#O ( LA ) -mEmerald} and pAGOC{ARP5’#O ( LA ) -mEmerald} and were used for germline transformation . The ’SiaTr open-reading frame sequence was amplified from the pTC-’SiaTr-GEM-T Easy vector with primer pair C13 , which introduced upstream an FseI and downstream a NotI site . The PCR product and the pAGOC{#P’#O ( LA ) -mEmerald} vector were digested accordingly and the insert was inserted into #O of the vector . The resulting vector was termed pAGOC{#P’SiaTr-mEmerald} and used as an intermediate vector for further cloning operations . The tubulin alpha 1-like protein promoter sequence was amplified from the pTC-ATub’-GEM-T Easy vector with primer pair C10 , which introduced upstream an AscI and downstream a BsmBI site . The PCR product was digested accordingly , and the pAGOC{#P’SiaTr-mEmerald} vector was digested with BtgZI , which led to compatible overhangs and allowed scarless insertion of the promoter sequence into #P of the intermediate vector . The resulting vector was termed pAGOC{ATub’SiaTr-mEmerald} and used for germline transformation . The ATub’H2B promoter/open-reading frame sequence was excised from pTC-ATub’H2B-GEM-T Easy with AscI and NotI and inserted into #P’#O of the accordingly digested pAGOC{#P’#O ( LA ) -mEmerald} vector . The resulting vector was termed pAGOC{ATub’H2B-mEmerald} and used for germline transformation . The HSP68’NLS-Cre recombinase promoter/open-reading frame sequence was excised from pTC-HSP68’H2B-GEM-T Easy with NheI and XhoI and inserted ( in reverse orientation ) into #3 of the accordingly digested pAGOC vector , replacing mO and the flanking Lox sites . The resulting vector was termed pAVOIAF{#1–#2–HSP68’NLS-Cre–mC} and used as an intermediate vector for further cloning operations . A hybrid sequence , which consists ( beside other elements ) of the mCerulean-based eye-specific transformation marker ( mCe ) that is composed of ( i ) the artificial 3×P3 promoter , ( ii ) the codon-optimized open-reading frame for mCerulean2 ( Markwardt et al . , 2011 ) and ( iii ) the SV40 poly ( A ) , was de novo synthetized and inserted into the unique SfiI site of pMK-RQ ( Thermo Fisher Scientific ) . The resulting vector was termed pGS[ACOS] . Next , mCe was amplified with primer pair C14 , which introduced upstream an AflII and downstream an AvrII site . The PCR product was digested accordingly and inserted into #4 of pAVOIAF{#1–#2–HSP68’NLS-Cre–mC} , replacing mC and the flanking Lox sites . The resulting vector was termed pICE{HSP68’NLS-Cre} and used for germline transformation to create the Cre recombinase-expressing helper lines . The ATub promoter and the piggyBac open-reading frame fragment were amplified from pTC-ATub’-GEM-T Easy and pBSII-IFP-ORF ( Yoshida et al . , 2009 ) with the C15 primer pairs and fused together in a secondary PCR reaction using both PCR products as a template as well as the promoter forward primer ( C15-1 ) and the open-reading frame reverse primer ( C15-4 ) . The primers introduced upstream a SalI and downstream a BglII site , respectively . The fusion PCR product was digested and then reintegrated into the accordingly digested pBSII-IFP-ORF vector . The resulting vector was termed pATub’piggyBac and used as the transposase-expressing helper vector during germline transformation . Approximately 500 F0 PWAS adults were incubated on 405 fine wheat flour ( 113061036 , Demeter , Darmstadt , Germany ) supplemented with 5% ( wt/wt ) inactive dry yeast ( 62–106 , Flystuff , San Diego , CA ) at 25°C and 70% relative humidity in light for 2 hr . After the incubation period , the adults were removed and the embryos ( around 700 to 900 ) were extracted from the flour and incubated another hour as stated above . Next , the embryos were briefly washed in 10% ( vol/vol ) sodium hypochlorite ( 425044–250 ML , Sigma Adlrich ) in autoclaved tap water for 10 s , stored in autoclaved tap water and lined up on microscopy slides within the next hour . The embryos were injected with a mixture of 500 ng/µl transformation-ready vector and 400 ng/µl pATub’piggyBac in injection buffer ( 5 mM KCl , 1 mM KH2PO4 in ddH2O , pH 8 . 8 ) . For injection , a microinjector ( FemtoJet , Eppendorf ) and 0 . 7 µm outer diameter capillaries ( Femtotips II , Eppendorf ) with an injection pressure of 400–800 hPa were used . After injection , the microscopy slides with embryos were placed on a 5 mm high 1% ( wt/vol ) broad range agarose ( T846 . 3 , Carl Roth ) in tap water ‘platform’ within Petri dishes and incubated at 32°C . After 3 days , hatched larvae , that is , F1 potential mosaics , were collected and raised individually in single wells of 24-well plates as described above . Germline transformation resulted in a total of seven lines with 21 sublines , which are summarized in Supplementary file 10 . All crossings were performed with single female-male pairs in small glass vials filled with 1 . 5 g or 2 . 5 g ( F4 cross ) of growth medium . Progeny were placed individually in wells of 24-well plates and scored for the presence of markers during pupal or adult stage by using a fluorescence stereo microscope ( SteREO Discovery . V8 , Zeiss ) with appropriate filter sets ( Supplementary file 11 ) . For each pair , images in the reflected light and fluorescence channels were taken in parallel with appropriate controls . The mating procedure is described within the results section . A one-sample/two tailed Student’s t-test was performed to determine whether the arithmetic means differ significantly from the theoretical Mendelian ratios . Insert numbers were determined by mating F2 hemizygotes with wild types and scoring the progeny , whereas a transgene distribution of 60% or less was interpreted as a single insertion . Homozygous viability was determined by mating two F3 hemizygotes and scoring the progeny , whereas a transgene distribution of 70% or more was interpreted as a homozygous viable line . During the experimental validation of the AGOC vector concept , four special cases of transgenesis occurred and were attended to as follows: ( i ) The homozygous viability crosses indicated that the transgenes of the proof-of-principle AGOC #3 , the functional AGOC{ATub’#O ( LA ) -mEmerald} #1 and the functional AGOC{ATub’H2B-mEmerald} #4 sublines are heterozygous/homozygous lethal ( Supplementary file 2 ) . However , F6 ( mO/mC ) heterozygotes were obtained for all three lines and the mating procedure could be performed successfully with the AGOC #3 and AGOC{ATub’#O ( LA ) -mEmerald} #1 sublines ( Table 1 , Supplementary file 6 and Supplementary file 7 ) . The mating procedure only aborted for the AGOC{ATub’#O ( LA ) -mEmerald} #4 subline , because the F6 ( mO/mC ) heterozygotes were sterile . Since the transgene of the proof-of-principle AGOC #3 subline might interfere with the sialin-like gene due to its insertion location ( Supplementary file 3 ) , and since both functional sublines mentioned above display a strong green fluorescence signal , hemizygous individuals of those three sublines are believed to be exposed to high stress levels , and that these levels are even higher in homozygotes . Thus , these transgenic sublines are believed to be essentially viable , but a certain percentage of the descendants fail to develop properly , which results in biased progeny ratios in the homozygous viability cross . ( ii ) In contrast to heterozygous/homozygous lethality , heterozygous/homozygous sterility cannot be estimated from the homozygous viability crosses . The AGOC{ATub’H2B-mEmerald} #4 subline was assumed to be heterozygous/homozygous lethal ( Supplementary file 2 ) , but mating a F5 ( mO ) post-recombination hemizygous female with a F5 ( mC ) post-recombination hemizygous male sibling resulted in F6 ( mO/mC ) heterozygotes . However , by mating F6 ( mO/mC ) heterozygous females with genotypically identical F6 males , no progeny was obtained ( n = 12 ) . Further , crossing F6 ( mO/mC ) heterozygous females and wild-type males did not result in any progeny ( n = 12 ) . To confirm sterility of both genders , F6 ( mO/mC ) heterozygous males were mated with wild-type females , which also did not result in any progeny ( n = 8 ) . ( iii ) The AGOC{Zen1’#O ( LA ) -mEmerald} #2 subline carries the transgene not on one of the nine autosomes , but on the X allosome . Thus , a slightly modified mating procedure was necessary to obtain F7 ( mO/mO ) and ( mC/mC ) homozygotes ( Figure 2—figure supplement 2 ) . ( iv ) The piggyBac transposon system is highly efficient in Tribolium , which results to a certain degree also from the 4 bp TTAA target sequence . However , due to this very short length of the targeting sequence , a certain probability for nested insertions is given , that is , the transgene inserts into another transformation vector , since these vectors also carry multiple TTAA target sequences . In the AGOC{Zen1’#O ( LA ) -mEmerald} #3 subline , an insertion of the transgene into the backbone of another vector occurred , and the nested transgene was subsequently inserted into the genome , as revealed by sequencing of the insertion junction ( Supplementary file 3 ) . This rare and undesired case is ‘corrected’ during the mating procedure of the AGOC vector concept , since within the F4 ( mCe; mO-mC ) double hemizygous generation , the Cre recombinase excises nearly one ‘stitched equivalent’ of the initial transformation vector from the genome . The F5 ( mO ) and ( mC ) hemizygous progeny then carries only one , but complete copy of the transgene . One of the most obvious special cases of transgenesis , heterozygous/homozygous lethality , did not occur . However , the AGOC vector concept would allow the determination of this special case with a high degree of certainty . When F5 ( mO ) post-recombination hemizygous females are mated with F5 ( mC ) post-recombination hemizygous male siblings and no F6 ( mO/mC ) heterozygotes are obtained , it can be assumed that the transgene is heterozygous/homozygous lethal . Multiple other transgenesis special cases are possible ( e . g . female- or male-only heterozygous/homozygous lethality or sterility , nested insertions of transgenes into transgenes , multiple inserts in close proximity on the same chromosome ) , but are not discussed here due to the very low probability of their occurrence . To determine the insertion junction of the piggyBac-based transgene , the inverse PCR approach was chosen ( Triglia et al . , 1988; Ochman et al . , 1988 ) . All inverse PCR primers are listed in Supplementary file 9 . At first , inverse PCR was performed for the junction at the 5’ piggyBac terminal repeat with the I-5’ primer pair with up to eight different restriction enzymes , and if unsuccessful , also at the 3’ piggyBac terminal repeat with the I-3’ primer pair with up to six different restriction enzymes . PCR products were extracted from the gel , A-tailed , ligated into pGEM-T Easy and sequenced . For each successful inverse PCR , a control PCR at the respective other side was performed . For the control PCR , location-specific primers were used to perform a standard PCR . Inverse PCR was successful for 10 out of 19 lines , for 4 out of 6 proof-of-principle lines and 6 out of 13 functional lines . The sequencing results were aligned to the Tribolium genome ( Richards et al . , 2008 ) via the BeetleBase ( Wang et al . , 2007; Kim et al . , 2010 ) ( RRID:SCR_001955 ) BLAST . Insertion junctions are listed in Supplementary file 3 . Long-term live imaging was performed with digitally scanned laser light-sheet-based fluorescence microscopy ( DSLM , LSFM ) ( Keller et al . , 2008; Keller and Stelzer , 2010 ) as described previously for Tribolium ( Strobl and Stelzer , 2014; Strobl et al . , 2015 ) . In brief , embryo collection was performed with the F7+ continuative ( mC/mC ) homozygous lines for 1 hr at 25°C , and embryos were incubated for 15 hr at 25°C . Sample preparation took approximately 1 hr at room temperature ( 23 ± 1°C ) , so that embryos were at the beginning of gastrulation . Embryos were recorded along four pair-wise orthogonal directions , that is , in the orientations 0° , 90° , 180° and 270° , with an interval of 30 min . All shown embryos survived the imaging procedure , developed to healthy and fertile adults , and when mated with wild types , produced only transgenic progeny that were also fertile . Metadata for the three datasets is provided in Supplementary file 12 .	Researchers frequently use model organisms , such as mice , zebrafish and various insect species , to understand biological processes – with the underlying idea that discoveries made can be applied to other species too . A common technique is genetic manipulation , in which a foreign gene is inserted into the chromosome of an organism . These introduced genes are called transgenes and the organisms carrying them are referred to as transgenic . Transgenic organisms are powerful tools to analyze biological processes or mimic human diseases . Many model organisms carry two homologous chromosomes – one inherited from each parent . Pairs of chromosomes carry genes in the same order , but do not necessarily have identical versions of those genes . Newly created transgenic organisms , however , carry the transgene on only one of the chromosomes . This can be a problem for researchers , as many experiments require individuals that carry the transgene on both . Unfortunately , only costly and error-prone methods can distinguish between these individuals . To overcome these drawbacks , Strobl et al . developed a concept called AGameOfClones and applied it to the red flour beetle Tribolium castaneum . In their approach , the transgene also expresses two marker-proteins with different fluorescent colors . After several generations of breeding , two versions of the transgene emerge – each retaining only one of the markers . This means that in the following generation , descendants that express both markers must be the offspring that carry the transgene on both of the chromosomes . The AGameOfClones concept has several major advantages: individuals with different markers can be easily identified , the procedure is cost-efficient and reliable , and it can be applied to nearly all model organisms . This will benefit breeding schemes and animal welfare since irrelevant individuals can be excluded as soon as the markers become detectable .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "developmental", "biology", "tools", "and", "resources" ]	2018	A universal vector concept for a direct genotyping of transgenic organisms and a systematic creation of homozygous lines
Most AAA+ remodeling motors denature proteins by pulling on the peptide termini of folded substrates , but it is not well-understood how motors produce grip when resisting a folded domain . Here , at single amino-acid resolution , we identify the determinants of grip by measuring how substrate tail sequences alter the unfolding activity of the unfoldase-protease ClpXP . The seven amino acids abutting a stable substrate domain are key , with residues 2–6 forming a core that contributes most significantly to grip . ClpX grips large hydrophobic and aromatic side chains strongly and small , polar , or charged side chains weakly . Multiple side chains interact with pore loops synergistically to strengthen grip . In combination with recent structures , our results support a mechanism in which unfolding grip is primarily mediated by non-specific van der Waal’s interactions between core side chains of the substrate tail and a subset of YVG loops at the top of the ClpX axial pore . Cells maintain homeostasis by balancing protein synthesis and degradation with growth . When nutrients are available , new proteins are constantly synthesized , whereas damaged , misfolded , or unneeded proteins are degraded . Regulated degradation typically requires a protein-unfolding motor of the AAA+ family ( ATPases associated with various cellular activities ) that associates with a self-compartmentalized protease ( e . g . , ClpX with ClpP , the 19S regulatory particle with the 20S proteasome ) or is genetically tethered to a protease ( e . g . , Lon , FtsH , Yme1 ) ( Sauer and Baker , 2011; Olivares et al . , 2016; Glynn , 2017 ) . When challenged with degrading a folded substrate , AAA+ motors couple ATP hydrolysis to mechanical motion that overcomes the resistance of the folded domain . Despite broad consensus on the overall mechanism of protein unfolding , it is largely unknown how interactions between a AAA+ motor and its substrate produce grip , the ability for the motor to maintain hold of the substrate while applying an unfolding force . ClpX is a ring-shaped AAA+ homohexamer that functions autonomously in protein remodeling in bacteria and eukaryotic organelles and also associates with ClpP tetradecamers to form the ATP-dependent ClpXP protease ( Baker and Sauer , 2012 ) . Substrates are targeted to ClpX or ClpXP by N- or C-terminal peptide tails ( also called degradation tags or degrons ) , which initially bind in the ClpX axial pore . Proteins marked with a sequence-defined degron or post-translational modification can be recruited to the AAA+ protease directly or with assistance from auxiliary adaptors ( Sauer and Baker , 2011; Trentini et al . , 2016 ) . For example , during rescue of stalled ribosomes in Escherichia coli , the 11-residue ssrA tag is appended to the C-terminus of abortive protein products ( Keiler et al . , 1996 ) , allowing ClpXP to recognize and degrade the attached protein ( Gottesman et al . , 1998; Farrell et al . , 2005 ) . The ssrA tag and other degron tails interact with ClpX loops that line the axial pore . A Tyr-Val-Gly ( YVG ) sequence in the pore-1 loop is critical for substrate binding , unfolding , and translocation ( Siddiqui et al . , 2004; Martin et al . , 2008a; Martin et al . , 2008b; Iosefson et al . , 2015 ) . Other AAA+ unfolding motors contain related pore-1 loops and mutation of these loops typically abolishes function ( Yamada-Inagawa et al . , 2003; Schlieker et al . , 2004; Hinnerwisch et al . , 2005; Park et al . , 2005 ) . In several AAA+ unfolding motors , the pore-1 loops adopt a spiral staircase conformation within the pore , which facilitates multivalent interaction with the bound peptide tail ( Monroe et al . , 2017; Gates et al . , 2017; Puchades et al . , 2017; de la Peña et al . , 2018; Majumder et al . , 2019; Dong et al . , 2019; White et al . , 2018 ) . ATP-dependent conformational changes are thought to draw the tail of a substrate into the pore until a folded domain too large to transit the pore impedes progress . For ClpXP , repeated cycles of ATP hydrolysis are then required to unfold the substrate and to translocate the polypeptide through the pore and into ClpP for degradation ( Kenniston et al . , 2003; Aubin-Tam et al . , 2011; Maillard et al . , 2011; Sen et al . , 2013; Cordova et al . , 2014 ) . How the amino acids in the bound substrate tail contribute to grip during unfolding remains poorly understood . When degrading unfolded substrates , the rate of substrate translocation through ClpXP is largely insensitive to amino-acid charge , size , or peptide-bond spacing ( Barkow et al . , 2009 ) . In contrast , when directly abutting a folded domain , sequences rich in glycine can result in failed unfolding by ClpXP or the 26S proteasome , leading to release of partially processed intermediates ( Lin and Ghosh , 1996; Levitskaya et al . , 1997; Sharipo et al . , 2001; Hoyt et al . , 2006; Daskalogianni et al . , 2008; Too et al . , 2013; Kraut , 2013; Vass and Chien , 2013 ) . Abortive unfolding caused by Gly-rich motifs occurs as a result of slower domain unfolding rather than rapid substrate dissociation , suggesting that these motifs bind normally but are gripped poorly during unfolding ( Kraut , 2013; Kraut et al . , 2012 ) . These results suggest that AAA+ motors struggle to efficiently grip sequences with very small side chains . Alternatively , sequence complexity rather than composition may dictate grip strength ( Tian et al . , 2005 ) . Here , we use green fluorescent protein ( GFP ) reporter substrates to interrogate the contributions of individual amino acids in the peptide tail to grip strength by ClpXP . In degradation assays performed in vivo and in vitro , we observe that substrate grip by ClpX is primarily mediated by interactions with a block of five amino acids , located two to six residues from the native GFP domain . Through systematic mutation , we characterize the ability of each amino acid to promote ClpX grip , and find that aromatic and large hydrophobic residues are gripped well , whereas charged and polar residues impair grip . Finally , we analyze synergistic contributions of multiple residues to unfolding , and show that contacts with more than one side chain lead to stronger grip and faster substrate unfolding . Our results provide unprecedented detail into the mechanism by which AAA+ motors grip terminal substrate tails during protein unfolding . To probe grip during substrate unfolding , we used Aequorea victoria GFP , as its native structure is highly kinetically stable , unfolding is rate limiting for ClpXP degradation of GFP-ssrA , and the pathway of mechanical unfolding of GFP-ssrA by ClpXP is well characterized ( Maillard et al . , 2011; Kim et al . , 2000; Nager et al . , 2011 ) . In our substrates , we truncated GFP at Ile-229 , the last amino acid that makes extensive native contacts in multiple crystal structures ( Ormö et al . , 1996; Yang et al . , 1996 ) , and added a 12-residue cassette of variable sequence followed by a partial ssrA degron to allow recognition by ClpXP ( Figure 1A ) . Given the length of the axial pore ( ~35 Å; Glynn et al . , 2009 ) , we reasoned that ClpX should only interact with residues within the cassette region during GFP unfolding . For studies of intracellular degradation , we used an E . coli B strain , which lacks the AAA+ Lon protease; deleted the chromosomal copies of clpP , clpX , and clpA , as ClpAP can also degrade ssrA-tagged substrates ( Gottesman et al . , 1998; Farrell et al . , 2005 ) ; and placed genes encoding ClpX∆N and ClpP on a plasmid under arabinose-inducible control ( Figure 1B; Guzman et al . , 1995 ) . Despite lacking a family-specific N-terminal domain , ClpXΔN supports ClpP-degradation of ssrA-tagged substrates as well as wild-type ClpX but does not interact with many other cellular substrates and adaptors ( Singh et al . , 2001; Wojtyra et al . , 2003; Flynn et al . , 2003; Martin et al . , 2005 ) . GFP substrates were cloned under transcriptional control of a constitutive ProD promoter ( Figure 1B; Davis et al . , 2011 ) . To control for ClpXP-independent degradation , we used a pBAD plasmid isogenic to the clpP/clpX∆N vector but lacking these genes ( Figure 1B ) . To determine the extent of intracellular GFP degradation , we measured GFP fluorescence after arabinose induction and growth for 35 min . We first used this system to characterize substrates with different high or low complexity sequences in the 12-residue cassette . Substrate expression levels were sensitive to the cassette sequence , possibly because of effects on mRNA stability or translation , varying by as much as 7-fold ( Figure 1C ) . To control for differences in expression when measuring degradation of different substrates , we normalized measurements in the strain expressing ClpXΔNP to a strain lacking it ( Figure 1D , Table 1 ) . In these experiments , low-complexity tails rich in acidic or basic residues promoted GFP degradation at levels comparable to a high-complexity sequence derived from the human titinI27 domain or a sequence of interspersed glycines and alanines ( called GA ) . By contrast , a cassette sequence of twelve glycines ( called Gly12 ) resulted in poor degradation . We purified N-terminally His6-tagged variants of these substrates and performed Michaelis-Menten analysis of steady-state ClpXΔNP degradation in vitro ( Figure 1E ) . Degradation with the Gly12 cassette was too slow to measure , but the GA and acidic cassettes promoted degradation with Vmax values similar to the titin sequence ( Figure 1F , Table 2 ) . The basic cassette sequence resulted in an intermediate rate of maximal degradation . The differences between our results in vivo and in vitro suggest that the endpoint assay in vivo has an upper limit and cannot differentiate rates once the pool of cellular GFP has been degraded . About 20% of the Gly12 substrate appeared to be degraded in vivo , whereas no degradation was seen in vitro . Maturation of the GFP chromophore lags protein folding ( Reid and Flynn , 1997 ) , and thus degradation of immature non-fluorescent GFP would not be detected in our cellular assay . It is possible that solutes or macromolecular crowding in the cell enhance ClpXP activity or make GFP easier to unfold . Nevertheless , 12 consecutive glycines inhibit ClpXP unfolding/degradation of GFP both in vivo and in vitro , whereas other low-complexity sequences do not . We substituted amino acids in the Gly12 cassette to identify residues/positions that might improve ClpX grip and thus rates of unfolding and degradation . We first positioned a three-residue Leu-Tyr-Val ( LYV ) sequence in a sliding window across an otherwise poly-Gly cassette ( Figure 2A ) . This tripeptide sequence was selected because its residues are large and hydrophobic , unlike the surrounding Gly residues . Placing the LYV sequence at positions 2–4 , 4–6 , or 6–8 ( numbered relative to the last residue of the folded domain ) improved GFP degradation to levels similar to the GA substrate , whereas this tripeptide at positions 8–10 and 10–12 had no substantial effect relative to the Gly12 parent ( Figure 2A , Table 1 ) . These results suggest that ClpX grips side chains within the first eight residues of the substrate tail during unfolding . To examine the contributions of individual residues to grip , we constructed another panel of substrates in which a single Tyr residue was placed at each of the first eight tail positions in otherwise all-glycine cassettes ( Figure 2B ) . We then tested degradation in vivo . Substrates with a single Tyr at positions 3 , 4 , and 5 were efficiently degraded , a Tyr at position 2 supported an intermediate level of degradation , and Tyr side chains at other cassette positions supported degradation similar to the Gly12 parent ( Figure 2B , Table 1 ) . Thus , four residues appear to contribute the most important grip contacts during unfolding , with tail positions 3–5 being most significant . When we measured degradation of purified substrates in vitro ( Figure 2C , Table 2 ) , single Tyr side chains at positions 2–6 facilitated GFP degradation , with the experimental Vmax values forming a roughly normal distribution centered around position 4 . Again , Tyr side chains at positions 3–5 were most important , Tyr residues at the flanking 2 and 6 positions had small effects , and Tyr side chains at positions 1 , 7 , or 8 had no discernable effect . Importantly , changing the position of the Tyr side chain altered the maximal rate of unfolding/degradation without substantially affecting KM for degradation or the ability of substrate to stimulate ATP hydrolysis ( Figure 2D , Figure 2—figure supplements 1–2 , Figure 2—source data 1 ) . As a result , substrates that were degraded slowly also exhibited a high ATP cost for degradation ( Figure 2E ) . In combination , these results support a model in which ClpX preferentially grips the side chains of residues at positions 3–5 during GFP unfolding . Moreover , gripping a single Tyr side chain at one of these positions is sufficient for robust unfolding and degradation of GFP . Next , we exploited this system to determine how different types of side chains affect ClpX grip . We constructed substrates in which each of the remaining 18 natural amino acids was placed at position 4 of a cassette with glycines at the other 11 positions . These substrates exhibited a wide range of susceptibility to ClpXP degradation in E . coli ( Figure 3A , Table 1 ) . In general , tails containing an aromatic or large/branched hydrophobic side chain ( Tyr , Phe , Val , Ile , Leu , or Met ) promoted the most efficient unfolding and degradation , whereas small and/or polar side chains were least efficient . The inhibitory effects of polarity and charge on grip were most obvious for side chains with similar shapes . For example , Val was one of the best side chains for grip , whereas the isosteric Thr side chain was very poor ( Figure 3B ) . Similarly , a polar Gln side chain resulted in better grip than an isosteric but negatively charged Glu side chain ( Figure 3B ) . We also determined steady-state kinetic parameters for degradation of a subset of purified substrates in vitro ( Figure 3C , Table 2 ) . These results largely mirrored results in vivo , with mid-sized or large hydrophobic and aromatic residues promoting the fastest rates of degradation ( Figure 3D ) . Again , Val supported much better degradation than Thr , and Gln promoted significantly faster degradation than Glu in degradation assays in vitro ( Figure 3E ) . Further , Ser failed to support GFP degradation while both Ala and Cys facilitated low-level degradation ( Figure 3E ) . The Ala-4 , Ser-4 , Cys-4 , Thr-4 , Val-4 , Glu-4 , and Gln-4 substrates at concentrations of 15 µM stimulated the rate of ClpX ATP hydrolysis ~3–4 fold compared to the absence of substrate ( Figure 2—figure supplement 2 , Figure 2—source data 1 ) . Thus , each substrate binds ClpX well at this concentration , supporting a model in which the large differences in maximal degradation arise from poor grip caused , at least in part , by differences in side-chain charge and polarity . The maximal degradation rates for these substrates ( Figure 3E ) were inversely correlated with their energetic efficiencies of degradation ( Figure 3F ) , indicating that poor grip results in non-productive ATP hydrolysis . Placing a single alanine at cassette position four with glycines at the remaining positions resulted in only marginally better degradation than the Gly12 substrate ( Figure 3A and C ) . By contrast , the GA cassette – with alanines at positions 1 , 3 , 7 , 9 , and 12 – supported efficient degradation ( Figure 1D and F ) , despite the fact that positions 1 , 7 , 9 , and 12 do not seem to be important determinants of grip ( Figure 4A ) . This discrepancy suggested that synergistic interactions between the ClpX pore and multiple side chains might allow substantially better grip . To test this model , we constructed a panel of substrates with one alanine at position 4 and a second alanine at position 1 , 2 , 3 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , or 12 ( Figure 4B ) . In our cellular assay , alanines at cassette positions 1/4 , 2/4 , 3/4 , and 4/5 supported robust degradation , alanines at positions 4/6 and 4/7 facilitated moderate degradation , and alanines at positions 4/8 , 4/9 , 4/10 , 4/11 and 4/12 were little better than the single alanine at position 4 ( Figure 4B ) . A single Ala at position 1 supported slightly better degradation in vivo than a single Ala at position 4 , but the Ala-1/4 substrate was degraded more efficiently ( Figure 4C ) . This difference was more pronounced in assays of degradation in vitro ( Figure 4D , Table 2 ) . Indeed , Vmax for degradation of the Ala-1/4 substrate ( 2 . 3 ± 0 . 2 min−1 ) was more that 10-fold greater than Vmax for the Ala-1 substrate ( 0 . 19 ± 0 . 04 min−1 ) or Ala-4 substrate ( 0 . 13 ± 0 . 06 min−1 ) . The non-additivity of these Vmax values provides direct evidence for synergy in grip . Because the pore loops of ClpX and other AAA+ motors interact with every other substrate residue in cryo-EM structures ( Monroe et al . , 2017; Gates et al . , 2017; Puchades et al . , 2017; de la Peña et al . , 2018; Majumder et al . , 2019; Dong et al . , 2019; White et al . , 2018 ) , we investigated whether a similar spacing of large side chains enhances grip . We designed panels of substrates with either Tyr or Val fixed at tail position 4 and a second residue of the same type at positions 1 , 2 , 3 , 5 , 6 , 7 , or 8 in an otherwise all-Gly cassette ( Figure 4E and F ) . Among the Tyr substrates , the Tyr-2/4 , Tyr-4/6 , and Tyr-4/8 substrates promoted faster GFP degradation than the parental Tyr-4 substrate , whereas the other substrates exhibited similar or slower degradation ( Table 2; Figure 4E ) . It is noteworthy that although the effect of multiple residues was irrelevant for the Ala-4/8 substrate ( Figure 4B ) , the Tyr-4/8 substrate was gripped well , suggesting that spacing of Tyr in multiples of two may allow ClpX to grip substrate in a preferred conformation . Among the Val substrates , Val-1/4 and Val-4/5 facilitated slightly faster GFP degradation than the parental Val-4 substrate , Val-2/4 and Val-3/4 were degraded at similar rates to Val-4 , and Val-4/6 , Val-4/7 , and Val-4/8 were degraded slightly slower ( Figure 4F; Table 2 ) . As the pattern of degradation rates for the branched Val residue ( Figure 4F ) is more similar to Ala ( Figure 4B and D ) than the aromatic Tyr ( Figure 4E ) , it is possible that the ClpX pore-1 loops interact with aromatic residues somewhat differently from non-aromatic residues . We tested an additional panel of substrates with a Tyr residue at position three and a second Tyr at positions 1 , 2 , 4 , 5 , 6 , 7 , or 8 ( Figure 4—figure supplement 1 ) . Unlike the other Tyr substrates , these substrates were generally degraded at rates slightly higher than those of the parental Tyr-3 substrate irrespective of spacing ( Table 2; Figure 4—figure supplement 1 ) . Thus , binding Tyr residues separated by multiples of two residues does not always enhance grip . To unfold target proteins , ClpX and other AAA +protein remodeling machines use cycles of ATP binding and hydrolysis to pull on the degron tail of a substrate , thereby transmitting force to the native domain , but how these machines interact with individual tail residues during unfolding was poorly understood . Here , we identify and quantify the abilities of different tail residues to promote substrate grip during unfolding by ClpXP . Our experiments are enabled by the observation that placing 12 Gly residues between native GFP and a degron eliminates ClpXP degradation in vitro and markedly slows degradation in vivo . GFP unfolding/degradation was not inhibited by 12-residue sequences containing mixtures of Gly and Ala ( GA cassette ) ; Gly , Lys , and Arg ( basic cassette ) ; or Gly , Asp , and Glu ( acidic cassette ) . Compared with these sequences , ClpXP probably grips Gly12 poorly because of the absence of β-carbons and distal side-chain atoms or increased backbone flexibility . We use ‘grip’ in a functional rather than strictly physical sense , although the two concepts are undoubtedly related . Ensemble and single-molecule experiments show that ClpXP can translocate an enormous number of different amino-acid sequences , including long Gly tracts , with only minor velocity differences ( Aubin-Tam et al . , 2011; Maillard et al . , 2011; Sen et al . , 2013; Cordova et al . , 2014; Barkow et al . , 2009 ) . For example , in assays requiring ATP-dependent translocation , ClpXP degraded peptide substrates containing Gly10 , [Val-Gly]5 , or [Phe-Gly]5 sequences at similar rates ( Barkow et al . , 2009 ) . However , if ClpX can translocate poly-Gly sequences , then why does Gly12 inhibit or slow unfolding/degradation ? When ClpX pulls on a native protein , Newtonian mechanics dictate that the folded domain resists with an opposing force , which would be absent during translocation of an unstructured polypeptide . Hence , when ATP hydrolysis is coupled to molecular motion during an unfolding power stroke , we imagine that ClpX's grip on the Gly12 sequence is insufficient to resist the opposition of the folded domain , causing an unproductive power stroke in which the pore-1 loops slip and fail to advance the substrate tail . In support of this model , we find that poor grip correlates with substantial increases in the ATP cost of degradation for the position-4 and Tyr-scan variants , an indication of slipping and futile power strokes . For example , degradation of one molecule of the Tyr-3 , Tyr-4 , and Tyr-5 substrates required hydrolysis of an average of ~180–240 ATPs , whereas degradation of the Tyr-2 or Tyr-6 substrates required hydrolysis of ~1900 and~3700 ATPs , respectively . These findings corroborate a previous report of futile power strokes during unsuccessful unfolding of a difficult substrate by ClpXP ( Kraut , 2013 ) . Furthermore , substrate-tail contacts with the axial pore that stimulate ATP hydrolysis by ClpX do not fully overlap with the contacts that determine grip . A Tyr-scan experiment shows that tail-position 4 is most important for grip , with flanking positions showing diminishing effects . We expect that amino-acid substitutions at positions 3 and 5 would show side-chain grip trends similar to those observed at position 4 . This is not true at tail-position 1 , where Tyr did not improve grip but Ala did , perhaps because this part of the tail interacts with different residues in ClpX or the folded GFP domain than downstream positions . A single Ala at tail-position four is gripped poorly but a second Ala at certain positions can improve unfolding/degradation . Contacts between the second Ala and the ClpX pore may contribute to stronger grip . Alternatively , the second Ala might affect ClpX contacts made by the first Ala by altering the substrate conformation . In our GFP substrate with a single Tyr at tail-position 4 , this side chain is likely to contact a pore-1 loop close to the folded GFP domain . In an extended chain , four residues would span ~12 Å , a distance that modeling suggests would allow interaction with either the highest or second highest pore-1 loop of ClpX ( Figure 5A; Puchades et al . , 2017; X . Fei , personal communication ) . This is also an area where the axial channel is most tightly constricted around substrate . The distribution of Tyr-effects at positions 2–6 could reflect interactions with different pore-1 loops or the probability that Tyr side chains at different positions contact one specific pore-1 loop . Optical trapping studies indicate that 5–8 residues are moved by a single ClpXP power stroke ( Aubin-Tam et al . , 2011; Maillard et al . , 2011 ) . Thus , once the Tyr side chain at position 4 is engaged by a pore-1 loop , one successful translocation event probably unfolds GFP . Indeed , although many unsuccessful power strokes and ATP hydrolysis events occur while ClpXP is attempting to unfold a stable domain , hydrolysis of a single ATP ultimately results in unfolding . We find it notable that the Tyr-scan distribution is only two residues wide at half height . Hence , one tyrosine at position 4 mediates robust unfolding , whereas one tyrosine at position 7 has no effect . If a position-4 side chain can contact a pore-1 loop high in the ClpX pore , then a position-7 side chain should be able to contact another pore-1 loop lower in the pore . If this model is correct , then it implies that physical contacts between substrate tail residues and the upper pore-1 loops of ClpX are far more important for grip than interaction with the lower loops . We can imagine several different mechanisms for the asymmetry in grip between pore-1 loops in the upper and lower sections of the ClpX pore . In one model , the stronger grip of upper pore-1 loops occurs because these loops maintain relatively static interactions with substrate throughout a power stroke ( Figure 5B , left ) . Several translocation models have been recently proposed for AAA+ unfoldases in which ATP-bound subunits with pore-1 loops oriented near the top of the pore move together as a rigid unit in response to ATP hydrolysis in a lower subunit ( Monroe et al . , 2017; Puchades et al . , 2017; de la Peña et al . , 2018; Dong et al . , 2019 ) . Furthermore , a previous study demonstrated that pore-1 loop mutations disrupt substrate unfolding most dramatically when they are in neighboring ClpX subunits , consistent with grip mediated by a clustered subset of pore-1 loops ( Iosefson et al . , 2015 ) . Alternatively , the substrate tail in the pore could absorb some unfolding force through elastic expansion , diverting part of the energy of each power stroke away from unfolding ( Figure 5B , right ) . Substrate interactions with the uppermost pore-1 loops would minimize the expansion length of the substrate tail , whereas interactions with lower pore-1 loops would allow the tail to absorb more force . In an otherwise all-Gly cassette context , we find that Val , Ile , Leu , Met , Phe , and Tyr at tail-position 4 all promote reasonable levels of grip . If we think of poly-Gly as a smooth and relatively featureless rope , then these larger and generally non-polar side chains can be viewed as knots in the rope that afford better grip . However , grip is not a simple function of side-chain size . For example , Trp supports slower GFP unfolding than the better-gripped residues , suggesting that there may be an upper limit on the size of a side chain that can be efficiently gripped . Polar atoms , especially those close to the peptide backbone of the substrate , weaken grip . For example , Val is one of the best residues in terms of grip , whereas Thr , which differs only by substituting a hydroxyl for a methyl group , barely supports unfolding . Similarly , Ser alone does not support GFP unfolding , but removing the hydroxyl group ( Ala ) or substituting a less-polar thiol group ( Cys ) restores low-level unfolding activity . ClpXP may grip polar side chains less tightly because oxygen or nitrogen atoms bearing partial or full charges are not fully solvated when they are in productive contact with a pore-1 loop and thus incur an energetic penalty . Our finding that large hydrophobic and aromatic side chains are gripped well by ClpX is consistent with a model in which van der Waal’s or hydrophobic interactions between the pore-1 loops and specific side chains in the tail are largely responsible for grip . Several recent cryo-EM structures of AAA+ proteases and protein-remodeling motors reveal a spiral arrangement of subunits in which aromatic residues in the pore-1 loops interact with substrate side chains spaced two residues apart ( Monroe et al . , 2017; Gates et al . , 2017; Puchades et al . , 2017; de la Peña et al . , 2018; Majumder et al . , 2019; Dong et al . , 2019; White et al . , 2018 ) . Our observation that substrate tails with two tyrosines can in some cases specifically enhance grip when spaced by a multiple of two residues is consistent with these structures . However , substrates with two Val residues do not exhibit the same periodic grip enhancement as Tyr , and multiple Ala residues together promote strong grip regardless of their relative spacing . It is clear that nonspecific interactions between the axial pore and substrate side chains are sufficient to promote strong grip independent of precise pore-1 loop intercalation . In specific cases , periodic side chain intercalation could enhance grip for aromatic side chains through the establishment of π-stacking networks with the pore-1 loop Tyr residues , possibly by optimizing the bound substrate conformation . A recent cryo-EM structure of the AAA+ motor NSF , which disassembles SNARE complexes following vesicle fusion , contains well-resolved density for substrate side chains , revealing interactions with the pore-1 loop tyrosine ( White et al . , 2018 ) . Although NSF disassembles SNARE complexes in a single round of ATP turnover in a mechanism distinct from ClpX ( Ryu et al . , 2015 ) , the structural similarities between NSF and many AAA+ unfoldase proteases suggests a common mode of substrate interaction and grip . The assembled SNARE complex is remarkably stable , and NSF likely requires strong grip to disassemble the complex . Consistent with our biochemical observations for ClpX , this structure indicates that the strongest substrate contacts are formed with pore-1 loops high in the upper ring of NSF , and that two substrate residues ( Met and Leu ) that are gripped well by ClpX participate in these interactions . A previous study found that a Gly15 sequence placed between GFP and an ssrA tag did not slow ClpXP degradation ( Barkow et al . , 2009 ) . However , their substrate contained six additional residues ( Thr1-His2-Gly3-Met4-Asp5-Glu6 ) between folded GFP and the Gly15 sequence . As we find that Met at position four supports robust unfolding , it is likely that interactions with the extra sequence mediate unfolding of the Gly15 substrate . Our findings suggest that evolutionary placement of tail residues that are gripped well by ClpX may tune degradation of substrates that unfold non-cooperatively or that have multiple-folded domains . Complete , processive unfolding of multi-domain substrates depends critically on interactions between ClpX and the peptide-tail remnants from unfolding and degradation of the previous domain . For example , ClpXP degradation of Domain III of C . crescentus DnaX is inhibited by a Gly-rich sequence between Domains III and IV , which acts as a partial processing mechanism essential for DNA replication ( Vass and Chien , 2013 ) . Gly-rich tracts also inhibit unfolding/degradation of E . coli DHFR , although multiple alanines in the tail do not improve degradation of this substrate ( Too et al . , 2013 ) . Thus , ClpX unfolding of different native substrates probably requires different degrees of grip strength , which could be mediated by more and/or better-gripped amino acids adjacent to the folded domain . ClpXP contains just two types of subunits , whereas the 26S proteasome consists of more than 30 subunit types ( Budenholzer et al . , 2017 ) . Nevertheless , our work is reminiscent of and reinforces studies of proteasomal degradation by Matouschek and colleagues . For example , they find that low-complexity sequences primarily composed of Gly , Ser , or Thr residues can inhibit proteasomal degradation ( Tian et al . , 2005 ) ; these residues individually are also insufficient to promote ClpXP degradation of GFP . Similarly , sequences that include Phe and Tyr residues can improve or rescue degradation by both the proteasome and ClpXP . These similarities may arise because the Rpt1-6 unfolding ring of the proteasome , despite containing six distinct subunits , has pore-1 loops very similar to those of ClpX . Given the structural similarities between many AAA+ protein remodeling machines , we expect that the principles underlying grip in ClpX reflect those of the broader family . An expression plasmid containing E . coli ClpP and E . coli ClpXΔN was constructed by cloning ClpP into the open reading frame downstream of the pBAD promoter in pBAD18 ( Guzman et al . , 1995 ) . A second ribosome binding site ( 5’-CAAGGAGAATAACG-3’ ) and the ClpX∆N coding sequence ( residues 62–424 ) was added downstream of the ClpP stop codon to produce a polycistronic expression construct . GFP substrates for cytoplasmic degradation assays were cloned downstream of the constitutive insulated ProD promoter in pSB3C5 ( Davis et al . , 2011 ) . His6-GFP substrates for purification were cloned into a pET4b derivative downstream of the pT7 promoter . For all substrates , the 12-residue variable cassette was encoded on an oligonucleotide and introduced upstream of a partial ssrA degron ( Gly-Ser-Glu-Asn-Tyr-Ala-Leu-Ala-Ala ) using PCR mutagenesis . The seven C-terminal residues of this degron are identical to those of the ssrA tag , but we removed the N-terminal part of the ssrA tag to preclude potential SspB inhibition ( Hersch et al . , 2004 ) . T7 Express ΔclpA ΔclpP ΔclpX was generated from the E . coli strain T7 Express ( New England Biolabs ) . The bicistronic clpP–clpX locus was removed by using lambda red recombineering ( Yu et al . , 2000 ) to replace the locus with an FRT-KanR cassette , which was subsequently removed by FLP recombinase expression . ClpA::FRT-KanR was then transduced into this strain with P1 phage from a ClpA::FRT-KanR strain in the Keio collection ( Baba et al . , 2006 ) , and the resistance marker was again removed with FLP recombinase . Modification of the correct loci was verified by PCR at each step in strain construction . His6-GFP-cassette-ssrA constructs were expressed as described ( Kim et al . , 2000 ) and purified by Ni-NTA affinity , Source 15Q anion exchange , and Superdex 200 size-exclusion chromatography . Purified substrates were assessed to be >99% pure by SDS-PAGE and were stored in 25 mM HEPES-KOH ( pH 7 . 5 ) , 150 mM KCl , 10% glycerol , and 500 μM dithiothreitol . The ProD-GFP plasmid encoding each substrate was transformed into T7 Express ΔclpA ΔclpP ΔclpX cells carrying either pBAD18 ( ClpP/ClpXΔN ) or pBAD18 ( null ) . After overnight growth at 30°C on LB agar plates supplemented with 100 μg/mL ampicillin and 34 μg/mL chloramphenicol , single colonies were picked into 5 mL of the same medium and antibiotics and cultures were grown overnight at 30°C . At the start of degradation assays , 50 µL of culture of either the ClpP/ClpXΔN expression strain or the null control strain for each substrate was inoculated into fresh 5 mL LB plus antibiotics and grown at 37°C to OD600 0 . 7–1 . 0 . The cultures were then centrifuged; resuspended at OD600 1 . 2 in fresh media plus antibiotics; and 500 μL was added to 1 mL of fresh media plus antibiotics supplemented with 120 mM L-arabinose , for a final concentration of 80 mM L-arabinose and OD600 of 0 . 4 . After 35 min of growth at 37°C , 1 mL of culture was removed , centrifuged , and resuspended in 600 μL of phosphate buffered saline ( pH 7 . 4 ) . Three 150 μL technical replicates of resuspended cells were transferred to wells of a clear-bottom black 96-well plate ( Greiner ) . Both the GFP fluorescence of the cell resuspension ( excitation 467 nm , emission 511 nm ) and the optical density ( absorbance 600 nm ) were measured on a SpectraMax M5 plate reader ( Molecular Devices ) . The GFP fluorescence for each ClpX∆NP sample and control sample was divided by the measured cell density to give normalized flprotease and flcontrol values respectively . Fraction degraded was calculated as:1− ( flprotease/flcontrol ) Each degradation assay was performed independently in multiple biological replicates , and the calculated value of fraction degraded was averaged across biological replicates . No obvious outliers were observed , and all values were included in the subsequent analysis . Degradation assays were performed at 37°C in 25 mM HEPES-KOH ( pH 7 . 5 ) , 5 mM MgCl2 , 200 mM KCl , 10% glycerol , with 0 . 1 μM ClpX∆N ( hexamer ) , 0 . 3 μM ClpP ( 14-mer ) , 5 mM ATP , 32 mM creatine phosphate ( Roche ) , and 0 . 08 mg/mL creatine kinase ( Millipore-Sigma ) . For the Thr-4 substrate , assays were performed with 0 . 5 μM ClpX∆N ( hexamer ) and 1 . 5 μM ClpP ( 14-mer ) to measure degradation rates more accurately and facilitate comparison with substrates that were degraded more rapidly . Degradation rates were measured by decrease in fluorescence ( excitation 467 nm; emission 511 nm ) in 20 μL reactions on a SpectraMax M5 plate reader . To control for signal loss from photobleaching , a parallel set of reactions was measured for each substrate without ClpX∆N or ClpP , and changes in GFP fluorescence in this experiment were subtracted from those in the degradation reaction . Each measurement included three technical replicates measured together in parallel , and the average values of these replicates were fit to a hyperbolic equation to determine KM and Vmax . Three independently-conducted biological replicates were performed in this manner for each substrate to determine average values ( ± S . D . ) for KM and Vmax . No obvious outliers were observed , and all values were included in the subsequent analysis . ATP hydrolysis rates were measured using a coupled-NADH oxidation assay as described ( Martin et al . , 2005 ) . Degradation efficiency ( ATP hydrolyzed per substrate degraded ) was calculated by dividing the rate of ATP hydrolysis at a near saturating substrate concentration ( 15 μM ) by Vmax for substrate degradation .	Proteins are the workhorses of the body , fulfilling many roles essential for life processes . These molecules are made up of hundreds or thousands of small units called amino acids , which attach to each other to form a long chain . The exact sequence of amino acids determines how the protein will then fold to acquire its final , three-dimentional shape . Enzymes called proteases can degrade unneeded or faulty proteins so that the amino acids can be recycled . For instance , in bacteria , the AAA+ protease ClpXP can recognize and ‘grab’ specific patterns of amino acids at the ends of a protein . This molecular machine then tugs on the segment and unfold the protein , the way a ball of yarn unwinds when pulled from one end . The unfurled protein is then fed into a different section of ClpXP , where it is chopped into short segments for recycling . ClpXP is the best-characterized enzyme amongst AAA+ proteases . However , it is still unclear how it can grip target proteins tightly enough to allow unfolding . To investigate , Bell et al . attached different patterns of 12 amino acids to the end of a folded protein . How well ClpXP grasped each of these proteins was then measured in bacteria and in test tubes . This revealed that ClpXP attaches to six to eight amino acids at a time , suggesting that only part of the enzyme clasps on the protein . Large amino acids are better gripped than small amino acids , similar to how a knotted string is easier to hold than a smooth rope . Amino acids that are electrically charged also interfere with ClpXP attaching to the protein . Finally , ClpXP grasps multiple amino acids at the same time , which dramatically increases grip strength . Many proteins , including some found in viruses , use ‘slippery’ patterns of amino acids to avoid being gripped and unfolded by proteases . By understanding how different patterns of amino acids are grasped , it may someday be possible to engineer enzymes able to target dangerous proteins .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2019	Interactions between a subset of substrate side chains and AAA+ motor pore loops determine grip during protein unfolding
Horizontal gene transfer ( HGT ) and gene loss result in rapid changes in the gene content of bacteria . While HGT aids bacteria to adapt to new environments , it also carries risks such as selfish genetic elements ( SGEs ) . Here , we use modelling to study how HGT of slightly beneficial genes impacts growth rates of bacterial populations , and if bacterial collectives can evolve to take up DNA despite selfish elements . We find four classes of slightly beneficial genes: indispensable , enrichable , rescuable , and unrescuable genes . Rescuable genes — genes with small fitness benefits that are lost from the population without HGT — can be collectively retained by a community that engages in costly HGT . While this ‘gene-sharing’ cannot evolve in well-mixed cultures , it does evolve in a spatial population like a biofilm . Despite enabling infection by harmful SGEs , the uptake of foreign DNA is evolutionarily maintained by the hosts , explaining the coexistence of bacteria and SGEs . Horizontal Gene Transfer ( HGT ) , the transmission of genetic material between unrelated individuals , is a major factor driving prokaryotic evolution ( Ochman et al . , 2000; Doolittle and Zhaxybayeva , 2009; Vogan and Higgs , 2011 ) . Recent estimates of the rate of HGT in closely related bacteria are staggeringly high ( Iranzo et al . , 2019; Sakoparnig , 2019 ) , with HGT possibly even outpacing gradual sequence evolution ( Hao and Golding , 2006; Puigbò et al . , 2014; Vos et al . , 2015 ) . Combining this with the fact that prokaryotes adapt mostly through rapid gene loss ( Kuo and Ochman , 2009; Morris et al . , 2012 ) , bacterial adaptation appears to be mainly driven by changes in gene content ( Snel et al . , 2002; Treangen and Rocha , 2011; Nowell et al . , 2014 ) . Rather than waiting for rare beneficial mutations to arise , taking up tried-and-true genes from a shared ‘mobile gene pool’ allows bacteria to adapt quickly to different ecological opportunities ( Jain et al . , 2003; Wiedenbeck and Cohan , 2011; Casacuberta and González , 2013; Mell and Redfield , 2014; Niehus et al . , 2015; Lopatkin et al . , 2016 ) . Indeed , many bacterial species show patterns consistent with this rapid turn-over of genes , where strains from a single niche contain a relatively small set of genes , while the set of genes found by sampling strains from various niches ( i . e . the pan-genome ) is much richer ( Welch et al . , 2002; Lefébure and Stanhope , 2007; Touchon et al . , 2009; Kim et al . , 2015 ) . Hence , genes appear to be rapidly lost from any individual lineage , but are retained in a much larger gene pool through HGT . When considering the effects of HGT on gene content , it is important to note that HGT can add novel genes to the genome . This might happen through plasmid transfer ( i . e . conjugation ) , but also through recombination . For example , when bacteria take up DNA from their environment ( i . e . transformation ) , genes inserted between two homologous regions may be integrated into the genome . The DNA that has been taken up may also carry mobile genetic elements ( MGEs ) , which can integrate into the genome without a requirement for sequence similarity . Such processes have been coined ‘additive HGT’ ( Thomas and Nielsen , 2005; Choi et al . , 2012; Soucy et al . , 2015 ) , which is distinct from ‘replacing HGT’ because of its ability to copy genes from one individual to another . On the one hand , DNA uptake may be beneficial for the cells , as it allows adaptations to new environments ( Casacuberta and González , 2013; Mell and Redfield , 2014; Lopatkin et al . , 2016 ) or the recovery of lost genes ( Vogan and Higgs , 2011; Takeuchi et al . , 2014 ) . On the other hand , DNA uptake also poses a risk in the form of Selfish Genetic Elements ( SGEs ) such as transposons , and may furthermore cause chromosome disruptions and cytotoxicity ( Baltrus , 2013 ) . Finally , the translocation of DNA and the required state of competence generally comes at energetic and metabolic costs for the cell in terms of expressing and operating DNA uptake machinery ( Ephrussi-Taylor and Freed , 1964; Bergé et al . , 2017; Villa et al . , 2019 ) . Hence , while picking up genes can be very beneficial for bacteria when adapting to a new environment , taking up foreign DNA is also a costly and highly risky endeavour ( Vogan and Higgs , 2011; Baltrus , 2013 ) . Given these disadvantages , is the uptake of DNA ever adaptive for bacteria when the environment does not change ? Can HGT be considered an evolved trait of bacteria , or is it only a side-effect of other unrelated processes like infection by SGEs or DNA repair ( Redfield , 2001 ) ? To address these questions , we here present and analyse a model of a bacterial population undergoing HGT of a single gene . We consider HGT through the uptake of genes from a shared pool of DNA , that is bacterial transformation . We assume that HGT is a costly process for the host cells , and that these costs are proportional to the rate of uptake . Such costs may reflect energetic costs for translocating DNA , growth impediments during the state of natural competence , or the various risks of incorporating foreign DNA into the genome . We show that this form of HGT has a positive impact on population growth rates by recovering slightly beneficial genes , which are hard to maintain in the population through selection alone . Based on whether or not the genes are lost from the population without HGT , and whether HGT can improve the population growth rate , we find that genes fall into one of five gene classes: ( i ) indispensable genes , that are never lost from the population , and for which HGT is therefore unnecessary and deleterious , ( ii ) enrichable genes , that are not lost from the population , but enriching the genes via HGT can nevertheless improve growth rates , ( iii ) rescuable genes , which are lost from the population without HGT , but can be rescued by HGT which improves population growth rates , ( iv ) unrescuable genes which are also lost from the population without HGT , but recovering them with HGT does not improve growth rates , and ( v ) selfish genetic elements , which confer a fitness penalty but can persist through HGT . For enrichable and rescuable genes , where HGT can increase population growth rates , we also investigate if HGT , that is the ability of cells to take up DNA , can evolve de novo . While the bacteria readily evolve to use HGT for enrichable genes , having sufficient donor cells to interact with , evolving HGT to ‘rescue’ rescuable genes faces a problem: HGT is needed for the gene to persist in the population , but sufficient donor cells are required to make HGT adaptive . This paradox is however resolved in a spatially structured population like a biofilm , as even a minority of donor cells can be locally abundant , giving rise to a localised ‘gene-sharing’ community that eventually overgrows the whole population . Finally , in this spatial eco-evolutionary context , HGT is evolutionarily maintained even when exploited by harmful genetic parasites , resulting in stable coexistence of bacteria and SGEs . Our model provides important insights and search images for how slightly beneficial genes may spread , or fail to spread , in an evolving microbial population . To better understand the impact of HGT , we next study how HGT impacts the population growth rate ( ϕ ) . The population growth rate in steady state is given by Equation 1 displayed below ( see full derivation in Supplementary Section 1 ) . The function is comprised of two parts; one where the population consists only of non-carriers ( if h≤l-b ) , and one where carriers survive and the gene persists within the population ( if h > l−b ) . When the gene persists , an optimal growth rate is found at ho⁢p⁢t=b⁢l/c-b ( see Supplementary material ) . ( 1 ) ϕ∗ ( h ) ={1−chifh≤ ( l−b ) ( genecannotpersist ) 1−ch+b−blb+hifh> ( l−b ) ( genepersists ) . By analysing Equation 1 , we find that we can distinguish distinct classes of genes depending on ( i ) whether HGT is required for the gene to persist within the population , and ( ii ) whether HGT is beneficial for the population growth rate ( Figure 2B ) . When genes are highly beneficial ( b > l/c ) , HGT is not required for the gene to persist , and HGT does not improve the population growth rate . In other words , although transferring these indispensable genes yields a small increase in the number of carrier cells , this does not outweigh the costs of HGT . When considering lower values of b , HGT is still not required for the gene to persist within the population , but transferring these enrichable genes is nevertheless beneficial for population growth rates . For even lower benefit ( b < l ) , HGT is a necessity for the gene to persist within the population , but the population growth rate can be improved by means of intermediate rates of HGT . We call these genes rescuable genes . If we consider genes with even smaller fitness effects ( b < 4cl/ ( 1+c ) 2 ) , HGT is still required for the survival of these genes , but the population growth rates are highest in the absence of HGT . Thus , despite being defined as a beneficial gene ( b > 0 ) , transferring these unrescuable genes is not beneficial . Finally , we can consider SGEs , genes with a negative effect on fitness ( b < 0 ) . These genetic parasites can only persist in the population at very high rates of HGT , but are of course never beneficial for the population growth rate . Figure 2C shows a bifurcation diagram that summarises how increasing or decreasing rates of HGT impact the population growth rate for these different classes . By analysing the simple model of cells undergoing HGT , we have found five distinct gene classes . For two of these classes , namely enrichable and rescuable genes , moderate rates of HGT improve the population growth rates . We next study ( i ) whether HGT of enrichable and rescuable genes is an evolutionarily stable strategy , and ( ii ) if bacteria can evolve this strategy de novo . To answer these questions , we consider two competing species: one that takes up DNA , and one that does not ( H⁢G⁢T+ and H⁢G⁢T- respectively , see Figure 1B ) . With this model , we have studied the evolution of DNA uptake by means of adaptive dynamics ( Metz et al . , 1995 ) . If H⁢G⁢T- cannot invade H⁢G⁢T+ , we call HGT an evolutionarily stable strategy , and if H⁢G⁢T+ can invade H⁢G⁢T- we call HGT evolvable . We found that HGT is an evolutionarily stable strategy for both enrichable and rescuable genes , but that HGT is evolvable only for enrichable genes ( see Supplementary Material for full analysis ) . In other words , invasion of H⁢G⁢T+-mutants is not possible with respect to rescuable genes . Even when we assume that the invading H⁢G⁢T+-mutant has the optimal rate of HGT , it cannot invade into a population of H⁢G⁢T- cells in steady state . These results were confirmed by numerical analysis , which indeed shows that H⁢G⁢T+ only invades when the founding population size of H⁢G⁢T+ ( C+/N+ ) is relatively large ( see Figure 3A ) . This failure to reach the alternative ( fitter ) evolutionary attractor is caused by positive frequency-dependent selection ( known as the Allee effect ) . Invading mutants , that is a small population of H⁢G⁢T+ cells , contain few carrier cells to act as donors for HGT . Moreover , since the resident population of H⁢G⁢T- is also not able to retain the rescuable genes , the resident population can also not serve as a donor ( see Figure 3B ) . As such , the costs of HGT for an invading H⁢G⁢T+-mutant do not outweigh the potential benefits . In summary , while HGT is an evolutionarily stable strategy , cells cannot evolve HGT to ‘rescue’ rescuable genes . So far , we have studied a well-mixed population of cells that undergoes all-against-all competition , and found that HGT is advantageous for slightly beneficial genes that ( i ) are not too beneficial , as these genes readily persist within the population without HGT , and ( ii ) are beneficial enough to compensate for the costly HGT . Next , we study the same dynamics of carrier and non-carrier cells in a spatially explicit , eco-evolutionary context . We do this by implementing an individual-based model ( IBM ) , where bacterial cells reside on a grid , interactions are local , and events like HGT and gene loss are implemented as stochastic processes ( see Materials and methods and Figure 1C ) . When the cells on this grid are sufficiently mixed each time step , the IBM should approximate the dynamics of the ODE model . However , when cellular mixing is minimal , the resulting spatially structured population is more analogous to that of a biofilm . What is the effect of this spatial structure ? We first analysed the IBM for a wide variety of values for b and h , and measured the average growth rates ϕ in the population . We can thus evaluate whether the aforementioned gene classes ( indispensable , enrichable , rescuable , unrescuable genes , and SGEs ) are found under the same conditions as in the ODE model . Figure 4A shows that , when the IBM is well-mixed , the gene classes indeed occur at values of b identical to the ODE model . However , the gene classes shift to higher values of b when mixing is decreased , making the range of benefits in which genes are classified as enrichable and rescuable much broader . In these biofilm populations , HGT was indeed found to be evolutionarily stable for this wider range of fitness-effects ( black outline ) , illustrating that it is not only the value of b , but also the ecological context in which a gene finds itself that determines whether or not HGT is adaptive . What causes these gene classes to shift depending on this spatial context ? How does an enrichable gene in the well-mixed system become rescuable in the spatially structured population , as though it is less beneficial ? Figure 4B shows how this can be intuitively understood by taking into account how , with low mixing , individuals in a spatial system mostly compete with their own kind ( i . e . progeny and conspecifics ) . Even when the majority of the population consists of non-carriers , carriers are still competing mostly with other carrier cells . Thus , the effective benefit of carrying the gene is lower in a biofilm , hence the gene becomes harder to maintain within the population . In Figure 4C is shown that , while carrier cells in well-mixed population experience a competitive advantage of 2% when carriers make up approximately half the population , carriers in a biofilm only reach a similar competitive advantage at very low carrier frequencies , that is when the carriers are almost extinct . At this point , the gene will readily be lost stochastically . The hampered ability of spatially structured populations to retain slightly beneficial genes , indeed changes how the population growth rate depends on the rate of HGT ( Figure 4D ) . The results described in the previous section illustrate that HGT ( i . e . the uptake of DNA ) is an evolutionarily stable strategy for a much broader range of b-values ( fitness effects of genes ) in a spatially structured population than in a well-mixed culture . Many more genes are furthermore classified as rescuable in these spatially structured populations , meaning that they can only persist through HGT . We have concluded in a previous section that HGT cannot evolve to ‘rescue’ these rescuable genes in populations that are well-mixed , fully deterministic , and by only considering a single H⁢G⁢T+ mutant type at a time . In the IBM on the other hand , the population is spatially structured , events are stochastic , and each individual cell has its own rate of DNA uptake . Can these different assumptions help to alleviate the Allee effect mediated by a lack of donor cells , which prevents the de novo evolution of HGT ? To answer the question posed above , we allowed the HGT-rate h ( i . e . the costly uptake of DNA ) of all individuals in the IBM to evolve ( see Materials and methods ) . When a non-carrier interacts with a ( local ) carrier , the h-value of this non-carrier ( i . e . the recipient ) determines the probability of accepting the gene . For simplicity , we will call individuals with an h-parameter greater than 0 . 02 H⁢G⁢T+ , and the others H⁢G⁢T- . We start with a non-carrier population of H⁢G⁢T- cells ( with h=0 . 00 ) , simulate this population for some time ( 20 , 000 time steps ) , and then allow cells to sporadically discover rescuable genes . Since rescuable genes cannot persist without HGT , the fate of a recently discovered gene depends on the ability of cells to engage in ( local ) HGT . Using this protocol , we investigated if the rescuable gene is able to spread through the evolution of HGT . We found that HGT never evolved for rescuable genes in well-mixed populations ( Figure 5A ) , consistent with our prior results in the well-mixed ODE model . Thus , we can conclude that the level of stochasticity in the IBM is insufficient to overcome the aforementioned Allee effect caused by a lack of donor cells . In the spatially structured population , HGT of rescuable genes does in fact emerge , therewith ‘rescuing’ the rescuable genes ( Figure 5B ) . Interestingly however , we found that HGT did not always evolve immediately after the influx of rescuable genes started ( yellow arrow ) , but nevertheless spread steadily once attained . To further elucidate the spread of genes , we barcoded each newly discovered gene with a unique ID , and visualised these on the spatial grid with different colours ( Figure 5C ) . Initially , rescuable genes fail to invade , even though different barcodes may locally persist for a while ( episode I ) . After some time , however , one gene ( green ) manages to persist within a local community of transferring cells ( episode II ) . This sets in motion a positive feedback mechanism , where the local abundance of the green gene alleviates the lack of donor cells , transforming nearby H⁢G⁢T+-mutants into carriers , and so on ( also see Video 1 ) . This emergent ‘gene-sharing’ community eventually overgrows the other cells , and the rescuable gene ultimately persists in up to 70% of the population . After the influx of rescuable gene is stopped ( episode III ) , the gene readily persists within the population , showing how this transferring community does not depend on the continuous influx of genes . In summary , HGT of rescuable genes through DNA uptake can only evolve if transfer happens within spatially localised sub-populations , and not under well-mixed conditions modelled by mass-action . Through a local ‘nucleation event’ , communities can reach the alternative stable state that can maintain the rescuable gene . Figure 5D summarises the outcome of de novo HGT evolution for a broad range of genes ( b-values ) with different levels of mixing , revealing how the uptake of DNA evolves for many more genes in a spatially structured population . While HGT of enrichable genes always evolved , HGT only evolved for rescuable genes in spatially structured populations . Finally , as expected from prior results , HGT never evolved for indispensable and unrescuable genes . We have shown that HGT can be adaptive and evolvable for bacteria in order to enrich or rescue slightly beneficial genes . We next investigated if HGT can be maintained under the pressure of harmful SGEs , genetic parasites that spread through horizontal transfer . Our earlier analysis has shown that , when the rate of HGT cannot evolve , SGEs ( genes with b<0 ) can persist within the population as long as h>l-b . However , when the rate of HGT is allowed to evolve , bacteria may lower their rate of DNA uptake to avoid these genetic parasites . Therefore , we next investigate whether bacteria in the IBM will maintain their ability to take up foreign DNA in the presence of SGEs . For this , we consider a population that evolved to engage in HGT of a rescuable gene ( b=0 . 03 ) , and expose this population to a low influx of SGEs which confer a fitness penalty ( β ) . We study if these SGEs , despite their fitness penalty , can persist within this bacterial population , and if HGT is evolutionarily maintained by the hosts . Figure 6A shows that , when the fitness penalty of the SGEs is small relative to the benefit of the rescuable gene ( hereafter called ‘weak SGEs’ , β=0 . 01 ) , these genetic parasites quickly rise to very high frequencies within the population . Although the host cells gradually evolve lower HGT rates in response ( from h≈0 . 05 it stabilises around h≈0 . 04 , also see Appendix 1—figure 5 ) HGT , the rescuable gene , and the SGEs are evolutionarily maintained . When the influx of SGEs is stopped , the cells ( and their beneficial gene ) stably coexists with these genetic parasites . Strikingly , if we introduce SGEs whose fitness penalty is greater than the benefit of the gene ( ‘strong SGEs’ , β=0 . 04 ) , we also observe the coexistence of cells , rescuable genes , and SGEs . By looking at the initial invasion dynamics ( Figure 6B ) , we can see that these strong SGEs cannot rise to very high frequencies . As the hosts evolve lower rates of DNA uptake , these genetic parasites are pushed to very low frequencies . However , the reduced threat of genetic parasites causes the host cells to once again increase their rates of DNA uptake , leading to a secondary outbreak of SGEs ( Figure 6B , from T = 300 , 000 onwards ) . It is interesting to note that , while the population growth rates ( ϕp⁢o⁢p ) clearly decrease due to this second infection , the growth rates along the line-of-descent ( ϕl⁢o⁢d , see Materials and methods ) remains largely unaffected . Thus , while a sub-set of the population has been infected , individuals in this infected strain will not be amongst the long-term ancestors . Counter-intuitively , strong SGEs only have a minor impact on bacterial growth rates , as they are only retained at very low frequencies . Weaker SGEs instead impose a significant burden on the population by rising to much higher frequencies . Indeed , when SGEs are very costly , the population purges them entirely by evolving lower rates of DNA uptake ( see Appendix 1—figure 5C ) . Finally , note how stopping the influx of SGEs does not impact the long-term coexistence of cells , beneficial genes , and these strong SGEs ( Appendix 1—figure 5C ) . To better understand the co-evolutionary process between SGEs and bacteria engaging in HGT of rescuable genes , Figure 6D shows long-term dynamics of barcoded SGEs in this spatial system . Although a diverse set of SGEs are initially discovered in parallel ( coloured by their unique barcode ) , eventually only a single barcode remains after the influx of SGEs is stopped . Moreover , it can also be seen how SGEs are either locally abundant , or entirely absent . Thus , spatially separated strains of bacteria experience opposing selection pressures for taking up DNA . Lower rates of uptake are favoured in the presence of these strong SGEs , but higher rates of uptake are favoured when these genetic parasites have ( locally ) died out . Indeed , this heterogeneity of SGEs is crucial for the strong SGEs to persist , as well-mixed populations can only retain weaker SGEs ( see Appendix 1—figure 5 ) . Interestingly , we also found that strong SGEs failed to persist when HGT was too localised ( e . g . only between neighbouring cells ) , as the SGEs then could not escape to a new pool of hosts that have high rates of DNA uptake ( Appendix 1—figure 6 ) . We conclude that , in a spatially structured population , strong SGEs can stably coexist in a bacterial population which maintains its ability to engage in HGT to ‘rescue’ rescuable genes . In nature , HGT can happen through a variety of mechanisms that each have their own potential advantages and disadvantages for the host cell ( Vogan and Higgs , 2011; Baltrus , 2013 ) . Bacteria do not always have full control over the rates at which HGT happens , especially when considering it as a side-effects of other processes ( Redfield , 2001 ) . However , it remains an intriguing question under which specific circumstances bacteria benefit from HGT , whether it is a side-effect or not . By abstracting away from the different mechanisms of HGT , and what it means for a gene to be ‘beneficial’ , we have revealed the conditions under which HGT is an adaptive trait for the host cells . In a similar spirit , earlier modelling by Vogan and Higgs , 2011 has shown that HGT can be adaptive with respect to genes that are frequently lost . However , in their work , natural selection eventually favoured improved replication accuracy , therewith decreasing the advantage of HGT . Other models have shown that HGT is beneficial to mitigate the effects of Muller , 1964 by decreasing assortment load ( Takeuchi et al . , 2014; Vig-Milkovics et al . , 2019 ) , analogous to the impact of sex and recombination on the balance between drift and selection ( Lynch et al . , 1995; Schultz and Lynch , 1997; Lynch et al . , 2016; Vos et al . , 2019 ) . Our work complements these aforementioned studies by showing that , however low the rate of gene loss may be , there may always be a class of slightly beneficial traits for which HGT is adaptive and evolvable . Although genes with such small fitness effects are very hard to detect experimentally ( Bataillon , 2000; Wiser and Lenski , 2015 ) , our model is a proof of principle that HGT may play a key role in preventing the loss of these genes , which may explain the differential rates of HGT as observed in the data ( Nogueira et al . , 2009; Rankin et al . , 2011; Madsen et al . , 2012; Novick and Doolittle , 2020 ) . With the upswing and improvement of experimental techniques like Hi-C metagenomics ( Beitel et al . , 2014; Burton et al . , 2014 ) and DNA barcoding ( Blundell and Levy , 2014; Nguyen Ba et al . , 2019 ) , we will soon have more insights into the eco-evolutionary dynamics of small-effect mutations ( Li et al . , 2018; Lerner et al . , 2019 ) and accessory genes ( Quistad et al . , 2019; Yaffe and Relman , 2020 ) , and we may learn when HGT can come to rescue a microbial population , and when it may be nothing more than a catastrophe . In this work , we study the dynamics of bacteria undergoing HGT of slightly beneficial genes and Selfish Genetic Elements ( SGEs ) . We do this by modelling the same processes with gradually increasing complexity , starting from simple Ordinary Differential Equations ( ODEs ) , and then evaluating the same dynamics in an Invididual-based Model ( IBM ) . A graphical representation of these models is found in the main text ( Figure 1 ) . The models consider the competition between cells of two types: carrier cells ( C ) that carry a gene , and non-carrier cells ( N ) . When carrier cells contain a beneficial gene ( i . e . it is a beneficial trait ) , they grow faster than the non-carrier cells ( N ) . However , carriers may lose this beneficial gene with a fixed rate l . Both cell types take up DNA with rate h , which comes with a cost c . This cost is equal for both cell types , reflecting for example the costs of expressing and operating the DNA uptake machinery , other metabolic burdens of the physiological state of natural competence , or the risks associated with taking up foreign DNA . Proportional to the density of available carrier cells , non-carriers can be transformed back into a carrier cell by means of ‘additive’ HGT . Both models use a chemostat assumption , where cells wash out at a rate proportional to the rate of growth , ensuring a constant population size in steady state . By modelling the dynamics described above by means of ODEs , we assume a well-mixed population of cells that compete according to all-against-all dynamics ( i . e . mass-action ) . Our equations describing the density of carrier ( C ) and non-carrier ( N ) cells are given in Equation 2 , where b is the benefit of the carried gene ( or burden if b<0 ) , l is the rate of gene loss , h is the rate at which cells engage in HGT , c is the continuous cost for engaging in HGT , and HGT transforms a non-carrier into a carrier when they interact ( h⁢C⁢N ) . This cost for HGT ( c ) is equal for both cell types , meaning that whatever the costs may entail , we assume they are continuously payed . Finally , the total amount of growth ( ϕ ) is subtracted from both populations , meaning that the population density in steady state is always 1 . ( 2 ) dCdt= ( 1−ch+b ) C⏟reproductionofC−lC⏟geneloss+hCN⏟HGT−ϕC⏟chemostatdNdt= ( 1−ch ) N⏟reproductionofN+lC⏟geneloss−hCN⏟HGT−ϕN⏟chemostatϕ= ( 1−ch+b ) C⏟totalgrowthofC+ ( 1−ch ) N⏟totalgrowthofNC+N=1 ( constantpopulationsize , ensuredbychemostatassumption . ) From the above model ( Equation 2 ) , we derived how the population growth rate ( ϕ ) depends on both b and h ( see Equation 1 in the main text ) , which shows the conditions under which HGT improves the total growth rate of the population . To analyse whether or not HGT could evolve , we extended the two-variable ODE model above ( of cells with the same h ) to a four-variable ODE model ( of two species with a different h , see Figure 1B and Equation 3 below ) . We use this extension to study whether or not a species with HGT ( C+ and N+ , h>0 ) could invade upon a species without HGT ( C- and N- , h=0 ) , and vice versa ( see Supplementary material for full analysis ) . Finally , we also extended the ODE model to study the impact on growth rates for cells that engage in HGT of both a beneficial gene and a Selfish Genetic Element ( SGE ) , which can be found in the Supplementary Material . ( 3 ) dC−dt= ( 1+b ) C−−lC−−ϕC−dN−dt=N−+lC−−ϕN−dC+dt= ( 1+b−ch ) C+−lC++hN+ ( C−+C+ ) −ϕC+dN+dt= ( 1−ch ) N++lC+−hN+ ( C−+C+ ) −ϕN+ϕ= ( 1+b ) C−+N−+ ( 1+b−ch ) C++ ( 1−ch ) N+ The individual-based model ( IBM ) describes the same dynamics as the ODE models , but differs in some important aspects . Firstly , individuals are discrete entities that live on a 2D grid , and reproduce locally . This allows us to study the model with and without spatial pattern formation by modifying the rate at which cells mix . When mixing is disabled or very limited , a spatially structured population like that of a biofilm will form , while an increased amount of cellular mixing will approximate a well-mixed culture . Under well-mixed conditions , individuals will interact with random individuals in the population ( approximating the all-against-all dynamics of the ODEs ) , while individuals will interact mostly with their conspecifics in case of the biofilm . We explicitly define a competition range ( focal cell plus its eight neighbouring grid points ) and a HGT range ( all cells within distance t ) which determine smaller samples of the total population with which individuals can interact . Each individual ( potentially ) has its own h-parameter , allowing us to study the evolution of HGT in an eco-evolutionary context ( see implementation of mutations below ) . As we primarily focus on the question if cells benefit from taking up genes from their environment or other cells , we assume that the h-parameter of the recipient cell determines the probability of HGT . While we tested if HGT could evolve de novo by studying the invasion critereon for H⁢G⁢T+ in the ODE model , we study the de novo evolution of HGT in the IBM by continuously introducing carriers at a low rate . With a rate f , genes with benefit b* are ( re ) discovered , allowing us to study how and if newly discovered genes/selfish elements spread through the population . Finally , note that processes such as gene loss , HGT , and competition are no longer deterministic like in the ODEs , but implemented as events that can stochastically happen at each simulated time step . To ensure the chance-events in the IBM ( reproduction , HGT , gene loss ) accurately represent the rates as used in the ODE , all probabilities were multiplied by a small constant ΔT = 0 . 1 . Throughout most of this study , the gene loss l was set to 0 . 02 and the cost for HGT was set to c=0 . 2 . In general , our results do not depend on the absolute value of these two parameters . Instead , our results depends on the relationship between these two parameters , meaning that for lower ( arguably more realistic ) rates of gene loss , the relevant parameter range will still be extensive if the costs for HGT are also lower . If the costs are very high , the range where HGT is adaptive ( i . e . enrichable and rescuable genes ) becomes much more narrow , but is nevertheless retained ( i . e . the gene classes discussed in Figure 2 simply shift to lower values of b ) . As these costs may entail a wide variety of different burdens on the cell ( operating and expressing DNA pumps , diverting resources to the physiolocial state of natural competence , the burden of expressing redundant gene copies , chromosome disruptions , cytotoxicity , and the risks of SGEs ) , we do not know quantitative information to judge the relevant value for this cost parameter . Instead , we argue that our model instead gives conceptual clarifications and demonstrations of novel possibilities . Parameters such as the benefit ( b ) , the HGT-rate ( h ) , the amount of mixing ( d ) , and the HGT distance ( t ) have been extensively sweeped , as discussed in the main text/Supplementary Material . In these cases , the used parameters are given in the captions of the relevant figures . When comparing the IBM with the ODE models ( e . g . occurrence of gene classes ) , evolution of h was disabled ( u=0 . 0 ) . For the de novo evolution of HGT , the initial population consisted only of non-carrier cells , but genes fluxed in at a low rate ( f=5⋅10−6 ) , while the initial level of HGT ( h=0 . 0 ) was allowed to evolve with f=5⋅10−5 with a uniform step size of m=0 . 05 . Finally , when testing whether HGT could be maintained , no influx of genes was present ( f=0 . 0 ) , but the initial population consisted of carrier-cells that already engage in HGT ( h=0 . 05 , see supplementary material for evolved rates of HGT ) . All experiments in the IBM with Selfish Genetic Elements were done with slightly lower costs ( c=0 . 1 ) , to compensate for the extra costs imposed by these genetic parasites . All the important parameters of our models are summarised in Table 1 . The analytical model was numerically analysed using grind . R by R . J . de Boer ( http://tbb . bio . uu . nl/rdb ) , an R script that uses the deSolve R-package ( Soetaert et al . , 2010 ) . The simulated model was implemented in Cash ( Cellular Automaton simulated hardware ) version 2 . 1 , an free and easy-to-use library to make simple spatially explicit simulations ( originally created by R . J . de Boer and A . D . Staritsk , further developed by Nobuto Takeuchi and Bram van Dijk ) . Visualisation of both models was done in R using ggplot ( Wickham , 2016 ) and plotly ( Inc PT , 2015 ) . Simulations were run in Linux Ubuntu 16 . 04 LTS using GNU parallel ( Tange , 2018 ) . Both the R-scripts for ODE analysis and the IBM code implemented in C , are available online https://github . com/bramvandijk88/HGT_Genes_And_SGEs ( van Dijk , 2020; copy arhived at https://github . com/elifesciences-publications/HGT_Genes_And_SGEs .	Most animals , including humans , inherit genes from their parents . However , bacteria and other microorganisms can also acquire genes from members of the same generation . This process , called horizontal gene transfer ( HGT for short ) , allows bacteria to quickly adapt to new environments . For example , rather than waiting for rare mutations to arise , bacteria can pick up ‘tried and true’ genes from their neighbours which allow them to exploit new resources or become resistant to antibiotics . But gene sharing comes at a cost . For instance , taking up DNA is an energetically costly process and exposes bacteria to so-called selfish genes which replicate at the expense of other more useful genes in the genome . Given the costs and the threat of selfish genes , it remained unclear whether HGT is still beneficial in a stable environment where no new resources or antibiotics are present . Here , van Dijk et al . used mathematical modelling to examine how gene sharing affects the growth rate of bacterial colonies living in a stable environment . The experiments showed that bacteria are able to take up new sequences of DNA even in the presence of selfish genes . This allows communities of bacteria to retain genes that provide a small benefit that would otherwise be lost from the population , even when taking up DNA imposes a cost upon the individual . van Dijk et al . found that this collective behaviour cannot evolve in well-mixed bacterial populations , but readily emerged in more structured populations , such as biofilms . This work demonstrates how HGT , a key component of bacterial evolution , has allowed bacteria to coexist with harmful selfish genes . It also provides insights into how genes persist and spread through bacterial communities , which has implications for our understanding of antibiotic resistance .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "computational", "and", "systems", "biology" ]	2020	Slightly beneficial genes are retained by bacteria evolving DNA uptake despite selfish elements
Plants have evolved intracellular immune receptors to detect pathogen proteins known as effectors . How these immune receptors detect effectors remains poorly understood . Here we describe the structural basis for direct recognition of AVR-Pik , an effector from the rice blast pathogen , by the rice intracellular NLR immune receptor Pik . AVR-PikD binds a dimer of the Pikp-1 HMA integrated domain with nanomolar affinity . The crystal structure of the Pikp-HMA/AVR-PikD complex enabled design of mutations to alter protein interaction in yeast and in vitro , and perturb effector-mediated response both in a rice cultivar containing Pikp and upon expression of AVR-PikD and Pikp in the model plant Nicotiana benthamiana . These data reveal the molecular details of a recognition event , mediated by a novel integrated domain in an NLR , which initiates a plant immune response and resistance to rice blast disease . Such studies underpin novel opportunities for engineering disease resistance to plant pathogens in staple food crops . Plant diseases are a continuous threat to crop production and a major constraint on achieving food security . Rice blast disease , caused by the fungal pathogen Magnaporthe oryzae , is the biggest pre-harvest biotic threat to global rice production ( Pennisi , 2010; Dean et al . , 2012; Liu et al . , 2014 ) . This disease can cause the loss of enough rice to feed 212–742 million people annually ( Fisher et al . , 2012 ) , and result in up to 100% yield loss in infected areas ( Dean et al . , 2012; Liu et al . , 2014 ) . The sustainability of rice production is critical , it is a staple food crop for greater than half the world's population . Approaches to controlling blast disease have mainly been via the deployment of rice resistance ( R ) genes , which encode intracellular immune receptors known as NLRs ( Nucleotide-binding , Leucine-rich-repeat ( LRR ) Receptors ) . NLRs are a conserved component of plants' innate immune systems and survey the host environment for perturbations caused by invading pathogens ( Dangl and Jones , 2001; Chisholm et al . , 2006; Jones and Dangl , 2006; van der Hoorn and Kamoun , 2008; Dodds and Rathjen , 2010 ) . Most NLRs respond to the presence or activities of translocated pathogen effectors , proteins delivered by adapted pathogens to affect the physiology of the host to benefit the parasite ( Dodds and Rathjen , 2010; Win et al . , 2012; Wirthmueller et al . , 2013 ) . The recognition event often results in a robust immune response and localised cell death , which limits disease caused by biotrophic pathogens on their hosts . Most NLRs comprise a multi-domain architecture with central nucleotide-binding ( NB-ARC ) and C-terminal LRR regions . They also usually contain N-terminal coiled-coil ( CC ) or TOLL/interleukin-1 receptor ( TIR ) domains ( Takken and Goverse , 2012 ) . In at least some cases , NLRs function in pairs to deliver disease resistance , and these pairs can be tightly linked genetically ( Eitas and Dangl , 2010; Cesari et al . , 2014; Le Roux et al . , 2015; Sarris et al . , 2015 ) . The protein:protein interactions that underlie NLR pair function are starting to be elucidated ( Cesari et al . , 2014; Williams et al . , 2014 ) , but many unknowns remain . Interestingly , most NLR pairs studied to date use one of the NLRs to detect the presence of specific effectors by direct binding ( Kanzaki et al . , 2012; Cesari et al . , 2013; Williams et al . , 2014; Zhai et al . , 2014 ) . One mechanism by which this can be achieved is via unconventional integrated domains in the NLRs ( known as integrated decoy or sensor domains ) that show evolutionary relationships to putative virulence targets ( Cesari et al . , 2014; Wu et al . , 2015 ) . Such domains can be integrated at different positions either before , in-between or after the standard NLR regions and are increasingly identified in NLRs of both model and crop plants ( Cesari et al . , 2014 ) . How these unusual integrated domains function in the direct molecular recognition of effectors , and how this results in initiation of immune signalling , are emerging as fundamental questions in plant NLR biology . To date , ∼100 NLRs in rice have been described to confer resistance to strains of M . oryzae , and 23 of these have been cloned ( Liu et al . , 2014 ) . Identification of the pathogen effectors ( also known as AVRs [avirulence proteins] ) that are recognised by these NLRs has lagged behind and only six have been cloned to date , AVR-Pia ( Yoshida et al . , 2009 ) , AVR-Pita ( Orbach et al . , 2000 ) , AVR-Pik ( Yoshida et al . , 2009 ) , AVR-Pii ( Yoshida et al . , 2009 ) , AVR-Piz-t ( Li et al . , 2009 ) and AVR1-CO39 ( Ribot et al . , 2013 ) . AVR-Pia and AVR1-CO39 are recognised by the RGA4/RGA5 NLR pair ( Okuyama et al . , 2011; Cesari et al . , 2013 ) through direct binding to a Heavy-Metal Associated domain ( HMA , also known as RATX1 ) integrated into RGA5 after the LRR ( Cesari et al . , 2013 ) . RGA4/RGA5 physically interact to prevent cell death mediated by RGA4 in the absence of AVR-Pia; the presence of the effector relieves this suppression ( Cesari et al . , 2014 ) . Intriguingly , the NLR pair Pik-1/Pik-2 , which recognises AVR-Pik ( Figure 1 ) ( Ashikawa et al . , 2008; Yoshida et al . , 2009 ) , also binds the effector via an HMA domain but this domain is integrated between the CC and NB-ARC regions of Pik-1 ( Figure 1B ) . The integrated HMA domains of RGA5 and Pik-1 appear to have evolved from a family of rice proteins that only contain the HMA domain ( Cesari et al . , 2013 , 2014; Wu et al . , 2015 ) . Interestingly , the rice protein Pi21 , a disease susceptibility factor , contains an HMA domain that is not part of an NLR protein ( Fukuoka et al . , 2009 ) . 10 . 7554/eLife . 08709 . 003Figure 1 . Schematic representations of ( A ) Magnaporthe oryzae AVR-Pik effector alleles with position of polymorphic residues shown , the effector domain is shown in green with the signal peptide ( SP ) in grey ( amino acids are denoted by their single letter codes ) , ( B ) Rice Pik resistance proteins , highlighting the position of integrated HMA domain in the classical plant NLR architecture of Pik-1 ( CC = coiled coil , HMA—Heavy Metal Associated domain , NB-ARC = Nucleotide-binding Apaf-1 , R-protein , CED4-shared domain , LRR = Leucine Rich Repeat domain ) , domain boundaries are numbered , based on the Pikp sequences . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 003 Both AVR-Pik ( Figure 1A ) and the HMA region of Pik-1 exhibit nucleotide polymorphisms between pathogen isolates and rice cultivars that result in changes at the amino acid level ( Yoshida et al . , 2009; Kanzaki et al . , 2012; Wu et al . , 2014; Zhai et al . , 2014 ) . These changes are most likely associated with co-evolutionary dynamics between M . oryzae and rice , predicted to play out at the molecular level via direct protein:protein interactions ( Kanzaki et al . , 2012 ) . The interaction of AVR-Pik allele AVR-PikD with the Pik-1 NLR Pikp-1 is thought to be the oldest in co-evolutionary time ( Kanzaki et al . , 2012 ) . Cultivars of rice containing the Pikp allele are resistant to M . oryzae isolates expressing AVR-PikD , but are susceptible to pathogen isolates expressing other AVR-Pik alleles ( Kanzaki et al . , 2012 ) . While the structural basis of function and recognition of plant pathogen effectors has advanced in recent years ( Wirthmueller et al . , 2013; Williams et al . , 2014 ) , only a few studies have focused on the M . oryzae/rice system . For example , the only structure known for a M . oryzae effector is that of AVR-Piz-t ( Zhang et al . , 2013 ) , which adopts a six-stranded β-sandwich structure and contains a single disulphide bond . To date , there is no available structural information on domains from rice NLRs , and no structural data from any system showing how plant pathogen effectors are directly recognised at the molecular level by an NLR . To better understand the mechanisms of direct recognition of effectors by NLRs , we have investigated the interaction between the M . oryzae effector AVR-Pik and the rice NLR Pikp-1 . We determined the affinity of interaction of AVR-PikD to the HMA domain of the rice NLR Pikp-1 ( Pikp-HMA ) in vitro , and compared the relative binding of AVR-Pik alleles AVR-PikE , AVR-PikA and AVR-PikC to this HMA . The crystal structure of AVR-PikD bound to Pikp-HMA was determined and this guided mutagenesis of the effector , targeting residues at the interface with Pikp-HMA . The binding of these mutants to the Pikp-HMA was tested in yeast and in vitro . We also used a combination of AVR-Pik alleles and AVR-PikD mutants in both the host rice and heterologous Nicotiana benthamiana systems to probe the degree to which AVR-NLR interactions mediate immunity-related readouts . Long after Harold Henry Flor proposed the gene-for-gene hypothesis of host–parasite interactions ( Flor , 1955 , 1971 ) , our study establishes the structural basis of direct recognition of a pathogen effector by a plant NLR . Previously , the full-length and CC domain ( containing the HMA ) of Pikp-1 have been shown to interact with AVR-PikD in yeast-2-hybrid ( Y2H ) assays ( Kanzaki et al . , 2012; Wu et al . , 2014; Zhai et al . , 2014 ) . Here we show that the Pikp-HMA domain alone selectively interacts with the AVR-Pik allele AVR-PikD in yeast ( Figure 2A , Figure 2—figure supplement 1 , Table 1 ) . Weak interaction was also observed with AVR-PikE ( as evidenced by limited growth on the selective–LTH plate and some blue colouration in the X-gal assay ) , but not AVR-PikA or AVR-PikC , which is consistent with previous experiments ( Kanzaki et al . , 2012 ) . Following expression , purification and verification of the proteins by intact mass spectrometry ( ‘Materials and methods’ , Figure 2—figure supplement 2 , Table 2 ) , we showed that AVR-PikD and Pikp-HMA form a stable complex in vitro that can be purified by analytical gel filtration ( Figure 2B ) . Using this qualitative assay , we also found that Pikp-HMA can form a complex with AVR-PikE and AVR-PikA , but not AVR-PikC ( Figure 2—figure supplement 3 ) . 10 . 7554/eLife . 08709 . 004Figure 2 . AVR-Pik effector alleles interact with the Pikp-HMA domain with different affinities . ( A ) Y2H assays showing the binding of effector alleles to the Pikp-HMA using two read-outs , growth on–Leu-Trp-His+3AT ( -LTH ) plates and the X-gal assay . ( B ) Analytical Gel Filtration traces depicting the retention volume of Pikp-HMA , AVR-PikD and the complex , with SDS-PAGE gels of relevant fractions ( similar results were obtained for AVR-PikE and AVR-PikA , but AVR-PikC did not bind [Figure 2—figure supplement 3] ) . ( C ) Binding curves derived from Surface Plasmon Resonance multi-cycle kinetics data for Pikp-HMA binding to AVR-Pik alleles , Kd values are shown ( NB = No Binding ) . The sensorgrams of the data used to derive these curves are shown in Figure 2—figure supplement 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 00410 . 7554/eLife . 08709 . 005Figure 2—figure supplement 1 . Confirmation of protein expression in yeast . Western blot demonstrating the expression of AVR-Pik alleles in yeast . The expected size of the AVR-Pik/GAL4-BD is 30 kDa , the GAL4-BD alone is 19 kDa . Ponceau staining confirms equivalent protein loading . The empty vector is presented on a separate image ( boxed ) as we found the expression of this domain alone was much greater than when fused to the effectors , and it was necessary to load 40-fold less protein so as not to overload the signal . Ponceau staining of the membrane was used to confirm protein loading . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 00510 . 7554/eLife . 08709 . 006Figure 2—figure supplement 2 . SDS-PAGE gels of purified proteins . AVR-Pik effector alleles , as labelled , and Pikp1-HMA domain . The Pikp-HMA domain was run on a separate gel to the effectors and is boxed to indicate this . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 00610 . 7554/eLife . 08709 . 007Figure 2—figure supplement 3 . Analytical gel filtration . Analytical gel filtration traces depicting the retention volume of AVR-PikE , AVR-PikA , AVR-PikC and Pikp-HMA individually and when effector and HMA are mixed ( ( A ) , ( B ) and ( C ) respectively ) . SDS-PAGE gels of the relevant fractions from the elution traces of the effector:Pikp-HMA are also shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 00710 . 7554/eLife . 08709 . 008Figure 2—figure supplement 4 . SPR sensorgrams . Multi-cycle kinetics data for the interaction of Pikp-HMA ( analyte ) with effectors ( A ) AVR-PikD , ( B ) AVR-PikE , ( C ) AVR-PikA , ( D ) AVR-PikC . This data was used to generate the binding curves shown in Figure 2C . Response units for each labelled protein concentration during the run are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 00810 . 7554/eLife . 08709 . 009Table 1 . Summary Table showing the outcomes of in vitro and in planta assays used to investigate the interactions and responses of AVR-Pik effectors with Pikp-dependent readoutsDOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 009AVR-PikDAVR-PikEAVR-PikAAVR-PikCAVR-PikDHis46GluAVR-PikDIle49GluAVR-PikDArg64AlaAVR-PikDAsp66ArgAVR-PikDAla67AspAVR-PikDPro47Ala/Gly48AspInteraction with Pikp-HMA in Y2H++++−−−+++−−+++Interaction with Pikp-HMA in SPR++++++−−++−−+++Recognition in Pikp+ rice plants++++ ( − ) ( − ) −+−−++++++CD response in Nicotiana benthamiana+++−−−−++−−−+++Y2H = yeast-2-hybrid , SPR = Surface Plasmon Resonance , Pikp+ = rice cv . K60 , CD = cell death . Parentheses depict results from ( Kanzaki et al . , 2012 ) . 10 . 7554/eLife . 08709 . 010Table 2 . Intact masses for proteins expressed and purified in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 010ProteinVectorMolecular Mass ( Da ) CalculatedObservedΔPikp-HMApOPINS3C7805 . 237804 . 97−0 . 26AVR-PikDpOPINS3C10 , 835 . 3110 , 832 . 95−2 . 36§AVR-PikDpOPINA†10 , 812 . 3310 , 809 . 99−2 . 34AVR-PikDpOPINE‡11 , 786 . 3311 , 784 . 16−2 . 17AVR-PikEpOPINS3C10 , 812 . 2710 , 809 . 91−2 . 36AVR-PikEpOPINE‡11 , 763 . 2911 , 760 . 96−2 . 33AVR-PikApOPINS3C10 , 844 . 2710 , 841 . 80−2 . 47AVR-PikApOPINE‡11 , 795 . 2911 , 793 . 01−2 . 28AVR-PikCpOPINS3C10 , 856 . 2810 , 853 . 72−2 . 56AVR-PikCpOPINE‡11 , 807 . 3011 , 804 . 97−2 . 33AVR-PikDHis46GlupOPINE‡11 , 778 . 3011 , 776 . 07−2 . 23AVR-PikDIle49GlupOPINE‡11 , 802 . 2811 , 800 . 04−2 . 24AVR-PikDArg64AlapOPINE‡11 , 701 . 2211 , 698 . 94−2 . 28AVR-PikDAsp66ArgpOPINE‡11 , 827 . 4311 , 825 . 31−2 . 12AVR-PikDAla67AsppOPINE‡11 , 830 . 3411 , 828 . 20−2 . 14AVR-PikDPro47Ala , Gly48AsppOPINE‡11 , 818 . 3211 , 816 . 20−2 . 12Non-native residues remaining after 3C cleavage: N-terminal Gly–Pro . †Non-native residues remaining: N-terminal Met . ‡Non-native residues remaining after 3C cleavage: N-terminal Gly–Pro; C-terminal Lys-His-His-His-His-His-His . §The measured mass of each AVR-Pik protein should be 2 . 0156 Da ( 2 × 1 . 0078 ) less than its calculated mass due to formation of the di-sulphide bond . Next we used Surface Plasmon Resonance ( SPR ) to determine the binding affinity of AVR-PikD to Pikp-HMA . The purified effector , with a non-cleavable 6xHis tag at the C-terminus , was immobilised on a Ni2+-NTA chip; Pikp-HMA was used as the analyte . Using a multi-cycle kinetics approach , we found that Pikp-HMA bound to immobilised AVR-PikD with a Kd of 31 ± 2 nM ( Figure 2C , Figure 2—figure supplement 4A , Table 1 ) . SPR studies were expanded to include AVR-PikE , AVR-PikA and AVR-PikC ( Figure 2C , Figure 2—figure supplement 4 , Table 1 ) . For AVR-PikE , even though this was not fully saturatable under the conditions of our assay , we obtained an apparent Kd of 367 ± 41 nM , a greater than 10-fold weaker binding compared to AVR-PikD ( Figure 2C ) . For AVR-PikA we could determine an apparent Kd of 710 ± 111 nM ( also not saturatable in the assay ) . We detected essentially no binding for AVR-PikC to Pikp-HMA in this assay ( Figure 2C ) . This is consistent with the Y2H data and also correlates with the published recognition specificity in planta , although M . oryzae isolates expressing AVR-PikE were reported to not be recognised by cultivars of rice expressing Pikp ( Kanzaki et al . , 2012 ) . Although we were able to express and purify AVR-PikD from Escherichia coli , we were unable to obtain crystals of this protein for structure determination by X-ray crystallography . However , following a co-expression strategy with 6xHis tagged Pikp-HMA and untagged AVR-PikD ( see ‘Materials and methods’ ) , we obtained crystals of this complex in multiple conditions . Optimisation of one of these conditions ( Figure 3—figure supplement 1 ) produced crystals diffracting X-rays to 1 . 6 Å resolution . The structure of the Pikp-HMA/AVR-PikD complex was solved using molecular replacement to position a Pikp-HMA dimer ( see below ) in the asymmetric unit , followed by automated rebuilding with the sequence of both proteins supplied . This was sufficient to produce an initial model containing both Pikp-HMA and AVR-PikD that could be used to complete structure determination ( see ‘Materials and methods’ , X-ray Data Collection and Refinement statistics are given in Table 3 ) . 10 . 7554/eLife . 08709 . 011Table 3 . X-ray data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 011Pikp-HMAPikp-HMA/AVR-PikDNativeIodideData collection Wavelength ( Å ) 1 . 202 . 000 . 90 Space groupP6522P6522P41212Cell dimensions a , b , c ( Å ) 54 . 65 , 54 . 65 , 235 . 2254 . 73 , 54 . 73 , 235 . 80118 . 41 , 118 . 41 , 35 . 81 α , β , γ , ( ° ) 90 . 00 , 90 . 00 , 120 . 0090 . 00 , 90 . 00 , 120 . 0090 . 00 , 90 . 00 , 90 . 00 Resolution ( Å ) 47 . 33–2 . 10 ( 2 . 15–2 . 10 ) 117 . 90–2 . 80 ( 2 . 87–2 . 80 ) 39 . 47–1 . 60 ( 1 . 64–1 . 60 ) Rmerge ( % ) 8 . 4 ( 117 . 6 ) 8 . 7 ( 45 . 8 ) 4 . 7 ( 65 . 1 ) I/σI32 . 3 ( 4 . 6 ) 34 . 7 ( 7 . 3 ) 32 . 3 ( 4 . 7 ) Completeness ( % ) Overall100 ( 99 . 9 ) 99 . 9 ( 98 . 9 ) 100 ( 100 ) Anomalous99 . 9 ( 99 . 4 ) Redundancy Overall45 ( 46 . 8 ) 32 . 8 ( 24 . 4 ) 17 . 7 ( 17 . 4 ) Anomalous19 . 4 ( 13 . 3 ) CC ( 1/2 ) ( % ) 100 ( 94 . 0 ) 100 ( 98 . 0 ) 100 ( 92 . 8 ) Refinement and model Resolution ( Å ) 47 . 33–2 . 10 ( 2 . 15–2 . 10 ) 39 . 47–1 . 60 ( 1 . 64–1 . 60 ) Reflections12356 ( 861 ) 32549 ( 2379 ) Rwork/Rfree ( % ) 20 . 2/22 . 9 ( 20 . 6/19 . 6 ) 17 . 8/20 . 5 ( 19 . 6/24 . 7 ) No . atoms Protein10631762 Water44138B-factors ( Å2 ) Protein29 . 9623 . 38 Water57 . 3134 . 07R . m . s deviations Bond lengths ( Å ) 0 . 0130 . 016 Bond angles ( ° ) 1 . 571 . 79Ramachandran plot ( % ) † Favoured97 . 198 . 7 Allowed2 . 91 . 3 Outliers00 MolProbity Score1 . 48 ( 98th percentile ) 1 . 21 ( 98th percentile ) The highest resolution shell is shown in parentheses . †As calculated by MolProbity . The structure of the Pikp-HMA/AVR-PikD complex reveals an intimate interface formed between these proteins that buries 18 . 7% of the effector's solvent accessible surface area ( 1031 . 0 Å2 , Figure 3A , B ) . The majority of the interaction is formed with a monomer of Pikp-HMA , with 87 . 5% of the effector's buried surface area ( 902 . 2 Å2 ) and nine residues contributing hydrogen bond and/or salt bridge interactions . This suggests that the AVR-PikD/Pikp-HMA monomer interaction most likely represents the biologically significant interface . No hydrogen bonds or salt bridge interactions are formed with the second monomer of the Pikp-HMA dimer . Further , due to steric clash that would occur , it is not possible for an AVR-PikD/Pikp-HMA heterotetramer ( 2:2 complex ) to assemble . All interface analysis was performed using PDBePISA ( Krissinel and Henrick , 2007 ) . 10 . 7554/eLife . 08709 . 012Figure 3 . Structure of the AVR-PikD/Pikp-HMA complex . ( A ) Schematic representation of the AVR-PikD/Pikp-HMA ( monomer ) , highlighting interfacing residues . The effector is shown in green cartoon , with side chains as sticks and green carbon atoms ( no surface ) . The Pikp-HMA is shown in blue cartoon , with side chains as sticks and blue carbon atoms; the molecular surface of this protein is also depicted . Effector residues selected for mutation are labelled , as are important interface residues of Pikp-HMA discussed in the text . Hydrogen bonds/salt-bridges are shown as dashed lines and the di-sulphide bond as yellow bars . ( B ) Buried surface areas of AVR-PikD ( left , purple ) and Pikp-HMA ( right , brown ) separated and shown from the perspective of the partner molecule . Cartoon and amino acid side chains shown are as for panel ( A ) . ( C ) Comparison of the Pikp-HMA ( monomer , blue ) with yeast Ccc2A ( wheat ) showing the conservation of the HMA fold . The copper ion bound to Ccc2a is shown as a red sphere . ( D ) Comparison of AVR-PikD ( green ) and AVR-Piz-t ( pink ) structures showing the conservation of the β-sandwich structure , and the N-terminal extension of AVR-PikD . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 01210 . 7554/eLife . 08709 . 013Figure 3—figure supplement 1 . Sample preparation for x-ray data collection . ( A ) Pikp-HMA , left: SDS-PAGE gel of sample used for crystallisation , right: crystal of Pikp-HMA . ( B ) Pikp-HMA/AVR-PikD complex , left: SDS-PAGE gel of sample used for crystallisation , right: crystals of Pikp-HMA/AVR-PikD complex . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 01310 . 7554/eLife . 08709 . 014Figure 3—figure supplement 2 . The structure of the Pikp-HMA dimer is conserved when bound to AVR-PikD . ( A ) Crystal structure of the Pikp-HMA dimer , monomers are coloured blue and magenta . ( B ) Crystal structure of the Pikp-HMA dimer in complex with AVR-PikD , coloured as above and with the effector in green ( Cys residues that form the di-sulphide bond are shown in yellow ) . ( C ) Crystal structure of the Pikp-HMA dimer in complex with AVR-PikD , with the Pikp-HMA dimer structure from ‘ ( A ) ’ overlaid , shown in grey . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 01410 . 7554/eLife . 08709 . 015Figure 3—figure supplement 3 . Polymorphic residue AVR-PikDHis46 is bound within a pocket on the Pikp-HMA surface . ( A ) AVR-PikDHis46 interacts via hydrogen bond and salt–bridge interactions with Pikp-HMA residues Ser218 and Glu230 and van der Waals interaction with Val232 . ( B ) Region of the Pikp-HMA structure , equivalent to that shown in ( A ) , determined from the uncomplexed crystal structure of Pikp-HMA . In both panels , each residue is shown overlaid with the electron density describing their position ( 2mFobs − DFcalc contoured at 1σ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 01510 . 7554/eLife . 08709 . 016Figure 3—figure supplement 4 . Amino acid sequence alignment of AVR-Pik alleles and Pik-HMA domains . ( A ) AVR-Pik alleles . The residues that are polymorphic in AVR-Pik alleles are marked by empty triangles . The residues of AVR-PikD targeted for mutation in this study are marked with filled triangles . ( B ) Pik-HMAs ( Pikp-HMA , Pikm-HMA and Pik-HMA ) . The residues of Pikp-HMA involved in hydrogen bond interactions with AVR-PikDHis46 are marked by empty triangles . The residues of Pik-HMA domains that may contribute to recognition specificity are marked with filled triangles . The secondary structure features presented above the alignment are derived from the structures of AVR-PikD and Pikp-HMA . The alignments were performed using Clustal Omega ( Sievers et al . , 2011 ) and presented using ESPript ( Robert and Gouet , 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 016 Each of the Pikp-HMA monomers adopts the HMA-domain fold ( Pfam: PF00403 ) , comprising a four-stranded antiparallel β-sheet and two α-helices packed in an α/β sandwich . The closest structural homologue of Pikp-HMA ( defined by PDBeFold [Krissinel and Henrick , 2004] ) is the HMA domain of yeast protein Ccc2A ( Banci et al . , 2001 ) , overlaying with an r . m . s . d . of 1 . 58 Å over 72 residues ( Figure 3C ) . Typically , HMA domains bind heavy metals , or lighter cations such as Cu1+ or Zn2+ , via two conserved Cys residues and are involved in metal transport or detoxification pathways ( Bull and Cox , 1994 ) . Interestingly , these Cys residues are not conserved in Pik-1 HMA domains , including Pikp-1 . Hence , the Pikp-HMA structure does not contain a metal ion , and the loop between β1 and α1 , which usually contains the metal-chelating Cys residues , is disordered . Further , this loop is positioned away from the interface with the effector ( Figure 3A , Figure 3—figure supplement 2 ) and does not contribute to complex formation . We were also able to obtain the crystal structure of Pikp-HMA in the absence of AVR-PikD ( see ‘Materials and methods’ , Figure 3—figure supplements 1 , 2A ) . The structure of the Pikp-HMA dimer in isolation is essentially identical to that found in the complex ( r . m . s . d . 0 . 67 Å over 69 residues , for the monomer bound to AVR-PikD ) , with the exceptions of a minor shift in the loop spanning residues Val222—Lys228 and the N-terminal four residues ( Figure 3—figure supplement 2C ) . AVR-PikD adopts a six-stranded β-sandwich structure , stabilised by a di-sulphide bond between Cys54 and Cys70 . The effector contains an N-terminal extension , comprising residues Arg31 to Pro52 , prior to the start of this fold ( Figure 3A , D ) . The extension is anchored to the β-sandwich at each end via a salt–bridge interaction involving the side chains of Asp45 and Arg110 and hydrogen bonds between both the main chain carbonyl of Arg39 and Glu38Oε1 and Arg64Nη1 . Database searches using PDBeFold reveals that a close structural homologue of AVR-PikD is AVR-Piz-t ( Zhang et al . , 2013 ) , another M . oryzae effector protein , despite there being essentially no sequence identity between these proteins ( r . m . s . d . = 2 . 33 Å over 58 aligned residues , Figure 3D ) . This suggests that sequence divergent translocated effectors of M . oryzae may share a conserved structural scaffold , despite very different sequences , which has striking parallels to RXLR-type effectors of plant pathogenic oomycetes ( Boutemy et al . , 2011; Win et al . , 2012 ) . Further , structural homology is also observed to ToxB , a protein toxin from Pyrenophora tritici-repentis , the causative agent of tan spot in wheat ( Nyarko et al . , 2014 ) . In each case , the identified structural homology only extends to the β-sandwich fold and the N-terminal extension of AVR-PikD appears to be unique . This raises the interesting possibility that candidate effectors from distant fungi could be identified by structure-guided sequence similarity searches . Three primary sites of interaction are apparent between Pikp-HMA and AVR-PikD . The first is dominated by main-chain hydrogen bonding between the C-terminal β-stand of Pikp-HMA and β3 of AVR-PikD , which results in formation of a continuous antiparallel β-sheet comprising the four β-strands of Pikp-HMA and β3–5 of AVR-PikD . The second involves the side chain of Pikp-HMAAsp224 , which forms a salt–bridge interaction with the side chain of AVR-PikDArg64 , and is also held in place by a hydrogen bond of its main chain NH group to the side chain of AVR-PikDAsp66 ( Figure 3A ) . The third interaction site centres on AVR-PikDHis46 , although has contributions from residues Asn42—Ile49 . This region forms part of the N-terminal extension and includes the polymorphic AVR-Pik residues 46 , 47 and 48 ( His46 , Pro47 and Gly48 in AVR-PikD , Figures 1A and 3A ) . AVR-PikDHis46 is bound in a pocket on Pikp-HMA via hydrogen bonds/salt bridge interactions between AVR-PikDHis46:Nδ1/Pikp-HMASer218:Oγ and AVR-PikDHis46:Nε2/Pikp-HMAGlu230:Oε1; also , Pikp-HMAVal232 packs on top of the AVR-PikDHis46 ring and contributes hydrophobic/van der Waals interactions ( Figure 3—figure supplement 3 ) . Finally , it is worth noting that there is an extensive network of buried solvent-mediated contacts between Pikp-HMA and AVR-PikD . Based on the Pikp-HMA/AVR-PikD structure , we designed four mutations in AVR-PikD predicted to perturb complex formation through generating steric clashes/introducing charged residues , or removing a salt–bridge interaction ( His46Glu , Ile49Glu , Asp66Arg and Arg64Ala ) , and two mutants to mimic other AVR-Pik alleles , but retain His46 ( Ala67Asp [based on AVR-PikC] , Pro47Ala/Gly48Asp [based on AVR-PikA] ) , Figure 3—figure supplement 4A . First , we screened these mutants for interaction with Pikp-HMA in the Y2H assay . We found that AVR-PikDHis46Glu , AVR-PikDArg64Ala and AVR-PikDAsp66Arg prevent the interaction ( as observed on the -LTH selective growth plate and in the X-gal assay ( Figure 4A , Figure 4—figure supplement 1 , Table 1 ) . However , AVR-PikDIle49Glu maintains an interaction and AVR-PikDAla67Asp and AVR-PikDPro47Ala/Gly48Asp showed intermediate binding ( weak interaction on -LTH selective growth plate and in the X-gal assay [Figure 4A] ) . 10 . 7554/eLife . 08709 . 017Figure 4 . Structure-based mutagenesis at the Pikp-HMA/AVR-PikD interface perturbs protein interactions in yeast and in vitro . ( A ) Y2H assays showing the binding of AVR-PikD mutants to Pikp-HMA using two read-outs , growth on–Leu-Trp-His+3AT ( -LTH ) plates and the X-gal assay . ( B ) Binding curves derived from Surface Plasmon Resonance single-cycle kinetics data for Pikp-HMA binding to AVR-PikD and AVR-PikD mutants , Kd values are shown where determined ( ND = Not Determined , NB = No Binding ) . The sensorgrams of the data used to derive these curves are shown in Figure 4—figure supplement 2B . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 01710 . 7554/eLife . 08709 . 018Figure 4—figure supplement 1 . Confirmation of protein expression in yeast . Western blot analysis demonstrating the expression of AVR-PikD and AVR-PikD mutants in yeast . The empty vector ( EV ) is presented on a separate image ( boxed ) as we found the expression of this domain alone was much greater than when fused to the effectors , and it was necessary to load 40-fold less protein to not overload the signal . Ponceau staining of the membrane was used to confirm protein loading . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 01810 . 7554/eLife . 08709 . 019Figure 4—figure supplement 2 . SDS-PAGE of AVR-PikD mutant proteins and SPR sensorgrams . ( A ) AVR-PikD mutants prepared for SPR analysis . ( B ) Single-cycle kinetics data for the interaction of Pikp-HMA ( analyte ) with AVR-PikD mutants . This data was used to generate the binding curves shown in Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 019 Next , we expressed and purified each of these AVR-PikD mutants ( as for wild-type and with C-terminal non-cleavable 6xHis tag ) and confirmed their identity by intact mass spectrometry ( Figure 4—figure supplement 2A , Table 2 ) . We then used SPR to determine the binding affinities between these mutants and the Pikp-HMA using a single-cycle kinetics approach ( [Karlsson et al . , 2006] Figure 4B , Figure 4—figure supplement 2B , Table 1 ) , having confirmed a similar affinity of Pikp-HMA for AVR-PikD ( Kd = 29 ± 3 . 5 nM ) using this approach . Consistent with the Y2H results , we could not measure any meaningful interaction of AVR-PikDHis46Glu , AVR-PikDArg64Ala and AVR-PikDAsp66Arg with Pikp-HMA ( Figure 4B ) . For AVR-PikDIle49Glu and AVR-PikDPro47Ala/Gly48Asp we were able to determine Kds of interaction of 99 ± 18 nM and 83 ± 16 nM respectively ( Figure 4B , Figure 4—figure supplement 2B ) . AVR-PikDAla67Asp showed a weaker response but we were unable to obtain a reliable Kd at the concentrations of Pikp-HMA used . AVR-PikDIle49Glu , AVR-PikDAla67Asp and AVR-PikDPro47Ala/Gly48Asp all interacted in the Y2H assay , with the latter two showing qualitatively weaker binding . To test the effects of the AVR-PikD mutations on pathogen virulence on rice plants expressing the Pikp gene , we transformed M . oryzae isolate Sasa2 with constructs encoding each of the six mutants above , with expression driven by the native AVR-PikD promoter . AVR-PikE was included in these experiments as it represents a naturally occurring point mutant at the important position 46 ( His46Asn ) . Each of the transformed M . oryzae lines were spot inoculated ( Kanzaki et al . , 2002 ) onto leaf blades of rice cultivars Nipponbare ( Pik− , lacks known Pik alleles ) and K60 ( which contains Pikp ) . The Nipponbare cultivar was susceptible to all of the M . oryzae lines , including Sasa2 wild type and empty vector control , as shown by the development of lesions around the inoculation sites ( Figure 5 ) . As expected , the K60 ( Pikp ) cultivar is resistant to the M . oryzae Sasa2 line expressing AVR-PikD ( Kanzaki et al . , 2012 ) . We observed that the K60 ( Pikp ) cultivar showed an intermediate phenotype between susceptible and resistant to M . oryzae Sasa2 lines expressing AVR-PikE . For the mutants , we found that the K60 ( Pikp ) cultivar was susceptible to M . oryzae Sasa2 lines expressing AVR-PikDHis46Glu , AVR-PikDArg64Ala and AVR-PikDAsp66Arg , but resistant to those expressing AVR-PikDAla67Asp and AVR-PikDPro47Ala/Gly48Asp ( Figure 5 , Table 1 ) . As for AVR-PikE , K60 ( Pikp ) shows an intermediate phenotype to Sasa2 lines expressing AVR-PikDIle49Glu ( Figure 5 ) . There is a correlation between AVR-PikD mutants that display the tightest binding affinities in vitro , and interact in the Y2H assay , with resistance in rice when delivered by M . oryzae ( partial phenotype in the case of AVR-PikDIle49Glu ) . Strains expressing AVR-PikD mutants that do not interact in vitro or in yeast ( AVR-PikDHis46Glu , AVR-PikDArg64Ala and AVR-PikDAsp66Arg ) are fully susceptible in rice . Expression of AVR-PikD and mutants in the transgenic M . oryzae during infection was confirmed by RT-PCR ( Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 08709 . 020Figure 5 . Structure-based mutagenesis at the Pikp-HMA/AVR-PikD interface leads to susceptibility in Pikp+ rice plants . Rice plants Pik− ( cv . Nipponbare ) and Pikp+ ( cv . K60 ) were spot-inoculated with M . oryzae Sasa2 expressing AVR-PikD , AVR-PikE and AVR-PikD mutants . The combinations resulting in resistant ( R ) , intermediate ( IM ) and susceptible ( S ) phenotype are labelled . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 02010 . 7554/eLife . 08709 . 021Figure 5—figure supplement 1 . RT-PCR . RT-PCR confirming AVR-PikD , AVR-PikE and AVR-PikD mutants are expressed during infection , M . oryzae actin was used as the positive control . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 021 To further investigate the link between recognition of AVR-PikD by Pikp-1/Pikp-2 and immunity-related signalling , we established a transient expression assay in N . benthamiana leaves using Agrobacterium tumefaciens to deliver these genes into plant cells ( henceforth Agroinfiltration , Figure 6A ) . In this system we required the co-expression of Pikp-1 , Pikp-2 and AVR-PikD to observe robust features of cell death , including necrotic tissue and accumulation of phenolic compounds that give rise to auto-fluorescence ( Figure 6B ) ( Bos et al . , 2006 ) . We did not observe significant effects in N . benthamiana leaves following expression of the individual proteins or any combination of protein pairs ( Figure 6B ) . Further , co-expression of AVR-PikE , AVR-PikA and AVR-PikC with Pikp-1 and Pikp-2 fails to elicit a cell death response ( Figure 6—figure supplements 1–4 ) . This demonstrates only the specific combination of Pikp-1 , Pikp-2 and AVR-PikD results in a robust response and is consistent with the observed pairings between rice cultivars with different Pik alleles and M . oryzae isolates with different AVR-Pik alleles ( Kanzaki et al . , 2012 ) . 10 . 7554/eLife . 08709 . 022Figure 6 . Pikp HR-like cell death in Nicotiana benthamiana requires co-delivery of AVR-PikD , Pikp-1 and Pikp-2 . ( A ) Western blots showing expression of AVR-PikD ( HA ) , Pikp-1 ( FLAG ) and Pikp-2 ( Myc ) in N . benthamiana . Blots were probed using the appropriate antibody for the tagged protein . ( B ) The Pikp HR-like cell death in N . benthamiana requires expression of Pikp-1 , Pikp-2 and AVR-PikD together . Expression of individual proteins or co-expression of any protein pair does not result in cell death . Images showing autofluorescence are horizontally flipped to present the same leaf orientation as white light images . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 02210 . 7554/eLife . 08709 . 023Figure 6—figure supplement 1 . Pikp HR-like cell death in N . benthamiana requires expression of Pikp-1 , Pikp-2 and AVR-PikD specifically . ( A ) Only the expression of Pikp-1 , Pikp-2 and AVR-PikD , not AVR-PikE , AVR-PikA or AVR-PikC , results in cell death in N . benthamiana . Images showing autofluorescence are horizontally flipped to present the same leaf orientation as white light images . ( B ) Western blots showing expression of AVR-Pik ( HA ) alleles , Pikp-1 ( FLAG ) and Pikp-2 ( Myc ) in N . benthamiana . Blots were probed using the appropriate antibody for the tagged protein . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 02310 . 7554/eLife . 08709 . 024Figure 6—figure supplement 2 . Expression of AVR-Pik alleles alone , or in any combination with Pikp-1 or Pikp-2 , does not result in HR-like cell death in N . benthamiana for ( A ) AVR-PikE , ( B ) AVR-PikA or ( C ) AVR-PikC . The Pikp-1 , Pikp-2 and AVR-PikD combination is included as a positive control . Images showing autofluorescence are horizontally flipped to present the same leaf orientation as white light images . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 02410 . 7554/eLife . 08709 . 025Figure 6—figure supplement 3 . Example images used for scoring HR-like cell death ( HR Index ) in N . benthamiana on expression of Pikp-1 , Pikp-2 and AVR-Pik alleles and AVR-PikD mutants . Views from the adaxial side of the leaves for white light images and abaxial side of the leaves for UV images were used . Images showing autofluorescence are horizontally flipped to present the same orientation as white light images . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 02510 . 7554/eLife . 08709 . 026Figure 6—figure supplement 4 . Box plots depicting HR Index for repeats of the assay shown in Figure 6 and Figure 6—figure supplement 1A . For each sample the number of repeats used were , Empty Vector ( EV ) and AVR-PikD: 156 , AVR-PikE: 68 , AVR-PikA: 69 , AVR-PikC: 70 . The scoring system for the HR Index is shown in Figure 6—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 026 To further correlate the cell death response with direct protein:protein interaction , we co-expressed Pikp-1 , Pikp-2 and the AVR-PikD mutants described above via Agroinfiltration in N . benthamiana leaves ( Figure 7—figure supplement 1 ) . Using this approach , we found that the AVR-PikDHis46Glu , AVR-PikDArg64Ala , AVR-PikDAsp66Arg and AVR-PikDAla67Asp mutations do not elicit a response , but AVR-PikDIle49Glu and AVR-PikDPro47Ala/Gly48Asp still promote cell death ( Figure 7 , Figure 7—figure supplement 2 , Table 1 ) . Interestingly , while AVR-PikDIle49Glu consistently generates a response in N . benthamiana and bound to Pikp-HMA in the SPR and Y2H assays , it displayed an intermediate phenotype in the M . oryzae inoculation assay . Further , while AVR-PikDAla67Asp did not elicit a response in N . benthamiana and showed only weak interaction by SPR , it did bind to Pikp-HMA in the Y2H assay , and induced resistance in the M . oryzae inoculation assay . These results suggest that differences in binding affinities between effectors and Pikp-HMA in vitro can occasionally result in subtly different readouts in plants . 10 . 7554/eLife . 08709 . 027Figure 7 . Structure based mutations at the AVR-PikD/Pikp-HMA interface leads to loss of HR-like cell death in N . benthamiana . Co-infiltration of Pikp-1 and Pikp-2 with AVR-PikD mutants His46Glu , Arg64Ala , Asp66Arg and Ala67Asp leads to loss of recognition and signalling in N . benthamiana . AVR-PikDIle49Glu and AVR-PikDPro47Ala/Gly48Asp retain recognition and signalling . Each infiltration site includes Pikp-1 and Pikp-2 with the AVR-PikD mutant indicated . Pikp-1 and Pikp-2 alone ( EV ) and with AVR-PikD are included as controls . Images showing autofluorescence are horizontally flipped to present the same leaf orientation as white light images . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 02710 . 7554/eLife . 08709 . 028Figure 7—figure supplement 1 . Western blots showing expression of AVR-Pik alleles , AVR-PikD mutants ( A ) and Pikp-1 , Pikp-2 ( B ) in N . benthamiana . Blots were probed using the appropriate antibody for the tagged protein . The same blots were previously shown in Figure 6—figure supplement 1B to show expression of Pikp-1 , Pikp-2 and AVR-Pik alleles . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 02810 . 7554/eLife . 08709 . 029Figure 7—figure supplement 2 . Box plots showing HR Index for repeats of the assay depicted in Figure 7 . For each sample the number of repeats used were , Empty Vector ( EV ) and AVR-PikD: 156 , AVR-PikDHis46Glu , AVR-PikDIle49Glu and AVR-PikDArg64Ala: 70 , AVR-PikDAsp66Arg and AVR-PikDAla67Asp: 69 , AVR-PikDPro47Ala/Gly48Asp: 68 . Note that the data presented here for EV and AVR-PikD are also presented in Figure 6—figure supplement 4 ( all N . benthamiana assay repeats were conducted at similar times ) . The scoring system for the HR Index is shown in Figure 6—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 08709 . 029 In plants , paired NLRs such as rice RGA5/RGA4 and Arabidopsis RRS1/RPS4 ( Narusaka et al . , 2009 ) , function through formation of homo- and hetero-protein complexes ( Cesari et al . , 2014; Williams et al . , 2014 ) . One member of the pair can constitutively activate an HR-like cell death on expression in plants ( RGA4 and RPS4 ) , and this activity is suppressed by the second ( RGA5 and RRS1 ) through the formation of hetero-complexes ( Cesari et al . , 2014; Williams et al . , 2014 ) . This suppression is relieved by co-expression of the cognate effectors and can result in signalling-competent NLR homo-complexes ( Cesari et al . , 2014 ) . In mammals , members of the NAIP ( NLR ) family act as sensors of pathogen signatures and associate with NLRC4 following perception to trigger signalling ( Kofoed and Vance , 2011; Zhao et al . , 2011 ) . Although it seems unlikely that assemblies of homo- and hetero-NLR complexes in their suppressed and activated states are universally conserved in mammals and plants , it appears that oligomerisation plays a key role in modulating activity . In contrast to RGA4 and RPS4 , expression of Pikp-2 does not constitutively activate cell death in Nicotiana . In our assays , the HR-like response requires co-expression of Pikp-1 and AVR-PikD with Pikp-2 , suggesting assembly of an active signalling complex requires all three proteins . Although the limits of our assays preclude a conclusive interpretation of this signalling complex , they do suggest that not all paired plant NLRs function within the confines of existing models . At present it is unknown whether or not Pikp-2 forms a heteromeric complex with Pikp-1 in an effector-dependent manner . Future work is required to dissect the underlying molecular interactions that promote signalling by the Pik NLRs . Phylogenetics suggests AVR-PikD is the ancestral AVR-Pik allele , and it is the only natural variant with a His at position 46 ( Kanzaki et al . , 2012 ) . In the Pikp-HMA/AVR-PikD complex , the AVR-PikDHis46 side chain is buried within a pocket on the Pikp-HMA surface that contributes hydrogen bonds/salt bridge interactions ( Figure 3A , Figure 3—figure supplement 3 ) . The AVR-PikDHis46Glu mutation prevents interaction with Pikp-HMA in vitro and in yeast , and response in planta either when delivered by M . oryzae into rice or on co-expression in N . benthamiana . This data supports AVR-Pik46 as a key site for recognition specificity , and that following introduction of Pikp into cultivated rice , M . oryzae evolved to evade recognition by mutating this residue . Interestingly , the only natural variant found at this position is a somewhat conservative His to Asn change ( giving rise to AVR-PikE ) . Conceptually , this residue could be accommodated at the Pikp-HMA/AVR-PikD interface without generating significant steric clashes , but the interactions formed at this site ( e . g . , hydrogen bonding pattern ) will be fundamentally different . This single amino acid polymorphism is sufficient to prevent AVR-Pik-dependent cell death in N . benthamiana , limit resistance in rice , and reduce the affinity of interaction with Pikp-HMA in vitro by an order of magnitude . This suggests a binding threshold is required to elicit a response in plants and the observed 10-fold reduced affinity of AVR-PikE for Pikp-HMA in vitro is sufficient for this . While this may be enough to explain why the conservative His46Asn mutation is the only one found in nature , it remains possible that other mutations would not be tolerated due to a trade-off with the effector's virulence activity . Of the other amino acid mutations explicitly designed to disrupt the Pikp-HMA/AVR-PikD interaction , AVR-PikDArg64Ala and AVR-PikDAsp66Arg prevented interaction in SPR and yeast , and responses in planta . These results , in addition to AVR-PikDHis46Glu , provide convincing evidence that the crystal structure of Pikp-HMA/AVR-PikD is consistent with the complex formed in plant cells . Interestingly , AVR-PikDIle49Glu retained interaction in SPR ( threefold reduction in affinity compared to AVR-PikD ) and yeast and elicited a response by Pikp in planta . Close inspection of the Pikp-HMA/AVR-PikD interface reveals how a Glu could be accommodated at this position . Of note is that AVR-PikDIle49 is close to the N-terminus of the Pikp-HMA domain and a repositioning of Pikp-HMAGly186 , for example in the context of the full-length protein , could create space for a Glu residue . AVR-PikA and AVR-PikC combine the AVR-PikDHis46Asn polymorphism with a Pro47Ala/Gly48Asp substitution and an Ala67Asp substitution respectively ( Figure 1A ) . AVR-PikC did not show measureable binding to Pikp-HMA in vitro or in yeast , or elicit responses in planta . In the Pikp-HMA/AVR-PikD structure , the side chain of AVR-PikDAla67 lies adjacent to the Pikp-HMAAsp224/AVR-PikDArg64 salt-bridge . An Asp at position 67 would likely perturb the Pikp-HMA/AVR-PikD interaction through disruption of this salt-bridge . While reducing the affinity to a level where a Kd cannot be determined in our SPR assay , and not eliciting a response in N . benthamiana , AVR-PikDAla67Asp still interacted with Pikp-HMA in yeast and was recognised in rice when the protein was delivered by M . oryzae . It is possible that the weaker AVR-PikDAla67Asp/Pikp-HMA interaction may be sufficient for generating a response in rice containing Pikp , but not sufficient in the N . benthamiana assay used here . AVR-PikA shows 23-fold apparent weaker affinity to Pikp-HMA compared to AVR-PikD in vitro , no interaction with Pikp-HMA in yeast , and elicited no response in planta . Despite being adjacent to AVR-PikDHis46 , in the Pikp-HMA/AVR-PikD complex the side chains of AVR-PikDPro47 and AVR-PikDGly48 do not contact the same Pikp-HMA monomer and Ala/Asp residues could be accommodated at these positions without disrupting the Pikp-HMA/AVR-Pik interface . Consistent with this , the AVR-PikDPro47Ala/Gly48Asp mutant showed only slightly reduced affinity for Pikp-HMA in vitro ( 2 . 7-fold ) , interacted with Pikp-HMA in yeast , and elicited responses in planta . This data suggests that AVR-PikDPro47Ala/Gly48Asp mutations may provide a minor additive benefit , but AVR-PikDHis46Asn is the dominant mutation contributing to the evasion of Pikp recognition . It is also plausible that these mutations contribute to the virulence activity of the effector in the context of an Asn at position 46 , rather than quantitatively contributing to the evasion of Pik recognition . Almost all amino acid variation between Pik-1 alleles lies within the HMA domain ( Figure 3—figure supplement 4B ) . Mapping the variation in Pik-HMAs at the Pikp-HMA/AVR-PikD interface suggests a ‘hot-spot’ centred on recognition of AVR-Pik46 . Pikp-1Glu230 ( Pikp numbering ) , that directly co-ordinates AVR-PikDHis46 , is a Val in both Pikm-1 and Pik-1 . Residue Pikp-1Val222 , whose side-chain extends towards Pikp-1Glu230 , is an Ala in Pikm-1 and Pik-1 . In the absence of specific binding data , none of these mutations in their own right , or when combined , would be predicted to preclude interaction with AVR-PikD ( as has been observed previously [Kanzaki et al . , 2012] ) , but would also not be predicted to explicitly enhance the binding of an Asn at position 46 ( as found in AVR-PikE , AVR-PikA and AVR-PikC , which are all recognised by Pikm ) . Three other Pik variable residues map at or close to the Pikp-HMA/AVR-PikD interface . Pikp-1Asp217 is adjacent to the invariant Ser218 ( which with Pikp-1Glu230 co-ordinates AVR-PikDHis46 ) and extends towards residue 48 of AVR-PikD . It is conceivable that a His residue at position 217 , as found in Pikm-1 and Pik-1 , may interact with residues in this region to extend recognition of AVR-Pik alleles , in particular Asp48 as found in AVR-PikA . Pikp-1Lys228 , a Glu in Pik*-1 and a Gln in Pikm-1 , and Pikp-1Glu253 , a Met in Pikm-1 , form hydrogen bonds with AVR-PikDAsp66 and AVR-PikDLys79 respectively . Interestingly , position 228 is one of two residues that have been suggested as diagnostic markers for Pik breeding in rice ( Costanzo and Jia , 2010 ) . Despite the variation in these Pik residues , as Asp66 and Lys79 are invariant in AVR-Pik alleles , it is difficult to appreciate how they contribute to recognition specificity . Teasing apart these seemingly fine-tuned recognition specificities between Pik-HMA and AVR-Pik alleles , and how these are balanced against the virulence-associated activity of the effectors , awaits further study . Recognition of AVR-Pik by Pik , and AVR-Pia and AVR1-CO39 by RGA5/RGA4 , is by direct binding to the HMA domains of Pik-1 and RGA5 . The position of the HMA domain , between the CC and NB-ARC region of Pik-1 and after the LRR in RGA5 , is one of the most striking differences between these functionally-related proteins . Interestingly , the integrated domain in the Arabidopsis NLR RRS1 , a domain with sequence similarity to WRKY transcription factors , is also positioned after the LRR . Cognate effectors AvrRps4 and PopP2 directly interact with this WRKY domain ( Cesari et al . , 2014; Le Roux et al . , 2015; Sarris et al . , 2015 ) . Together , these proteins reveal that functional integrated domains can occupy different positions in NLRs . How does the binding of effectors to HMA domains trigger immunity-related signalling ? The absence of obvious enzymatic activity in the effectors , and a lack of large conformation changes in Pikp-HMA in the AVR-PikD-bound and unbound states , supports the hypothesis that effector binding promotes domain re-arrangements in NLR complexes ( Cesari et al . , 2014 ) . This could be by ( 1 ) direct competition for a shared binding surface on the HMA between the effectors and either an intra- or inter-molecular contact with another NLR domain , ( 2 ) the effectors may ‘bridge’ contacts between HMAs and other NLR domains to stabilise interactions , ( 3 ) effector binding disrupts or promotes interaction of NLRs with other , as yet unknown , molecules , ( 4 ) subtle changes within the HMAs , in the context of the full length proteins , promotes NLR domain rearrangements . Any of these scenarios could break existing and/or promote new interactions within NLR complexes . They could also promote presentation of new molecular surfaces that could interact with downstream components to initiate immunity-related signalling . Due to the different positions of the Pik-1 and RGA5 HMA domains , the conformational changes underlying transduction of direct effector binding to immunity-related signalling are likely to be different , but the intra- and/or inter-molecular complexes mediating output maybe conserved . In the future , transferring unconventional integrated domains to the different positions within and between NLRs will determine the importance of domain location , and whether these positions can accommodate novel integrated domains with the potential to deliver new-to-nature resistance capabilities . The Proquest two-hybrid system ( Life Technologies , United Kingdom ) was used to detect protein–protein interactions , according to manufacturer instructions , with only minor modifications . Briefly , DNA encoding Pikp-HMA in pDEST22 was co-transformed with either the individual AVR-Pik alleles or the AVR-PikD mutants in pDEST32 , into chemically competent Saccharomyces cerevisiae MaV203 cells . Single colonies grown on selection plates were resuspended in 100 μl H2O and 2 μl were spotted on SC-Leu-Trp ( as growth control ) and SC-Leu-Trp-His+10 mM 3AT ( His auxotrophy assay ) . Photographs of colonies on SC-Leu-Trp-His+10 mM 3AT plates were taken after incubation for 24 hr at 28°C and 16 hr at room temperature . For the X-gal assay , 2 μl of resuspended cells were spotted on a Hybond N membrane ( GE Healthcare ) on a YAPD plate . After 24 hr the membrane was removed and place on top of 2 layers of 3 MM paper ( GE Healthcare ) soaked with 10 ml buffer Z supplemented with 15 mg X-gal in 100 μl dimethylformamide and 60 μl 2-mercaptoethanol . After incubation for 24 hr at 37°C the membrane was air dried and a picture taken . All pictures are representative of at least three experimental repeats , with consistent results . To confirm protein expression in yeast , total protein extracts from transformed colonies were produced according to the urea/SDS method as described in the Clontech Yeast Protocols Handbook . Aliquots of 4 μl were separated by SDS-PAGE and transferred to PVDF membrane . Due to the high level of expression of the GAL-4-DB domain from the empty pDEST32 , a 10 μl of a dilution 1:100 of the original extract was loaded on SDS-PAGE gel for this sample . Membranes were probed with anti-GAL4-DBD HRP-conjugated antibody ( Santa Cruz Biotechnology , Dallas , TX , United States ) and developed with a mix of 500 μl of SuperSignal West Pico Chemiluminescent Substrate and 800 μl of SuperSignal West Femto Maximum Sensitivity Substrate ( Life Technologies ) following standard procedures . M . oryzae strains Sasa2 , Sasa2 with pex31-D fragment ( AVR-PikD ) or with AVR-PikE used in this study are stored at the Iwate Biotechnology Research Center ( Yoshida et al . , 2009; Kanzaki et al . , 2012 ) . To obtain protoplasts , hyphae of each Sasa2 strain were incubated for 3 days in 200 ml of YG medium ( 0 . 5% yeast extract and 2% glucose , wt/vol ) . Protoplast preparation and transformation were performed as described previously ( Takano et al . , 2001 ) . Bialaphos-resistant transformants were selected on plates with 250 µg/ml of bialaphos ( Wako Pure Chemicals ) . Rice leaf blade spot inoculations were performed with M . oryzae strains ( Kanzaki et al . , 2002 ) . Disease lesions were photographed 14 days post inoculation . Rice seedlings ( cvs . Nipponbare and K60 ) at the fourth leaf stage were used for inoculation . The assays were repeated at least 3 times with similar results . For RT-PCR , total RNA was extracted from disease lesions of rice cv . Nipponbare leaves using Purelink Plant RNA Reagent , which was subsequently treated with TURBO DNase ( Life Technologies ) . From 2 µg of the DNase-treated RNA of each sample , single-strand cDNA was synthesized using oligo ( dT ) primer and ReverTra Ace ( Toyobo ) . To confirm the gene expression of AVR-Pik and the M . oryzae actin gene ( Mo-Actin ) , these genes were amplified by PCR with primers given in ( Supplementary file 1 ) . For agroinfiltration in N . benthamiana , A . tumefaciens strain GV3101 was transformed with the relevant binary constructs . Leaves of 4 weeks old N . benthamiana plants were agroinfiltrated using a needleless syringe . The total OD600 of infiltrated cultures was 1 . 0 with ratios used 1 . 5:1 . 5:6:1 for Pikp-1:Pikp-2:effector:P19 . When one or more constructs were not present , total OD600 was maintained with appropriate amount of empty vector . Photos showing cell death were taken 4 dpi from the adaxial side of the leaves for white light images and abaxial side of the leaves for UV images . Pictures are representative of four independent experiments , with internal repeats . Data for the box plots presented in Figure 6—figure supplement 4 and Figure 7—figure supplement 2 are from three independent experiments with internal repeats . The HR index was scored according to the scale presented in Figure 6—figure supplement 3 . For extraction of total protein from samples , leaf disks were taken at 2 dpi and homogenised in extraction buffer ( 250 mM Tris–HCl , pH 7 . 5 , 2 . 5 mM EDTA , 0 . 1% ascorbic acid , 1 mM PMSF , 0 . 1% [vol/vol] Protease Inhibitor Cocktail for plant cell and tissue extracts [SIGMA , St . Louis , MO , United States] ) . Supernatants were centrifuged and separated on 10–20% precast e-PAGEL gels prior to transfer onto Immobilon Transfer Membranes ( Millipore , Germany ) . The blots were blocked in 2% ECL Advanse Blocking Agent ( GE Healthcare ) in TTBS ( 10 mM Tris–HCl , pH 7 . 5 , 100 mM NaCl , 0 . 1% Tween 20 [vol/vol] ) for 1 hr at room temperature with gentle agitation . For immunodetection , blots were probed with anti-HA ( 3F10 ) -HRP ( Roche , Switzerland ) , anti-FLAG M2-HRP ( SIGMA ) or anti-Myc-tag ( HRP-DirecT ) ( MBL , Woburn , MA , United States ) in a 1:10 , 000 dilution in TTBS for 2 hr . After washing the membrane for 3 × 10 min , the reactions were detected using ChemiLumi One Super or Ultra ( Nacalai Tesque , Japan ) and a Luminescent Image Analyzer LAS-4000 ( Fujifilm , Japan ) . Protein structures , and the data used to derive these , have been deposited at the PDB with accession numbers 5a6p ( Pikp-HMA ) , 5a6w ( Pikp-HMA/AVR-PikD complex ) .	Plant diseases reduce harvests of the world's most important food crops including wheat , rice , potato , and corn . These diseases are important for both global food security and local subsistence farming . To fight these diseases , crops ( like all plants ) have an immune system that can detect the telltale molecules produced by disease-causing microbes ( also known as pathogens ) and mount a defence response to protect the plant . Nucleotide-binding , leucine-rich repeat receptors ( or NLRs for short ) are plant proteins that survey the inside of plant cells looking for these telltale molecules . These receptors have played a central role in efforts to breed disease resistance into crop plants for decades , but little is known about how they work . Maqbool , Saitoh et al . have now used a range of biochemical , structural biology and activity-based assays to study how one NLR from rice directly interacts with a molecule from the rice blast fungus . This fungus causes the most important disease of rice ( called rice blast ) , and the fungal molecule in question is also known as an ‘effector’ protein . A technique called X-ray crystallography was used to reveal the three-dimensional structure of the effector bound to part of the NLR called the ‘integrated HMA domain’ . Biochemical techniques were then used to measure how strongly the effector ( and other related effectors ) interacted with this domain of the NLR . These results , combined with a close examination of the three-dimensional structure , allowed a set of changes to be made to the effector that stopped it interacting with the NLR protein domain in the laboratory . Maqbool , Saitoh et al . then performed experiments in rice plants and showed that changes to the effector that stopped it interacting with the NLR domain also stopped the effector from triggering a defence response in plants . Similar results were also obtained in experiments that used the model plant Nicotiana benthamiana . In the middle of the 20th century , an American plant pathologist called Harold Henry Flor proposed that the outcomes of interactions between plants and disease-causing microbes were based on interactions between specific biological molecules . The findings of Maqbool , Saitoh et al . provide a new structural basis for this model . A detailed picture of these molecular interactions will allow researchers to engineer tailored NLRs that detect a wider range of pathogen molecules . In the future such an approach could contribute to efforts to protect the world's most important crops from plant diseases .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	Structural basis of pathogen recognition by an integrated HMA domain in a plant NLR immune receptor
Cdc42 is a signaling protein important for reorganization of actin cytoskeleton and morphogenesis of cells . However , the functional role of Cdc42 in synaptic plasticity and in behaviors such as learning and memory are not well understood . Here we report that postnatal forebrain deletion of Cdc42 leads to deficits in synaptic plasticity and in remote memory recall using conditional knockout of Cdc42 . We found that deletion of Cdc42 impaired LTP in the Schaffer collateral synapses and postsynaptic structural plasticity of dendritic spines in CA1 pyramidal neurons in the hippocampus . Additionally , loss of Cdc42 did not affect memory acquisition , but instead significantly impaired remote memory recall . Together these results indicate that the postnatal functions of Cdc42 may be crucial for the synaptic plasticity in hippocampal neurons , which contribute to the capacity for remote memory recall . The synapse is a highly dynamic structure exhibiting constant turn over and remodeling . This synaptic morphoring is driven by fibrous actin ( F-actin ) , which creates the underlying cytoskeletal scaffold for neuronal structures ( Dillon and Goda , 2005; Korobova and Svitkina , 2010; Koleske , 2013 ) . Particularly , the dynamic actin turnover in dendritic spines , the postsynaptic portion of the excitatory synapse , produces morphological and functional changes in synapses ( Matus , 1999; Star et al . , 2002; Kim et al . , 2013 ) , that are thought to be a fundamental basis of synaptic plasticity underlying learning and memory ( Kim and Lisman , 1999; Krucker et al . , 2000; Fukazawa et al . , 2003; Okamoto et al . , 2004; Kim et al . , 2013 ) . Within the spine , small GTPases of the Rho family , such as Cdc42 , Rac , and RhoA exert distinct roles for actin remodeling by regulating actin organization by regulating many downstream factors including WAVE , WASP , Arp2/3 and cofilin ( Hall , 1998; Jaffe and Hall , 2005; Soderling et al . , 2007 ) . In vitro studies have shown that both Cdc42 and Rac promote the formation and maintenance of dendritic spines , whereas RhoA may negatively regulate spinogenesis ( Nakayama et al . , 2000; Scott et al . , 2003; Ahnert-Hilger et al . , 2004; Newey et al . , 2005 ) . These GTPases are regulated by more than 60 activators ( guanine nucleotide exchange factors or GEFs ) and inactivators ( GTPase activating proteins or GAPs ) ( Van Aelst and D'Souza-Schorey , 1997; Saneyoshi et al . , 2010 ) . Previous studies using two-photon fluorescence lifetime imaging microscopy ( 2pFLIM ) have demonstrated continuous activation of Cdc42 for more than 30 min within single dendritic spines undergoing structural plasticity . This activation is restricted to the stimulated spine heads and shows a steep signal gradient of active Cdc42 at the spine necks ( Murakoshi et al . , 2011; Yasuda and Murakoshi , 2011 ) , suggesting that Cdc42 may be intimately involved in long-term maintenance of structural spine plasticity during the sustained phase of spine enlargement . It is generally believed that the Cdc42 pathway plays a key role in neurite outgrowth ( Mueller , 1999; Luo , 2000; Aoki et al . , 2004 ) , neuronal polarity ( Schwamborn and Puschel , 2004; Garvalov et al . , 2007 ) , neuronal migration ( Wong et al . , 2001 ) , and dendritic morphogenesis ( Scott et al . , 2003 ) by activating the WASP-family and PAK-family pathways ( Kreis and Barnier , 2009 ) during the development of neuronal networks . Little , however , is known about the postnatal roles of Cdc42 in dendritic spines under physiological conditions . Additionally , the effects of deletion of Cdc42 on behavioral characteristics such as learning and memory have not been examined . In the present study , by using Cdc42 conditional KO mice , we evaluate how postnatal disruption of Cdc42 in excitatory neurons affects structural plasticity of dendritic spines and synaptic plasticity . To confirm the consequential effects of the Cdc42 deletion , we further investigate the spine morphology of knockout neurons and conduct a variety of behavioral tests with the mutant mice . To investigate the neuronal morphology and the behavioral outcomes provoked by postnatal Cdc42 disruption in cortical excitatory neurons in vivo , we crossed Cdc42f/f mice with the Camk2a-Cre line , which drives Cre recombinase within pyramidal neurons of the forebrain including the hippocampus and the cerebral cortex by p16-p19 ( Tsien et al . , 1996 ) . The total amount of Cdc42 protein in the hippocampus was markedly decreased in Cdc42f/f: Camk2a-Cre mice ( p120 ) compared to that of the control littermates as shown in input lanes ( Figure 1A ) . The CRIB pulldown assay , which was performed using with same amounts of hippocampal lysates from p120 mice , clearly displayed a substantial reduction of GTP-bound , active Cdc42 in the mutant mice , whereas , the total and active form levels of Rac1 were similar both in mutants and controls ( Figure 1A ) . Coomassie staining confirmed an equal loading of total protein for both groups ( Figure 1A ) . Analyses of band intensities from both genotypes revealed that ∼80% of total Cdc42 protein was lost in mutant hippocampi compared to those of littermate controls ( Figure 1B ) ( t ( 1 , 6 ) = 13 . 895 , p<0 . 0001 ) . Because the Cdc42f/f: Camk2a-Cre mice used in this study express Cre recombinase selectively in excitatory neurons , the remaining ∼20% of Cdc42 protein detected in mutant hippocampi most likely originates from other cell types such as glia or inhibitory interneurons that also express Cdc42 ( Etienne-Manneville and Hall , 2001 ) . 10 . 7554/eLife . 02839 . 003Figure 1 . Loss of hippocampal Cdc42 in Cdc42f/f: Camk2a-Cre mice . ( A ) Top and middle panels , representative western blots of Cdc42 and Rac1 levels for input ( left two lanes ) and GST-CRIB pulldowns ( right two lanes ) . Bottom panel is a representative coomassie stain showing equivalent amounts of total protein ( left two lanes ) or GST-CRIB fusion protein ( right two lanes ) . ( B and C ) Graphs depicting the quantification of ( B ) Cdc42 or ( C ) Rac1 GTPases from western blot analysis of GST-CRIB pulldowns from hippocampal lysates . n = 4 for each group . p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 02839 . 003 We next tested whether the loss of total Cdc42 led to a decrease of the active form of Cdc42 and/or a compensatory change in active Rac levels . CRIB pulldown assays revealed an 85% decrease ( 15% remained ) of active Cdc42 proteins in mutant animals ( Figure 1B ) ( t ( 1 , 6 ) = 20 . 725 , p<0 . 0001 ) . In contrast , there were no changes in total Rac1 and active Rac1 proteins in hippocampi of the mutant mice ( Figure 1C ) , suggesting a selective alteration of Cdc42 in excitatory neurons of the mutant hippocampus . Cdc42 is known as one of the critical factors regulating dendritic spine morphogenesis in cultured neurons at an early development stage ( Nakayama et al . , 2000 ) . These in vitro observations led us to conduct morphological analyses for KO neurons in vivo to determine whether Cdc42 function in the post-developmental neuron is critical for spine maintenance under physiological conditions , using the Cdc42f/f: Camk2a-Cre mice . To analyze the expression pattern of Cre recombinase in the Camk2-Cre mouse at the adult stage ( p60 ) , the Camk2a-Cre mouse was crossed with the Rosa26-lox-stop-lox-tdTomato reporter mouse ( Figure 2A ) . The cre-induced tdTomato expression was detected in cortical areas including the medial prefrontal cortex ( mPFC ) ( Figure 2B ) , anterior cingulate cortex ( ACC ) ( Figure 2C ) , and CA1 region of hippocampus ( Figure 2D ) . To ascertain the long-term effects of Cdc42 deletion in vivo , we prepared brains from p120 Cdc42f/f: Camk2a-Cre mice and conducted Goli-Cox staining ( Figure 2E–G ) . Morphological analysis of neurons in hippocampal CA1 revealed a slight ( 8% ) , yet statistically significant decrease of spine density in the CA1 pyramidal neurons of Cdc42f/f: Camk2a-Cre mice ( t ( 1 , 8 ) = 2 . 447 , p<0 . 05 ) ( Figure 2M ) . No differences in spine density were found in either the mPFC or the ACC ( Figure 2H , I , K , L ) , suggesting that hippocampal neurons are more susceptible to the Cdc42 disruption for maintenance of hippocampal spines . 10 . 7554/eLife . 02839 . 004Figure 2 . Analysis of dendritic spines in Cdc42f/f: Camk2a-Cre mice . ( A ) Schematic of breeding for the analysis of Camk2a-cre expression analysis in B–D . ( B–D ) Representative images of cre-dependent tdTomato expression in ( B ) medial pre-frontal cortex ( mPFC ) , ( C ) anterior cingulate cortex ( ACC ) , and ( D ) Hippocampus ( Hip ) . ( E–G ) Representative images of golgi stained tissue sections from ( E ) the mPFC , ( F ) the ACC , and ( G ) CA1 hippocampal region . ( H–J ) Representative images of individual dendritic segments from the ( H ) the mPFC , ( I ) ACC , or ( J ) CA1 hippocampal region ( top panels ) control Cdc42f/f or ( bottom panels ) cKO Cdc42f/f: Camk2a-Cre mice . ( K–M ) Graphs depicting the quantitative analysis of spines per 100 micron of dendritic seqements for each genotype from each region in H–J . n = 5 for each group . p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 02839 . 004 Spine enlargement is thought to be the structural basis of LTP and learning and memory ( Matsuzaki et al . , 2004; Murakoshi , et al . , 2011; Kim et al . , 2013 ) . Previously we showed that Cdc42 is required for structural and functional plasticity of dendritic spines using shRNA targeted to Cdc42 in rat hippocampus ( Murakoshi et al . , 2011 ) . To further test the roles of Cdc42 in spine plasticity , we transfected neurons of Cdc42f/f mice with EGFP-Cre or EGFP ( control ) together with mCherry as a volume marker using ballistic gene transfer , and measured spine volume change induced by glutamate uncaging in CA1 pyramidal neurons ( Figure 3A–D ) . In control neurons , stimulated spines showed sustained volume change lasted more than 30 min ( 60 ± 19% at 20–30 min ) ( Figure 3A , B ) . However , in neurons expressing EGFP-Cre , the volume change was significantly impaired ( 8 ± 5% , p<0 . 05 ) . In another set of experiments , we expressed mEGFP-Cdc42 together with mCherry and EGFP-Cre . We found that structural plasticity in these neurons was similar to those paired neurons in which EGFP and mEGFP were expressed instead of EGFP-Cre and mEGFP-Cdc42 ( 64 ± 20% vs 63 ± 23% ) ( Figure 3C , D ) . Thus , the impaired structural plasticity in neurons expressing EGFP-Cre was caused by removal of Cdc42 and exogenous Cdc42 can rescue the effect . From these experiments , we concluded that Cdc42 is necessary for spine structural plasticity , consistent with our previous report ( Murakoshi et al . , 2011 ) . 10 . 7554/eLife . 02839 . 005Figure 3 . Impaired structural and functional synaptic plasticity in Cdc42f/f: Camk2a-Cre mice . ( A ) Single spine volume changes in response to glutamate uncaging for ( open circle ) WT control Cdc42f/f or ( closed circle ) cKO Cdc42−/− mice . ( B ) Mean responses for minutes 20–30 between each genotype from ( A ) showing a significant impairment in cKO spines . N = 15 spines/15 slices for each group , p<0 . 05 . ( C ) Same as in ( A ) except the cKO neurons are co-transfected with a Cdc42 expression construct ( rescue ) . ( D ) Mean responses for minutes 20–30 between WT and cKO Cdc42−/− rescue spines are shown . N = 9 spines/9 slices for each group . ( E ) Graph depicting changes in fEPSP slope in response to high-frequency stimulation ( HFS; 100 pulses at 100 Hz; three times with 20 s intervals ) of the SC–CA1 pathway ( time zero ) . ( F ) Mean fEPSP potentiation for minutes 30–40 was significantly reduced in Cdc42−/− hippocampal slices when compared to WT littermates . N = 7 and 8 slices for WT and Cdc42−/− , respectively . p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 02839 . 005 To further examine whether Cdc42 is necessary for hippocampal LTP , we measured fEPSP in the CA1 region of hippocampal slices taken from Cdc42f/f: Camk2a-Cre mice and their litter-mate Cdc42f/f mice at P21-P28 ( Figure 3E ) . We found that , while control Cdc42f/f mice displayed a robust LTP in response to HFS ( 49 ± 18% at 30–40 min ) , LTP induction was significantly impaired in Cdc42f/f: Camk2a-Cre mice ( −2 ± 9% , p<0 . 05 ) ( Figure 3F ) . These results indicate that Cdc42 is necessary for hippocampal LTP as well as spine structural plasticity . Previously we showed that proper regulation of actin cytoskeletal remodeling in the forebrain is critical for a variety of behaviors using the Camk2a-Cre line to delete ArpC3 ( Kim et al . , 2013 ) . Although prior work clearly shows that Cdc42 is activated downstream of NMDA receptors during LTP , the importance of Cdc42 signaling in postnatal neurons for behavioral responses is unknown . To address this , a battery of behavioral tests was conducted to evaluate the Cdc42f/f: Camk2a-Cre mice from the age of p120 . We weighed mice at p120 to check for possible gross health impairments during postnatal development . Neither weight loss nor gain was noted in Cdc42f/f: Camk2a-Cre mice when compared to their littermate controls ( Figure 4A ) , suggesting normal health and development of cKO mice . 10 . 7554/eLife . 02839 . 006Figure 4 . Behaviors unaffected by loss of Cdc42 . ( A ) Average body weight of Cdc42f/f ( control , open bar ) or Cdc42f/f: Camk2a-Cre mice ( cKO , black bar ) . ( B ) Percent alternation in the Y-maze for both genotypes . Dashed line indicates the expected percent correct alternation that would be observed by chance . ( C–E ) Analysis of Open Field exploration behavior for ( C ) distance traveled , ( D ) stereotypy , or ( E ) duration spent in the center of the field . ( F–H ) Analysis of zero maze exploration behavior for ( F ) latency to enter the open arm , ( G ) number of entries to both the closed and open arms of the maze , and ( H ) total time spent in each arm . n = 14 for each group . DOI: http://dx . doi . org/10 . 7554/eLife . 02839 . 006 Y-maze spontaneous alteration was analyzed to evaluate the role of Cdc42 forebrain signaling in working memory . The tests revealed that WT and cKO mice engaged in similar numbers of alternations in the 3-way maze , suggesting an intact working memory of Cdc42 cKO ( Figure 4B ) . Locomotor activities were examined by open field test ( OFT ) . In this test , Cdc42 cKO mice displayed normal levels of locomotor ( Figure 4C ) and repetitive activities ( Figure 4D ) that were statistically indistinguishable from their control littermates , demonstrating that cKO mice do no exhibit locomotor behavioral abnormalities . The hippocampus has been reported to play a pivotal role in the processing of emotional information through circuitry connections with other brain regions such as the amygdala ( LeDoux , 2000; Engin and Treit , 2007 ) . Because our Cdc42f/f: Camk2a-Cre mice showed defects in the synaptic plasticity and reduction of dendritic spines in the hippocampus , we hypothesized the Cdc42 cKO mice might exhibit an anxiety phenotype . Analysis of the OFT data however revealed that WT and cKO mice spent similar amounts of time in the central and marginal areas of the arena compared with their littermate controls , suggesting normal anxiety levels in the cKO mice ( Figure 4E ) . Anxiety was further evaluated in the elevated zero maze ( EZM ) test . In this test , cKO mice displayed no significant differences in latencies to the open arms ( Figure 4F ) , number of entries to the open arms ( Figure 4G ) , and duration in the open arms of the maze ( Figure 4H ) when compared to those of WT controls . Together these data confirmed that the Cdc42 cKO mice show normal anxiety levels . Synaptic plasticity and neuronal morphology are intimately related to cognitive function ( Martin et al . , 2000; Segal , 2005; Kim et al . , 2013 ) . Based on our findings that Cdc42 deletion leads to defects in structural/synaptic plasticity and spine loss in hippocampus , we also suspected that the behavioral outcomes of Cdc42f/f: Camk2a-Cre mice may be abnormal in certain aspects of hippocampus-dependent cognitive tasks . To test this hypothesis we conducted a variety of behavioral analyses of episodic memory . Contextual memory capability was tested by a fear conditioning paradigm in which mice learn to predict aversive events ( mild electric shock ) ( Figure 5A ) . Following an aversive stimulus in a conditioning chamber , control Cdc42f/f and cKO Cdc42f/f: Camk2a-Cre mice showed similar freezing rates when placed in the conditioning chamber for the first 4 days ( Figure 5B ) . ANOVA with Repeated Measure ( RMANOVA ) for freezing rates of control Cdc42f/f and cKO Cdc42f/f: Camk2a-Cre mice revealed no effects of test-day or genotype and no interaction between test-day and genotype . Bonferroni corrected pair-wise comparisons revealed no significant differences between both genotypes from day 1 trough day 4 , suggesting that long-term memory retention is not affected in the cKO mice . However , a test of remote memory , which was conducted at 45 days after the conditioning , showed a marked decrease of freezing rates when compared to their WT littermates ( p<0 . 001; Bonferroni corrected pair-wise comparisons ) , even though the Cdc42 cKO mice clearly were not impaired in the initial 4 days ( Figure 5B ) . These data suggested the Cdc42 cKO mice have pronounced deficits in remote memory recall . Moreover , the normal anxiety level in the cKO mice ( Figure 4E–H ) support the fear conditioning data in that freezing rate ( an indicator of memory ) was not affected by a difference of innate anxiety levels between each genotype . 10 . 7554/eLife . 02839 . 007Figure 5 . Cdc42 cKO mice exhibit reduced memory recall in the fear conditioning learning and memory paradigm . ( A ) Schematic of the fear conditioning protocol in which the mice receive a mild aversive foot-shock on day 1 ( D1 ) in a conditioning chamber . Freezing upon placement in the chamber ( without shock ) was assessed during acquisition ( day 1 ) or for long-term ( days 2–4 ) or remote memory ( day 45 ) . ( B ) Graph depicting the average percent time spent freezing at each time point for each genotype . n = 14 for each group . p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 02839 . 007 We also performed the Morris water maze task to test the spatial memory of the Cdc42 cKO mice ( Figure 6A ) . WT and cKO mice explored a similar distance ( Figure 6B ) and expended a similar amount time ( data not shown ) to find the hidden platform during the 8 days of learning sessions . RMANOVA for the distances explored until reaching the hidden platform for WT and cKO mice revealed significant main effects of test-day ( F ( 7 , 112 ) = 28 . 858 , p<0 . 0001 ) , indicating successful spatial learning ( a reduction of swimming distance to reach the hidden platform for each subsequent test day ) of both genotypes ( Figure 6B ) . However , we found no genotype effect and no interaction between test-day and genotype . Bonferroni corrected pair-wise comparisons found no significant differences between both genotypes from day 1 through day 8 . RMANOVA for the time spent to find the hidden platform also revealed a similar result ( F ( 7 , 112 ) = 32 . 793 , p<0 . 0001 [main effect of test-day] for duration , no genotype effect and no interaction between test-day and genotype ) ( figure not shown ) . No difference of swim speed was found between WT and cKO mice throughout the acquisition trials ( data not shown ) . These data suggest that cKO mice have normal swimming abilities and spatial memory acquisition capabilities when compared to WT controls . Probe trials conducted every other day during the learning sessions ( day 2 , 4 , 6 , 8 ) showed similar long-term memory performances for both genotypes . ANOVA followed by Bonferroni pair-wise comparisons ( among four quadrants ) for each probe test revealed that both WT and cKO mice traveled a significantly greater distance in the target quadrant ( NE ) compared to other three quadrants from day 4 ( Figure 6C , D ) ( ps < 0 . 05 ) , indicating a normal capability of long-term memory formation in cKO mice . 10 . 7554/eLife . 02839 . 008Figure 6 . Cdc42 is essential for normal memory recall in the water maze test . ( A ) Schematic of the water maze testing schedule showing the acquisition phase during days ( D ) 1–8 , platform reversal phase during D9-16 , and remote probe trial test on D37 . No significant differences were observed between control Cdc42f/f and Cdc42f/f: Camk2a-Cre cKO mice during water maze acquisition phase as measured by ( B ) total swim distance to the platform , or in distance moved in the target vs non-target quadrants for ( C ) control or Cdc42 cKO mice . ( E ) Swim distances to the platform during acquisition was also unaffected in the Cdc42 cKO mice during reversal learning . ( F ) Control mice spent significantly more time in the target vs non-target quadrants during the remote memory probe trial on day 37 , however ( G ) Cdc42 cKO mice did not distinguish between these quadrants . There were significant differences between the control and Cdc42 mice in both the ( H ) distance moved and ( I ) duration of time spent within the target ( SW ) quadrant during the remote memory probe trial . n = 8 for WT; n = 10 for cKO . ps < 0 . 05 , # = no significant difference from target quadrant . DOI: http://dx . doi . org/10 . 7554/eLife . 02839 . 008 Next we performed reversal water maze trainings to test how quickly the mice learned the new location by relocating the platform to the opposite quadrant of the water tank ( SW ) . In this paradigm , cKO mice showed acquisition performances throughout the re-learning sessions ( day 9–day 16 ) that were similar to those of WT controls . RMANOVA for the distances explored and for the time spent to reach the hidden platform revealed significant main effects of test-day for swim distance ( F ( 7 , 112 ) = 23 . 612 , p<0 . 0001 ) ( Figure 6E ) and for swim duration ( F ( 7 , 112 ) = 23 . 602 , p<0 . 0001 ) ( figure not shown ) , indicating both genotypes successfully re-learned the new location of the hidden platform . There were no effects of genotype and no interactions between test-day and genotype either in distance and duration , and no significant differences were found between both genotypes as tested by Bonferroni corrected pair-wise comparisons , further indicating normal re-learning capabilities of the cKO mice . In the reversal probe tests of long-term memory formation , however , Cdc42 cKO mice showed a slight delay in long-term memory formation for the new location compared to their littermate controls ( Figure 6F , G ) . ANOVA followed by pair-wise comparisons using Bonferroni post-hoc analyses for each probe test showed that the WT mice traveled a significantly greater distance in the new target quadrant ( SW ) compared to each of the other three quadrants beginning on day 12 ( Figure 6F ) ( ps < 0 . 05 ) . In contrast , the cKO mice did not travel a significantly greater distance in the target vs the adjacent SE non-target quadrant on day 12 or day 14 ( Figure 6G ) ( p=0 . 614 for day 12; p=0 . 236 for day 14 ) . The cKO mice only distinguished the target from each non-target quadrant by day 16 , as measured by the distance traveled in the target vs each non-target quadrant ( ps < 0 . 05 ) . Probe trials at 21 days after the last reversal probe test ( day 37 ) were next carried out to evaluate the remote memory capability of the cKO mice ( bar graphs in right side of each Figure 6F , G ) . In these trials , WT mice traveled more time in target quadrant ( SW ) compared to other three quadrants . ANOVA revealed significant main effects for distance ( F ( 3 , 28 ) = 29 . 437 , p<0 . 0001 ) ( Figure 6F ) and duration ( F ( 3 , 28 ) = 34 . 363 , p<0 . 0001 ) ( figure not shown ) , suggesting that WT mice successfully recalled the location of the platform . Bonferroni post-hoc tests for each comparisons ( SW vs NE , NW , and SE ) found that traveling distances and durations of WT mice in the target quadrant ( SW ) were significantly higher than those of other three quadrants ( ps<0 . 001 for distance , ps<0 . 0001 for duration [figure not shown] ) . By contrast , cKO mice traveled similar distances and durations among the four quadrants . ANOVA followed by Bonferroni post-hoc tests revealed no main effects and no individual differences among the four quadrants , suggesting a remote memory deficit of cKO mice ( Figure 6G ) . We also examined the differences in remote memory between both genotypes . Cdc42 cKO mice traveled significantly shorter distances ( t ( 1 , 16 ) = 3 . 238 , p<0 . 01 ) ( Figure 6H ) and spent less time ( t ( 1 , 16 ) = 3 . 208 , p<0 . 01 ) ( Figure 6I ) within the target quadrant compared to those of WT controls ( independent t tests ) , again supporting a selective deficit in remote memory . Together , both the fear conditioning and water maze tests indicate that postnatal deletion of Cdc42 in forebrain excitatory neurons does not cause defects in initial memory acquisition or long-term memory retention . Instead the results of these tests show that during reversal learning loss of Cdc42 leads to a slight delay in long-term memory formation and a pervasive deficit in remote memory recall . Actin dynamics are known to subserve the activity-dependent morphological alterations of spines , which is necessary for synaptic plasticity ( Fukazawa et al . , 2003 ) . Cdc42 is one of the small GTPase proteins present in the dendritic spine , and is well known for its actin remodeling functions . Because our previous study demonstrated an intensive activation of Cdc42 specifically within the spine head following single spine glutamate uncaging stimuli ( Murakoshi et al . , 2011 ) , we expected that Cdc42 cKO neurons may show specific defects in activity-dependent spine enlargement . In line with this hypothesis , glutamate uncaging-induced spine plasticity was markedly disrupted in both the transient and sustained phase of spine enlargement upon postnatal deletion of Cdc42 . Electrophysiological LTP with Cdc42f/f: Camk2a-Cre mice also confirmed a plasticity deficit upon Cdc42 deletion , indicating that Cdc42 activation is pivotal for both activity-dependent morphogenesis of spines and in the functional synaptic plasticity . These results suggest that Cdc42 exerts its morphing activities not only in the developing neurons ( neuronal migration and establishment of polarity etc ) but also in mature neurons . Since we found a robust defect of glutamate uncaging-inducing spine enlargement in the Cdc42 cKO mice , and because Cdc42 is generally believed to play a key role in the morphogenesis of variety cells by regulating actin structure , we expected a substantial effect of Cdc42 deletion on the maintenance of spine density following long-term exposure to cre-recombinase within forebrain excitatory neurons . However , in contrary to this hypothesis , the spine loss in cKO neurons was very mild: only 8% reduction in spine density was found in hippocampal pyramidal neurons . Moreover , there was no detectable change in spine density in medial prefrontal cortex or anterior cingulate cortex of the cKO mice . This stands in contrast to our prior findings in ArpC3f/f: Camk2a-Cre mice , which exhibit a loss of approximately 50% of spines in both the hippocampus and medial prefrontal cortex ( Kim et al . , 2013 ) . Comparing these results suggest that although Camk2a-Cre drives the loss of floxed alleles in both regions , Cdc42 signaling is specifically involved in rapid time-scale spine morphing/maintaining processes during the neuronal activation but , unlike Arp2/3 , it is not strongly involved in the long-term maintenance of spine morphology in mature neurons . This observation suggests functional characteristics of Cdc42 strongly dependent on the developmental status of neuron . During the early developmental stages , Cdc42 is reported to be essential for the morphogenesis and polarity establishment of neurons that leads to a permanent change of cell shapes and locations ( Mueller , 1999; Luo , 2000; Wong et al . , 2001; Aoki et al . , 2004; Schwamborn and Puschel , 2004; Garvalov et al . , 2007 ) . However , during the postnatal period , Cdc42 activity may be functionally restricted to manage activity-dependent changes of actin dynamics that governs aspects of synaptic plasticity rather than gross maintenance of spines . Even though the hippocampi of Cdc42f/f: Camk2a-Cre mice showed robust deficits in both structural and functional measurements of synaptic plasticity , the cKO mice surprisingly displayed normal performances in many behavioral tests , including those for working memory and memory acquisition . Remote memory as determined by both the fear conditioning and water maze tests , however , was significantly diminished in the Cdc42 cKO mice . How does the loss of Cdc42 lead to the deficit in the process of remote memory ? Two explanations are possible to account for the remote memory defect of the cKO mice: a memory storage defect , or a remote memory retention and retrieval problem . Normal performances in learning and intact working memory of the cKO mice suggest that memory formation and storage processing appears not to be altered by Cdc42 deletion , although it should be noted we did observe a slight delay in long-term memory formation that was specific to the water maze reversal test . We speculate the retention or the retrieval capacities for the remote memory are most likely affected in Cdc42f/f: Camk2a-Cre mice . Notably , a line of studies have found that hippocampal lesions impair recent memory such as short- and long-term memories , whereas the lesions do not affect remote memory ( Squire and Alvarez , 1995; Knowlton and Fanselow , 1998 ) . This ‘graded retrograde amnesia’ supports an idea that remote memory retrieval may be independent of hippocampal functions . This finding is opposite to the behavioral phenotype observed in Cdc42f/f: Camk2a-Cre mice . However , other studies using human patients who have medial temporal lobe ( MTL ) damages find these patients exhibit a memory loss without any temporal gradient ( Rosenbaum et al . , 2001; Moscovitch et al . , 2006; Winocur et al . , 2010 ) , revealing that the retrograde amnesia is not always graded . Moreover , a recent study showed the hippocampus is tightly involved in the retrieval of remote memory . Remote memory recall was affected by a temporal and precise optogenetic inhibition of CA1 excitatory neurons in hippocampus ( Goshen et al . , 2011 ) . If Cdc42 is critically involved in the neuronal activation during the process of remote memory retrieval , Cdc42 loss may result in a remote memory defect with normal learning and working memory . Rac and Cdc42 share similar downstream pathways: both Rac and Cdc42 can remodel actin via the p21 Kinase ( PAK ) and LIMK pathway to inactivate cofilin-mediated actin severing or by stimulating the WAVE1/ N-WASP pathway to activate Arp2/3 dependent actin polymerization . Both pathways are implicated in synaptic plasticity and are thought to be critical for processes important for learning and memory . Thus it has remained unclear as to whether Rac or Cdc42 can be distinguished from each other at the level of behavioral phenotypes in intact animals . It is of interest , therefore , to compare the results reported here with the analogous Rac1f/f: CamKllα-Cre mice previously characterized ( Haditsch et al . , 2009 ) . Loss of Rac1 results in impaired hippocampal LTP similar to our findings in the Cdc42 cKO mice . Surprisingly , however , the behavioral impairments are quite different . Loss of Rac1 results in impaired working memory , but has no effect on long-term or remote memory . This is in stark contrast to the Cdc42 cKO mice which display normal working memory , but are impaired in remote memory . Together , these findings reveal that although the biochemical pathways that modulate actin remodeling evoked by synaptic activation of Rac1 and Cdc42 may overlap , their functions during learning and memory are clearly distinguishable from the other , with little overlap . Further work will be required to define how the physiologic relevance of Rac1 and Cdc42 are functionally segregated during these processes despite their similar biochemical pathways . Conditional Cdc42 knockout animals were generously provided by Dr Cord Brakebusch ( University of Copenhagan ) . Cdc42 conditional knockout mice ( Cdc42f/f ) were crossed with the Camk2a-Cre line ( stock no . 005359; The Jackson Laboratory; Bar Harbor , ME ) for synaptic plasticity studies , biochemical assays , Golgi-cox staining , and behavioral tests . To analyze the expression pattern of Cre recombinase , the Rosa26-lox-stop-lox-tdTomato reporter line ( generously provided by Dr Fan Wang ) was crossed with the Camk2a-Cre line . Littermate male and female mice from heterozygous pairings were used in all experiments . All mice were housed in the Duke University's Division of Laboratory Animal Resources facilities and all procedures were conducted as approved by the Duke University Institutional Animal Care and Use Committee in accordance with National Institutes of Health guidelines . Glutamate uncaging was performed as described ( Murakoshi et al . , 2011 ) . P6 Cdc42f/f pups were deeply anesthetized with isoflurane and decapitated; the hippocampus was rapidly dissected into medium containing ( mM ) : HEPES 25 , NaHCO3 2 , Sucrose 248 , glucose 10 , KCl 4 , MgCl2 5 , CaCl2 1 . Then , 350 µm slices were cut with a tissue chopper ( Ted Pella , Inc . ; Redding , CA ) and transferred to the surface of membrane inserts ( EMD Millipore; Darmstadt , Germany ) , and placed in culture media containing ( mM ) : L-glutamine 1 , CaCl2 1 , MgSO4 2 , D-glucose 12 . 9 , NaHCO3 5 . 2 , Na-HEPES 30 , insulin 0 . 001 , Ascorbic acid 0 . 53 , 20% heat-inactivated horse serum , 80% HEPES-based MEM 8 . 4 g/l ( pH 7 . 35 , 320 Osm ) . Slice-containing plates were maintained in a 37°C incubator with 5% CO2 . 5–10 days after incubation , cultures were transfected biolistically with a gene gun . To make bullets , 40–50 µg DNA were mixed with 8–12 mg 1 . 6 µm gold particles ( Bio-Rad; Hercules , CA ) . Amount of DNA used was: 20 µg EGFP + 20 µg mCherry ( Control ) ; 20 µg EGFP-Cre + 20 µg mCherry ( Cdc42 knockout ) ; 20 µg EGFP + 20 µg mCherry + 10 µg monomeric EGFP ( mEGF ) ( Control for rescue ) ; 20 µg EGFP-Cre + 20 µg mCherry + 10 µg mEGFP-Cdc42 ( rescue ) . Uncaging experiments were performed 3–4 days after transfection . A Ti:Sapphire laser was tuned to 720 nm to uncage the caged glutamate in artificial cerebral spinal fluid ( ACSF ) that contained ( mM ) : NaCl 130 . 0 , KCl 2 . 5 , NaHCO3 2 , 0 , NaH2PO4 1 . 25 , glucose 25 . 0 , CaCl2 4 . 0 , tetrodotoxin 0 . 001 and MNI-caged L-glutamate 2 . 0 at 25–27°C . Structural plasticity associated with LTP was induced by 4–7 ms , 5–8 mW uncaging pulses applied at 0 . 5 Hz for 30 pulses . Spines of neurons expressing mCherry and EGFP were imaged with a 1030 nm ytterbium-doped laser ( Amplitude ) . EGFP-Cre expression was confirmed with the presence of strong fluorescence in the nucleus . Volume change was monitored by measuring the change in the intensity of EGFP or mCherry fluorescence over time . Hippocampal slices , 400 µm thick , were prepared from Cdc42f/f: Camk2a-Cre mice and their litter-mate control Cdc42f/f mice ( P21-P28 ) . Field EPSP ( fEPSP ) was measured with a pipette filled with 1 M NaCl located at the dendritic level ( ∼100 µm away from the somatic layer ) , while stimulating axons with a bipolar tungsten electrode located at striatum radiatum ( single pulses , 50–100 µs , 30 s intervals ) . Stimulation strength was adjusted so that fEPSP amplitude is less than 50% saturation . LTP was induced with three sets of high frequency stimulation ( HFS; 100 Hz , 1 s , 20 s intervals ) . The experiments were performed in artificial CSF ( ACSF ) containing 2 mM MgCl2 and 2 mM CaCl2 at room temperature . Persons who performed experiments and analyses were blinded from genotypes until the whole experiments and analyses are finished and statistical significance was calculated . Golgi-Cox staining procedures were performed as described ( Kim et al . , 2013 ) . Mice were deeply anesthetized with isoflurane and then transcardially perfused with 4% PFA . Brains were removed and treated with solutions A and B from the FD Rapid GolgiStain Kit ( FD Neuro Technologies , Columbia , MD ) for 2 weeks , and then treated with solution C for 7 days . Sections ( 100 μm thick ) were cut by cryostat and transferred to solution C and incubated for 24 hr at 4°C . After brief rinsing with distilled water , floating sections were stained consecutively with solutions D and E for 30 min and then transferred to a 0 . 05% gelatin solution . Sections were mounted on glass slides , dehydrated through a graded series of ethanol concentrations , and then mounted with Permount . Images were collected by a MicroPublisher 5 . 0 MP color camera ( QImaging; Surrey , BC , Canada ) on a Zeiss Axio Imager microscope under a 100 × oil-immersion objective using MetaMorph 7 . 6 . 5 software . For quantification , spine density from segments of secondary or tertiary branches of CA1 pyramidal neurons in the stratum radiatum of the hippocampus were measured . Cdc42 activity was measured using glutathione S-transferase ( GST ) -Cdc42 and Rac1 interactive binding domain ( CRIB ) binding assay as described ( Kim et al . , 2009 ) . Cdc42f/f: Camk2a-Cre and control Cdc42f/f mice were deeply anesthetized with isoflurane and hippocampi were rapidly removed , homogenized with 15 strokes using a Teflon-glass homogenizer in ice-cold lysis buffer ( mM ) : NaCl 150 , HEPES 25 ( pH 7 . 4 ) , EDTA 1 , EGTA 0 . 5 , 0 . 5% NP-40 that contained , protease/phosphatase inhibitors and then sonicated . The homogenate was centrifuged at 10000×g for 5 min at 4°C and then the supernatant was collected . The protein concentration was determined with the Bradford protein assay ( Bio-Rad ) . For GST-CRIB binding assay , GST protein fused to CRIB domain from human PAK1B was expressed in E . coli BL-21 cells and used for the assay . Identical amounts of purified GST-CRIB proteins were pre-incubated with 20 µl of glutathione-Sepharose beads ( 4 Fast Flow; Amersham Biosciences; Pittsburgh , PA ) in NETN buffer ( 20 mM Tris–HCl , pH 8 . 0; 100 mM NaCl , 1 mM EDTA , 0 . 5% NP-40 ) , then washed twice each with 600 μl of NETN and 600 μl of lysis buffer . The hippocampal lysates were incubated with the GST-CRIB protein-bound glutathione-Sepharose beads for 2 hr at 4°C on a rotator ( 15 rpm ) . The beads were collected by centrifugation ( 1500×g/1 min ) , and the supernatant was removed and the pellet was rinsed with lysis buffer . The bound proteins were eluted from the beads by boiling the samples in SDS loading buffer ( 1 M Tris–HCl [pH 6 . 8] , 10% [vol/vol] SDS , 50% [vol/vol] glycerol , 5% [vol/vol] 2-mercaptoethanol , and 1% [vol/vol] bromophenol blue ) . For Western blotting , 10 μg of samples were electrophoresed through 12% SDS-PAGE ( Bio-Rad ) and transferred onto a nitrocellulose membrane ( Whatman; Pittsburgh , PA ) , and nonspecific sites were blocked with 5% nonfat dry milk in TRIS-buffered saline ( TBS; pH 7 . 4 ) containing 0 . 05% Tween-20 . For detection , the membranes were probed with rabbit anti-Cdc42 polyclonal antibody ( Santa Cruz; Dallas , TX ) and mouse anti-Rac1 monoclonal antibody ( BD; San Jose , CA ) for 24 hr at 4°C . After washing , the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies ( GE Healthcare Life Sciences; Pittsburgh , PA ) for 2 hr , washed , and then developed using the ECL system ( Thermo Scientific; Waltham , MA ) . Membranes were then exposed to autoradiography films ( Genesee Scientific; San Diego , CA ) . Fear conditioning was conducted as described ( Porton et al . , 2010 ) . Med-Associates mouse fear conditioning chambers were used for conditioning and testing . The tests consist of three sessions: conditioning ( day 1 ) , long-term fear memory tests ( day 2-day 5 ) , and remote fear memory test ( day 45 ) . Following a 2 min acclimation in the conditioning chamber , mice received a 0 . 4 mA scrambled foot shock for 2 s . Each mouse remained in the chamber for an additional 30 s before being placed into its home cage . Fear memory testing was conducted daily for 4 days by placing the mice in the conditioning chamber for 5 min in the absence of foot shock . The remote memory was tested at 45 days after the conditioning . Freezing rate was analyzed by trained observers who were blind to the genotypes of mice using Noldus Observer ( Noldus Information Technology; Wageningen , Netherlands ) software . Freezing was defined by criteria previously described for mice as the absence of all visible movement except that required for respiration ( Anagnostaras et al . , 2000 ) . Morris water maze task was conducted as described ( Porton et al . , 2010 ) . A 120 cm diameter water tank was used . Opaque water in the tank was maintained at 25°C . The water pool was divided into four quadrants ( NE , NW , SE , and SW ) . A 12 cm diameter round platform was submerged 1 cm below the water surface and 20 cm apart from the wall of the water tank at the NE quadrant . Testing consisted of three sessions: acquisition and probe trials ( day 1-day 8 ) , reversal acquisition and reversal probe trials ( day 9-day 16 ) , and remote probe trials ( day 37 ) . 1 week prior to testing , all mice were handled daily for 5 min and then were placed in a pan of shallow water ( 1 cm ) for 30 s to acclimate them to water . On the seventh day after handling , each mouse was placed onto the hidden platform in the NE quadrant for 20 s and then allowed to swim freely for 60 s before being returned to the platform for 15 s . Acquisition testing consisted of 32 trials given across 8 days with four trials administered per day . Trials were run in pairs , with each pair separated by 60 min . Probe trials were conducted without platform at the end of days 2 , 4 , 6 and 8 . Reversal acquisition and reversal probe tests were conducted same way as the acquisition/probe tests described above , but the platform location was moved from NE to SW . Remote test was conducted at day 37 . For each trial , the release point for the animals was randomized across seven equally spaced points along the perimeter of the maze . All test trials were 1 min in duration . The swim distance , swim time were determined by Noldus Ethovision ( Noldus Information Technology ) . Y-maze test was performed as described ( Kim et al . , 2013 ) . Spontaneous alternation in a Y-maze was conducted under indirect illumination ( 80–90 lux ) in a 3-arm Y-maze . The mouse was placed into the center arm of the maze and permitted free exploration for 5 min . Entry into an arm was defined as the mouse being more than 1 body length into that arm , with both hind-paws past the entrance to that arm . An arm alternation was defined as three successive entries into each of the different arms . Alternation , calculated as the total number of alternations divided by the total number of arm entries minus 2 , was expressed as a percentage . OFT was performed as described ( Kim et al . , 2013 ) . Mice were placed into an open field ( AccuScan Instruments ) and their activities were monitored over 1 hr under 350 lux illumination using VersaMax software ( AccuScan Instruments; Columbus , OH ) . Locomotor ( distance traveled ) , rearing ( vertical beam-breaks ) , stereotypical activities ( repetitive beam-breaks <1 s ) , and anxiety level ( duration in center area of arena ) were measured in 5-min time-bins . EZM was performed as described ( Pogorelov et al . , 2005 ) . The zero maze is a 5 . 5 cm-wide circular running platform elevated 43 cm from the floor . The inside diameter of the maze is 34 cm with two opposite quadrants were enclosed by 11 cm-high walls . Mice were placed into a closed quadrant and permitted to investigate the maze for 5 min under 50–60 lux illumination . The behaviors were recorded and analyzed by Noldus Observer ( Noldus Information Technology ) . The scored behaviors included percent of time spent in open quadrants and total number of transitions between quadrants , and latency to enter the open quadrants . All data are expressed as means ± SEM and all statistics were analyzed using SPSS software ( SPSS 20 ) . Independent t tests were used for analysis of differences between two groups . When comparing more than two groups , ANOVA followed by Bonferroni post-hoc analyses was used . To monitor changes over time , repeated-measures ANOVA ( RMANOVA ) were run followed by Bonferroni corrected pair-wise comparisons . A p<0 . 05 was considered statistically significant .	Neurons communicate with one another at junctions called synapses , which are typically formed between the dendrite of one neuron and the axon terminus of another . The dendrites are protrusions coming out of the cell body that receive inputs from other cells; the axon is a cable-like structure that enables neurons to contact other cells . In excitatory neurons in part of the brain called the hippocampus , the dendrites are themselves covered in structures called spines , so most synapses are formed between an axon terminus ( belonging to the presynaptic cell ) and a dendritic spine ( on the postsynaptic cell ) . The hippocampus is necessary for the formation of long-term memories . The strength of a synapse can increase or decrease over time—a property that is called synaptic plasticity . Changes in the strength of synapses are thought to underlie learning and memory , and long-lasting changes in synaptic strength involve increases or decreases in the number and size of dendritic spines . Such changes are possible because spines have an internal skeleton that can be assembled and disassembled in a matter of minutes . This ‘remodeling’ process is regulated by a family of enzymes called small GTPases . One of these , known as Cdc42 , has been shown to promote the formation and maintenance of spines in cell culture , but its role in synaptic plasticity , learning and memory remains unknown . Now , Kim , Wang et al . have used genetically modified mice who have had Cdc42 deleted from excitatory neurons in their forebrain to examine the functions of this enzyme in living animals . These ‘knockout’ mice showed a small but statistically significant reduction in the number of dendritic spines in the hippocampus . They also showed smaller changes in spine volume and impaired long-term synaptic plasticity in the hippocampus . When the mice performed long-term memory tests where they learnt to associate a specific set of visual cues with an impending electric shock , the knockout mice performed well for up to a few days . However , when tested again on the same task 45 days later , the knockout mice did not perform as well as normal mice . This is surprising , given the presumed role of long-term synaptic plasticity in learning and memory , and indicates that Cdc42 is required for ‘remote memory’ , the form of memory lasting for many days . Similar results were obtained with another memory test using a water maze , where the animals have to remember the location of a hidden platform . Normal mice remember the location for more than 30 days . In contrast , the knockout mice could only remember the location for a few days . As well as providing the first demonstration of the role of Cdc42 in synaptic plasticity in live animals , the work of Kim , Wang et al . has provided new insights into the functions of this enzyme in memory . Further work is required to determine how Cdc42 interacts with other proteins present at synapses .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Loss of Cdc42 leads to defects in synaptic plasticity and remote memory recall
Glia in the central nervous system engulf neuron fragments to remodel synapses and recycle photoreceptor outer segments . Whether glia passively clear shed neuronal debris or actively prune neuron fragments is unknown . How pruning of single-neuron endings impacts animal behavior is also unclear . Here , we report our discovery of glia-directed neuron pruning in Caenorhabditis elegans . Adult C . elegans AMsh glia engulf sensory endings of the AFD thermosensory neuron by repurposing components of the conserved apoptotic corpse phagocytosis machinery . The phosphatidylserine ( PS ) flippase TAT-1/ATP8A functions with glial PS-receptor PSR-1/PSR and PAT-2/α-integrin to initiate engulfment . This activates glial CED-10/Rac1 GTPase through the ternary GEF complex of CED-2/CrkII , CED-5/DOCK180 , CED-12/ELMO . Execution of phagocytosis uses the actin-remodeler WSP-1/nWASp . This process dynamically tracks AFD activity and is regulated by temperature , the AFD sensory input . Importantly , glial CED-10 levels regulate engulfment rates downstream of neuron activity , and engulfment-defective mutants exhibit altered AFD-ending shape and thermosensory behavior . Our findings reveal a molecular pathway underlying glia-dependent engulfment in a peripheral sense-organ and demonstrate that glia actively engulf neuron fragments , with profound consequences on neuron shape and animal sensory behavior . To interpret its environment accurately and respond with appropriate behaviors , an animal’s nervous system needs to faithfully transmit information from the periphery and through neuron–neuron contacts within the neural network . Precision in this information transfer and processing depends partly on neuron-receptive endings ( NREs ) , specialized subcellular structures where a neuron receives input from either the external environment or other neurons ( Bourne and Harris , 2008; Harms and Dunaevsky , 2007; Shaham , 2010; Singhvi et al . , 2016 ) . In the peripheral nervous system ( PNS ) , sensory NREs house the sensory transduction machinery and appropriate NRE shape is important for sensory information capture . In the central nervous system ( CNS ) , the size and number of interneuron NREs ( dendritic spines ) help determine the connectome and thereby the path of information transfer ( Bargmann and Marder , 2013; Eroglu and Barres , 2010; Nimchinsky et al . , 2002 ) . While remodeling of NRE shape has been suggested to be important for experiential learning and memory ( Bourne and Harris , 2008; Harms and Dunaevsky , 2007 ) , directly correlating these subcellular changes with animal behavior has been challenging . Glia are a major cell type of the nervous system and approximate neurons in number ( von Bartheld et al . , 2016 ) . They have been proposed to actively modulate development , homeostasis , and remodeling of neural circuits , and are thought to influence NRE shape and numbers ( Allen and Eroglu , 2017; Stogsdill and Eroglu , 2017; Zuchero and Barres , 2015 ) . One mechanism by which glia may do so is by engulfment of neuron fragments , including NREs ( Freeman , 2015; Schafer and Stevens , 2013; Wilton et al . , 2019 ) . Aberrant neuron fragment uptake by glia is implicated in neurodevelopmental as well as neurodegenerative diseases , including Alzheimer’s dementia , autism , and epilepsy ( Chung et al . , 2015; Henstridge et al . , 2019; Neniskyte and Gross , 2017; Schafer and Stevens , 2013; Vilalta and Brown , 2018; Wilton et al . , 2019 ) . Fundamental questions about the roles and mechanisms of glia-dependent phagocytosis remain open . Whether glia initiate engulfment or passively respond to neuron shedding is unclear . Furthermore , correlating glia-dependent remodeling at single synapse or NREs with changes in animal behavior remains challenging in most systems ( Koeppen et al . , 2018; Wang et al . , 2020 ) . Also , glial engulfment mechanisms have been primarily dissected in the context of injury or development , and their impact on adult neural functions remains less understood . Finally , whether glia-dependent engulfment occurs in the peripheral nervous system ( PNS ) or dictates normal sensory functions has not been extensively explored . The nervous system of the adult Caenorhabditis elegans hermaphrodite comprises 302 neurons and 56 glial cells ( Singhvi and Shaham , 2019; Sulston et al . , 1983; White et al . , 1986 ) . These arise from invariant developmental lineages , form invariant glia–neuron contacts , and each neuron performs defined functions to enable specific animal behaviors . These features allow single-cell and molecular analyses of individual glia–neuron interactions with exquisite precision ( Singhvi et al . , 2016; Singhvi and Shaham , 2019 ) . Here , we describe our discovery that the C . elegans AMsh glial cell engulfs NRE fragments of the major thermosensory neuron of the animal , AFD . Thus , this critical glial function is conserved in the nematode and across sense-organ glia . We find that engulfment requires the phospholipid transporter TAT-1/ATP8A , α-integrin PAT-2 , and glial phosphatidylserine receptor PSR-1 . PSR-1 engages a conserved ternary GEF complex ( CED-2/CrkII , CED-5/DOCK180 , CED-12/ELMO1 ) to activate CED-10/Rac1 GTPase . The actin remodeling factor WSP-1/nWASp , a known effector of CED-10 , acts in AMsh glia to regulate engulfment . We also show that glial engulfment rates are regulated by temperature and track AFD neuron activity . Importantly , glial CED-10/Rac1 acts downstream of neuron activity , and CED-10 expression levels dictate NRE engulfment rates . Finally , perturbation of glial engulfment leads to defects in AFD–NRE shape and associated animal thermosensory behavior . Our studies show that glia actively regulate engulfment by repurposing components of the apoptotic phagocytosis machinery . Importantly , while cell corpse engulfment is an all-or-none process , glia-dependent engulfment of AFD endings can be dynamically regulated . We propose that other glia may similarly deploy regulated phagocytosis to tune sensory NREs and synapses , and to dynamically modulate adult animal behaviors . Glia of the nematode C . elegans share molecular , morphological , and functional features with vertebrate sense-organ glia and astrocytes ( Bacaj et al . , 2008a; Katz et al . , 2018; Katz et al . , 2019; Lee et al . , 2021; Singhvi and Shaham , 2019; Wallace et al . , 2016 ) . In previous studies , we established the AMsh glia–AFD neuron pair as a tractable experimental platform to define molecular mechanisms of single glia–neuron interactions ( Singhvi et al . , 2016; Singhvi and Shaham , 2019; Wallace et al . , 2016 ) . The AFD–NRE comprises ~45 actin-based microvilli and a single microtubule-based cilium that are embedded in the AMsh glial cell . An adherens junction between the AFD–NRE base and the AMsh glial cell isolates this glia–NRE compartment ( Figure 1A , B; Doroquez et al . , 2014; Perkins et al . , 1986 ) . Upon imaging fluorescently labeled AFD–NREs in transgenic animal strains , we consistently observed labeled fragments disconnected from the neuron ( Figure 1C , C’ , Video 1 ) . Our previous reconstructions-based FIB-SEM serial section data had also revealed AFD–NRE fragments disconnected from the rest of the AFD neuron ( marked yellow , Video 1 ) in Singhvi et al . , 2016 . We examined this further using two-color imaging , which revealed that many of these fragments reside within the AMsh glial process and cell body ( Figure 1D–F’ , Video 2 ) . To confirm that these glial puncta do not reflect spurious reporter protein misexpression in glia but rather derive from the AFD , we ablated AFD neurons early in larval development and looked for puncta on the first day of adulthood . Upon ablation of one of the two bilateral AFD neurons by laser microsurgery in first larval stage ( L1 ) animals , fragment formation was blocked on the operated side , but not on the unoperated side , or in mock-ablated animals ( Figure 1G , H ) . Similar results were seen with stochastic genetic ablation of AFD using the pro-apoptotic BH3-domain protein EGL-1 , expressed using an embryonic AFD-specific promoter ( Figure 1I ) . We conclude , therefore , that AMsh glia engulf fragments of the AFD–NRE in C . elegans . 3D super-resolution microscopy studies revealed that the average size of AFD-derived glial puncta is 541 ± 145 nm along their long ( yz ) axis ( Figure 2A ) . These fragments are an order of magnitude smaller than recently described exophers extruded from neurons exposed to cellular stress ( ~3 . 8 µm in diameter ) and larger than ciliary extracellular vesicles ( ~150 nm ) ( Chung et al . , 2013; Melentijevic et al . , 2017; Wang et al . , 2014 ) . This size is of the same order of magnitude as the sizes of individual AFD–NRE microvilli or cilia as measured by electron microscopy ( Figure 2B , Figure 2—figure supplement 1A ) and ( Doroquez et al . , 2014 ) . Engulfment of neuronal fragments by glia has been suggested to refine neuronal circuit connectivity during neural development ( Chung et al . , 2013; Wilton et al . , 2019 ) . Post development , glial engulfment is thought to regulate animal behaviors and memory ( Koeppen et al . , 2018; Wang et al . , 2020 ) . To determine when C . elegans AMsh glia initiate engulfment of AFD–NRE fragments , we counted engulfed NRE puncta at different life stages . We found that these puncta are rarely found in embryos or early larval stages , but are easily detected in L4 larvae and increase in numbers during adulthood ( Figure 2C , D ) . Thus , consistent with L1 laser ablation studies ( Figure 1G–I ) , engulfment of AFD–NREs by glia occurs after development of the AFD–NRE is largely complete . We found that ~65% of 1-day old adult animals expressing the AFD–NRE-specific gcy-8:GFP raised at 20°C have AMsh glia containing >10 puncta , and another ~32% of animals have 1–9 puncta/glia ( n = 171 ) ( Figure 2C ) ( see Materials and methods for binning details ) . The AMsh glial cell of 1-day-old adults has on average 14 ± 1 puncta ( n = 78 ) ( Figure 2D ) . Using time-lapse microscopy , we found that individual puncta separate from the NRE at a frequency of 0 . 8 ± 0 . 3 events/min and travel at 1 . 05 ± 0 . 1 µm/s down the glial process towards the cell body , consistent with motor–protein-dependent retrograde trafficking ( quantifications of videos from n = 5 animals ) ( Figure 2—figure supplement 1B , Videos 1 and 2; Maday et al . , 2014; Paschal et al . , 1987 ) . Finally , age-matched animals raised at different cultivation temperatures differ in glia puncta accumulation ( Figure 2E ) . AFD–NREs comprise multiple microvilli and a single cilium ( Figure 1B ) . The size of puncta we observed ( 541 ± 145 nm , Figure 2A ) was similar to the diameters of both the microvilli ( 214 ± 30 nm ) ( Figure 2B , Figure 2—figure supplement 1A ) and AFD cilium ( 264 ± 13 nm ) ( Doroquez et al . , 2014 ) , precluding easy inference of the source of these puncta . To distinguish which organelle was engulfed , we undertook two approaches . First , we labeled each organelle with specific fluorescent tags and examined uptake by AMsh glia . To probe microvilli , we examined transgenic animals labeled with either of four AFD-microvilli-specific proteins with fluorescent tags , SRTX-1 , GCY-8 , GCY-18 , and GCY-23 ( Colosimo et al . , 2004; Inada et al . , 2006 ) . We found that all four transgenic strains consistently show fluorescent puncta in glia ( Figure 1 , Figure 3A ) . Time-lapse microscopy of one of these ( Psrtx-1:SRTX-1:GFP ) also revealed that fragments originate from the AFD–NRE microvilli ( Figure 2—figure supplement 1B , Videos 1 and 2 ) . To label cilia , we generated transgenic animals with the ciliary protein DYF-11/TRAF31B1 fluorescently tagged and expressed under an AFD-specific promoter and confirmed that PAFD:DYF-11:GFP localizes to AFD cilia ( Figure 3B ) . However , we found no DYF-11:GFP puncta in AMsh glia ( Figure 3A ) . In a complementary approach , we examined mutants lacking either microvilli or cilia . The development of AFD , including its microvilli ( but not cilia ) , requires the terminal selector transcription factor TTX-1/Otx1/orthodenticle ( Hobert , 2016; Satterlee et al . , 2001 ) . We found that ttx-1 ( p767 ) mutants lack AFD–NRE puncta in AMsh glia ( Figure 3C–E ) . Cilia development requires the IFT-B early assembly proteins DYF-11/TRAF31B1 and OSM-6/IFT52 . Both are expressed in most , if not all , ciliated neurons , and mutations in the respective genes exhibit defective amphid cilia ( Bacaj et al . , 2008a; Collet et al . , 1998; Kunitomo and Iino , 2008; Li et al . , 2008; Perkins et al . , 1986; Starich et al . , 1995 ) . In contrast to ttx-1 mutants , glia puncta were present in animals mutant for either dyf-11 ( mn392 ) or osm-6 ( p811 ) ( Figure 3C–E ) . In fact and on the contrary , we found that dyf-11 cilia-defective mutants accumulate more glial puncta that wild-type animals ( dyf-11: 38 ± 3 puncta , n = 27 vs . wild type: 14 ± 1 , n = 78 , Figure 3C , E; and a larger fraction of dyf-11 and osm-6 mutants exhibit >10 puncta/glia [dyf-11: 95% , n = 61 animals , osm-6: 100% animals , n = 82 , vs . wild type: 65% , n = 171]; Figure 3D ) . This indicates that cilia are likely not the primary source of glia puncta . Data from all these approaches taken together suggest that that the observed puncta in AMsh glia derive from AFD–NRE microvilli as the primary , if not sole , source . What molecular mechanism drives AFD–NRE microvilli engulfment ? In other contexts , neurons expose the membrane phospholipid phosphatidylserine ( PS ) on the outer leaflet of the plasma membrane as a signal for glial phagocytosis ( Hakim-Mishnaevski et al . , 2019; Li et al . , 2020; Nomura-Komoike et al . , 2020; Raiders et al . , 2021; Scott‐Hewitt et al . , 2020 ) . However , the underlying molecular mechanisms that regulate this exposure in neurons are unclear . Apoptotic corpse phagocytosis , including in C . elegans , is also mediated by PS exposure ( Figure 4A ) . PS exposure in apoptotic cells is promoted partially by the Xkr8 factor CED-8 , which is cleaved by the caspase CED-3 to promote PS presentation for cell corpse phagocytosis ( Bevers and Williamson , 2016; Wang et al . , 2007 ) . However , mutations in neither ced-8 ( Figure 4B ) nor ced-3 ( data not shown ) affect glial NRE uptake . Likewise , mutations in scrm-1 , encoding a scramblase-promoting PS exposure ( Wang et al . , 2007 ) , only mildly decrease AFD–NRE engulfment ( Figure 4B ) . However , a presumptive null mutation in tat-1 , an ortholog of mammalian translocase ATP8A required for PS sequestration to the plasma membrane inner leaflet ( Andersen et al . , 2016 ) , results in increased apoptotic cell corpse engulfment ( Darland-Ransom et al . , 2008; Hong et al . , 2004 ) and AFD–NRE engulfment ( Figure 4B , E ) . Thus , common and context-specific mechanisms control apoptotic and NRE engulfment . Importantly , re-expression of wild-type tat-1 cDNA under an AFD-specific promoter fully rescues the tat-1 engulfment defect ( Figure 4B ) . We conclude that cell-autonomous function of the PS-flippase TAT-1 in the AFD neuron regulates engulfment of AFD–NRE fragments by AMsh glia . How is PS on the AFD membrane recognized by AMsh glia ? To address this question , we examined mutants in receptors required for C . elegans apoptotic cell engulfment ( Figure 4A ) . CED-1/Draper/MEGF10 is required for removal of neuron debris in many contexts ( Cherra and Jin , 2016; Mangahas and Zhou , 2005; Nichols et al . , 2016 ) , including by glia in other species ( Chung et al . , 2013; Freeman , 2015; Hamon et al . , 2006; Raiders et al . , 2021 ) . Surprisingly , two independent ced-1 loss-of-function alleles do not block NRE fragment uptake ( Figure 4C ) . Similarly , disrupting CED-6/GULP and CED-7/ABCA1 , which function with CED-1/MEGF10 in C . elegans apoptotic phagocytosis and in other species ( Flannagan et al . , 2012; Hamon et al . , 2006; Morizawa et al . , 2017; Reddien and Horvitz , 2004; Zhou et al . , 2001 ) , does not block engulfment either ( Figure 4C ) . Further , mutations in tyrosine kinases related to MeRTK , required for astroglial engulfment of neuronal debris in vertebrates ( Chung et al . , 2013 ) , also seem to not be required for AMsh engulfment of AFD–NRE ( Figure 4—figure supplement 1A; Popovici et al . , 1999 ) . Loss of the conserved phosphatidylserine receptor PSR-1/PSR has defects in apoptotic cell corpse engulfment in C . elegans and zebrafish ( Hong et al . , 2004; Wang et al . , 2003 ) . Remarkably , deletion of psr-1 dramatically reduces AFD–NRE engulfment by AMsh glia ( Figure 4D , E ) . Expression of the PSR-1C long isoform in AMsh glia rescues psr-1 mutant defects significantly ( Figure 4D ) , suggesting that PSR-1 acts in glia to promote NRE uptake . Consistent with this function , a GFP:PSR-1 translational reporter expressed under an AMsh-glia-specific promoter localizes to glial membranes , including those around AFD–NRE microvilli ( Figure 4F , F’ ) . If PSR-1 recognizes PS on AFD–NRE membranes to mediate engulfment , we reasoned it should act downstream of TAT-1 . We therefore constructed and analyzed psr-1; tat-1 double mutants . Unlike tat-1 single mutants that show increased NRE engulfment , psr-1; tat-1 animals exhibit reduced engulfment similar to psr-1 single mutants ( Figure 4D ) . Thus , PSR-1 acts downstream of TAT-1 . The transthyretin protein TTR-52 mediates binding between PS and PSR-1 ( Neumann et al . , 2015; Wang et al . , 2010 ) . Supporting the PSR-1 results , we found that a mutation in ttr-52 also reduces NRE uptake to a similar extent as mutations in psr-1 ( Figure 4D ) . In addition , we found that psr-1; ttr-52 double mutants show no significant enhancement of puncta defects compared to either single mutant , suggesting that PSR-1 and TTR-52 function within the same pathway for PS recognition by AMsh glia ( Figure 4D ) . Although psr-1 loss reduces puncta numbers ( and by inference , NRE engulfment ) dramatically , we noted that neuronal fragment uptake is not completely eliminated ( Figure 4D ) . This suggested that another receptor may be involved . Integrins function with MeRTK to promote photoreceptor cell outer segment engulfment by retinal RPE glia ( Mao and Finnemann , 2012 ) , and the C . elegans genome encodes two α-integrin subunits , INA-1 and PAT-2 , both of which are implicated in apoptotic cell phagocytosis in C . elegans ( Hsieh et al . , 2012; Neukomm et al . , 2014; Sáenz-Narciso et al . , 2016 ) . We found that while a mutation in ina-1 has no effect on NRE engulfment ( Figure 4—figure supplement 1A ) , loss of PAT-2 by RNA interference ( RNAi ) significantly blocks AFD–NRE phagocytosis ( Figure 4G ) . Further , pat-2 RNAi strongly enhances glia engulfment defects of psr-1 mutants ( Figure 4G ) . Thus , PAT-2/α-integrin and PSR-1 appear to act together for glial engulfment of AFD–NRE . Curiously , not only do mutations in ced-1 not block the appearance of puncta in glia , we found that ced-1 ( e1754 ) strong loss-of-function mutant animals actually exhibit enhanced puncta numbers compared to wild-type animals ( Figure 4C ) . We found that pat-2 RNAi did not block this enhanced engulfment defect of ced-1 ( e1754 ) animals ( Figure 4—figure supplement 1B ) , suggesting that PAT-2 and CED-1 likely do not function synergistically as PS-receptors for glia-dependent phagocytosis . In line with this , while psr-1 ced-1 double mutant animals exhibit a slightly higher fraction of animals with no puncta , ced-1 in fact suppresses the synergistic engulfment defects seen in psr-1; pat-2 ( RNAi ) animals ( Figure 4—figure supplement 1B ) . This suggests that either ced-1 has a minor role in engulfment as a PS-receptor or its role in this glia-dependent phagocytosis is non-canonical . To examine this further , we also asked if ttr-52 acts with ced-1 . The ced-1;ttr-52 double mutant had the same increased glia puncta as ced-1 single mutants , suggesting that ced-1 acts genetically downstream of ttr-52 ( Figure 4—figure supplement 1C ) . Finally , the ced-1; ttr-52; psr-1 triple mutant also phenocopied ced-1 single mutants in having increased number of glia puncta , suggesting again that CED-1 acts downstream of PSR-1 and TTR-52 . These data raise the possibility that in NRE engulfment CED-1 may instead act in phagolysosome maturation downstream of PS recognition , as has been observed for CED-1 in other contexts ( Yu et al . , 2006 ) . The ternary complex of CED-2/CrkII , CED-5/DOCK1 , and CED-12/ELMO1 acts downstream of PSR-1 for apoptotic cell engulfment ( Reddien and Horvitz , 2004; Wang et al . , 2003 ) . We found that animals bearing mutations in ced-2 , ced-5 , or ced-12 exhibit reduced AFD–NRE puncta in AMsh glia ( Figure 5A ) . Furthermore , expression of the CED-12B isoform in AMsh glia is sufficient to rescue ced-12 mutant defects ( Figure 5A ) . We conclude , therefore , that the CED-2/CED-5/CED-12 complex also likely regulates engulfment of AFD–NREs . CED-2/CED-5/CED-12 act as a GEF for the Rac1 GTPase CED-10 , a major downstream effector of a number of apoptotic phagocytosis pathways ( Flannagan et al . , 2012; Reddien and Horvitz , 2004; Wang and Yang , 2016; Figure 4A ) . CED-10 is also implicated in engulfment of photoreceptor outer segments by RPE glia-like cells in mammals and debris of injured axons by glia in Drosophila ( Kevany and Palczewski , 2010; Lu et al . , 2014; Nichols et al . , 2016 ) . We found that two loss-of-function mutations in ced-10 , or overexpression of dominant-negative CED-10T17N , block nearly all engulfment of AFD–NRE fragments by AMsh glia ( Figure 5B–D ) . Specifically , in two different alleles , very few puncta are observed in glia ( ced-10 ( n3246 ) [3 . 08 ± 0 . 79 , n = 39] and ced-10 ( n1993 ) [2 . 4 ± 0 . 6 puncta , n = 24 animals] vs . wild type [14 ± 1 puncta , n = 78 animals] ) . Furthermore , barely any mutant animal had >10 puncta ( ced-10 ( n3246 ) = 0 . 81% , n = 124; and ced10 ( n1993 ) = 2 . 78% , n = 72; compared to wild type = 64% , n = 171 ) . Expressing CED-10 only in AMsh glia completely restores engulfment to ced-10 loss-of-function mutants ( Figure 5B–D ) . To determine how CED-10 functions with respect to CED-2/CED-5/CED-12 and PSR-1 , we generated psr-1; ced-10 and ced-12; ced-10 double mutants . Both strains show strong defects in puncta numbers reminiscent of ced-10 single mutants ( Figure 5E ) . Furthermore , transgenic expression of CED-10 is sufficient to overcome the partial loss of NRE engulfment in psr-1 mutants ( Figure 5E ) . Our data are consistent with the interpretation that , like in cell corpse engulfment , CED-10/Rac1 GTPase likely functions in glia downstream of CED-2/CED-5/CED-12 and PSR-1 , to promote AMsh glial engulfment of NREs . This activation is specific as mutations in another CED-10 activator , UNC-73/TRIO , do not affect NRE uptake ( Figure 4—figure supplement 1A; Lundquist et al . , 2001; Sáenz-Narciso et al . , 2016 ) . Unexpectedly , expression of constitutive active CED-10G12V also results in reduced engulfed puncta ( Figure 5D ) . This may indicate that a GTPase cycle is needed for engulfment to proceed ( Bernards and Settleman , 2004; Sáenz-Narciso et al . , 2016; Singhvi et al . , 2011; Takai et al . , 2001; Teuliere et al . , 2014 ) . Alternatively , it may be that this form of the protein promotes hyperefficient engulfment , which does not leave much NRE to be engulfed . Supporting the latter model , the AFD–NRE is significantly shorter in CED-10G12V mutants ( see below ) . Furthermore , overexpression of wild-type CED-10 , but not of wild-type PSR-1 or CED-12 , increases NRE engulfment Figure 4D , Figure 5A , D ) . Glial CED-10 is , therefore , both necessary and sufficient to regulate the rate at which AMsh glial engulf AFD–NRE fragments . During apoptotic cell engulfment , CED-10 executes phagocytic arm extension by mediating actin remodeling ( Wang and Yang , 2016 ) . We , therefore , examined animals bearing a loss-of-function mutation in wsp-1 , which encodes an actin polymerization factor , and found a block in NRE engulfment ( Figure 5—figure supplement 1A ) . As with overexpression of CED-10 , increasing levels of WSP-1 specifically in AMsh glia also lead to increased NRE engulfment ( Figure 5—figure supplement 1A ) . These results suggest that CED-10-dependent actin remodeling is the rate-limiting step for the engulfment of AFD–NREs by glia . Previous studies showed that cyclic-nucleotide-gated ( CNG ) ion channels localize to the AFD cilium base and are required for AFD neuron firing in response to temperature stimuli ( Cho et al . , 2004; Ramot et al . , 2008; Satterlee et al . , 2004 ) . These channels are mis-localized in cilia-defective mutants ( Nguyen et al . , 2014 ) . Independently , it has been shown that cilia-defective mutants exhibit deficits in thermotaxis behavior ( Tan et al . , 2007 ) . Since we found that cilia-defective mutants have increased engulfment ( Figure 3 ) , these taken together prompted us to examine the role for neuron activity in glial engulfment directly . We examined animals defective in TAX-2 , the sole CNG β-subunit in the C . elegans genome , or in TAX-4 and CNG-3 , α-subunits that function together in AFD ( Cho et al . , 2004; Hellman and Shen , 2011; Satterlee et al . , 2004 ) for engulfment defects . Glia in mutant animals accumulate extra puncta ( tax-2: 28 . 1 ± 2 puncta , n = 37; tax-4; cng-3 double mutants: 23 . 8 ± 2 . 3 puncta , n = 17 ) ( Figure 6A–C ) , and in tax-2 mutants , a larger fraction of the animal population has >10 puncta ( tax-2 , 99% , n = 92 animals; wild type , 65% , n = 171 animals ) ( Figure 6C ) . Conversely , we assessed the consequence of increasing the levels of cGMP , which promotes CNG channel opening , by mutating the cGMP degrading enzymes PDE-1 and PDE-5 expressed in AFD neurons ( Ramot et al . , 2008; Singhvi et al . , 2016 ) . We found that pde-1; pde-5 double mutant animals have reduced glia puncta numbers compared to wild type ( 7 . 1 ± 1 . 4 , n = 11 vs . 14 ± 1 , n = 78 ) ( Figure 6B , C ) . Finally , acute and cell-specific chemogenetic silencing of AFD using a histamine-gated chloride channel ( Pokala et al . , 2014 ) expressed under an AFD-specific promoter leads to puncta enrichment in AMsh glia within 24 hr ( Figure 6E , F ) . Thus , AFD activity levels reciprocally affect AFD–NRE engulfment levels and can do so acutely . Accumulation of glial puncta in AFD activity mutants could result from increased engulfment rates or , alternatively , from decreased puncta degradation . We favor the former model as we found that the increase in puncta number seen in tax-2 mutant glia is entirely suppressed by loss of CED-10 ( Figure 6B , C ) . Likewise , we also observed significant suppression in tax-2; psr-1 ( tm469 ) double mutants compared to tax-2 alone; and this suppression is enhanced further by pat-2 ( RNAi ) ( Figure 6C , D ) . Loss of ced-10 also suppresses excess engulfment following acute chemogenetic silencing of AFD ( Figure 6F ) . Our findings are therefore consistent with neuron activity controlling NRE engulfment through the CED-10 pathway . What might be the function of AFD–NRE engulfment by glia ? To test this , we examined AFD–NRE shape by 3D super-resolution imaging of transgenic mutants bearing a tagged reporter that specifically marks AFD–NRE microvilli . We found that ced-10 loss of function , or AMsh glia-specific overexpression of dominant negative CED-10T17N , results in elongated AFD–NRE microvilli ( Figure 7A , B , Figure 7—figure supplement 1B ) . By contrast , overexpressing wild-type CED-10 , which has excess puncta , produces shorter AFD–NRE microvilli , and this defect worsens with age ( Figure 7A , B ) . Furthermore , overexpressing GTP-locked CED10G12V also leads to shorter AFD–NRE microvilli even though it paradoxically has reduced number of puncta in glia ( Figure 7—figure supplement 1A , B ) consistent with the idea that engulfment in this strain may be so efficient that no NREs remain to be engulfed . Thus , AMsh glial engulfment of NRE fragments is important for regulating the AFD–NRE microvilli length . When placed on a temperature gradient , C . elegans seek their temperature of cultivation , Tc ( Hedgecock and Russell , 1975; Figure 7C–F , wild-type data in black line ) . This animal behavior depends on thermosensory transduction at the AFD–NRE ( Goodman and Sengupta , 2018; Mori and Ohshima , 1995 ) . Previous studies have shown that animals with defects in AFD–NRE shape also exhibit defects in this thermosensory behavior . Consistent with this , we found that ced-10 mutants exhibit altered thermosensory behavior . While wild-type animals reared at 25°C migrate to their Tc = 25°C on a linear temperature gradient , ced-10 mutants prefer cooler temperatures ( Figure 7C , D ) . Furthermore , animals carrying integrated transgenes overexpressing CED-10 only in AMsh glia also exhibit athermotactic defects regardless of the cultivation temperatures ( Figure 7E , F ) . We conclude , therefore , that AFD–NRE engulfment by AMsh glia is required for appropriate animal thermotaxis behaviors . The behavior defects we observed are consistent with the thesis that reduced neuron activity drives glial engulfment . The athermotactic behavior of CED-10 overexpression strains mimics similar defects of tax-2 or tax-2; tax-4 double mutant animals , and both manipulations lead to increased puncta and reduced neuron activity ( Figure 7—figure supplement 1C , D; Cho et al . , 2004; Satterlee et al . , 2004 ) . Likewise , the cryophilic behavior of ced-10 mutants , which have reduced glia puncta , is similar to that observed in other mutants with increased AFD cGMP levels ( Singhvi et al . , 2016 ) . We favor the model that activity-dependent glial engulfment of NRE is one mechanism by which AMsh glia and AFD coordinate regulation of NRE shape and animal thermosensory behavior . Our studies reveal a fundamental distinction between glia-dependent phagocytosis and other modes of engulfment . Apoptotic cell phagocytosis , glial clearance of injury-induced neuronal debris , and related engulfment events are all-or-none phenomena: engulfment either occurs or does not . By contrast , we show here that in AMsh glia engulfment rate is dynamically tuned throughout animal life to modulate NRE morphology , impacting animal behavior . The molecular parallels between the engulfment machinery in the peripheral sense-organ AMsh glia and other CNS glial engulfment lead us to posit that controlled phagocytosis may similarly regulate glial engulfment in other settings . Accompanying this more versatile engulfment program is a shift in the relevance of specific engulfment receptors . Apoptotic phagocytosis in C . elegans relies predominantly on CED-1 , with the PS-receptor PSR-1 playing a minor role ( Wang et al . , 2003; Wang and Yang , 2016 ) . Surprisingly , while CED-1 is dispensable for pruning by AMsh glia , we identified PSR-1/PS-receptor as a novel regulator of glial pruning . Why do CED-1 and PSR-1 have differing valence in apoptotic phagocytosis and glial pruning ? One possibility is that this difference in receptors reflects the size of particles engulfed . Supporting this notion , engulfment of small cell process debris of the C . elegans tail-spike cell is also independent of CED-1 ( Ghose et al . , 2018 ) . We identified PSR-1 and integrins as a PS-receptor driving AMsh glial engulfment of AFD–NRE . Other PS-receptors that have been shown to regulate glial engulfment across species include CED-1/MEGF10/Draper , MerTK , and GPR56 , and it is likely that yet others await identification ( Chung et al . , 2013; Freeman , 2015; Hilu-Dadia and Kurant , 2020; Kevany and Palczewski , 2010; Li et al . , 2020; Nomura-Komoike et al . , 2020; Raiders et al . , 2021; Tasdemir-Yilmaz and Freeman , 2014; Vecino et al . , 2016 ) . This then raises the question of why one analogous glial function of pruning would require different receptors . We speculate that this may reflect the molecular heterogeneity across glia and/or the context of engulfment ( Raiders et al . , 2021 ) . PS exposure has emerged as a classic engulfment signal for both apoptotic phagocytosis and glial pruning , but how this is regulated remains enigmatic . We identify this as a conserved feature in C . elegans glial engulfment and implicate the phospholipid transporter TAT-1/ATP8A in this process . TAT-1 is a member of the type 4 family P4 ATPases , which flip PS from exoplasmic to cytoplasmic membrane leaflets ( Andersen et al . , 2016 ) . We note that murine P4-ATPases ATP8A1 and ATP8A2 are expressed in the nervous system , and knockout mice exhibit deficient hippocampal learning , sensory deficits , cerebellar ataxia , mental retardation , and spinal cord degeneration , and shortened photoreceptor NRE length ( Coleman et al . , 2014 ) . Given this intriguing parallel , it will be interesting to probe whether ATP8A similarly modulates glial pruning in mammals . We also identify the PS bridging molecule TTR-52 as a regulator of pruning . It is also implicated in apoptotic phagocytosis and nerve regeneration ( Neumann et al . , 2015; Wang et al . , 2010 ) . Retinal RPE glia and cortical astrocytes also require PS-bridging opsonins ( Gas6 and MFGE8 ) to engulf neuron fragments ( Bellesi et al . , 2017; Kevany and Palczewski , 2010 ) . Whether all glia require PS opsonization for pruning remains to be determined . Our finding that proper animal behavior requires precise levels of NRE engulfment by glia suggests that engulfment must proceed with extraordinary specificity so that behavior is optimal . Indeed , we find that AMsh glia prune AFD–NRE with subcellular precision . While AFD’s actin-rich microvilli are removed by glia , its adjacent microtubule-based cilium is not . Aberrantly excessive/reduced pruning correlate with disease in mammals , hinting that similar subcellular precision in marking fragments/endings for engulfment might be involved ( Chung et al . , 2015; Wilton et al . , 2019 ) . How this precision is regulated will be fascinating to explore . A role for pruning in normal neural functions has so far been investigated for central nervous system glia ( astrocytes , microglia , retinal glia ) . Peripheral glia of the inner ear are known to activate phagocytosis only in injury settings ( Bird et al . , 2010 ) . Our studies demonstrate that pruning of sensory neuron endings by glia is required for accurate sensory perception . Thus , glial pruning is conserved in both CNS and PNS and is executed for normal neural functions by analogous molecular mechanisms . While these studies identify glial pruning as a mechanism to control NRE shape in response to activity states , we note that it is likely that AMsh glia and AFD neuron cooperate through multiple mechanisms to regulate AFD–NRE shape and animal thermosensory behaviors , including some that we previously identified ( Singhvi et al . , 2016; Wallace et al . , 2016 ) . Such regulatory complexity might reflect the fact that appropriate thermosensory behaviors are critical for animal survival . An outstanding question in understanding the role of glia is whether glia actively prune NREs and neuron fragments or passively clear shed debris . Three lines of evidence in this study lead us to conclude that AMsh glia actively drive engulfment rather than passively clearing debris . ( 1 ) Our finding that glial CED-10 levels can modulate engulfment rates , NRE shape , and animal behavior suggests that this process can be triggered by glia . ( 2 ) While both CED-10 overexpression and ttx-1 mutants have short NRE ( Satterlee et al . , 2001; Figure 4—figure supplement 1B ) , unlike animals overexpressing CED-10 , ttx-1 mutants have fewer puncta , not more ( Figure 2A ) . Thus , short NRE shape can derive from independent mechanisms . ( 3 ) While both ced-10 and tax-2 mutants have longer , disorganized NRE ( Satterlee et al . , 2004; Singhvi et al . , 2016 ) , tax-2 mutants have more puncta , not fewer . If glial pruning only passively cleared debris , we would have expected the opposite . Furthermore , that engulfment tracks neuron activity and modulating this process impacts animal behavior also suggests a physiological role for this process . In summary , our findings reveal glial engulfment as an active regulator of neural functions . Importantly , they directly and causally link pruning of individual neuron endings to animal behavior at single-molecule and single-cell resolution . This raises the possibility that engulfment may be a general mechanism by which glia dynamically modulate sensory perception and neural functions , across modalities , systems , and species . C . elegans animals were cultured as previously described ( Brenner , 1974; Stiernagle , 2006 ) . Bristol N2 strain was used as wild type . For all experiments , animals were raised at 20°C for at least two generations without starvation , picked as L4 larvae onto fresh plate , and assayed 1 day later unless otherwise noted . Germline transformations by microinjection to generate unstable extrachromosomal array transgenes were carried out using standard protocols ( Fire et al . , 1990; Mello et al . , 1991; Stinchcomb et al . , 1985 ) . Integration of extrachromosomal arrays was performed using UV+ trimethyl psoralen . All transgenic arrays were generated with 5 ng/µl Pelt-2:mCherry , 20 ng/µl Pmig-24:Venus , or 20 ng/µL Punc-122:RFP as co-injection markers ( Abraham et al . , 2007; Armenti et al . , 2014; Miyabayashi et al . , 1999 ) . Further information on all genetic strains and reagents is available upon request . Animals were immobilized using either 2 mM tetramizole or 100 nm polystyrene beads ( Bangs Laboratories , catalog # PS02004 ) . Images were collected on a Deltavision Elite RoHS wide-field deconvolution system with Ultimate Focus ( GE ) , a PlanApo 60×/1 . 42 NA or OLY 100×/1 . 40 NA oil-immersion objective and a DV Elite CMOS Camera . Super-resolution microscopy images were collected on the Leica VT-iSIM microscope or the Leica SP8 confocal with Lightning . Images were processed on ImageJ , Adobe Photoshop CC , or Adobe Illustrator CC . Binning categories for population analyses were based on preliminary analyses of population distribution of puncta numbers/animal in wild type , and mutants with excess puncta ( tax-2 ) or reduced puncta mutants ( ced-10 , psr-1 ) . Preliminary analyses of these strains suggested that the bin intervals ( 0 , 1–9 , or 10+ puncta ) are the most robust , conservative , and rapid assessment of phenotypes . Higher than 10 puncta/cell were not readily resolved without post-processing and therefore binned together in population scores . Some genotypes were selected for further post-hoc single-cell puncta quantification analyses . For this , glia puncta numbers were quantified using Analyze Particles function in ImageJ on deconvolved images . Individual puncta size measurements were done on yz orthogonal rendering of optical sections using 3D objects counter plug-in in ImageJ . Adult hermaphrodites were fixed in 0 . 8% glutaraldehyde−0 . 8% osmium tetroxide−0 . 1 M cacodylate buffer ( pH 7 . 4 ) for 1 hr at 4°C in the dark , and then quickly rinsed several times with 0 . 1 M cacodylate buffer . Animal heads were decapitated and fixed in 1% osmium tetroxide−0 . 1 M cacodylate buffer overnight at 4°C , quickly rinsed several times in 0 . 1 M cacodylate buffer , and dehydrated through a graded ethanol series . The samples were then embedded in Eponate 12 resin ( Ted Pella , Inc , Redding , CA ) and polymerized overnight in a 60°C oven . 70 nm ultrathin serial sections were collected onto pioloform-coated slot grids from the anterior tip of the animal to a distance of approximately 7 µm . Sections were examined on a JEOL 1400 TEM ( JEOL , Tokyo , Japan ) at an accelerating voltage of 120 kV . Images were acquired with a Gatan Rio 4k × 4k detector ( Gatan , Inc , Pleasanton , CA ) . Microvilli size measurements were done with ImageJ Measure Function on electron micrograph thin sections . Population puncta scoring was statistically analyzed using Fisher’s exact statistical test in GraphPad Prism 8 . Puncta images were quantified using Analyze Particles function in Image J and analyzed with a one-Way ANOVA with multiple-comparison test in GraphPad Prism 8 . For chemogenetic silencing assays , 10 mM histamine ( Sigma , catalog # H7250 ) was added to NGM agar plates . L4 larval stage transgenic worms expressing HisCl1 in AFD were grown for 24 hr on either normal or histamine plates and assayed as day 1 adults ( Pokala et al . , 2014 ) . Plasmids expressing double-stranded RNA were obtained from the Ahringer Library ( Fraser et al . , 2000; Kamath and Ahringer , 2003 ) . The L4440 empty vector was used as negative control . RNAi was performed by feeding synchronized L1 animals RNAi bacteria ( Timmons , 2004 ) . L4 larva were moved to a fresh plate with RNAi bacteria and scored 24 hr later for glial puncta ( nsIs483 ) or AFD–NRE defects ( nsIs645 ) . Thermotaxis assays were performed on a 17−26°C linear temperature gradient , designed as previously described ( Hedgecock and Russell , 1975; Mori and Ohshima , 1995 ) . Animals were synchronized and the staged progeny were tested on the first day of adulthood . Briefly , animals were washed twice with S-Basal and spotted onto the center of a 10 cm plate warmed to room temperature and containing 12 ml of NGM agar . The plate was placed onto the temperature gradient ( 17–26°C ) with the addition of 5 ml glycerol to its bottom to improve thermal conductivity . At the end of 45 min , the plate was inverted over chloroform to kill the animals and allow easy counting of animals in each bin . The plates have an imprinted 6 × 6 square pattern , which formed the basis of the six temperature bins . Each data point is the average of 3–8 assays with ~150 worms/assay .	Neurons are tree-shaped cells that receive information through endings connected to neighbouring cells or the environment . Controlling the size , number and location of these endings is necessary to ensure that circuits of neurons get precisely the right amount of input from their surroundings . Glial cells form a large portion of the nervous system , and they are tasked with supporting , cleaning and protecting neurons . In humans , part of their duties is to ‘eat’ ( or prune ) unnecessary neuron endings . In fact , this role is so important that defects in glial pruning are associated with conditions such as Alzheimer’s disease . Yet it is still unknown how pruning takes place , and in particular whether it is the neuron or the glial cell that initiates the process . To investigate this question , Raiders et al . enlisted the common laboratory animal Caenorhabditis elegans , a tiny worm with a simple nervous system where each neuron has been meticulously mapped out . First , the experiments showed that glial cells in C . elegans actually prune the endings of sensory neurons . Focusing on a single glia-neuron pair then revealed that the glial cell could trim the endings of a living neuron by redeploying the same molecular machinery it uses to clear dead cell debris . Compared to this debris-clearing activity , however , the glial cell takes a more nuanced approach to pruning: specifically , it can adjust the amount of trimming based on the activity load of the neuron . When Raiders et al . disrupted the glial pruning for a single temperature-sensing neuron , the worm lost its normal temperature preferences; this demonstrated how the pruning activity of a single glial cell can be linked to behavior . Taken together the experiments showcase how C . elegans can be used to study glial pruning . Further work using this model could help to understand how disease emerges when glial cells cannot perform their role , and to spot the genetic factors that put certain individuals at increased risk for neurological and sensory disorders .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "neuroscience" ]	2021	Glia actively sculpt sensory neurons by controlled phagocytosis to tune animal behavior
Swine influenza presents a substantial disease burden for pig populations worldwide and poses a potential pandemic threat to humans . There is considerable diversity in both H1 and H3 influenza viruses circulating in swine due to the frequent introductions of viruses from humans and birds coupled with geographic segregation of global swine populations . Much of this diversity is characterized genetically but the antigenic diversity of these viruses is poorly understood . Critically , the antigenic diversity shapes the risk profile of swine influenza viruses in terms of their epizootic and pandemic potential . Here , using the most comprehensive set of swine influenza virus antigenic data compiled to date , we quantify the antigenic diversity of swine influenza viruses on a multi-continental scale . The substantial antigenic diversity of recently circulating viruses in different parts of the world adds complexity to the risk profiles for the movement of swine and the potential for swine-derived infections in humans . Swine have been hypothesized to be a mixing vessel for influenza viruses and were the source of the 2009 H1N1 human pandemic virus ( Garten et al . , 2009; Smith et al . , 2009 ) . The directionality and relative threat of transmission is generally assumed to be from swine into humans with attention only recently turning to introductions in the opposite direction – from humans into swine – and the role that such anthroponotic introductions play in influenza virus evolution ( Nelson et al . , 2012; Terebuh et al . , 2010 ) . Understanding the dynamics of influenza viruses at the interface between humans and swine is key for designing optimal surveillance and control strategies . Currently there are three dominant subtypes of influenza A viruses circulating enzootically in swine: H1N1 , H1N2 , and H3N2 . The evolutionary history of these viruses reflects multiple introductions of influenza viruses into swine from other species ( Nelson et al . , 2012; Vincent et al . , 2014; Grøntvedt et al . , 2013; Cappuccio et al . , 2011; Ottis et al . , 1982; Pereda , 2010; Njabo et al . , 2012; Karasin et al . , 2006; Hofshagen , 2009; Holyoake et al . , 2011; Forgie et al . , 2011; Pensaert et al . , 1981 ) . Much of the work on characterizing swine influenza viruses and their ancestry has been carried out in the U . S and in Europe ( Vincent et al . , 2014; Marozin et al . , 2002; Nelson et al . , 2014; Simon et al . , 2014; Brown et al . , 1993; Brown et al . , 1997; Brown et al . , 1998; Webby et al . , 2000; Webby et al . , 2004; Vincent et al . , 2009a; Vincent et al . , 2009b; Kyriakis et al . , 2011; Watson et al . , 2015 ) with evidence for numerous introductions of human seasonal H1 and H3 influenza viruses into swine since 1990 . This pattern continued with the 2009 human pandemic H1N1 virus ( H1N1pdm09 ) with at least 49 separate human-to-swine transmission events , some of which appear to have become established in pig populations ( Nelson et al . , 2012 ) . In addition to viruses from humans , avian influenza viruses also occasionally infect swine . Notably , in 1979 , epidemics of an avian H1N1 influenza virus lineage were reported in Belgian swine leading to the establishment of an ‘avian-like’ H1N1 virus lineage in Europe . This Eurasian avian-like 1C lineage ( Zell et al . , 2013 ) has continued to circulate in swine until the present day , and been introduced into other geographic areas through international movement of swine ( Pensaert et al . , 1981; Zhu et al . , 2013; Nelson et al . , 2015c ) . Unsurprisingly , genetic diversity and multi-species origins of influenza viruses in swine are not restricted solely to pig populations in the U . S . and Europe . China has the largest swine population of any country and shows similar patterns of frequent introductions of both H1 and H3 influenza viruses from both avian and human hosts as well as introductions of swine lineages from other geographic regions ( Chen et al . , 2014; Vijaykrishna et al . , 2011; Zhu et al . , 2011 ) . Once established in swine , influenza viruses of human or avian origin pose a substantial threat to swine populations’ health and may spread to other geographically segregated populations ( Nelson et al . , 2015c ) . Similarly , these viruses also pose a threat for introduction or re-introduction into the human population ( Nelson et al . , 2014 ) , with associated outbreak and pandemic risks . This risk is exemplified by the re-introduction of the H3N2v ( variant ) virus with surface glycoprotein genes of human origin to over 300 people in 2011–12 from an H3N2 virus that has been endemic in swine in the U . S . since the late 1990s ( Centers for Disease Control and Prevention , 2011; Bowman et al . , 2014 ) . The influenza virus hemagglutinin ( HA ) surface glycoprotein is the primary target of the humoral immune response , undergoes antigenic drift over time , and is the major antigenic constituent in influenza vaccines used in both humans and livestock . The genetic heterogeneity among the HAs of influenza viruses circulating in swine in certain geographic regions has been relatively well studied and characterized ( Smith et al . , 2009; Nelson et al . , 2012; Cappuccio et al . , 2011; Nelson et al . , 2015c; Vijaykrishna et al . , 2011; Kitikoon et al . , 2013; Anderson et al . , 2015; Takemae et al . , 2008; 2013; Perera et al . , 2013; Pereda et al . , 2011 ) . However , the relationships between genetic diversity and antigenic diversity among contemporary swine influenza viruses and their antigenic similarity to human seasonal influenza strains are unknown , largely owing to a lack of antigenic data . Such antigenic data are critical in assessing the phenotype of the virus , any potential cross-immunity with viruses circulating in humans and in swine , and the risk of re-emergence from these hosts . These analyses are also critical in the design of antigenically well-matched vaccines . In order to help fill this knowledge gap , we compiled the largest and most geographically comprehensive antigenic dataset of swine influenza viruses to date . We then used a Bayesian framework ( Bedford et al . , 2014 ) to quantify the antigenic diversity of influenza viruses circulating in humans and swine thus generating a framework for exploring the antigenic component of risk of introduction of influenza A viruses from swine into humans and to swine in other geographic regions based on the antigenic characteristics of swine influenza viruses circulating in different parts of the world . We used hemagglutination inhibition ( HI ) assays to antigenically characterize a geographically diverse collection of H1 viruses from swine and humans from 1930–2013 ( n=194 ) and H3 viruses from swine and humans from 1968–2013 ( n=379 ) and genetically sequenced all antigenically characterized viruses . Of these , 101 H1 and 73 H3 swine influenza viruses were newly characterized for this study . In order to accurately quantify the antigenic variation of influenza viruses from swine we first determined the appropriate number of dimensions for accurately representing their antigenic relationships . Previous work on quantifying the antigenic evolution of seasonal influenza viruses showed that the evolution of H3 viruses in humans , as measured in HI assays , could be accurately visualized in two dimensions ( 2D ) using antigenic cartography ( Smith et al . , 2004 ) . In contrast , previous work on swine and equine H3 influenza viruses found that three dimensions ( 3D ) were necessary to accurately visualize antigenic evolution using antigenic cartography ( Lewis et al . , 2011; 2014; de Jong et al . , 2007; Lorusso et al . , 2011 ) . In this study we extensively tested the dimensionality of H1 and H3 influenza virus antigenic maps and found that the antigenic variation of both virus subtypes was most accurately represented in three dimensions . To investigate the combined antigenic and genetic evolution of our study viruses we implemented a Bayesian multi-dimensional scaling ( BMDS ) method for inferring antigenic and genetic relationships ( Bedford et al . , 2014 ) . As of 2013 , three major H1 lineages circulated in swine: 1 . the classical 1A lineage – descended from the ancestors of the 1918 human influenza pandemic , was first detected in swine in the 1930’s , and gave rise to the human H1N1pdm09 viruses and swine H1N1pdm09-like sub-lineage viruses; 2 . the human seasonal-like 1B lineage – resulting from multiple introductions from human seasonal H1 viruses; 3 . the Eurasian avian-like 1C lineage – arising from an introduction from wild birds into swine and first detected in Europe in the 1980’s ( Figure 1A , Zell et al . , 2013 ) . 10 . 7554/eLife . 12217 . 003Figure 1 . Evolutionary relationships of H1 ( A , B ) and H3 ( C , D ) influenza viruses circulating in swine and humans inferred by Bayesian Multi-dimensional scaling ( BMDS ) . Each colored ball represents a single virus . Viruses are colored by lineage ( A , C ) and by geography ( B , D ) . Lines connecting each virus represent inferred phylogenetic relationships . Distances for antigenic dimensions are measured in antigenic units ( AU ) and each unit is equivalent to a two-fold dilution in HI assay data . Antigenic distance can be interpreted as a measure of antigenic similarity – viruses close to one another are more antigenically similar than viruses further apart . Interactive visualizations are available at https://phylogeography . github . io/influenzaH1/ and https://phylogeography . github . io/influenzaH3/ . Source data and GIF files for rotational views of 3D antigenic maps in Figure 1 have been deposited in Dryad ( Lewis et al . , 2016 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 00310 . 7554/eLife . 12217 . 004Figure 1—figure supplement 1 . Figure 1A , B colored by H1 genetic sub lineages in the Bayesian MCC tree . DOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 00410 . 7554/eLife . 12217 . 005Figure 1—figure supplement 2 . Bayesian H1 MCC tree with taxa labels and posterior support valuesDOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 00510 . 7554/eLife . 12217 . 006Figure 1—figure supplement 3 . Figure 1C , D colored by H3 genetic sub lineages in the Bayesian MCC tree . DOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 00610 . 7554/eLife . 12217 . 007Figure 1—figure supplement 4 . Bayesian H3 MCC tree with taxa labels and posterior support valuesDOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 00710 . 7554/eLife . 12217 . 008Figure 1—figure supplement 5 . Maximum likelihood phylogenetic trees colored by lineage . DOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 00810 . 7554/eLife . 12217 . 009Figure 1—figure supplement 6 . Maximum likelihood H1 phylogenetic tree with taxa labels and bootstrap support valuesDOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 00910 . 7554/eLife . 12217 . 010Figure 1—figure supplement 7 . Maximum likelihood H3 phylogenetic tree with taxa labels and bootstrap support values . DOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 010 Our combined antigenic and genetic analyses of H1 viruses in swine showed an inconsistent relationship between antigenic clustering and genetic lineage through time with occasional emergence of new antigenic variants . In all geographic regions for which we had antigenic data we found unique patterns of antigenic evolution within individual lineages . In Asia , we observed co-circulation of classical 1A , human seasonal-like 1B , and Eurasian avian-like 1C lineage viruses . However all three lineages were not detected in any one country . In Hong Kong – where samples were derived at slaughter from swine sourced from China – classical swine 1A lineage including the H1N1pdm09 sub-lineage , and Eurasian avian-like 1C lineage viruses were detected ( Figure 1A , light red ) . For the viruses from Hong Kong , the mean pairwise antigenic distance ( MPD ) between the H1N1pdm09 sub-lineage viruses and the two classical 1A lineage clades was 5 . 0 and 6 . 0 AU away ( Figure 1—figure supplement 1: Figure 1—figure supplement 2 , Supplementary file 2 ) . The Classical swine 1A lineage viruses collected in Japan ( Figure 1A , dark red ) differed from the Hong Kong H1 viruses with a MPD of 5 . 6 AU . Within the Hong Kong Eurasian avian-like 1C lineage viruses there were two distinct phylogenetic groups – group 1 and group 2 ( as previously reported in Yang et al . ( 2016 ) and Figure 1—figure supplement 1 , Supplementary file 2 ) . The Group 1 and Group 2 viruses were 6 . 5 AU and 4 . 4 AU ( MPD ) away from the H1N1pdm09 sub-lineage viruses . The MPD between all Asian and European Eurasian avian-like 1C lineage strains was 5 . 2 AU . All of these distances are sufficiently large that it would be surprising if prior infection or un-adjuvanted vaccines would provide any protection among these virus groups . In Europe , the Eurasian avian-like 1C lineage co-circulated with the H1N1pdm09 sub-lineage of classical 1A viruses and the human seasonal-like 1B lineage . The human seasonal-like 1B lineage viruses were first detected in European swine in the mid-1990’s but have a putative human seasonal ancestor from the 1980’s . The human seasonal-like 1B viruses differed substantially from the Eurasian avian-like swine 1C viruses and the H1N1pdm09 sub-lineage viruses ( MPD: 6 . 7 AU and 10 . 4 AU respectively ) . The H1N1pdm2009 sub-lineage viruses also differed substantially from the human seasonal-like 1B lineage viruses ( MPD:10 . 6 AU ) . In Canada , only the classical swine 1A lineages viruses and H1N1pdm09 sub-lineage viruses were detected . Most of the viruses within the classical swine 1A lineage were antigenically similar to each other but differed from H1N1pdm09 sublineages viruses with a MPD of 6 . 5 AU . However , the alpha sub-lineage of classical 1A lineage viruses ( Figure 1B green , Supplementary file 2 ) antigenically diverged in Canadian swine to form a separate antigenic group with a MPD of 7 . 0 AU from the H1N1pdm09 strains and 4 . 2 AU from the other Canadian classical swine 1A lineage viruses . In the USA , the classical swine 1A lineage viruses co-circulated with the H1N1pdm09 sub-lineage and the human seasonal-like 1B lineage viruses . In the USA the human seasonal-like 1B lineage viruses evolved two sub-lineages – Delta 1 and Delta 2 ( Figure 1—figure supplement 1 , Supplementary file 2 ) – likely as the result of two separate but nearly contemporaneous introductions into swine first detected in the early 2000’s . The Delta 1 and 2 sub-lineage viruses antigenically differed from the H1N1pdm2009 lineage viruses with an MPD of 7 . 8 AU . Representatives of alpha sub-lineage of the classical swine 1A lineage in the USA were 7 . 0 AU ( MPD ) from the H1N1pdm2009 sub-lineage viruses . The beta and gamma sub-lineages of the classical swine 1A lineage were 6 . 6 and 4 . 7 AU ( MPD ) respectively from the H1N1pdm2009 viruses . To assess the risk of virus introduction associated with moving live swine between continents we compared the MPDs of swine influenza H1 viruses circulating in different parts of the world . The human seasonal-like 1B viruses circulating in European swine were on average 11 . 7 AU ( MPD ) from the human seasonal-like 1B viruses in USA swine . The Eurasian avian-like 1C lineage viruses were 6 . 8 AU ( MPD ) from the human seasonal-like 1B lineage viruses in the USA . Just as for the H1 viruses , we used BMDS ( Bedford et al . , 2014 ) to investigate the antigenic and genetic evolution of H3 viruses ( Figure 1C , D ) . In European swine there were four H3 introductions included in our antigenic characterization ( Figure 1—figure supplement 3: Figure 1—figure supplement 4 ) . Of these , the European 3A lineage ( A/Port Chalmers/1973-like ) introduction was the most commonly detected lineage , showing evidence for three sequential antigenic clusters and evolving on a markedly different path from the putative ancestor in humans ( Figure 1C , D [blue] ) . A separate introduction from humans into swine of an antigenically distinct 1970’s-like human seasonal virus also circulated in European swine in the 1980s and 1990s ( Figure 1—figure supplement 1 [Europe 2 Introduction] ) . The earliest strain in the European 3A lineage was 6 . 5 AU from the A/Port Chalmers/1973 reference strain ( Figure 1C large grey ) and the most recent strain was 9 . 8 AU ( Supplementary file 2 ) . The distance from our representative of recent human seasonal H3 viruses ( A/Victoria/361/2011 ) to the H3 European virus in swine and was 14 . 9 AU ( MPD ) . In the USA , we found evidence for three H3 virus introduction events from humans into swine in the mid-1990’s , one of which resulted in the establishment of the 3B lineage ( Figure 1C , D ) which became the most frequently detected lineage in swine through 2013 , and gave rise to the H3N2v viruses which infected >300 humans in the US from 2011–2012 ( Centers for Disease Control and Prevention , 2012 ) . We only antigenically characterizes a single representative of the H3N2v viruses – A/Indiana/8/2011 – and found that it differed from A/Victoria/361/2011 by 6 . 4 AU ( Figure 1 pink ) . The MPD between the swine H3 viruses circulating in the USA and in Europe was 12 . 6 AU . In Asia we found evidence for multiple introductions of H3 viruses and subsequent antigenic drift from the six representative viruses in our dataset including an introduction likely mediated by translocation of a virus circulating in USA swine to Asian swine ( Nelson et al . , 2015d ) . Two separate introductions from humans were detected in Japan that resulted in circulation in swine beyond the time of circulation of the ancestral strain in humans . The introductions were ~5 . 2 AU from A/Victoria/361/2011 and 10 . 9 AU ( MPD ) from the H3 strains circulating in European swine . The most recent seasonal human H3 introduction detected in Asia was from Vietnam and this virus was 3 . 8 AU away from A/Victoria/361/2011 ( Supplementary file 2 ) . To investigate the rates of swine influenza virus antigenic drift away from the most recent common ancestor for each genetic lineage in each geographic region we used TimeSlicer available in SPREAD ( Bielejec et al . , 2011 ) . For H1 viruses in swine , the classical 1A lineage evolved at a mean rate of 0 . 15 AU per year from the earliest antigenically characterized representative in the dataset ( A/swine/Iowa/15/1930 ) ( Figure 2A , Supplementary file 3 ) . The human seasonal-like 1B viruses introduced into European swine evolved at a mean rate of 0 . 17 AU per year ( Figure 2B , Supplementary file 3 ) . The introductions of human seasonal H1 viruses into USA swine that gave rise to the Delta-1 and Delta-2 lineages showed slightly different rates of antigenic drift – 0 . 63 for Delta-1 and 0 . 85 for Delta-2 – both significantly faster than the human seasonal introduction in European swine ( Figure 2B , Supplementary file 3 ) but they were only observed over a relatively short time period . The Eurasian avian-like swine 1C lineage evolved at a mean rate of 0 . 15 AU per year ( Figure 2C , Supplementary file 3 ) . With the exception of the human seasonal-like 1B lineage viruses in the USA , the overall antigenic diversity of H1 viruses in swine remained remarkably constant over the study period ( Figure 2A , B , C ) . The antigenic diversity of the Delta-1 and Delta-2 sublineages in the USA began to expand rapidly around 2009 for reasons that remain to be elucidated ( Figure 2B ) . 10 . 7554/eLife . 12217 . 011Figure 2 . Time series of year-to-year rates of antigenic drift distance and antigenic diversity of H1 and H3 viruses in swine by genetic lineage . Solid colored lines represent year-to-year antigenic drift distance , where drift for year i is measured as the mean of Euclidean distances among strains in a phylogenetic lineage in year i compared to the mean of Euclidean distances among strains of that phylogenetic lineage from the previous year ( i–1 ) . The dotted line represents antigenic diversity among H1 and H3 strains by lineage through time . For the solid and dotted lines , the shaded region represents the range of the highest posterior density estimates . Multiple introductions which circulate for >5 years of the human seasonal-like swine H1 lineage in European ( purple ) and USA ( gold ) swine were calculated separately . Source data for Figure 2 has been deposited in Dryad ( Lewis et al . , 2016 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12217 . 011 For H3 viruses , we found that the 3B lineages viruses in USA swine evolved antigenically at a faster rate than the H3 viruses introduced into European swine ( 0 . 49 AU per year vs 0 . 28 AU per year [Figure 2D , E , Supplementary file 3] ) . Additionally , the antigenic diversity of H3 viruses in USA swine , particularly since 2010 , has increased at a faster rate than that of H3 viruses in Europeans swine . Thus sustained circulation of antigenic drift variants within US swine has substantially increased the observed antigenic diversity in this lineage relative to the observed antigenic diversity in other geographic regions . In addition to the BMDS analyses of viruses for which we had both antigenic and genetic data , we also performed large-scale phylogenetic analyses of publicly available swine influenza virus sequence . Through these analyses we inferred potentially important gaps in the global antigenic data that suggest our antigenic dataset underrepresents the true global antigenic diversity of swine influenza viruses . For H1 viruses that circulated in swine as of 2013 , we find that the major genetic lineages have been captured in this study ( Figure 1—figure supplement 5: Figure 1—figure supplement 6 ) . However , new genetic evidence recently became available to demonstrate additional human seasonal introductions into swine in previously under-surveilled regions such as Central and South America ( Nelson et al . , 2015a; Dibárbora et al . , 2013; Nelson et al . , 2015b ) ( Figure 1—figure supplement 5 ) . Since genetic lineage does not always predict antigenic phenotype , we cannot make any inferences about their antigenicity . Even in cases where we have antigenically-characterized clade representatives , the long branch lengths associated with some introductions could conceal antigenic drift away from the characterized clade representatives . The markedly different evolutionary patterns seen in the H3 lineages characterized in swine and the inferred antigenic profile of the viruses introduced to swine , suggests that the true diversity of H3 viruses circulating in swine is likely to be substantially greater than the diversity found in our BMDS analyses . In particular , we identified 19 additional examples of introductions of seasonal H3 viruses from humans in 10 countries ( Figure 1—figure supplement 5 – black tips: Figure 1—figure supplement 7 ) . Based on the likely time of these introductions from humans into swine we estimate that the introduced viruses represent at least 12 distinct antigenic variants based on the antigenic variation of their human seasonal influenza ancestors . Similar to the H1 genetic lineages described above , these additional H3 lineages also require antigenic testing to interpret the impact of these introductions on the global antigenic diversity of IAV in swine . In addition to the challenge the observed antigenic diversity poses to swine populations , the antigenic diversity of swine influenza viruses also creates substantial risks for human populations . The outbreak and pandemic potential of swine influenza viruses in humans is at least partially determined by human population immunity against swine viruses . This component of risk can be inferred from the antigenic distance of swine influenza viruses to seasonal viruses or vaccine strains in humans to which humans are likely to have immunity . Importantly , individuals are unlikely to have immunity to viruses that only circulated before they were born ( Fonville et al . , 2014 ) . Since most of the current swine influenza viruses are the product of human seasonal influenza virus introductions into swine , we anticipate at least some cross-protective immunity in the human population that could potentially interfere with re-introduction of these viruses . For example , the H1N1pdm09 viruses circulating in both humans and swine are very similar antigenically and likely induce at least some cross-immunity in both hosts ( Figure 1A , B ) . However , for the H1 1C , H3 3A , and H3 3B human seasonal lineages in swine , the risk of re-introduction into the human population increases with the proportion of the human population that were born after the human precursor virus circulated in humans and is amplified by antigenic evolution of these viruses in swine . Thus , in terms of antigenic similarity and likely prior exposure , earlier introduced lineages of human H1 and H3 viruses – particularly those with precursors antigenically similar to the H3 1968 pandemic strain – pose the greatest current risk to humans because of the low or negligible predicted levels of cross immunity in individuals born since the 1970s . In comparison , the H3N2v viruses , which ultimately derive from a human seasonal virus introduction in the 1990s , have predominantly infected people born after the mid-1990s . In this study we characterized the antigenic diversity H1 and H3 viruses circulating in both swine and humans on a multi-continental scale . The observed diversity was largely driven by frequent introductions of variants from humans into swine and subsequent antigenic evolution within swine combined with long-term geographic segregation of multiple virus lineages . Much of the antigenic evolution of human-origin viruses in swine was via long and divergent evolutionary paths from those observed in humans . These divergent paths create threats for the re-introduction of viruses into humans as well as the possibility for antigenically novel strains in humans to be introduced into swine . Interestingly , the introduction of viruses from birds to swine has made only one major contribution to the antigenic diversity of influenza viruses in swine . This could be at least partially due to the antigenic similarity of the avian-like and the classical swine 1A lineages found here , although an exhaustive test of antigenic similarity between avian and swine H1 and H3 viruses could not be performed as part of this study and to the best of our knowledge has not been reported elsewhere . Successful introduction of avian influenza viruses into swine will also likely be influenced by other factors including mutations conferring adaption for mammalian transmission ( Herfst et al . , 2012; Imai et al . , 2012 ) and other physiological and ecological factors . In addition to multiple introduction events and subsequent antigenic drift , the patterns of antigenic diversity and heterogeneity among geographic regions are also influenced by the relative geographic isolation of swine populations resulting in region-specific patterns of virus evolution and circulation . The observed patterns of antigenic heterogeneity among geographic regions are also likely influenced by livestock production system differences , particularly regional differences in the movement of swine within and between animal holdings . USA producers commonly move swine among otherwise geographically isolated areas at different ages after weaning , whereas in the EU there is negligible intra- or inter-country movement of animals during a single production cycle . Such livestock movement patterns have been shown to influence the evolution of influenza viruses in swine in the USA , with the Midwest serving as an ecological sink for swine influenza virus with the source of genetic diversity located in the south east and south central states for some HA lineages ( Nelson et al . , 2011 ) . Vaccination is used extensively as a means to control influenza in swine in the USA and sporadically globally . Control strategies vary by region with some countries having no swine influenza vaccine usage , while others produce autologous ( herd specific ) vaccine for individual producers , which may be used consecutively with commercially-manufactured multi-strain vaccine when a new outbreak strain requires an update to vaccine composition . Most autologous and commercial products are multivalent due to the diversity of circulating strains . However , there is no formalized system for matching vaccine strains with circulating strains , nor validated protocols for standardization and effective vaccine use . The antigenic diversity quantified here is sufficiently large that it is highly unlikely that one strain per subtype would be efficacious globally , or even within a given region . Multiple within-subtype strains would likely be required to produce a vaccine that afforded adequate protection against most circulating variants using current inactivated vaccine production processes and application protocols . Interestingly , H3 viruses circulating in European swine showed lower antigenic diversity than other regions . This limited antigenic diversity , combined with the use of oil adjuvanted vaccines , could explain why the prototype A/Port Chalmers/1973 ( H3N2 ) -based vaccine induced protection against contemporary H3N2 swine influenza viruses isolated up to 35 years later ( de Jong et al . , 2007; Van Reeth and Ma , 2013; De Vleeschauwer et al . , 2015 ) . However , movement of swine among geographic regions could lead to the introduction of novel variants into a new sub-population of swine despite current import-export quarantine procedures between some countries . Therefore a more structured approach to vaccine strain selection in swine , where epidemiological , antigenic and genetic data are considered at a global level by the World Organization for Animal Health ( OIE ) , similar to the process undertaken for creating equine influenza vaccine recommendations , could provide an evidence-based rationale for swine vaccine strain selection and improved vaccine efficacy in swine . Age-related susceptibility within the human population could be one of the limiting factors in onward spread of recent human infections with swine viruses . However there remains a need for focused surveillance in areas with high swine population density and situations where humans and swine have opportunities for close contact to better assess the incidence of cross-species transmission and risk of onward transmission within the human population . Neutralizing immunity to influenza viruses in both humans and swine is largely directed towards the influenza HA and the antigenic dissimilarity among influenza HAs within and between these two hosts is undoubtedly important for influenza epidemiology . However , the potential for a virus to emerge and spread widely in human or swine populations is likely to be a multi-genic trait comprising factors beyond just antigenicity ( Trock et al . , 2012 ) . These factors include receptor binding ( Matos-Patrón et al . , 2015; Elderfield et al . , 2014 ) , adaptations to the matrix , non-structural , and neuraminidase proteins ( Elderfield et al . , 2014 ) as well as other factors differing between human and swine influenza viruses yet to be identified . There is also a need to better understand the competition dynamics of different influenza virus variants and subtypes in both human and swine populations and how these dynamics affect the potential for invasion by novel viruses . Quantifying the public health risk of circulating swine influenza viruses in terms of immunological protection could be improved by assessing human population immunity to a selection of the antigenically-diverse swine viruses characterized in this study . A globally coordinated and systematic pipeline of antigenic and genetic analyses of relative antigenic distance between human and swine strains for subsequent testing of highly divergent swine strains against human sera ( age-stratified ) would add valuable information to the current WHO-led emergent pandemic risk assessment and inform the design of strategic vaccine stockpiles for human populations most at risk of infection with swine influenza virus , whether by age , geographic location or other factors . Using hemagglutination inhibition ( HI ) assays , we characterized the antigenic properties of swine influenza A H1 and H3 virus strains circulating 1 ) in Europe , as part of the third program of the European Surveillance Network for Influenza in Pigs ( ESNIP3 ) ( www . esnip3 . com ) , and 2 ) in Asia in collaboration with laboratories in Hong Kong , Japan , Vietnam and Thailand . We also expanded previously published datasets for North America . To this HI assay dataset we added previously published swine and human HI assay data for H1 and H3 influenza viruses ( Lewis et al . , 2014; de Jong et al . , 2007; Lorusso et al . , 2011; Koel et al . , 2013; Nfon et al . , 2011 ) , resulting in an antigenic dataset consisting of 194 swine and human H1 influenza viruses and 379 swine and human H3N2 influenza viruses . Viruses were propagated in Madin-Darby canine kidney ( MDCK ) cells or embryonated fowls’ eggs . Harvested cell culture supernatant or allantoic ﬂuid was clarified by centrifugation . For antisera production virus was concentrated by ultracentrifugation over a 20% w/v sucrose cushion . Virus pellets were resuspended overnight at 4°C in sterile phosphate buffered saline at pH 7 . 4 and stored at - 70°C for use in immunizing naïve pigs . Swine antisera to influenza A viruses were generated by the United States Department of Agriculture National Animal Disease Centre , Ames , Iowa as previously described ( Lorusso et al . , 2011 ) . Additional sera were generated by the French Agency for Food , Environmental and Occupational Health & Safety , Ploufragan , France , and by the Technical University of Denmark , Copenhagen , Denmark . To generate these additional antisera each animal was first inoculated intranasally with live influenza virus ( 7 to 8 log10 EID50 in a volume of 3 ml; 1 . 5 ml per nostril ) . Three weeks later equal volumes of the same inoculating virus and Freund’s complete adjuvant ( total volume 3 ml ) or inoculating virus and Montanide ISA206 ( Seppic , Givaudan-Lavirotte , France ) ( total volume 2 ml ) were administered by intramuscular/intradermal injection The animals were killed and bled two weeks after the last immunization . For sera raised in the USA , pigs were cared for in compliance with the Institutional Animal Care and Use Committee of the National Animal Disease Centre and for sera raised in the EU , all animal work was done in accordance with the local rules and procedures with ethical permissions granted following the codes of practice for performing scientific studies using animals . HI assays using swine antisera were performed to compare the antigenic properties of swine and human influenza H1 and H3 viruses . Prior to HI testing , sera used for testing most H1 and all H3 viruses were treated with receptor-destroying enzyme ( Sigma-Aldrich , MO , USA ) and sera used for testing H1 viruses from the USA were treated with kaolin ( Fisher Scientific , Pittsburg , PA , USA ) . All sera were then heat inactivated at 56°C for 30 min and adsorbed with 50% turkey red blood cells ( RBC ) to remove nonspecific inhibitors of hemagglutination . HI assays were performed by testing reference antisera raised to swine influenza viruses against selected H1 and H3 swine and human influenza viruses according to standard techniques . Serial 2-fold dilutions starting at 1:10 were tested for their ability to inhibit the agglutination of 0 . 5% turkey RBC with four hemagglutinating units of swine and human H1 and H3 viruses . To combine HI data generated in different laboratories we used a standard reference panel consisting of viruses and swine antisera raised to selected reference strains and shared among laboratories ( Lewis et al . , 2014; Lorusso et al . , 2011 ) . Our initial quantitative analyses of the antigenic properties of swine and human influenza H1 and H3 viruses used antigenic cartography methods as previously described for human ( H3 ) and swine influenza A ( H3 ) and ( H1 ) viruses ( Smith et al . , 2004; Lewis et al . , 2014; de Jong et al . , 2007; Lorusso et al . , 2011; Nfon et al . , 2011 ) . To identify the most appropriate dimensionality of the antigenic maps for each virus subtype , we made first constructed antigenic maps of H1 and H3 viruses in 1 , 2 , 3 , 4 , and 5 dimensions and compared the HI titer differences to the resultant antigenic map distances for each dimension . Increasing from 1 to 2 , and from 2 to 3 dimensions , resulted in significant improvement in correlation between HI data and map distances indicating that 3 dimensional ( 3D ) maps provide a more accurate representation of the underlying HI data ( Supplementary file 1 ) . Further increasing map dimensionality resulted in smaller and statistically insignificant improvements in the quality of the fit of the map to the underlying HI data . Based on these results , we concluded that 3D maps were sufficient for capturing the antigenic variation of both H1 and H3 viruses in our combined swine and human datasets . Next we removed a random 10% , 20% , 30% , 40% , and 50% of the HI data and re-made the antigenic maps to test the robustness of the antigenic maps and the resolution at which they could visualize antigenic distances among viruses . The precision of point positioning in the 3D antigenic maps was 0 . 68 antigenic units ( AU ) for H1 and 0 . 85 AU for H3 . One AU is equivalent to a two-fold difference in HI assay titer . Three AUs are considered sufficient antigenic difference to warrant an update of the human influenza vaccine . For all isolates with both HA sequence data and HI measurements available , we implemented a Bayesian multidimensional scaling ( BMDS ) cartographic model to jointly infer antigenic and phylogenetic relationships of the viruses as described by Bedford et al . ( 2014 ) In brief , time-resolved phylogenies were estimated for a total of 194 H1 and 379 H3 sequences using the Bayesian Markov chain Monte Carlo ( MCMC ) method implemented in BEAST ( Drummond et al . , 2012 ) . We incorporated the Hasegawa-Kishino-Yano ( HKY ) model of nucleotide substitution model ( Hasegawa et al . , 1985 ) with codon partitioning at all three positions , and a Bayesian Skygrid coalescent model ( Gill et al . , 2013 ) . A strict molecular clock and an uncorrelated lognormal relaxed clock were used for H3 and H1 , respectively , based on linear regression analyses in Path-o-Gen v1 . 4 ( http://tree . bio . ed . ac . uk/software/pathogen/ ) . Ancestral sequence states were reconstructed using the ‘renaissance counting’ method ( Lemey et al . , 2012 ) . Two independent MCMC chains were run for 200 million states with sampling every 20 , 000 states and a burn-in of 20 million states , and then combined and further subsampled every 180 , 000 states for a total of 2000 trees after assessing convergence in Tracer v1 . 6 ( http://tree . bio . ed . ac . uk/software/tracer/ ) . Along with the matching HI measurements for the virus isolates against post-infection ferret and swine antisera , these subsets of 2000 trees were used as an empirical tree distribution for the subsequent BMDS analysis to simultaneously model antigenic locations and evolutionary history . The BMDS approach implements a Bayesian analog of antigenic cartographic models from Smith et al . ( Smith et al . , 2004 ) that arrange virus and serum locations in N-dimensions , such that Euclidean distances between locations are inversely proportional to serological cross-reactivity ( Bedford et al . , 2014 ) . We generated the antigenic maps by imposing a weakly informative prior on expected antigenic locations such that antigenic distance increases with sampling time along one dimension and incorporated genetic data by modelling changes in antigenic phenotype as a diffusion process along the viral phylogeny , i . e . expected virus locations co-vary with genetic relatedness . Virus avidities and serum potencies were also estimated to account for experimental variation in serum and virus reactivity . MCMC was used to sample virus and serum locations in either two or three antigenic dimensions , as well as virus avidities , serum potencies , antigenic drift rate , MDS precision , virus and serum location precisions , and phylogenetic patterns . MCMC chains were run for 500 million states with sampling every 200 , 000 states , and checked for convergence by high ESS values in Tracer . Following burn-in of 100 million states , we obtained a total of 2 , 000 trees from which the maximum clade credibility tree was summarized in TreeAnnotator v1 . 8 . 2 . We used TimeSlicer available in SPREAD ( Bielejec et al . , 2011 ) to quantify the rates of antigenic evolution and levels of standing diversity over the posterior distribution of trees . This was achieved by ‘slicing’ through the BMDS-derived antigenic map location-tagged phylogenies at year-intervals , imputing the unobserved ancestral antigenic locations for branches that intersect those time points , and summarizing the drift front distance ( in AU ) from the most recent common ancestor ( MRCA ) and the mean pairwise distance between locations ( in AU ) for each lineage at that slice time . The overall antigenic drift rate was calculated by taking the total imputed drift front distance from the lineage MRCA divided by the total time elapsed , measured in AU per year . 95% high posterior density ( HPD ) estimates were used to measure the uncertainty in these inferences from the posterior sample of trees . H1 and H3 influenza A hemagglutinin ( HA1 ) sequences representing the swine reference strains as well as viruses from swine populations in all available geographic regions were compiled from the NIAID Influenza Research Database ( IRD ) through the web site at http://www . fludb . org on 22/9/2014 ( Squires et al . , 2012 ) and combined with then unpublished sequences from European swine influenza viruses generated as part of the ESNIP3 project ( Watson et al . , 2015 ) . We added HA1 human seasonal H1 and H3 influenza virus sequences from IRD that represented the entire time period of the study and all human vaccine strain selections . Nucleotide alignments of the HA1 domain for both subtypes were generated using default settings in MAFFT with subsequent manual correction . For each alignment the sequences were curated for poorly curated data by inferring maximum likelihood phylogenetic trees in IQ-Tree ( Nguyen et al . , 2015 ) having performed the model test procedure to select the best-fit model . The provisional ML trees were then analyzed using Path-O-Gen and outlier sequences in the root-to-tip regression plots were identified and removed . ML phylogenetic trees were then inferred again using IQ-TREE as above and ultrafast bootstrap analyses performed ( Minh et al . , 2013 ) .	Influenza viruses , commonly called flu , infect millions of people and animals every year and occasionally causes pandemics in humans . The immune system can neutralise flu viruses by recognising the proteins on the virus surface , generically referred to as antigens . These antigens change as flu viruses evolve to escape detection by the immune system . These changes tend to be relatively small such that exposure to one flu virus generates immunity that is still effective against other related flu viruses . However , over time , the accumulation of these small changes can result in larger differences such that prior infections no longer provide protection against the new virus . Influenza A viruses infect a wide variety of birds and mammals . Viruses can also transmit from one species to another , which may result in the introduction of viruses with antigens that are new to the recipient species and which have the potential to cause substantial outbreaks . Pig flu viruses have long been considered to be a potential risk for human pandemic viruses and were the source of the 2009 pandemic H1N1 virus . Importantly , humans often transmit flu viruses to pigs . Understanding the dynamics and consequences of this two-way transmission is important for designing effective strategies to detect and respond to new strains of flu . Influenza A viruses of the H1 and H3 subtypes circulate widely in pigs . However , it was poorly understood how closely related swine and human viruses circulating in different regions were to one another and how much the antigens varied between the different viruses . Lewis , Russell et al . have now analysed the antigenic variation of hundreds of H1 and H3 viruses from pigs on multiple continents . The antigenic diversity of recent swine flu viruses resembles the diversity of H1 and H3 viruses observed in humans over the last 40 years . A key factor driving the diversity of the H1 and H3 viruses in pigs is the frequent introduction of human viruses to pigs . In contrast , only one flu virus from a bird had contributed to the observed antigenic diversity in pigs in a substantial way . Once in pigs , human-derived flu viruses continue to evolve their antigens . This results in a tremendous diversity of flu viruses that can be transmitted to other pigs and also to humans . These flu viruses could pose a serious risk to public health because they are no longer similar to the current human flu strains . These findings have important implications not only for developing flu vaccines for pigs but also for informing the development of more-effective surveillance and disease-control strategies to prevent the spread of new flu variants .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "epidemiology", "and", "global", "health", "microbiology", "and", "infectious", "disease" ]	2016	The global antigenic diversity of swine influenza A viruses
Ciliated surfaces harbouring synchronously beating cilia can generate fluid flow or drive locomotion . In ciliary swimmers , ciliary beating , arrests , and changes in beat frequency are often coordinated across extended or discontinuous surfaces . To understand how such coordination is achieved , we studied the ciliated larvae of Platynereis dumerilii , a marine annelid . Platynereis larvae have segmental multiciliated cells that regularly display spontaneous coordinated ciliary arrests . We used whole-body connectomics , activity imaging , transgenesis , and neuron ablation to characterize the ciliomotor circuitry . We identified cholinergic , serotonergic , and catecholaminergic ciliomotor neurons . The synchronous rhythmic activation of cholinergic cells drives the coordinated arrests of all cilia . The serotonergic cells are active when cilia are beating . Serotonin inhibits the cholinergic rhythm , and increases ciliary beat frequency . Based on their connectivity and alternating activity , the catecholaminergic cells may generate the rhythm . The ciliomotor circuitry thus constitutes a stop-and-go pacemaker system for the whole-body coordination of ciliary locomotion . Multiciliated surfaces characterised by many beating cilia are widespread in eukaryotes . Such surfaces can effectively generate flow in many different contexts , including the driving of solute transport in reef corals ( Shapiro et al . , 2014 ) , moving mucus and particles in mucociliary epithelia ( Kramer-Zucker et al . , 2005; Walentek et al . , 2014 ) , driving the cerebrospinal fluid ( Faubel et al . , 2016 ) , or moving the ovum in the mammalian oviduct ( Halbert et al . , 1989 ) . Multiciliated surfaces can also drive locomotion , as in ciliates , colonial green algae , or marine invertebrate larvae ( Tamm , 1972; Chia et al . , 1984; Brumley et al . , 2015 ) . A universal feature of multiciliated surfaces is the long-range beat synchronisation of individual cilia into metachronal waves ( Tamm , 1984; Brumley et al . , 2012; Knight-Jones , 1954; Tamm , 1972 ) . Theoretical studies indicate that metachronal waves are energetically efficient and have a higher efficiency of flow generation than non-metachronal beating ( Osterman and Vilfan , 2011; Gueron and Levit-Gurevich , 1999; Elgeti and Gompper , 2013 ) . Metachronal coordination requires the orientation of ciliary beating planes during development ( Mitchell et al . , 2007; Park et al . , 2008; Mitchell et al . , 2009; Guirao et al . , 2010; Vladar et al . , 2012; Kunimoto et al . , 2012 ) . If cilia are oriented , synchronisation emerges by hydrodynamic coupling as adjacent cilia engage by flow ( Brumley et al . , 2014; Elgeti and Gompper , 2013 ) . In addition , basal body coupling can also contribute to beat synchronisation as demonstrated in green algae ( Wan and Goldstein , 2016 ) . However , biophysical mechanisms alone cannot explain all aspects of ciliary beat synchronisation . For example , the flow-networks generated by ependymal cilia show complex reorganisation that is under circadian regulation ( Faubel et al . , 2016 ) . In mucociliary surfaces , changes in ciliary beat frequency are coordinated to regulate flow rates . Some of these changes are due to locally secreted diffusible molecules , including serotonin ( Maruyama et al . , 1984; König et al . , 2009; Walentek et al . , 2014 ) and neuropeptides ( Conductier et al . , 2013 ) . However , long-range ciliary coordination is often under neuronal control ( Doran et al . , 2004; Tamm , 2014; Arkett et al . , 1987; Kuang and Goldberg , 2001 ) . Ciliary swimmers display the most elaborate forms of neuronal ciliary coordination . In the ctenophore Pleurobrachia , prey-capture triggers a rapid synchronised beating of all eight ciliary comb-rows resulting in fast forward swimming ( Tamm , 2014 ) . In gastropod veliger larvae , the normal beating of velar cilia is periodically interrupted by coordinated , velum-wide ciliary arrests , triggered by calcium-spikes ( Arkett et al . , 1987 ) . Similarly , annelid larvae show regular and coordinated arrests of the entire prototroch ciliary band ( Conzelmann et al . , 2011 ) . Such coordinated changes in ciliary activity , often triggered throughout the whole body , require neuronal control . The neuronal circuits coordinating whole-body ciliary activity have not been described in any animal . Here we reconstruct and functionally analyse the complete ciliomotor circuitry in the planktonic nectochaete larva of the marine annelid Platynereis dumerilii . Platynereis has emerged as a powerful model to study neural cell types and development ( Tomer et al . , 2010; Marlow et al . , 2014; Tessmar-Raible et al . , 2007 ) and the whole-body circuit bases of larval behaviour , including conical-scanning and visual phototaxis ( Jékely et al . , 2008; Randel et al . , 2014 ) . Platynereis nectochaete larvae ( Fischer et al . , 2010 ) have three trunk segments and use segmentally arranged ciliary bands to swim and muscles to turn while swimming ( Randel et al . , 2014 ) or to crawl on the substrate . Ciliary beating is also under neuromodulatory control and can be influenced by several neuropeptides and melatonin ( Conzelmann et al . , 2011; Tosches et al . , 2014 ) . We used whole-body connectomics , transgenic neuron labelling , and calcium imaging to reconstruct and functionally analyse the entire ciliomotor system in Platynereis larvae . We identified a ciliomotor system consisting of interconnected catecholaminergic , cholinergic , and serotonergic ciliomotor neurons . These neurons form a pacemaker system responsible for the whole-body coordination of alternating episodes of ciliary arrests and ciliary beating to regulate larval swimming . Platynereis nectochaete larvae have segmentally arranged locomotor ciliated cells that form distinct ciliated fields ( Figure 1A–D ) . The main ciliary band ( prototroch ) is located between the head and the trunk . The head has two patches of locomotor cilia ( akrotroch ) and multiciliated cells that cover the olfactory pit of the nuchal organs ( nuchal cilia ) . There is also an unpaired multiciliated cell , the crescent cell , located anteriorly in the head in the middle of the sensory region of the apical organ . The three chaetigerous ( with chaetae ) trunk segments each have four fields of cilia forming the paratroch ciliary bands ( paratroch I-III ) . There is an additional non-chaetigerous cryptic segment between the head and the trunk segments ( Steinmetz et al . , 2011 ) that harbours ventrally the metatroch ciliary band . Younger larvae have four posterior telotroch cells that merge with the paratroch III ( eight cells ) during development to form a posterior ciliary band of 12 cells ( Starunov et al . , 2015 ) ( referred to as paratroch III ) . 10 . 7554/eLife . 26000 . 003Figure 1 . Coordinated ciliary beating and arrests in Platynereis larvae . ( A–B , D ) Scanning electron micrographs of 72 hr-post-fertilization larvae , ( A ) ventral view , ( B ) lateral view , ( D ) anterior view , close up . The structures dorsal to the crescent cell are non-motile sensory cilia . ( C ) Serial TEM reconstruction of ciliary band cells , lateral view . Nuclei of ciliary band cells are shown as spheres . The nervous system ( brain and ventral nerve cord ) is shown in light grey . ( E ) Kymographs of spontaneous ciliary activity from an immobilized 72 hr-post-fertilization larva . Arrowheads indicate the beginning of ciliary arrests . ( F ) Ventral view of the left half of a 72 hr-post-fertilization larva in a calcium-imaging experiment ( left panel ) with the corresponding differential interference contrast ( DIC ) image ( right panel ) . ( G ) GCaMP6s signal from the prototroch and paratroch ciliary bands . A kymograph of ciliary activity in paratroch III is shown below . ( H ) GCaMP6s signal recorded from the prototroch of a 48 hr-post-fertilization larva at 45 frames per second . White areas in the kymograph correspond to periods of arrest . The boxed area is shown enlarged in the inset . Scale bars , 50 μm ( A ) , 10 μm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 003 To investigate beat coordination across the ciliary bands in Platynereis , we imaged ciliary activity of mechanically immobilized 3-day-old nectochaete larvae . We found that spontaneous arrests of ciliary beating occurred synchronously across all ciliary bands ( Figure 1A–E and Video 1 ) . Cilia first resumed beating in the prototroch followed in synchrony by the paratroch cilia ( with a delay of 19 ± 3 msec relative to the prototroch cilia ) . Arrests of prototroch cilia were also in synchrony with arrests in the akrotroch ( Video 2 ) and the nuchal cilia ( data not shown ) . 10 . 7554/eLife . 26000 . 004Video 1 . Coordinated arrests of cilia on all ciliary band cells in a 72 hr-post-fertilization Platynereis larva . Differential interference contrast ( DIC ) imaging , recorded and played at 60 frames per second ( fps ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 00410 . 7554/eLife . 26000 . 005Video 2 . Coordinated arrests of akrotroch and paratroch but not crescent cilia . DIC imaging of cilia in the Platynereis larval head . Recorded and played at 30 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 005 We next combined differential interference contrast ( DIC ) imaging with calcium imaging using a ubiquitously expressed fluorescent calcium indicator , GCaMP6s . In Platynereis larvae , regular arrests are triggered by bursts of spikes that can be recorded from the ciliary band cells ( Conzelmann et al . , 2011; Tosches et al . , 2014 ) . With calcium imaging , we detected periodic synchronous increases in calcium signals in cells of each ciliary band that corresponded with the arrests of cilia ( Figure 1G , Video 3 ) . 10 . 7554/eLife . 26000 . 006Video 3 . Calcium and DIC imaging of ciliary bands in a 72 hr-post-fertilization Platynereis larva . The GCaMP6s signal ( left panel ) and DIC signal ( right panel ) are shown for the left side of the larva ( ventral view ) . Regions of interest used to record calcium signals are indicated . Recorded at 3 fps , played at 25 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 006 The only cilia that did not beat and arrest synchronously were the cilia of the crescent cell ( Figure 1D ) . Although calcium signals also increase in the crescent cell simultaneously with the other multiciliated cells ( data not shown ) , the cilia of this cell were beating when the other cilia were arrested ( Video 2 ) . Crescent cilia stopped beating after the locomotor cilia resumed beating ( Video 2 ) . The crescent cell has a lower density of cilia that beat more irregularly and do not organise into metachronal waves ( Video 2 and Figure 1D ) . To investigate in more detail how intracellular calcium concentrations [Ca2+]i in the prototroch cells relate to the arrests of cilia , we imaged calcium activity in close-ups from the prototroch cells at higher frame rates ( Video 4 ) . We found that ciliary activity did not correlate with the absolute levels of [Ca2+]i , but with the time derivative of calcium d[Ca2+]i/dt ( Figure 1H ) . As long as calcium levels increase , the cilia are arrested . As soon as calcium levels start to drop , the cilia resume their metachronal beating . A similar case was reported in sea urchin sperm where d[Ca2+]i/dt rather than absolute [Ca2+]i controls ciliary waveform during chemotaxis ( Alvarez et al . , 2012 ) . 10 . 7554/eLife . 26000 . 007Video 4 . Calcium and DIC imaging of the ciliary band in a 48 hr-post-fertilization Platynereis larva . The video was recorded at 45 fps , here only every 10th frame is shown at 45 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 007 To understand the mechanism of ciliary beat synchronisation across ciliary bands and to get a comprehensive view of ciliomotor circuits in the Platynereis larva , we employed serial-section transmission electron microscopy ( ssTEM ) . We used a whole-body ssTEM dataset of a 3 day old larva ( Randel et al . , 2015; Shahidi et al . , 2015; Shahidi and Jékely , 2017 ) to identify and reconstruct all neurons innervating ciliary band cells . First , we identified all multiciliated cells in this larva ( Figure 1C ) . There are 23 cells in the prototroch that are arranged in an anterior and posterior tier , with one unpaired cell in the position of 11 o’clock ( Randel et al . , 2014 ) . In the head , there are eight akrotroch cells , six ciliated cells in the nuchal organs , and an unpaired crescent cell . The metatroch consists of eight cells . In the first , second , and third trunk segments there are eight , 14 , and 12 paratroch cells , respectively . We next identified and reconstructed all neurons that were presynaptic to any of the multiciliated cells in the larva ( Jékely and Verasztó , 2017 ) . We define these neurons as ciliomotor neurons . Combining transgenic labelling , immunohistochemistry , and serial multiplex immunogold labelling ( siGOLD ) ( Shahidi et al . , 2015 ) , we could assign a neurotransmitter and/or a neuropeptide to most of the ciliomotor neurons . We identified three main ciliomotor systems , consisting of either cholinergic , serotonergic , or mixed catecholaminergic/peptidergic neurons . We describe these in detail below . We identified 11 cholinergic ciliomotor neurons in the larva ( Figure 2 , Figure 2—figure supplement 1 ) . These include the previously described six ventral decussating cholinergic motorneurons ( vMNs , including MNr1 , MNr2 , MNr3 , MNl1 , MNl2 , MNl3 ) and two rhabdomeric photoreceptor cells of the larval eyespots ( eyespotPRCR3 ) ( Figure 2—figure supplement 1 ) ( Randel et al . , 2014 , 2015; Jékely et al . , 2008 ) . In addition to these neurons , we found three large biaxonal ciliomotor neurons , including a single ciliomotor neuron in the head , the MC neuron , and two trunk ciliomotor neurons that we call the Loop neurons ( Figure 2A–D , Video 5 ) . 10 . 7554/eLife . 26000 . 008Figure 2 . Anatomy and connectivity of biaxonal cholinergic ciliomotor neurons . ( A ) ssTEM reconstruction of the MC neuron ( blue ) , ventral view . Ciliated cells are shown in grey . Circles represent position of cell body , lines represent axonal track . Presynaptic sites along the axon are marked in red . ( B ) The MC neuron in anterior view . ( C ) ssTEM reconstruction of the Loop neurons , ventral view . ( D ) The right Loop neuron in lateral view . ( E ) Synaptic connectivity of the MC neuron and the prototroch cells . ( F ) Synaptic connectivity of the Loop neurons and ciliary band cells . ( G ) Transgenic labelling of the right Loop neuron with an sGCβ1:palmi-3xHA-tdTomato construct . Expression was detected with anti-HA antibody staining . Right panel: ventral view , left panel: lateral view . Asterisks indicate background signal in the spinning glands . In the network graphs , nodes represent individual or grouped cells , as indicated by the numbers in square brackets . The edges show the direction of signalling from presynaptic to postsynaptic cells . Edges are weighted by the number of synapses . The number of synapses is indicated on the edges . Scale bar , 40 μm ( G ) . Abbreviation: prn , prototroch ring nerve . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 00810 . 7554/eLife . 26000 . 009Figure 2—figure supplement 1 . Anatomy and connectivity of eyespotPRCR3 , ventral MN , and MNant ciliomotor neurons . ( A ) ssTEM reconstruction of the eyespotPRCR3 neurons , anterior view . Ciliated cells are shown in grey or in teal if they receive synapses from eyespotPRCR3 . ( B ) Synaptic connectivity of the eyespotPRCR3 neurons and prototroch . ( C ) ssTEM reconstruction of the MNr1 neurons , ventral view . ( D ) Synaptic connectivity of the MNr1 neurons , ciliary band cells , and muscles . ( E ) ssTEM reconstruction of the MNr2 neurons , ventral view . ( F ) Synaptic connectivity of the MNr2 neurons , ciliary band cells , and muscles . ( G ) ssTEM reconstruction of the MNr3 neurons , ventral view . ( H ) Synaptic connectivity of the MNr3 neurons and ciliary band cells . ( I–L ) ssTEM reconstruction of the MNant neurons , ( I ) anterior view , ( J , K ) ventral view . ( L ) Synaptic connectivity of the MNant neurons and ciliary band cells . In the network graphs , nodes represent individual or grouped cells , as indicated by the numbers in square brackets . The edges show the direction of signalling from presynaptic to postsynaptic cells . Edges are weighted by the number of synapses . The number of synapses is indicated on the edges . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 00910 . 7554/eLife . 26000 . 010Figure 2—figure supplement 2 . sGCbeta1 is a cholinergic marker . Average registered gene expression pattern of ( A ) ChAT and ( B ) sGCbeta1 . ( C ) Overlap of the ChAT and sGCbeta1 gene expression domains . ( D–F ) Transgenic labeling of MNl3 neuron with the sGCbeta1:palmi-3xHA-tdTomato construct . Larvae in ( D , E ) were counterstained for acetylated tubulin ( white ) . Expression was detected with anti-HA antibody staining . Scale bar in ( D , F ) 30 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 01010 . 7554/eLife . 26000 . 011Video 5 . The Loop ciliomotor neurons . ssTEM reconstruction of the Loopl and Loopr neurons . Ciliary bands and the ventral nerve cord are shown in grey . Synapses are marked in red . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 011 The MC neuron has its soma in the centre of the larval head and projects two axons that run left and right towards the ciliary band . The axons branch at the prototroch and form dorsal and ventral branches that run along the prototroch ring nerve , a nerve running at the inside of the prototroch ring . The MC neuron is the most synapse-rich neuron in the entire Platynereis larval connectome ( unpublished results ) ( Randel et al . , 2015; Shahidi et al . , 2015; Randel et al . , 2014 ) . We identified 341 presynaptic sites in the MC neuron . 335 of these target the ciliated cells of the prototroch ciliary band . The MC neuron is the only neuron in the body that is presynaptic to each of the 23 prototroch cells ( 3–25 synapses per prototroch cell ) ( Figure 2B , E , Supplementary file 1 ) . The somas of the Loop neurons are located in the first trunk segment . These neurons project one axon anteriorly and one posteriorly ( Figure 2C , D , Video 5 ) . The anterior axon joins the prototroch ring nerve and continues in the dorsal side of the larva where it runs posteriorly along a thin dorsal nerve . The posterior axon runs at the lateral edge of the ventral nerve cord , spans all three trunk segments and in the third segment runs to the dorsal side where it turns anteriorly . The anterior and posterior axons meet on the dorsal side . The Loop neurons thus loop around the entire body spanning the head and three segments in both the ventral and dorsal side . The Loop neurons have additional side branches in the head and in every segment . These project to and synapse on cells of the different ciliary bands . A single Loop neuron synapses on every ipsilateral paratroch cell , the akrotroch , the ciliated cells of the nuchal organ , and the lateral cells of the prototroch ( Figure 2F ) . The Loop neurons also synapse on the crescent cell and represent its only synaptic input . Overall , the two Loop neurons innervate all locomotor ciliary fields in the larva , with the exception of the metatroch where we could not find any synaptic input at either side in the reconstructed specimen . In the prototroch , only some of the most dorsally and ventrally located prototroch cells ( at 5 , 6 , and 12 o’clock position ) lack synaptic input from the Loop neurons ( Figure 2F , Supplementary file 1 ) . The anatomy and connectivity of the ventral cholinergic motorneurons ( vMN ) and the eyespot photoreceptors have already been described ( Jékely et al . , 2008; Randel et al . , 2014 , 2015 ) . We include them here for completion , and because we now have a more complete and revised reconstruction of the large intersegmental vMNs ( Figure 2—figure supplement 1 ) . The vMNs form three left-right pairs ( Figure 2—figure supplement 1 ) . One pair ( MNr1 and MNl1 ) projects to the ventral nerve chord and innervates ventral paratroch cells and ventral longitudinal muscles . The second pair ( MNr2 and MNl2 ) projects dorsally and posteriorly along the trunk dorsal nerve and innervates lateral prototroch cells , dorsal paratroch cells , and dorsal longitudinal muscles . ( Note that we swapped the name of MNl1 and MNl2 relative to ( Randel et al . , 2015 ) to reflect bilateral pairs ) . The third pair ( MNr3 and MNl3 ) is a purely ciliomotor pair ( no muscle targets ) . Both cells synapse on lateral prototroch cells ( only the posterior ring ) and project to the ventral nerve cord . The eyespotPRCR3 cells synapse on two ipsilateral prototroch cells of the posterior prototroch belt ( Jékely et al . , 2008; Randel et al . , 2013 ) , representing two of the three direct sensory-ciliomotor neurons in the larva ( Figure 2—figure supplement 1 ) ( see below ) . The cholinergic identity of the above neurons is supported by pharmacological experiments and by the expression of cholinergic markers ( Jékely et al . , 2008; Randel et al . , 2014; Tosches et al . , 2014; Denes et al . , 2007 ) ( Figure 2—figure supplement 2 ) . To further test this at the single neuron level , we developed a transgenic reporter construct for cholinergic neurons . We could not obtain a working promoter construct for the canonical cholinergic markers vesicular acetylcholine transporter ( VAChT ) and choline acetyltransferase ( ChAT ) , however , we identified a soluble guanylyl cyclase-β gene ( sGCβ1 ) that coexpresses with ChAT as determined by in situ hybridisation combined with whole-body gene expression registration ( Asadulina et al . , 2012 ) ( Figure 2—figure supplement 2 ) . We cloned a 12 kB fragment directly upstream of the start site of sGCβ1 and fused it with a tdTomato fluorescent reporter . The reporter was tagged with three tandem haemagglutinin ( HA ) tags and a palmitoylation sequence . We used zygote microinjection and mosaic transient transgenesis to label individual cholinergic neurons . In transgenic animals , we could label the Loop neuron ( Figure 2G ) and the ventral MNl3 neuron ( Figure 2—figure supplement 2 ) , as identified by position and morphology , confirming their cholinergic identity . We could not label the MC neuron with this transgene , but its cholinergic identity is supported by VAChT expression in the cell body area of this cell ( Figure 2—figure supplement 2 ) ( Jékely et al . , 2008 ) and also by pharmacological evidence ( see below ) . In addition to these 11 cholinergic cells , there are two biaxonal MNant cells , described before ( Randel et al . , 2015 ) , that are likely cholinergic . These cells have their cell bodies below the ciliary photoreceptor cells in a region of VAChT expression , but we could not verify their cholinergic identity at a single-cell resolution . The MNant cells form synapses on the contralateral half of the prototroch , the metatroch , and the paratrochs ( Randel et al . , 2015 ) ( Figure 2—figure supplement 1 ) . We identified five serotonergic ciliomotor neurons in the larva ( Figure 3 ) . There are two serotonergic ciliomotor neurons in the head ( Ser-h1 ) , with their somas lateral to the ciliary photoreceptor cells ( Figure 3A , B and Figure 3—figure supplement 1 ) . These neurons cross the midline and project along the contralateral side of the ciliary band , forming several synapses on the ciliated cells . The Ser-h1 cells also synapse reciprocally on each other in the brain neuropil . 10 . 7554/eLife . 26000 . 012Figure 3 . Anatomy and connectivity of serotonergic ciliomotor neurons . ( A–B ) ssTEM reconstruction of the right ( A ) and left ( B ) head serotonergic ciliomotor ( Ser-h1 ) neuron , anterior view . ( C ) Transgenic labelling of the right Ser-h1 neuron with the TrpH:palmi-3xHA-tdTomato construct , anterior view . Expression was detected with anti-HA antibody staining . ( D–E ) ssTEM reconstruction of the right ( D ) and left ( E ) trunk serotonergic ciliomotor ( Ser-tr1 ) cell , anterior view . ( F ) Transgenic labelling of the left Ser-tr1 neuron with the TrpH:palmi-3xHA-tdTomato construct , ventral view . ( G ) ssTEM reconstruction of the pygPBbicil serotonergic sensory-ciliomotor neuron . Ciliated cells are shown in grey . Circles represent position of cell body , lines represent axonal track . Presynaptic sites along the axon are marked in red . ( H ) Synaptic connectivity of the serotonergic ciliomotor neurons and the ciliary band cells . Nodes represent individual or grouped cells , as indicated by the numbers in square brackets . Edges show the direction of signalling from presynaptic to postsynaptic cells . Edges are weighted by the number of synapses . Synapse number is indicated . Edges with one synapse are not shown . Scale bars , 30 μm ( C , F ) . Abbreviation: prn , prototroch ring nerve . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 01210 . 7554/eLife . 26000 . 013Figure 3—figure supplement 1 . Serotonergic neurons in Platynereis larvae . ( A–C ) Immunostaining for serotonin and acetylated tubulin in ( A ) 48 hr-post-fertilization , ( B ) 72 hr-post-fertilization , and ( C ) 50 hr-post-fertilization larvae . Anterior ( A , B ) and ventral views ( C ) . Scale bars , 30 μm ( A , C ) , 20 μm ( B ) . Abbreviations: cPRC , ciliary photoreceptor cell . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 01310 . 7554/eLife . 26000 . 014Figure 3—figure supplement 2 . Parallel axons of the Loop and Ser-tr1 neurons . ( A ) Electron micrographs of Loop and Ser-tr1 neuron axon cross-sections . Axon outlines are shown . The Loop neurons contain several dense core vesicles . The Ser-tr1 neurons have a clearer cytoplasm and no dense core vesicles . ( B ) ssTEM reconstruction of the Loop and Ser-tr1 neurons , ventral view . Loop and Ser-tr1 axons run in parallel throughout the body . Scale bar in ( A ) 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 014 In the trunk , there are two large serotonergic biaxonal neurons ( Ser-tr1 ) with their somas in the first segment , near the somas of the cholinergic Loop neurons , anterior to the first commissure of the ventral nerve cord ( Figure 3D , E and Figure 3—figure supplement 1 ) . The Ser-tr1 and the Loop neurons represent the largest neurons in the entire Platynereis connectome ( between 1 , 600–1 , 850 µm total cable length each ) . One of the axons of the Ser-tr1 cells projects anteriorly and innervates ipsilateral prototroch cells . The other axon crosses the midline in the first segment and its axon loops around the body , running in close proximity to the axons of the contralateral Loop neuron . The Ser-tr1 neurons also innervate every paratroch cell as well as the metatroch ( Figure 3H , Supplementary file 1 ) . The axons of the Ser-tr1 neurons run parallel to the axons of the Loop neurons throughout the body and are ultrastructurally distinct from the Loop neurons . The Ser-tr1 axons are characterised by a less electron-dense cytoplasm and the lack of dense core vesicles ( abundant in the Loop neurons ) ( Figure 3—figure supplement 2 ) . The Ser-tr1 cells were identified as serotonergic by comparing cell body positions and neuron morphologies to cells stained with a serotonin antibody in 2-day-old and 3-day-old larvae ( Figure 3—figure supplement 1 ) ( Fischer et al . , 2010 ) . We also used mosaic transient transgenesis with a construct containing a 5 kB upstream regulatory sequence of the tryptophan hydroxylase gene ( TrpH ) , a marker of serotonergic neurons . We could label both the head Ser-h1 neurons and the Ser-tr1 cells ( Figure 3C , F ) . The unique morphology of these cells and the correspondence of the TEM reconstruction ( Video 6 , Video 7 ) to the transgenic labelling confirm the serotonergic identity of these cells . 10 . 7554/eLife . 26000 . 015Video 6 . The Ser-tr1 ciliomotor neurons . ssTEM reconstruction of the Ser-tr1l and Ser-tr1r neurons . Ciliary bands are shown in grey . The metatroch is shown in darker grey , synapses are marked in red . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 01510 . 7554/eLife . 26000 . 016Video 7 . The Ser-tr1l neuron labelled by transient transgenesis . The TrpH:palmy-3xHA-tdTomato was used to label the Ser-tr1l neuron . The reporter was visualised by anti-HA antibody staining . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 016 We also identified a giant serotonergic sensory-motor neuron ( pygPBbicil ) with a cell body in the pygidium and a unique morphology ( Figure 3G ) . This cell has two sensory cilia and two axons that run anterior along the ventral nerve chord . The axons then innervate the prototroch cilia in the head and send a branch to the anterior nervous system . The pygPBbicil neuron is identifiable by serotonin immunolabeling ( Figure 3—figure supplement 1 ) and also expresses the neuropeptide FMRFamide as determined by serial multiplex immunogold ( siGOLD ) labelling ( Shahidi et al . , 2015 ) . This cell , together with the eyespot photoreceptor cells , is the third sensory neuron that directly synapses on ciliated cells . There are three molecularly distinct , asymmetric peptidergic cells ( cMNvl , cMNd , cMNATO ) with their cell bodies at the level of the prototroch ciliary band ( Figure 4 ) . ( cMNvl was referred to in [Shahidi et al . , 2015] as cMNPDF-vcl1 ) . The two ventral-lateral cells ( cMNvl , cMNATO ) express tyrosine hydroxylase , a marker of dopaminergic neurons ( Figure 4—figure supplement 1 ) . The cMNATO cell also expresses dopamine-β-hydroxylase , indicating that it is noradrenergic ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 26000 . 017Figure 4 . Anatomy and connectivity of catecholaminergic/peptidergic ciliomotor neurons . ( A–C ) ssTEM reconstruction of the cMNd ( green ) ( A ) , cMNATO ( orange ) ( B ) , and cMNvl ( blue ) ( C ) ciliomotor neurons , anterior view . Ciliated cells are shown in grey . Circles represent position of cell bodies , lines represent axonal tracks . Presynaptic sites along the axon are marked in red . ( D ) Synaptic connectivity of the cMN neurons with the prototroch and the MC cell . Nodes represent individual or grouped cells , as indicated by the numbers in square brackets . Edges show the direction of signalling from presynaptic to postsynaptic cells . Edges are weighted by the number of synapses and indicate the direction of signalling . Synapse number is indicated . Abbreviation: prn , prototroch ring nerve . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 01710 . 7554/eLife . 26000 . 018Figure 4—figure supplement 1 . Catecholaminergic and neuropeptide marker expression in cMN neurons . ( A–C ) In situ hybridisation for TyrH ( A ) , dbh ( B ) , and allatotropin proneuropeptide ( C ) . ( D ) Immunostaining with an anti-allatotropin antibody . The samples were counterstained for acetylated tubulin ( white ) . Scale bar in ( A ) 30 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 018 The cMN cells express different combinations of neuropeptides . The neurons cMNvl and cMNd express pigment dispersing factor ( PDF ) , as determined by siGOLD ( Shahidi et al . , 2015 ) . The cMNATO cell expresses allatotropin and its full morphology is revealed by immunostaining with an allatotropin antibody ( Figure 4—figure supplement 1 ) . The cMNvl neuron also expresses FLamide and FVamide , cMNATO expresses DLamide , and cMNd expresses the neuropeptides L11 , and LYamide , as revealed by immunostainings ( Conzelmann et al . , 2011 ) and in situ hybridisation ( Figure 4—figure supplement 1 ) . The cMN cells synapse on prototroch cells and the MC cell and also synapse reciprocally among themselves ( Figure 4D , Supplementary file 1 ) . To analyse the overall connectivity of the ciliomotor system ( Video 8 , Supplementary file 1 ) , we next performed modularity analysis to identify more strongly connected communities of cells ( Blondel et al . , 2008 ) . We could subdivide the network of ciliomotor neurons and multiciliated cells into three communities each containing cells that were more strongly connected among each other ( Figure 5A , Supplementary file 1 ) . The three modules can be distinguished primarily based on the ciliated cells they contain ( Figure 5B and Supplementary file 1 ) . Module 1 includes ciliomotor neurons that only innervate prototroch cells ( MC , cMN , Ser-h1 , and pygPBbicil , MNl3 , MNr3 , eyespotPRCR3 , and most prototroch cells ) . Module 2 is composed of the right ( ipsilaterally-projecting ) Loopr neuron and the left ( contralaterally projecting ) Ser-tr1l , MNantl , MNl1 , and MNl2 neurons . These neurons share ciliated target cells that are primarily on the right body side ( Figure 5A , B ) . Module 3 includes neurons providing innervation primarily to cilia on the left body side ( Loopl , Ser-tr1r , MNantr , MNr1 , and MNr2 ) . Module 2 and 3 thus contain the corresponding set of left-right neuron pairs and their ciliary targets . 10 . 7554/eLife . 26000 . 019Figure 5 . Overall connectivity of ciliomotor neurons and ciliary bands . ( A ) Synaptic connectivity graph of all ciliomotor neurons and ciliary band cells . The network is partitioned into three modules . The neurons belonging to each module are shown in the anatomical reconstructions . The ciliary band cells are grouped into anatomical units . ( B ) Matrix representation of the connectivity of ciliomotor neurons to all ciliated cells . Colour intensity is proportional to the number of synapses . ( C ) Matrix representation of the connectivity of ciliomotor neurons amongst themselves . Colour intensity is proportional to the number of synapses . ( D–E ) Graph representations of the ciliomotor circuit . Nodes with <4 synapses are not shown . In ( D ) nodes are grouped by cell type ( teal , cholinergic; red , serotonergic; orange , cMN; grey , ciliated cells ) . In ( E ) all cMN , serotonergic , cholinergic neurons , and ciliary band cells are represented as one node each . In ( D , E ) edges with <4 synapses are not shown . Abbreviations: proto , prototroch; meta , metatroch; para , paratroch; akro , akrotroch . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 01910 . 7554/eLife . 26000 . 020Video 8 . The entire ciliomotor circuit of the Platynereis larva . Neuron reconstructions of all ciliomotor neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 020 There are also many interconnections between ciliomotor neurons expressing distinct transmitters ( cholinergic , serotonergic , and catecholaminergic/peptidergic ) ( Figure 5C–E , Supplementary file 1 ) . The cMN neurons are presynaptic to the MC neuron , the MNant and the Loop neurons , and show reciprocal connectivity with the Ser-h1 and pygPBbicil cells . The serotonergic and cholinergic systems are also interconnected . The Ser-h1 cells are presynaptic to the MC neuron and the Ser-tr1 and Loop neurons are reciprocally connected . The pygPBbicil neuron is also presynaptic to the MC cell ( Figure 5C–E ) . To functionally characterise the ciliomotor system , we performed calcium-imaging experiments in immobilized larvae ubiquitously expressing GCaMP6s . Since 3-day-old larvae are difficult to immobilize without compromising neuronal activity , we used 2-day-old larvae for activity imaging . At this stage , the musculature is not yet well developed . However , several ciliomotor neurons are already present as evidenced by immunostaining or in situ hybridisation ( cMN , serotonergic , cholinergic cells ) ( Figure 3—figure supplement 1 , Figure 4—figure supplement 1 ) . Calcium imaging revealed that the cMN neurons display spontaneous rhythmic activity patterns . The somas of the cMN cells are close to the prototroch cells at unique positions allowing their unambiguous identification . We found that the cMNvl and cMNd neurons showed rhythmic increases in the GCaMP signal in synchrony with the ciliary band ( Figure 6 ) . In contrast , the activity of the cMNATO was negatively correlated with the two other cMN neurons and the ciliary band ( Figure 6C , D , F , Figure 6—source data 1 ) . We could detect the rhythmic activation of the cMN neurons in larvae from 24 hr-post-fertilisation ( hpf ) onwards . 10 . 7554/eLife . 26000 . 021Figure 6 . Activity of cMN neurons . ( A ) ssTEM reconstruction ( left ) and correlation ( Pearson's r ) map of neuronal activity ( right ) of cMNvl , anterior view . Correlation values were calculated relative to the prototroch ( outlined ) . ( B ) Calcium-imaging trace of cMNvl and the prototroch ( C ) ssTEM reconstruction ( left ) and correlation ( Pearson’s r ) map of neuronal activity ( right ) of cMNATO . Correlation values were calculated relative to the prototroch ( outlined ) . ( D ) Calcium imaging trace of cMNATO and the prototroch . ( E ) ssTEM reconstruction ( left ) and correlation ( Pearson’s r ) map of neuronal activity ( right ) of cMNd . Correlation values were calculated relative to the prototroch ( outlined ) . ( F ) Calcium-imaging trace of cMNd and the prototroch . ( G ) Correlation of GCaMP signals of cMN neurons with GCaMP signals measured from the prototroch ciliary band ( CB ) . Data points represent measurements from different larvae , n > 12 larvae . Mean and standard deviation are shown . All sample medians are different from 0 with p-values≤0 . 0002 as determined by Wilcoxon Signed Rank Test . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 02110 . 7554/eLife . 26000 . 022Figure 6—source data 1 . Source data for Figure 6G with correlation values of neuronal activity . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 022 Next , we imaged the activity of the cholinergic MC and Loop neurons . We could identify the MC neuron based on its position and biaxonal morphology as revealed by the GCaMP signal ( Video 9 ) . We found that the activity of the MC neuron was periodic and strongly correlated with the activity of the prototroch cells and with ciliary arrests ( Figure 7A , D , Figure 7—source data 1 ) . Recording from 96 different larvae ( Figure 7—figure supplement 1 , Figure 7—figure supplement 1—source data 1 ) followed by Fourier analysis of the calcium signals revealed a median cycle length of 71 s of the periodic activation of the MC neuron . In 43% of larvae the MC neuron had a cycle length between 50–100 s ( Figure 7—figure supplement 1 ) . 10 . 7554/eLife . 26000 . 023Figure 7 . Activity of cholinergic neurons . ( A–B ) GCaMP6s signal revealing cell morphologies of the active MC ( anterior view ) ( A ) and Loop ( ventral/posterior view ) ( B ) neurons in 2 days-post-fertilisation larvae . ( C ) Correlation ( Pearson’s r ) map of neuronal activity of the MC , vMNsync and vMNasync neurons . Correlation values were calculated relative to the prototroch . ( D ) Correlation of GCaMP signals of cholinergic neurons with GCaMP signals measured from the prototroch ciliary band ( CB ) . Data points represent measurements from different larvae , n > 12 larvae . Mean and standard deviation are shown . All sample medians are different from 0 with p-values≤0 . 0034 as determined by Wilcoxon Signed Rank Test . ( E–F ) Effect of MC neuron ablation on prototroch activity and ciliary closures . Red bar represents the time of the ablation . GCaMP signals in the MC neuron , in a vMNsync neuron , and in the prototroch before ( E ) and after ( F ) MC neuron ablation . ( G ) Number of ciliary arrests in the prototroch before and after MC neuron ablation . ( H ) Calcium signals measured from the MC neuron and the prototroch following the addition of 50 µM alpha-bungarotoxin . Abbreviation: prn , prototroch ring nerve . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 02310 . 7554/eLife . 26000 . 024Figure 7—source data 1 . Source data for Figure 7D with correlation values of neuronal activity . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 02410 . 7554/eLife . 26000 . 025Figure 7—source data 2 . Source data for Figure 7G with ciliary closures before and after MC neuron ablation . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 02510 . 7554/eLife . 26000 . 026Figure 7—figure supplement 1 . Characterisation of MC neuron activity . ( A ) Normalised calcium signals in the MC neuron ( n = 96 larvae ) . ( B ) Cycle length of MC neurons as determined by Fourier analysis of the traces shown in ( A ) . ( C ) Frequency distribution of cycle length of MC neurons , same data as in panel ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 02610 . 7554/eLife . 26000 . 027Figure 7–figure supplement source data 1 . Source data for Figure 7—figure supplement 1A with calcium imaging data of the MC neuron from 96 individual larvae . The data are organized in rows . The first row is the time stamp ( sec ) . These data are shown in 0–1 normalized form in Figure 7—figure supplement 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 02710 . 7554/eLife . 26000 . 028Figure 7—figure supplement 2 . Effect of MC neuron ablation on prototroch activity and ciliary closures . ( A–F ) Calcium signals in the MC neuron , a vMNrsync neuron , and in the prototroch , before and after MC neuron ablation . Kymographs showing arrests of prototroch cilia are also shown . Arrowheads indicate the beginning of arrests . ( A–F ) show examples of MC neuron ablations in different larvae . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 02810 . 7554/eLife . 26000 . 029Video 9 . Calcium imaging of the MC neuron in a 30 hr-post-fertilisation larva . Ubiquitously expressed GCaMP6s was used to image neuronal activity . The MC neuron , a vMNsync , a vMNasync neuron , and prototroch cilia are labelled . The video was recorded at 1 . 5 fps and is shown at 45 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 029 In the trunk , we also identified two cells in the first segment that were activated synchronously with the ciliary band cells and thus the MC neuron ( though not imaged simultaneously for technical reasons ) . These two trunk cells had somas in the approximate position of the two Loop neurons with ipsilaterally projecting axons , and could thus be identified as the Loop neurons ( Figure 7B , D , see also Figure 8—figure supplement 1A ) . In the connectome graph , the Loop and the MC neurons provide the main cholinergic motor input to the ciliary bands . Their coordinated activation and high number of synapses suggest that these cells are responsible for triggering the coordinated arrests of locomotor cilia across the body . To test this experimentally , we focused on the MC cell . First , we ablated the MC cell with a pulsed laser in 24–36 hra-post-fertilisation larvae . MC-cell ablation eliminated the rhythmic activity in the prototroch cells and thus most ciliary arrests ( Figure 7E , F and Figure 7—figure supplement 2 , Figure 7—source data 2 ) . This shows that the MC cell is essential for triggering the rhythmic arrests of the prototroch cilia . To test whether acetylcholine is involved in rhythm generation and triggering ciliary arrests , we incubated larvae in alpha-bungarotoxin , a specific blocker of nicotinic acetylcholine receptors . When we imaged alpha-bungarotoxin-treated larvae , we observed the decoupling of the activity of the MC neuron and the ciliary band . While the MC neuron continued to rhythmically activate , the calcium signals were progressively reduced and then disappeared in the ciliary band ( Figure 7H ) . Ciliary arrests were also eliminated ( data not shown ) . This indicates that cholinergic signalling is not required for the generation of the neuronal rhythm but is required for ciliary arrests . The ventral cholinergic motorneurons ( vMN ) were previously shown to rhythmically activate , and were suggested to trigger ciliary arrests ( Tosches et al . , 2014 ) . To directly correlate the activity of these cells to ciliary activity we imaged from the ventral MN cells and the prototroch cells simultaneously . We could identify three ventral MN cells that were activated either synchronously or asynchronously ( designated vMNsync and vMNasync ) with the ciliary band ( Figure 7C–E ) . We cannot unambiguously link these three cells to the six ventral MN cells identified by EM , but based on cell body positions and their distinct cell lineage as identified by blastomere injections ( Tosches et al . , 2014 ) , the ventral-most cells likely correspond to MNr3 and MNl3 . Following MC neuron ablation , the activity of the prototroch correlated with the activity of the vMNsync neuron suggesting that this cholinergic neuron can maintain a lower rate of prototroch activation . The cholinergic and serotonergic ciliomotor neurons have many synaptic connections , suggesting that their activity is interdependent ( Figure 5 ) . Correlation analysis of the calcium-imaging videos revealed that the head serotonergic cells ( Ser-h1 ) , identified by position and morphology , showed anti-correlated activity with the ciliated cells and the MC neuron ( Figure 8A , B , E , Figure 8—source data 1 ) . 10 . 7554/eLife . 26000 . 030Figure 8 . Activity of serotonergic neurons . ( A ) ssTEM reconstruction ( left ) and correlation ( Pearson’s r ) map of neuronal activity ( right ) of the Ser-h1 neurons , anterior view . Correlation values were calculated relative to the prototroch . ( B ) Calcium signals measured from a Ser-h1 neuron and the MC cell . ( C ) ssTEM reconstruction ( left ) and neuronal activity correlation ( right ) of the Ser-tr1 neurons , ventral view . Correlation values were calculated relative to the prototroch . ( D ) Calcium signals measured from a Ser-tr1 neuron and a Loop neuron . ( E ) Correlation of GCaMP signals of serotonergic neurons with GCaMP signals measured from the prototroch ciliary band ( CB ) . Data points represent measurements from different larvae , n > 9 . Mean and standard deviation are shown . All sample medians are different from 0 with p-values≤0 . 002 as determined by Wilcoxon Signed Rank Test . ( F ) Ciliary beat frequency of prototroch cilia in the absence and presence of serotonin . ( G ) Ciliary closures of prototroch cilia in the absence and presence of serotonin . P-values of an unpaired ( F ) and paired ( G ) t-test are shown . In ( F , G ) n > 9 larvae . Samples in ( F , G ) passed the D'Agostino and Pearson omnibus normality test ( alpha = 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 03010 . 7554/eLife . 26000 . 031Figure 8—source data 1 . Source data for Figure 8E with correlation values of neuronal activity . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 03110 . 7554/eLife . 26000 . 032Figure 8—source data 2 . Source data for Figure 8F . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 03210 . 7554/eLife . 26000 . 033Figure 8—source data 3 . Source data for Figure 8G . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 03310 . 7554/eLife . 26000 . 034Figure 8—figure supplement 1 . In vivo labelling of a Ser-tr1 cell by photoactivation during calcium imaging . ( A ) Correlation map of neuronal activity of a Ser-tr1 and a Loop neuron , ventral view . ( B ) Calcium signals measured from the Ser-tr1 and Loop neurons shown in ( A ) . ( C ) Serotonin antibody staining ( cyan ) and mCherry1 fluorescence ( orange ) in the Ser-tr1 cell following local photoactivation of PAmCherry1 during calcium imaging ( in the larva shown in A and B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 034 In the ventral nerve cord , we also identified two neurons whose activity was anti-correlated with the activity of the ciliary band . ( Figure 8C–E ) . The GCaMP signal did not reveal the axonal morphology of these neurons , but their cell-body position corresponded to the position of the Ser-tr1 cells . To directly test the identity of these cells , we performed in vivo photoactivation experiments . We co-injected GCaMP mRNA with an mRNA encoding a photoactivatable mCherry fluorescent protein ( PAmCherry1 ) . During calcium imaging , we identified the Ser-tr1 candidate neuron based on its activity and then photoactivated PAmCherry1 in this neuron . The larvae were subsequently fixed and stained for serotonin . In all cases ( n = 11 larvae ) we labelled the serotonergic Ser-tr1 neuron , together with a few out-of-focus cells that were labelled more weakly ( Figure 8—figure supplement 1 ) . To further explore how serotonin influences ciliary activity , we treated larvae with serotonin . Serotonin increased ciliary beat frequency and inhibited ciliary arrests compared to untreated larvae ( Figure 8F , G , Figure 8—source data 2 , Figure 8—source data 3 ) . The three-segmented Platynereis larva has many distinct cell groups with locomotor cilia that efficiently propel the larva in water . Periods of swimming are interrupted by ciliary arrests . Here we investigated how spontaneous ciliary arrests and beating are coordinated across ciliated fields . With whole-body connectomics and calcium imaging , we described the anatomy , connectivity , and activity of the entire ciliomotor system in the Platynereis larva . In trochophore larvae , which do not yet have a functioning musculature , this system represents the entire motor part of the nervous system . One of the most fascinating aspects of this ciliomotor system is its rhythmic pattern of activation consisting of two alternating phases . In one phase , a dopaminergic , a peptidergic , and several cholinergic neurons are active together with the ciliary band cells . In the other phase , anti-correlated with the cholinergic rhythm , one noradrenergic neuron and the serotonergic neurons are active . Our connectome and imaging data indicate the rhythm is generated by the interactions of the ciliomotor neurons themselves . This is supported by the presence of many and often reciprocal synaptic connections between the ciliomotor motorneurons . We cannot exclude the possibility that interneurons connecting distinct ciliomotor neurons are involved in the generation of the rhythmic pattern and the synchronisation of ciliomotor activity . We could identify such interneurons in the connectome ( unpublished results ) . However , we could not observe rhythmic activity for any of these interneurons by calcium imaging . We could subdivide the ciliomotor circuit into three connectivity modules . All three modules contain both serotonergic and cholinergic neurons and have largely non-overlapping ciliary targets ( prototroch , left and right ciliated fields , respectively ) . Module 1 is special in that besides the serotonergic and cholinergic neurons it also contains the three unpaired peptidergic cMN neurons , two of which are also catecholaminergic . These neurons are likely involved in the generation of the rhythmic pattern . Three observations support this . First , together with the MC neuron , the cMN cells are the first neurons that show a rhythmic pattern during development . Second , they contain three cells with distinct transmitter profiles suggesting the possibility of antagonistic reciprocal signalling . Third , they are presynaptic or reciprocally connected with all the other ciliomotor neurons . For example , the cMN neurons are presynaptic to both the MC and the Loop neurons that are themselves only weakly connected , yet activate synchronously . We can also exclude a role for cholinergic neurons in rhythm generation since blocking cholinergic transmission did not eliminate the rhythmic activation of the MC neuron . The precise delineation of the pacemaker neurons will require further cell ablation studies . The function of the rhythmically active ciliomotor circuitry is to trigger the coordinated arrests of all cilia in the body and likely to trigger the coordinated resumption of beating . Targeted neuron ablation revealed a critical role for the cholinergic MC neuron in coordinated arrests of prototroch cilia suggesting that this neuron relays the rhythmic activation of the cMN neurons to trigger arrests . Our imaging and connectome data suggest that the MC neuron together with the two Loop neurons is able to trigger coordinated , body-wide ciliary arrests . These three cells innervate all multiciliated cells in the body ( with the exception of the metatroch ) . Arrests of metatroch cilia may be triggered by the MNant neurons , the only putatively cholinergic neurons that innervate the metatroch . The metatroch is a special ciliary band that in some feeding trochophore larvae beats in opposition to the prototroch to collect food particles ( Rouse , 1999; Pernet and Strathmann , 2011 ) . The distinct pattern of innervation for the metatroch may reflect the unique evolutionary history of this ciliary band . The crescent cell represents another specialised multiciliated cell . This cell shows a beat pattern alternating with the beating of locomotor cilia . Crescent cilia thus beat when the larvae do not swim . The only synaptic contacts to the crescent cell come from the Loop neurons . The crescent cell also shows increases in calcium signals , in synchrony with the ciliary bands . However , crescent cilia respond differently to stimulation than locomotor cilia and start beating instead of arresting . The crescent cell is in the middle of the apical organ , a sensory area . Beating of the crescent cilia may serve to generate water flow to facilitate chemosensory sampling in the apical organ , similar to the role of cilia-generated flows in olfaction in fish ( Reiten et al . , 2017 ) . Parallel and upstream to the cholinergic system , we identified four serotonergic ciliomotor neurons that also collectively innervate all ciliary bands . Serotonergic cells are also presynaptic to the cholinergic cells and serotonin application suppresses closures and increases ciliary beat frequency . Determining the site of serotonin action would require further studies . One possibility is that serotonin influences the pacemaker activity thereby indirectly affecting closures . Serotonin may also act directly on ciliated cells via serotonergic synapses to increase beat frequency . The activation of the serotonergic cells corresponds to periods of swimming and thus these cells are likely responsible for the resumption of beating . The monoamines and neuropeptides we mapped to the ciliomotor neurons may act at synapses or extrasynaptically ( Bentley et al . , 2016 ) . Understanding the details of signalling will require the identification of the receptor expressing cells . We recently characterised several neuropeptide receptors , adrenergic receptors and other monoamine receptors in Platynereis ( Bauknecht and Jékely , 2017 , 2015 ) . We also developed a method to dissect chemical connectivity of neuropeptide signalling with cellular resolution ( Williams et al . , 2017 ) . These approaches will help to dissect the mechanisms of rhythm generation and the modulation of the rhythm and ciliary activity . What is the relevance of coordinated ciliary arrests for larval behaviour ? It was shown that in freely-swimming Platynereis larvae the alternation of periods of upward swimming ( when cilia beat ) and sinking ( when cilia are arrested ) is important for maintaining a constant depth in the water column of the negatively buoyant larvae ( Conzelmann et al . , 2011 ) . It is remarkable that the Platynereis larval nervous system exhibits a rhythmic autonomic activity already in the 1-day-old stage , when the nervous system only consists of <20 differentiated neurons . This ongoing activity is present in all larvae and is not induced by sensory input . Early larval behaviour is thus best described in operant terms ( Brembs , 2017 ) . This autonomic activity can then be modified by sensory input . Several neuromodulators , including neuropeptides and melatonin were shown to either inhibit or induce ciliary arrests , and thereby shift larvae upwards or downwards in the water column ( Conzelmann et al . , 2011 , 2013; Tosches et al . , 2014 ) . Many neuropeptides are expressed in sensory-neurosecretory cells , suggesting that sensory cues can influence larval swimming by neuroendocrine signalling ( Conzelmann et al . , 2013; Tessmar-Raible et al . , 2007 ) . The serotonergic and cholinergic ciliomotor neurons are also postsynaptic to functionally distinct sensory neuron pathways that can trigger ciliary arrests or induce swimming by synaptic signalling ( unpublished results ) . The ciliomotor system we describe here represents a novel pacemaker motor circuit that forms a stop-and-go system for ciliary swimming ( Figure 9 ) . Remarkably , some of the neurons in this circuit ( Loop and Ser-tr1 neurons ) span twice the entire body length to coordinately regulate segmental cilia . Furthermore , many cells have a biaxonal morphology , a feature that is very rare in Platynereis neurons belonging to other circuits ( Randel et al . , 2014; Shahidi et al . , 2015 ) . Comparative morphological studies using histological techniques suggest that similar ciliomotor systems , sometimes with large biaxonal neurons ( Temereva and Wanninger , 2012 ) , are widespread in ciliated larvae . Serotonergic neurons innervating ciliary bands have been described in several groups ( Hay-Schmidt , 2000 ) , including annelid ( Voronezhskaya et al . , 2003 ) , mollusc ( Friedrich et al . , 2002; Kempf et al . , 1997 ) , phoronid ( Temereva and Wanninger , 2012 ) , and bryozoan larvae ( Wanninger et al . , 2005 ) , and likely represent an ancestral character in lophotrochozoans ( Wanninger et al . , 2005 ) . Serotonergic innervation of ciliary bands is also present in deuterostomes , including echinoderm and hemichordate larvae ( Hay-Schmidt , 2000 ) . Likewise , catecholamine-containing processes innervate ciliary bands in mollusc ( Croll et al . , 1997 ) , phoronid ( Hay-Schmidt , 1990b ) , echinoderm , hemichordate ( Dautov and Nezlin , 1992; Nezlin , 2000 ) , and nemertean larvae ( Hay-Schmidt , 1990a ) . Cholinergic nerves also often associate with ciliary bands in echinoderm , hemichordate ( Dautov and Nezlin , 1992; Nezlin , 2000 ) , and mollusc larvae ( Raineri , 1995 ) . 10 . 7554/eLife . 26000 . 035Figure 9 . The stop-and-go ciliomotor system of the Platynereis larva . ( A ) ssTEM reconstruction of ciliomotor neurons that are activated synchronously ( orange ) and asynchronously ( teal ) with the ciliary bands . ( B ) Schematic of the circuitry , consisting of a pacemaker system , as well as cholinergic and serotonergic ciliomotor neurons that span the entire body and innervate all multiciliated cells . Abbreviation: prn , prototroch ring nerve . DOI: http://dx . doi . org/10 . 7554/eLife . 26000 . 035 One intriguing ( although controversial ) scenario is that ciliated larvae and their nervous systems , including the apical sensory organ ( Marlow et al . , 2014 ) and the ciliomotor systems , are homologous across bilaterians or even eumetazoans . If this is the case , this could mean that the Platynereis ciliomotor system represents an ancient system that evolved close to the origin of the nervous system ( Jékely , 2011 ) . The internal coordination of multiciliated surfaces may thus have been one of the early functions of nervous systems ( Jékely et al . , 2015 ) . In-depth comparative studies are needed to understand how larval locomotor systems are related and whether ciliary coordination has a common origin or evolved multiple times independently . Platynereis dumerilii larvae were reared at 18°C in a 16 hr light 8 hr dark cycle until the behavioural experiments . Occasionally animals were kept overnight at 10°C to slow down growth . Larvae were raised to sexual maturity according to established breeding procedures ( Hauenschild and Fischer , 1969 ) . We found that nectochaete larvae showed very similar ciliary activity to trochophore larvae , but younger larvae displayed fewer muscle contractions . For this reason , we often used younger animals for recordings . Experiments were conducted at 22°C , most often between 36 and 52 hr-post-fertilisation . Whole mount in situ hybridization , and confocal imaging of the animals were performed as described previously ( Asadulina et al . , 2012 ) . The custom-made Platynereis allatotropin antibody was described in Shahidi et al . , 2015 . Immunostainings were done as described before ( Conzelmann and Jékely , 2012 ) , with the following modifications . Larvae were fixed with 4% formaldehyde in PTW ( PBS + 0 . 1% Tween-20 ) for 15 min at room temperature and rinsed three times with PTW . Larvae were blocked in PTWST ( PTW , 5% sheep serum , 0 . 1% Triton-X 100 ) at room temperature , incubated overnight in 1:250 rabbit anti-HA antibody ( rabbit , RRID: AB_1549585 ) and 1:250 anti acetylated-tubulin antibody ( mouse , RRID: AB_477585 ) in PTWST . Larvae were washed five times for 20–60 min in PTW , blocked for 1 hr in PTWST at room temperature and incubated for 2 hr at room temperature in anti-mouse Alexa Fluor-488 1:250 ( goat , RRID: AB_2534069 ) and anti-rabbit Alexa Fluor-633 1:250 ( goat , RRID: AB_2535731 ) , in PTWST . Larvae were washed five times for 20–60 min in PTW and transferred gradually to 97% 2 , 2’-thiodiethanol ( TDE ) ( 166782 , Sigma-Aldrich , St . Louis , USA ) in steps of 25% TDE/PTW dilutions . For the whole-mount in situ and immunostaining samples we generated projections using Imaris 8 . 0 . 2 ( Bitplane , Zürich ) . Neuron reconstruction , visualization , and synapse annotation were done in Catmaid ( Schneider-Mizell et al . , 2016; Saalfeld et al . , 2009 ) as described before ( Randel et al . , 2015; Shahidi et al . , 2015 ) . The ciliomotor neurons described here include all neurons that form at least five synapses combined on any of the multiciliated cells in the whole larval body . Six neurite fragments could not be linked to a cell body , these were omitted from the analyses . Neurons were named based on transmitter content or morphology and position , ( e . g . , dorsal ciliomotor , cMNd , ventral left ciliomotor , cMNvl ) . The ciliated troch cells were named following the morphological literature . We numbered prototroch cells based on their position ( e . g . 2 o’clock ) and whether they belong to the anterior or posterior prototroch ring ( e . g . , proto5P is the posterior prototroch cell at 5 o’clock position in the ring ) . We grouped paratroch cells from the same larval segment into four clusters , based on their position ( e . g . , paraIIIvl1 is a paratroch cell in the third segment in the ventral-left cluster ) . Cells are labelled based on body side of their soma ( left [l] or right [r] ) . Network analyses were done in Catmaid ( RRID:SCR_006278 ) and Gephi 0 . 8 . 2 ( RRID:SCR_004293 ) . Modules were detected in Gephi with an algorithm described in ( Blondel et al . , 2008 ) with randomization on , using edge weights and a resolution of 1 . 4 . Force-field-based clustering was performed using the Force Atlas 2 Plugin . Cable length was measured in Catmaid using skeleton smoothing with Gaussian convolution with a sigma of 6 μm . Ciliary beating and ciliary arrests were imaged with a Zeiss Axioimager microscope ( Carl Zeiss , Jena , Germany ) and a DMK 21BF04 camera ( The Imaging Source , Bremen , Germany ) at 60 frames per second . Experiments were done at room temperature in filtered natural seawater . Larvae were immobilized between a slide and a coverslip spaced with adhesive tape . Calcium-imaging experiments were performed with GCaMP6s ( Chen et al . , 2013 ) mRNAs ( 1000 μg/μL ) injected in one-cell stage as described previously ( Randel et al . , 2014 ) . Injected larvae were mounted following the same protocol as ciliary beating experiments . Confocal images were taken on an Olympus Fluoview-1200 ( with a UPLSAPO 60X water-immersion objective , NA 1 . 2 ) , and a Leica TCS SP8 ( with a 40X water-immersion objective , NA 1 . 1 ) confocal microscopes . Movies were acquired with a temporal resolution between 0 . 8 and 45 Hz . Responses to drugs were imaged 5–20 min after application to the bath and without any change of imaging settings . Laser ablation was performed on the Olympus FV1200 confocal microscope equipped with a SIM scanner ( Olympus Corporation , Tokyo , Japan ) , as described in ( Randel et al . , 2014 ) . Larvae were immobilized between a slide and a coverslip . A 351 nm pulsed laser ( Teem Photonics , Grenoble , France ) at 90–95% power was used , coupled via air to the SIM scanner for controlled ablation in a region of interest ( ROI ) . 95% corresponds to a beam power of 297 . 5 µwatts as measured with a microscope slide power sensor ( S170C; Thorlabs , Newton , USA ) . PAmCherry1 ( Addgene plasmid 31928 ) was cloned into a Platynereis expression vector ( pUC57-T7-RPP2- ( NLS ) pPAmCherry1 ) ( Randel et al . , 2014 ) with an NLS-tag before the N-terminus . PAmCherry1 RNA was synthesized in vitro and co-injected with GCaMP6s RNA . Photoactivation in a region of interest was performed with a 405 nm laser ( 0 . 2% power ) with 2–3 200 msec pulses using the SIM scanner of an Olympus FV12000 confocal microscope . Photo-activated larvae were recovered from the slide , fixed and processed for immunostaining with serotonin ( rabbit , RRID: AB_572263 ) and acetylated tubulin ( mouse , RRID: AB_477585 ) antibodies . GCaMP6s movies were analysed with FIJI ( Schindelin et al . , 2012 ) ( RRID:SCR_002285 ) and a custom Python script , as described in ( Gühmann et al . , 2015 ) with modifications ( Verasztó , 2017 ) . The same ROI was used to quantify fluorescence before and after drug application . Data are presented as ΔF/F0 . For the calculation of the normalized ΔF/F0 with a time-dependent baseline , F0 was set as the minimum standard deviation of fluorescence in a 10-frame-window in every experiment . Videos were motion-corrected in FIJI with moco ( Dubbs et al . , 2016 ) and Descriptor based registration ( https://github . com/fiji/Descriptor_based_registration ) . Correlation analyses and fast Fourier transformation to compute the discrete Fourier transform of activation sequences were done using FIJI and Python ( Verasztó , 2017; a copy is archived at https://github . com/elifesciences-publications/Veraszto_et_al_2017 ) . We only analysed segments of videos without movement artefacts . The ROI was defined manually and was correlated with every pixel of the time-series . Finally , a single image was created with the Pearson correlation coefficients and a [−1 , 1] heatmap plot with two colours .	The oceans contain a wide variety of microscopic organisms including bacteria , algae and animal larvae . Many of the microscopic animals that live in water use thousands of beating hair-like projections called cilia instead of muscles to swim around in the water . Understanding how these animals move will aid our understanding of how ocean processes , such as the daily migration of plankton to and from the surface of the water , are regulated . The larvae of a ragworm called Platynereis use cilia to move around . Like other animals , Platynereis has a nervous system containing neurons that form networks to control the body . It is possible that the nervous system is involved in coordinating the activity of the cilia to allow the larvae to manoeuvre in the water , but it was not clear how this could work . Here , Veraszto et al . investigated how Platynereis is able to swim . The experiments show that the larvae can coordinate their cilia so that they all stop beating at the same time and fold into to the body . Then the larvae can stimulate all of their cilia to resume beating . Veraszto et al . used a technique called electron microscopy to study how the nervous system connects to the cilia . This revealed that several giant neurons span the entire length of the larva and connect to cells that bear cilia . When these neurons were active , all the cilia in the body closed . When a different group of neurons in the larva was active , all of the cilia resumed beating . Together , these two groups of neurons were ultimately responsible for the swimming motions of the larvae . Together , the findings of Veraszto et al . show that a few neurons in the nervous system of the larvae provide a sophisticated system for controlling how the larvae swim around . This suggests that the microscopic animals found in marine environments are a lot more sophisticated than previously appreciated . A next challenge is to find out how the neurons that control cilia connect to the rest of the animal’s nervous system and how different cues influence when the larva swims or stops swimming . This would help us understand how the environment influences the distribution of animal larvae in the oceans and how this may change in the future .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2017	Ciliomotor circuitry underlying whole-body coordination of ciliary activity in the Platynereis larva
In Huntington’s disease , a mutated version of the huntingtin protein leads to cell death . Mutant huntingtin is known to aggregate , a process that can be inhibited by the eukaryotic chaperonin TRiC ( TCP1-ring complex ) in vitro and in vivo . A structural understanding of the genesis of aggregates and their modulation by cellular chaperones could facilitate the development of therapies but has been hindered by the heterogeneity of amyloid aggregates . Using cryo-electron microscopy ( cryoEM ) and single particle cryo-electron tomography ( SPT ) we characterize the growth of fibrillar aggregates of mutant huntingtin exon 1 containing an expanded polyglutamine tract with 51 residues ( mhttQ51 ) , and resolve 3-D structures of the chaperonin TRiC interacting with mhttQ51 . We find that TRiC caps mhttQ51 fibril tips via the apical domains of its subunits , and also encapsulates smaller mhtt oligomers within its chamber . These two complementary mechanisms provide a structural description for TRiC’s inhibition of mhttQ51 aggregation in vitro . The presence of huntingtin aggregates in the brain correlates strongly with functional deficits in individuals with Huntington’s disease ( HD ) ( DiFiglia et al . , 1997 ) . These aggregates originate from a mutation in the huntingtin gene that causes an expansion of glutamine ( Q ) repeats in the huntingtin protein ( Zoghbi and Orr , 2009 ) . This polyglutamine expansion ( Q>36 ) contributes to mutant huntingtin’s ( mhtt ) tendency to aggregate within the cell ( Nekooki-Machida et al . , 2009 ) . Variants of mhtt exon1 containing an expanded polyglutamine ( polyQ ) tract greater than 36 residues are commonly used as models to investigate mutant huntingtin’s properties in vitro and in vivo . Understanding how polyQ aggregation is modulated by cellular factors , including molecular chaperones , might lead to therapeutic strategies to treat polyQ pathogenesis . Some chaperones , such as the eukaryotic chaperonin TRiC , are key regulators of protein aggregation ( Muchowski and Wacker , 2005; Liebman and Meredith , 2010; Voisine et al . , 2010 ) . In fact , TRiC modulates and suppresses polyglutamine aggregation , including that of mhtt variants having polyQ tracts of different length . TRiCs inhibition of mhtt aggregation has been biochemically characterized for several variants both in vitro and in vivo ( Kitamura et al . , 2006; Tam et al . , 2006 ) . TRiC is a ∼1 MDa double-ring complex that interacts with ∼10% of all cytosolic proteins and is stringently required for the proper folding of many essential proteins ( Yam et al . , 2008 ) . TRiC can assume different conformations in different biochemical states ( Cong et al . , 2012 ) and can bind substrates like actin and tubulin through different sites , either at its apical tips , which are conformationally flexible in the apo state ( Llorca et al . , 1999 ) , or inside the chamber ( Dekker et al . , 2011; Muñoz et al . , 2011 ) . TRiC’s interactions with polyQ aggregates are also likely to be variable , as seen with other chaperone/substrate complexes ( Kim et al . , 2002 ) . There is no existing structural model for how TRiC suppresses mhtt aggregation , and most structural biology techniques are not suited to characterize conformationally heterogeneous specimens such as mhtt aggregates in complex with TRiC . However , cryo-electron microscopy ( cryoEM ) can preserve specimens in close-to-solution conformations because the samples remain frozen-hydrated without chemical fixation or staining during the entire microscopy session ( Dubochet et al . , 1988 ) . Additionally , cryo-electron tomography ( cryoET ) allows for visualization of entire protein aggregates in 3-D . Furthermore , 3-D subtomogram processing approaches ( Walz et al . , 1997; Böhm et al . , 2000; Bartesaghi et al . , 2008; Förster et al . , 2008; Schmid and Booth , 2008 ) , also known as single particle tomography ( SPT ) , have facilitated the determination of the structure of heterogeneous specimens by classifying and averaging subvolumes containing a 3-D view of individual macromolecular complexes , potentially with different conformations and compositions . Indeed , SPT is an emerging technique in structural biology that can resolve the structure of heterogeneous macromolecular complexes at nanometer resolution . We imaged mhtt exon1 containing 51 glutamine repeats ( mhttQ51 ) ( ‘Materials and methods’ under ‘In vitro GST-Q51 aggregation assay’ ) using 2-D cryoEM ( ‘Materials and methods’ under ‘Cryo electron microscopy’ ) to assess the progression of its aggregation over time and how TRiC affects it ( Figure 1 ) . Amongst the mhtt variants whose susceptibility to TRiC has been previously characterized in vitro and in vivo ( Behrends et al . , 2006; Tam et al . , 2006 ) , our choice of the mhttQ51 variant was guided by the fact that mutations yielding a polyQ tract longer than ∼58Q are rare in patients ( Myers , 2004 ) , and N-terminal fragments of similar length ( Q53 ) have been isolated from brain tissue ( Lunkes et al . , 2002 ) . We do not observe any regular repeat in the images of the aggregates , in contrast to more ordered aggregates such as beta amyloid and alpha synuclein ( Lührs et al . , 2005; Trexler and Rhoades , 2010 ) . The heterogeneity of mhttQ51 fibrils precluded us from resolving their structure in detail since all current structural techniques require some underlying homogeneity in fibrillar structures to achieve high resolution . 10 . 7554/eLife . 00710 . 003Figure 1 . The presence of TRiC inhibits the progression of mhttQ51 aggregation . 2-D cryoEM images of ( A ) mhttQ51 at 45 min , ( B ) 4 hr , and ( C ) 24 hr post-initiation of aggregation ( PIA ) , and of mhttQ51 incubated with TRiC and imaged at ( D ) 45 min ( E ) 4 hr and ( F ) 24 hr PIA . ( G , H ) Higher magnification images of mhttQ51 + TRiC at 4 hr PIA . Some TRiCs are seemingly attached to mhtt fibrils ( red boxes ) , while others are seemingly freestanding ( blue boxes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 003 At 45 min post-initiation of aggregation ( PIA ) in the absence of TRiC , we see a few isolated mhttQ51 fibrils ( Figure 1A ) . Even at early time points , mhttQ51 fibers exhibit extensive heterogeneity , varying both in length and width , with the tips appearing more tapered than the bulk of the fiber . By a process that is not yet fully understood , thin fibrils progress to form thicker fibers , and eventually form branched-out aggregates or ‘sheaves’ with a dense central core by 4 hr PIA ( Figure 1B ) . At longer incubation times , we observe a dramatic increase in the number and thickness of such sheaves , characterized as large fiber bundles densely packed in their central regions and splayed out at both ends . The number of sheaves and their thickness appears to plateau at 24 hr PIA ( Figure 1C ) . On the other hand , in the presence of TRiC , not even thin mhttQ51 fibrils are visible at early time points ( 45 min PIA; Figure 1D ) . At 4 hr and 24 hr PIA ( Figure 1E , F ) , sparse fibrils and a few small sheaves are present , but there are no aggregates as large as those seen in the absence of TRiC at the same time points ( Figure 1B , C ) . TRiC does not completely decorate mhttQ51 fibrils , but rather binds at discrete locations , which are often discernible fibril tips ( Figure 1G–H ) . In a 2-D cryoEM image , the entire volume of the actual specimen is projected onto a plane; therefore , some TRiCs seemingly close to a fibril in 2-D might be floating above or below it and not bound at all . Since 2-D cryoEM images cannot unambiguously show whether TRiC is directly bound to mhttQ5 , we performed cryoET experiments . We collected and reconstructed 20 tilt series of mhttQ51 incubated with TRiC at 4 hr PIA ( ‘Materials and methods’ under ‘Cryo electron tomography and tomogram annotation’ ) . We also reconstructed tilt series of TRiC in the absence of mhttQ51 as a control . Figure 2A is a slice extracted from one of our mhtt + TRiC tomograms , low-pass filtered for visualization . Figure 2B is a re-projection of the entire tomogram ( Video 1 ) and Figure 2C a manual 3-D annotation of its features , which shows an aggregate of mhttQ51 fibrils in yellow , freestanding TRiC particles in blue , and putatively fibril-bound TRiCs in magenta . As in our 2-D images , our 3-D tomograms also show that TRiC does not fully decorate mhtt fibrils , but rather seems to bind at discrete locations , which often are discernible fibril tips . Our visual estimates from these 3-D tomograms indicate that 37% ( 120 of 326 ) of fibril-bound TRiCs locate to obvious and clearly discernible tips . Conversely , about half of the distinguishable fibril tips seem clearly capped by TRiC , but many tips ( capped or uncapped ) might be obscured by the dense tangles , or may exist as short stubs budding from the side of thicker fibers , which might not be discernible . In fact , given the low signal to noise ratio ( S/N ) of individual tomograms , their inherently anisotropic resolution , and ‘missing wedge’ distortions present in all cryoET data , it is challenging to visually discern all TRiC molecules and mhtt fibril tips that might be present in our crowded fibrillar tangles . In addition , some fibril-bound TRiCs might be bound to fibril tips barely ‘budding out’ from the sides of fibril bundles , too short to be discerned . Therefore , the number of TRiCs that locate to tips is likely to be underestimated when counting only those that are on visually discernible tips . 10 . 7554/eLife . 00710 . 004Figure 2 . TRiC caps the tips of mhttQ51 fibrils . ( A ) 2-D slice through a tomogram of TRiC incubated with mhttQ51 imaged at 4 hr post initiation of aggregation . ( B ) 2-D reprojection of the entire tomogram and ( C ) 3-D annotation of its features , showing mhtt fibrils in yellow , freestanding TRiCs in blue , and fibril-bound TRiCs in magenta . Zoomed-in field shows fibril-bound TRiC localizing mostly to discernible tips of mhtt fibrils ( seemingly detached magenta TRiC is attached to a fibril on another z-slice of the tomogram ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 00410 . 7554/eLife . 00710 . 005Video 1 . Annotation of 3-D tomogram of mhtt aggregate in the presence of TRiC . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 005 In both our 2-D images ( Figure 1G–H ) and 3-D tomograms ( Figure 2 , Video 1 ) , we identified two distinct populations of particles similar in size and shape to TRiC . Particles in one population seem to be attached to mhttQ51 fibrils ( ‘fibril-bound TRiC’ ) , while those in the other do not ( ‘freestanding TRiC’ ) . To circumvent some of the difficulties presented by the structural heterogeneity of our specimens and to improve the resolvability of features , we extracted subtomograms from the fibril-bound and freestanding TRiC populations and processed them separately to obtain 3-D averages . We developed new algorithms recently available in EMAN2 ( http://blake . bcm . edu/emanwiki/SPT/Spt ) and used them for all our SPT processing . Given TRiC’s aggressive inhibitory effect on mhtt aggregation and the tendency of mhtt aggregates to localize to the carbon instead of the ice in cryoEM grid holes , finding sizable fibrillar aggregates in complex with TRiC proved challenging . We processed a total of 326 fibril-bound TRiC subvolumes extracted from our tomograms using a completely bias-free subvolume alignment and averaging strategy ( ‘Materials and methods’ under ‘Single particle tomography [SPT]’ ) . This represented about one quarter of 1216 TRiC subvolumes extracted from all our mhtt-TRiC tomograms combined . Figure 3 shows examples of fibril-bound TRiC averages using subvolumes from separate tomograms . Although the heterogeneity of our fibril-bound TRiC sets precluded them from converging to a single structure and thwarted the determination of specific binding sites between TRiC and mhttQ51 fibrils , all our datasets yielded consistent barrel-shaped averages clearly resembling TRiC , with extra mass at the ends of the barrel . These extra densities are attributable to conformational averages of heterogeneous mhttQ51 fibrils . A few averages suggest a mode in which the fibrils might bind predominantly to the tip of a single CCT subunit , while others show mass splayed across the ring , making additional contacts . 10 . 7554/eLife . 00710 . 006Figure 3 . TRiC’s apical tips bind mhtt-fibrils . SPT symmetry-free and model-free averages of fibril-bound TRiC reveal extra density at the chaperonin’s apical tips . Each panel shows an end-on and a side view isosurface for a 3-D average of TRiC in complex with mhttQ51 fibrils , and corresponding 2-D projections ( black/white ) . Independent sets from single tomograms are labeled A , B , C , D and E . Within each , the numbering ( i , ii and iii ) indicates mutually exclusive averages . The numbers of subvolumes contributing to each average are ( A . i . ) 7 , ( A . ii . ) 10 , ( A . iii . ) 13 , ( B . i . ) 7 , ( B . ii . ) 10 , ( B . iii . ) 5 , ( C ) 12 , ( D ) 8 and ( E ) 12 . All the averages show overall TRiC-like morphology with extra densities protruding beyond the apical domains of the chaperonin . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 006 We also sought to understand whether , and how , freestanding TRiC might interact with smaller mhttQ51 oligomers . We extracted a total of 890 freestanding TRiC subvolumes from our tomograms and performed SPT ( i . e . , subvolume alignment , classification , and averaging ) . These were about three quarters of the 1216 TRiC subvolumes extracted from all our mhtt-TRiC tomograms combined . When processing the subvolumes from each tomogram separately , SPT averages were consistent with the results from the combined data set , indicating a high level of self-consistency in the results . Our statistical analyses ( ‘Materials and methods’ under ‘Criteria to discriminate between cavity-empty and cavity-occupied freestanding TRIC’ , Figure 4 ) suggest that , while the freestanding TRiC subvolumes exhibit a high but continuous level of variability ( Figure 4A ) , they can be divided into two statistically separable populations ( Figure 4B ) . These populations appear to correspond to cavity-empty and cavity-occupied TRiC . 10 . 7554/eLife . 00710 . 007Figure 4 . Freestanding TRiCs are classifiable into cavity-empty and cavity-occupied . ( A ) Ranked correlation scores for freestanding TRiCs against a hollow TRiC reference yield a continuous fast-decaying plot , suggesting structural heterogeneity . ( B ) Mean-density distributions for cavity-empty TRiC ( blue ) and cavity-occupied TRiC ( green ) sets , masking the particles to ( i ) include all the density in each particle , ( ii ) include the cavity of each molecule only , and ( iii ) include TRiC density but exclude the cavity . Corresponding t-scores for the difference ( or shift ) between the blue and green distributions are shown for each masking condition . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 007 The final cavity-empty TRiC average from the combined dataset highly resembles apo-TRiC and was produced using the same bias-free approach applied to fibril-bound TRiC ( ‘Materials and methods’ under ‘Single particle tomography [SPT]’ ) . The average is pseudo eightfold symmetrical , barrel-shaped , and hollow ( Figure 5A ) , and accounts for almost one third of the entire 1216 TRiC subvolumes extracted from all of our mhtt-TRiC tomograms combined . 10 . 7554/eLife . 00710 . 008Figure 5 . A subpopulation of freestanding TRiC contains intra-cavity extra density . SPT symmetry-free and template-free processing of freestanding TRiC demonstrates pseudo-eightfold symmetry in an average of ( A ) 356 cavity-empty TRiCs and ( B ) 356 cavity-occupied TRiCs . 2-D projections for each average are shown in black/white . The plots show volumetric rotational correlation analyses to test for C8 ( blue ) and D8 ( green ) symmetry in each average . The cavity size in B is significantly reduced , due to internalized density . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 008 Applying the same bias-free approach to the cavity-occupied TRiC set ( also about one third of the entire 1216 TRiC subvolumes ) produced an average with significant extra density lining the interior of the chaperonin’s cavity , attributable to a poorly localized substrate ( Figure 5B ) . In order to validate our results for freestanding TRiC in the presence of mhtt , we performed control experiments using TRiC in the absence of mhtt ( Figure 6 ) . When this data was processed identically , we could not identify any subset with significant extra density inside the cavity . This strongly suggests that the large encapsulated density seen in our mhtt-TRiC data derives from the incubation of TRiC with mhtt , rather than residual substrate remaining from the TRiC purification process . While we do not have an experimental assay at this time that can conclusively identify the internal density as mhtt , all other answers seem unlikely . 10 . 7554/eLife . 00710 . 009Figure 6 . Control TRiC in the absence of mhtt . End-on and cut-away views for ( A ) the published cryoEM structure of apo-TRiC ( EMDB 1960 ) , low-pass filtered to match the resolution of freestanding TRiCs , and ( B ) model-free SPT TRiC average of 208 subvolumes from TRiC-only tomograms or ‘control TRiC’ ( no mhtt added ) . Corresponding projections for the control are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 009 Computing the Fourier shell correlation ( FSC ) between the cavity-empty and cavity-occupied TRiC averages gives a value of ∼52 Å using the FSC = 0 . 333 threshold . As these maps are not expected to be identical , this test gives a measure of self-consistency and provides the minimum resolution achieved in the two maps independently . This threshold is appropriate when the two maps are being compared , but not averaged , as a 0 . 5 threshold is appropriate when a map is being compared to a high-resolution reference ( Rosenthal and Henderson , 2003 ) . Comparing the maps to the published apo-TRiC structure ( Cong et al . , 2012 ) using the FSC = 0 . 5 cutoff ( ‘Materials and methods’ under ‘Assessment of the quality of bias-free freestanding TRiC averages’; Figure 7 ) , yields similar values , ∼48 Å and ∼46 Å for cavity-occupied and cavity-empty TRiC respectively . 10 . 7554/eLife . 00710 . 010Figure 7 . FSC curves for freestanding TRiC averages . FSC of cavity-empty TRiC ( maroon ) and cavity-occupied TRiC ( blue ) with the published structure of apo-TRiC . The FSC of cavity-empty TRiC with cavity-occupied TRiC is shown in green . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 010 To further localize the interior density in the cavity-occupied TRiC data set , we adopted a hollow-template-guided alignment strategy ( ‘Materials and methods’ under ‘Localization of density within cavity-occupied TRiC SPT average’ ) . This additional analysis results in a TRiC average with a large density asymmetrically located within the cavity ( Figure 8A ) . The interior mass is clearly seen in a difference map between the cavity-empty and the cavity-occupied averages ( Figure 8B ) . Of note , the template was completely hollow , and therefore model bias cannot account for the extra density we see inside . The TRiC part of this cavity-occupied average is also pseudo eightfold symmetric ( Figure 8A ) and the dip in the rotational correlation plot ( or ‘u-shaped’ path of the overall curves ) can be explained by the presence of internal mass , localized to one side of the cavity when viewed end-on ( ‘Materials and methods’ under ‘Localization of density within cavity-occupied TRiC SPT average’ , ‘Estimation of the size of mhttQ51 oligomers encapsulated by TRiC’ ) . 10 . 7554/eLife . 00710 . 011Figure 8 . Localized intra-cavity extra density within TRiC . ( A ) Hollow-template-guided average of 356 cavity-occupied TRiCs , 2-D projections ( black/white ) and rotational correlation analysis to test for C8 ( blue ) and D8 ( green ) symmetry , showing a ‘dip’ in self-correlation at azimuth equal to 180° due to the presence of localized inner mass . ( B ) End-on ( left ) and cut-away side ( right ) views of the superimposition of the cavity-empty TRiC average ( blue wire ) from Figure 5A , and the difference map ( yellow ) between it and the cavity-occupied average in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 011 Our study of mhtt fibrillogenesis in the absence and presence of TRiC provides new insights into how this chaperonin inhibits mhttQ51 aggregation . Our aggregation and suppression of aggregation reactions followed well characterized conditions , previously described using filter trap assays that detect mhttQ51 amyloid aggregates ( Tam et al . , 2006 ) . The mhtt concentration in our experiments was chosen to allow mhtt aggregation to occur in a reasonably short time in the absence of TRiC ( Scherzinger et al . , 1999 ) , and the ratio of TRiC to mhtt was chosen to yield initial-stage fibril aggregates and yet thwart large-aggregate growth ( Behrends et al . , 2006 ) . It is worth noting however that the TRiC:mhtt ratio under normal physiological conditions has not been carefully explored . Nonetheless , increasing cellular levels of TRiC has a beneficial effect in suppressing mhtt toxicity and aggregation ( Behrends et al . , 2006; Kitamura et al . , 2006; Tam et al . , 2006 ) . In our 2-D cryoEM images , the formation of groups of sheaves as seen in Figure 1C could correspond to the large inclusions observed in cells expressing mhtt . On the other hand , the smaller aggregates ( Figures 1E , F and 3 ) , which are not detectable by filter-trap assays , could correspond to the multiple , smaller foci seen in cells overexpressing TRiC or some of its subunits ( Behrends et al . , 2006; Tam et al . , 2006 ) . Importantly , overexpression of TRiC is known to suppress mhtt toxicity in a variety of systems ( Behrends et al . , 2006; Kitamura et al . , 2006; Tam et al . , 2006 ) and thus our observations provide a structural basis for the observed effect of TRiC . We envision the progression of mhttQ51 aggregation as a process whereby , after the nucleation stage , small and thin irregular fibrils with exposed hydrophobic moieties ( ‘fibril tips’ ) drive the growth of the aggregates into large sheaves . TRiC is known to bind hydrophobic motifs via its apical substrate-binding domains ( Spiess et al . , 2006 ) , and has also been shown to crosslink with the N-terminal N17 motif of mhtt , which drives mhtt oligomerization ( Tam et al . , 2009 ) . It is therefore likely that the binding we demonstrate here between mhttQ51 fibrils and TRiC ( Figures 1–3 ) , which thwarts the progression of fibers into sheaves , may involve mhttQ51 hydrophobic moieties that include the N17 domain . Interestingly , this domain has been proposed to function as a ‘molecular switch’ ( Greiner and Yang , 2011 ) controlled by multiple post-translational modifications that can determine the fate of mhtt exon-1 either by rendering it innocuous or enhancing its pathogenicity . By virtue of its interaction with the N17 domain , TRiC might provide a protective effect against deleterious post-translational modifications . Although we do not observe a unique structure for fibril-bound TRiC , we demonstrate that TRiC binds mhtt fibrils via its apical tips to cap them . This fibril-capping mechanism could explain why inhibition of aggregation is observed in vitro at substoichiometric ratios of TRiC to mhtt . It may also explain why mhtt aggregates look different in vivo when TRiC is overexpressed , and become SDS soluble ( Tam et al . , 2006 ) . It is conceivable that TRiC might bind fibrils through one of its ends while the binding sites at the other end remain available to snatch mhttQ51 oligomers lurking in their proximity . On the other hand , internalization of smaller mhttQ51 oligomers would also contribute to inhibition of sheaf formation by decreasing the effective concentration of oligomers in the reaction mixture . Our freestanding TRiC averages , free from template and symmetry bias ( Figure 5 ) , reveal a subpopulation with extra density inside the chaperonin’s cavity . Although other substrates have also been shown to bind deep inside the cavity of apo-TRiC ( Muñoz et al . , 2011 ) , the finding that mhtt exon-1 might also be encapsulated is unprecedented . While it is conceivable for native substrates to be copurified with apo-TRiC , their presence within the chaperonin’s cavity is rare . For instance , our controls show that extra density is only detectable in ∼10% of the particles in the absence of added mhtt . On the other hand , we see significant extra density in ∼40–50% of the freestanding TRiC particles from our TRiC + mhtt tomograms . Furthermore , previous studies have demonstrated that TRiC does not bind GST ( Feldman et al . , 2003 ) . Therefore , it is unlikely that either native substrates or the GST moiety that is cleaved from mhttQ51 to initiate aggregation could account for the internalized mass we observe ( Figures 5B and 8 ) . Rather , this density likely represents mhtt . Given the modest resolution of our cavity-occupied average , it is challenging to measure the encapsulated mass . However , our estimates suggest an average size of ∼60 kDa ( ‘Materials and methods’ under ‘Estimation of the size of mhttQ51 oligomers encapsulated by TRiC’ ) , which would correspond to an oligomer of about six mhttQ51 exon1 monomers . The mhttQ51 oligomers that might be encapsulated by TRiC could be analogous to precursor forms of the 100–230 kDa toxic oligomers previously reported ( Behrends et al . , 2006 ) . Furthermore , it is possible that the extra density we see inside TRiC ( Figures 5B and 8 ) corresponds to an average of encapsulated mhtt oligomers of different sizes and conformations . While previous biochemical studies showed the inhibitory effect of TRiC on mhttQ51 aggregation , the structural basis of this process was not explained . The compositional and conformational heterogeneity of such a specimen precludes resolving nanometer resolution structures from it with any technique in structural biology other than SPT . Indeed , our approach demonstrates TRiC’s association with mhttQ51 fibrillar aggregates , and our results suggest TRiC might also sequester small mhttQ51 oligomers to effectively inhibit mhtt aggregation ( Figure 9 , Video 1 ) . This novel mechanism contrasts with those typically proposed for the action of chaperones in suppressing aggregation of misfolded proteins ( Hendrick and Hartl , 1995; Young et al . , 2004 ) by binding monomeric forms . However , since our analysis lacks the sensitivity to distinguish TRiCs that might be bound to one mhtt monomer only , we cannot rule out that such a binding mode may co-exist with the ones presented here . 10 . 7554/eLife . 00710 . 012Figure 9 . Model for the progression of mhttQ51 aggregation in the absence and presence of TRiC . In the absence of TRiC ( top row ) , mhttQ51 ( orange ) aggregation progresses from monomers and oligomers into fibrils and large sheaves . On the other hand , the presence of TRiC ( bottom row ) prevents the progression of mhtt fibrils into the large bundled ‘sheaf’ form by capping ( purple ) fibril tips and encapsulating ( blue ) mhtt oligomers . TRiCs that are not bound to mhtt are shown in gray . DOI: http://dx . doi . org/10 . 7554/eLife . 00710 . 012 While our observation that TRiC can bind substrates via its apical domains is consistent with the current understanding of TRiC–substrate interactions , our data is unprecedented in demonstrating the direct binding of fibrils ( a substrate too large to be internalized ) . The suggestion that TRiC might cap fibril growth as well as internalize smaller oligomers indicates a broad scope for this chaperonin’s action on protein aggregates . The ability of TRiC to recognize and handle both types of proposed mhtt toxic species observed in Huntington’s disease ( DiFiglia et al . , 1997; Arrasate and Finkbeiner , 2011; Nucifora et al . , 2012 ) , namely oligomers and fibrillar aggregates , could account for the reduced toxicity of mhtQ51 and other mhtt species when TRiC is overexpressed in cells ( Behrends et al . , 2006; Tam et al . , 2006 ) . TRiC’s interaction with mhtt aggregates might render them more amenable to processing by cellular pathways of quality control , such as autophagy and proteolytic degradation by the proteasome ( Sarkar et al . , 2009 ) . Since mhtt oligomers and fibrillar aggregates are strongly implicated in the pathogenesis of Huntington’s disease ( Sathasivam et al . , 2010 ) , the interactions that we describe between TRiC and mhttQ51 provide a structural mechanism for the chaperonin’s aggressive inhibition of mhtt aggregation in vitro and suggest a potential for TRiC-inspired therapeutics ( Sontag et al . , 2013 ) . Aggregation of GST-huntingtin exon 1 fusion proteins with 51 glutamine repeats ( GST-mhttQ51 ) at a concentration of 6 µM was initiated in vitro by adding AcTEV protease ( Invitrogen ) , both alone and in presence of 1 . 2 µM TRiC at a t = 0 hr ( Tam et al . , 2006 ) in TEV Buffer ( Invitrogen , Grand Island , NY ) . Samples were subjected to spin-column equilibration using Bio-Spin Columns with Bio-Gel P-30 ( Bio-Rad ) . Samples were incubated at 30°C and aliquots vitrified at t= 0 hr , 0 . 75 hr , 4 hr , and 24 hr . All aliquots were applied to 200-mesh holey carbon Quantifoil copper grids ( rinsed in PBS , Biological Industries ) and plunge-frozen into liquid ethane at liquid nitrogen temperature ( unfixed , unstained ) , as described ( Dubochet et al . , 1988 ) . 2-D images were collected using the standard minimum dose system on a 200KV JEOL electron microscope ( JEM 2100 , LaB6 gun ) , with a sampling of 2 . 86 Å/pixel , a target underfocus of 4 μm and a total dose of ∼20 e/Å2 . We collected 20 tilt series of mhttQ51 + TRiC at t = 4 hr using SerialEM ( Mastronarde , 2005 ) on a JEM2100 electron microscope equipped with a LaB6 gun operating at 200 kV , from −60° to 60° , at 2° increments , target underfocus of 5 μm , sampling of 4 . 4 Å/pixel and a cumulative dose of ∼62 e/Å2 . 6 of the 20 tilt series appeared appropriate for tomographic reconstruction and were reconstructed into tomograms using IMOD ( Kremer et al . , 1996 ) . The 14 unused tilt series either displayed charging , thick ice , or suboptimal sample distribution . MhttQ51 fibrils were annotated in yellow , while fibril-bound TRiC were annotated in magenta and freestanding TRiC in blue using AVIZO 6 . 1 software ( Visage Imaging GmbH ) ( Figure 2C ) . All SPT processing and averaging was done with subtomograms extracted from the original raw tomograms ( unbinned ) using e2spt_boxer . py in EMAN2 ( Tang et al . , 2007 ) . To avoid missing wedge bias , we used cross-correlation map normalization ( Schmid and Booth , 2008 ) . Our bias-free subtomogram processing uses unsupervised hierarchical ascendant classification and averaging ( HAC ) , a technique involving pairwise alignment of sub-tomograms and clustering based on similarity ( Bartesaghi et al . , 2008; Förster et al . , 2008; Schmid and Booth , 2008 ) , without the aid of a reference and without imposing symmetry at any point . For fibril-bound TRiC , we visually inspected the new averages generated in each iteration to find representative examples of structures with extra density at the ends of the chaperonin ( Figure 3 ) . For each freestanding TRiC population ( cavity-empty and cavity-occupied ) as well as control TRiCs , we allowed HAC to converge to one final average of all the particles ( Figure 5A , B ) . We noticed significant but poorly localized extra density inside the chamber of the cavity-occupied population . We tested our hypothesis that some freestanding TRiCs might be empty while others might contain mhttQ51 oligomers by aligning the particles against a hollow D8 symmetrical apo-TRiC reference ( Booth et al . , 2008 ) and ranking them according to their correlation with it . Cavity-empty TRiCs would have a higher correlation to the hollow reference than cavity-occupied ones . Indeed , our classification method separated freestanding TRiCs into two groups , and our statistical analyses show that the difference in mass within the cavity between the groups is significant ( Figure 4B ) . When processing the subvolumes in each tomogram , we divided the set into two halves: TRiCs with highest correlation against the reference or ‘cavity-empty group’ , and TRiCs with lowest correlation against the reference or ‘cavity-occupied group’ . Indeed , the average of the first group was a hollow apo-TRiC-like structure , while the average from the second showed a large but poorly localized mass within the chaperonin’s cavity . The larger set from grouping the freestanding TRiC subtomograms from all our tomograms ( 890 subtomograms ) allowed for a more scrupulous classification , although there is no clear ‘cutoff’ in the plot of ranked correlation coefficients from aligning the particles against the D8 TRiC reference ( Figure 4A ) . To discriminate between cavity-empty and cavity-occupied TRiC , we divided the ranked subtomograms into quintiles ( subsets of 178 ) and averaged the particles in each quintile . The average from each of the two highest quintiles had an empty cavity , while the cavity for each average from the two lowest quintiles contained a large density . The average from the middle quintile was neither clearly empty nor significantly occupied; therefore , we did not subject those subtomograms to further analysis . We made a final cavity-empty set from the two top quintiles and a final cavity-occupied set from the two bottom quintiles , each with 356 subtomograms . That is , each freestanding set ( cavity-empty and cavity-occupied ) was comprised of 40% of all freestanding TRiCs . In the context of all TRiC subvolumes extracted , each of these sets represented about one third of all particles picked ( i . e . , 356 of 1216 ) . We tested the statistical significance of our classification by performing t-tests on histograms ( Figure 4B ) that compared the mean density distribution of the cavity-empty group against that of the cavity-occupied group under three separate masking conditions , to examine the density variations across different regions of the subvolumes . The first mask included all the density in each subvolume ( mask A ) , the second mask included only the density in each particle’s cavity ( mask B ) , and the third mask was the difference between the first two masks , thus corresponding to TRiC’s density in each subvolume ( mask C ) . For each masking condition , we plotted the mean density distribution of each set ( cavity-empty and cavity-occupied ) as separate histograms in one plot . The displacement between the cavity-empty and the cavity-occupied histograms using mask A is statistically significant beyond a 99 . 9% confidence interval ( CI ) , according to its t-value . This means that the cavity-empty particles are significantly different from the cavity-occupied particles when the entire density in the subvolumes is considered . The displacement between the histograms using mask B is also statistically significant beyond a 99 . 9% confidence interval ( CI ) , and the difference is much larger according to the t-value . This means that the difference between the particles in each group is very large when only the cavity is considered . Lastly , there was no significant displacement between the histograms using mask C , as confirmed by its low t-value , far below the 95% CI value . This means that when the cavity is excluded , the particles in each group are not significantly different . Therefore , our analyses show that the statistically significant differences between the cavity-empty and cavity-occupied TRiC sets are due to the central chamber of the chaperonin rather than the chaperonin itself . This validated our classification and justified further processing the cavity-empty group independently from the cavity-occupied one . TRiC has eight highly similar subunits; therefore , if the densities from our tomograms picked as TRiC were in fact TRiC and our methodology averaged them correctly , our maps would display pseudo-eightfold symmetry . Indeed , our rotational correlation plots on the template-free and symmetry-free TRiC averages reveal the presence of pseudo-eightfold symmetry in them ( Figure 5 ) . We assessed the resolution of the cavity-empty structure as 46 Å by computing an FSC with the apo-TRiC map ( Cong et al . , 2012 ) using the FSC = 0 . 5 threshold , as is appropriate when comparing a map to a much higher resolution structure , as formulated previously ( Rosenthal and Henderson , 2003 ) ( Figure 7 ) . Because TRiC’s eight subunits differ the most in their apical tips and their C-termini within the cavity , both containing substrate-binding sites , it is unlikely that any internalized oligomers would bind to all subunits uniformly or line up TRiC’s cavity ( Figure 5B ) . Therefore , we applied a hollow-template-guided approach to further localize the inner mass within the chamber of the cavity-occupied TRiC average . This consisted of aligning the raw cavity-occupied TRiCs to a hollow D8 symmetric apo-TRiC reference ( Booth et al . , 2008 ) to generate an initial average . We then seeded iterative refinement of the raw subtomograms using this initial average . During all the iterations , the particles were masked to include only the cavity and were allowed only to visit D8 related orientations . Since the first round of alignment was against the D8 reference , restricting subsequent iterations to visit D8-related positions preserved the initial alignment of the particles against the reference while allowing the internal density to converge within the frame of TRiC after 17 iterations . While this procedure localized the internalized density in our cavity-occupied average , it had little effect on the cavity of the cavity-empty set . Of note , the ‘dip’ in the overall shape of the rotational correlation plot for the cavity-occupied average with the localized density can be explained by the presence of a large mass inside ( Figure 8A ) . Applying the same classification and processing strategy to fibril-bound and control TRiCs as applied to freestanding TRiC did not yield any averages with comparable density inside the cavity . The normalized cavity-occupied ( Figure 8A ) and cavity-empty ( Figure 5A ) TRiC averages were aligned to one another , and the power spectrum of one structure was filtered to match that of the other . Then , the cavity-empty average was subtracted from the cavity-occupied average to generate a difference map ( Figure 8B ) . We estimated the mass of the mhttQ51 density internalized by TRiC using three different methods . First , we applied a mask to our final cavity-empty average ( Figure 5A ) that included the TRiC density only , thus excluding the cavity . We calculated the mean density inside this mask and multiplied this value by the volume of the mask . This yielded the final total mass of TRiC ( without the cavity ) in arbitrary units . Then we applied a mask to the difference map between the final cavity-empty average ( Figure 5A ) and the final cavity-occupied average ( Figure 8A ) that would include the region of the cavity only . We also computed the total mass within this mask , in arbitrary units . Using the actual mass of TRiC ( ∼950 kDa ) to scale the masses in arbitrary units , we estimated the mass of the density in the cavity to be ∼55 kDa . For our second method of estimation , we first averaged the density values of the histograms in Figure 4B ( iii ) . These histograms correspond to the density distribution of TRiC particles when the cavity is excluded , and therefore the average value in arbitrary units is proportional to TRiC’s mass . On the other hand , the histograms in Figure 4B ( ii ) correspond to the density distribution within the cavity of TRiC particles . The average density of the histogram from the cavity-empty group ( blue ) was considered to be ‘background’ , in arbitrary units . Therefore , we subtracted this value from the average of the cavity-occupied histogram ( green ) . This final value for average density in the cavity region in arbitrary units should be proportional to the average mass of encapsulated oligomers . Knowing the actual mass of TRiC ( ∼950 kDa ) allowed for the scaling of masses in arbitrary units , yielding a value of ∼65 kDa for the mass within the cavity . Taken together , our two estimates indicate an average size of ∼60 kDa for mhttQ51 oligomers encapsulated by TRiC . Finally , in the rotational correlation plots for the cavity-empty average ( Figure 5A ) and the cavity-occupied average before the substrate was localized ( Figure 5B ) , the seven peaks in correlation ( excluding the first peak ) have a somewhat uniform height . On the other hand , the plot dips at azimuth equal to 180° for the cavity-occupied average ( Figure 8A ) . The first peak is bound to be higher in any rotational correlation plot because self-correlation will always be at its maximum when no rotations have occurred ( all voxels in the volume correlate well; even those beyond the molecule but within the box ) . The difference in height ( or correlation ) between the next two highest peaks ( two and eight ) with the middle peak ( five ) in the rotational correlation plot of the cavity-occupied average arises from the mass of the encapsulated density: when the mass coincides perfectly with its copy , there is high correlation; when it is in opposite alignment to itself , the correlation dips . Therefore , using the mass of TRiC ( ∼950 kDa ) to scale this difference indicates a substrate size of ∼50 kDA . This is slightly underestimated , since the difference would ideally be calculated between the first and the middle peaks , but the first peak is unusable due to the perfect correlation of all voxels at the default position before any rotations occur . The Electron Microscopy Data Bank accession numbers for the TRiC-mhttQ51 structures reported in this paper are as follows: For the cavity-empty TRiC average as shown in Figure 5A , the EMDB accession number is EMD-5531 . For the cavity-occupied TRiC average as shown in Figure 8A , the EMDB accession number is EMD-5530 . For two of the fibril-bound TRiC averages shown in Figure 3A , the EMDB accession numbers are EMD-5532 and EMD-5533 .	Huntington’s disease is an inheritable neurodegenerative disorder that typically begins in mid-adulthood . It initially affects muscle coordination and progresses to include psychiatric symptoms and cognitive decline , leading to premature death . The disease is caused by a mutation in the huntingtin gene , which codes for the huntingtin protein , and all individuals who inherit a pathogenic form of the mutant gene will eventually develop the condition . The huntingtin gene contains a series of repeats of the tri-nucleotide sequence CAG , which encodes for the amino acid glutamine . The number of repeats varies between individuals but if it exceeds 36 , the huntingtin protein starts to form aggregates in the brain . Aggregation occurs when soluble protein precursors , known as oligomers , combine to form structures called fibrils , which in turn assemble into larger clusters . This phenomenon also occurs in several other tri-nucleotide diseases , each of which involves a mutated gene with an excess of tri-nucleotide repeats . Inside cells , proteins called chaperones regulate the folding of other proteins and help to prevent aggregate formation . A chaperone protein known as TRiC , which interacts with approximately 10% of proteins in the cytosol , has been shown to inhibit the aggregation of mutant huntingtin proteins . However , it has not been possible to map the structural interactions between TRiC and huntingtin to date . Now , Shahmoradian and Galaz-Montoya et al . have used cryo-electron tomography , combined with 3-D mapping and computer-aided reconstruction , to reveal the structure of a molecular complex consisting of TRiC and a pathogenic mutant huntingtin protein containing 51 CAG repeats . By imaging this system at different time points during the aggregation of mutant huntingtin , it was possible to characterize how the aggregates changed over time . They found that their shape differs in the presence and absence of TRiC , and that the chaperone interacts both with soluble huntingtin molecules—sequestering them so that they cannot join together—and with the tips of fibrils , preventing them from growing longer . By providing the first direct demonstration of how TRiC inhibits the aggregation of mutant huntingtin , the results of Shahmoradian and Galaz-Montoya et al . could aid in the design of TRiC-based drugs to be used in the treatment of Huntington’s disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2013	TRiC’s tricks inhibit huntingtin aggregation
Bruton's tyrosine kinase ( Btk ) , a Tec-family tyrosine kinase , is essential for B-cell function . We present crystallographic and biochemical analyses of Btk , which together reveal molecular details of its autoinhibition and activation . Autoinhibited Btk adopts a compact conformation like that of inactive c-Src and c-Abl . A lipid-binding PH-TH module , unique to Tec kinases , acts in conjunction with the SH2 and SH3 domains to stabilize the inactive conformation . In addition to the expected activation of Btk by membranes containing phosphatidylinositol triphosphate ( PIP3 ) , we found that inositol hexakisphosphate ( IP6 ) , a soluble signaling molecule found in both animal and plant cells , also activates Btk . This activation is a consequence of a transient PH-TH dimerization induced by IP6 , which promotes transphosphorylation of the kinase domains . Sequence comparisons with other Tec-family kinases suggest that activation by IP6 is unique to Btk . Activation of B- and T-lymphocytes relies on a chain of tyrosine phosphorylation events initiated by B-cell and T-cell receptors , respectively ( Cambier et al . , 1993; Weiss and Littman , 1994; Kurosaki , 2011 ) . These phosphorylation-dependent signals are generated by three kinds of non-receptor tyrosine kinases , linked to the receptor by membrane co-localization ( Dal Porto et al . , 2004; Kurosaki , 2011 ) . Src-family kinases ( principally Lyn and Fyn in B-cells , Lck and Fyn in T-cells ) , resident at the plasma membrane , respond first to receptor activation . They phosphorylate pairs of tyrosine residues in ITAM motifs in the cytoplasmic tails of activated receptors ( Chu et al . , 1998; Mócsai et al . , 2010; Wang et al . , 2010 ) , which in turn recruit the tandem-SH2 domain tyrosine kinases Syk ( in B-cells ) and ZAP-70 ( in T-cells ) . Membrane localization of Syk or ZAP-70 leads to the recruitment and activation of Tec-family tyrosine kinases , Btk ( in B-cells ) and Itk ( in T-cells ) ( Andreotti et al . , 2010 ) . Btk is present in all hematopoietic cells , except for T-cells and natural killer cells ( Tsukada et al . , 1993; Lindvall et al . , 2005 ) . It is also part of several other receptor-mediated signaling networks , including those initiated by Toll-like receptors ( Jefferies et al . , 2003 ) , Fc receptors ( Kawakami et al . , 1994 ) , and G-protein coupled receptors ( Tsukada et al . , 1994; Bence et al . , 1997 ) . Btk is critical for the generation of calcium flux in response to receptor activation , because it activates phospholipase C-γ2 ( PLC-γ2 ) , which produces the second messengers diaceylglycerol ( DAG ) and inositol ( 1 , 4 , 5 ) -trisphosphate ( IP3 ) . Btk dysfunction is the cause of X-linked agammaglobulinemia , a severe disease of primary immunodeficiency ( Tsukada et al . , 1993 ) . The Tec family kinases have a lipid-interaction module attached to the N terminus of a Src-like module ( i . e . , concatenation of an SH3 domain , an SH2 domain and a tyrosine kinase domain ) ( Figure 1A ) . Like the non-receptor tyrosine kinase c-Abl , the Tec kinases lack a C-terminal tyrosine phosphorylation site that in Src kinases engages the SH2 domain in an autoinhibitory interaction ( Sicheri et al . , 1997; Xu et al . , 1997; Nagar et al . , 2003 ) . The lipid-interaction module of the Tec kinases is an extended plekstrin homology ( PH ) domain fused to a Tec homology ( TH ) domain ( Hyvönen and Saraste , 1997 ) . The PH-TH binds PIP3 in membranes , or the soluble head group of PIP3 , inositol tetrakisphosphate ( IP4 ) , in solution . A flexible linker connects the PH-TH module to the Src-like module; its length varies from 26 residues in Itk , the Tec family kinase in T-cells , to 45 residues in Btk . 10 . 7554/eLife . 06074 . 003Figure 1 . Crystal structure of the Src-like module of Btk . ( A ) Domain architectures of Btk , c-Abl , and c-Src . ( B ) Model for the Src-like module of Btk , based on the crystal structure of the domain-swapped dimer . ( C ) Comparison of the Src-like modules of Btk and c-Abl ( PDB: 1OPK ) ( Nagar et al . , 2003 ) , superimposed on the C lobes of the kinase domains . There is a relative rotation of about 15° for the SH2 domains and 20° for the SH3 domains for the two structures . ( D ) Details of the SH3/SH2-kinase linker interface and the kinase catalytic cleft in Btk . Left panel: a cluster of hydrophobic residues on the SH3 domain packs against the polyproline type-II helix formed by the SH2-kinase linker . This type of interaction is seen in most SH3-peptide ligand complexes . Right panel: the activation loop of Btk folds into the mouth of the catalytic cleft , blocking part of the ATP-binding cleft . Glu 445 , in helix αC , forms an ion pair with Lys 430 in the active conformation but is prevented from doing so in this inactive conformation by the activation loop and helix αC . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 00310 . 7554/eLife . 06074 . 017Figure 1—source data 1 . Structures of fragments of Tec family kinases in the Protein Data Bank . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 01710 . 7554/eLife . 06074 . 004Figure 1—figure supplement 1 . Structural details of the Src-like module of Btk . ( A ) The Src-like module of Btk forms a domain-swapped dimer in the crystal lattice , with one molecule per asymmetric unit . The SH3 domain and the kinase domain are intact , but the SH2 domain forms a domain-swapped dimer with another molecule . ( B ) Molecular details of the domain-swapped SH2 dimer in Btk and Grb2 ( Schiering et al . , 2000 ) . The SH2 domain opens up at a position within the β-sheet , so that helix αB of one polypeptide chain packs against the β-sheet of the other . ( C ) Interactions between the SH2-kinase linker and the kinase domain . Leu 390 of the linker stabilizes the inactive conformation of the kinase domain by being sandwiched between residues Trp 421 and Tyr 461 in the kinase domain . ( D ) The ‘electrostatic switch’ in the Btk kinase domain . The active conformation of the Btk kinase domain is modeled from the crystal structure of Lck ( PDB: 3LCK ) ( Yamaguchi and Hendrickson , 1996 ) using Modeller ( Eswar et al . , 2006 ) . Activation entails a shuffling of salt-bridge and polar hydrogen-bond partners . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 004 Tyrosine phosphorylation controls Btk activity . B-cell activation leads to rapid phosphorylation of Tyr 551 in the kinase domain and Tyr 223 in the SH3 domain of Btk ( Park et al . , 1996; Rawlings et al . , 1996 ) . Tyr 551 corresponds to the canonical phosphorylation site in the activation loop of tyrosine kinases; as expected , its phosphorylation increases Btk catalytic activity ( Rawlings et al . , 1996 ) . Phosphorylation of Tyr 223 has been reported to have no apparent influence on Btk catalytic activity ( Park et al . , 1996; Middendorp et al . , 2003 ) . It lies within the peptide-binding groove of the SH3 domain , and its phosphorylation alters SH3 domain selectivity among binding partners ( Morrogh et al . , 1999; Middendorp et al . , 2003 ) . Binding of a Tec-kinase PH domain to the membrane lipid PIP3 , produced as a consequence of receptor activation , provides a simple control mechanism for switching on the kinase . Membrane recruitment triggers trans-autophosphorylation , through increased local concentration , or trans-phosphorylation by membrane-bound Src-family kinases , thereby activating the Tec kinase ( Rawlings et al . , 1996 ) . We present here an analysis of two crystal structures of Btk , which jointly provide molecular details of its autoinhibition . One structure is of the Src-like module , in an assembled and inactive conformation , resembling those of inactive c-Abl and of the Src kinases c-Src and Hck ( Sicheri et al . , 1997; Xu et al . , 1997; Nagar et al . , 2003 ) . The other is of a Btk construct that lacks the SH3 and SH2 modules and that links the PH-TH module directly to the kinase domain . We find that the PH-TH module stabilizes the inactive conformation of the kinase by interacting with the N lobe . The inactive conformation of Btk is thus a compact one , in which the PH-TH module cooperates with the SH2 and SH3 domains to maintain an inactive structure . Several structures of isolated domains of Btk have been determined previously , as summarized in Figure 1—source data 1 , which also lists structures of fragments of other Tec family kinases . We also report a set of findings that reveal an unanticipated role for the PH-TH module in activating the kinase domain . We find that inositol hexakisphosphate ( IP6 ) , a soluble signaling molecule ubiquitous in animal and plant cells ( Szwergold et al . , 1987; Vallejo et al . , 1987 ) , can activate Btk strongly . We have determined a crystal structure of the PH-TH module of Btk bound to IP6 and identified a previously undetected binding site on the PH-TH module , distinct from the canonical PIP3/IP4 site . The new site is critical for activation by IP6 . We show that Btk activation by IP6 is a consequence of a transient , IP6-induced PH-TH dimerization , which promotes transphosphorylation of the kinase domains . The possibility that PH dimerization might be important for Btk activation was first proposed by Saraste and colleagues ( Hyvönen and Saraste , 1997 ) , and we now show that the dimer interface identified previously is essential for IP6-dependent Btk activation . IP6-triggered transient dimerization of the PH domain may provide a membrane-independent mechanism for activating Btk in the nucleus or cytoplasm or may have a role in feedback stimulation of Btk that has been activated by PIP3 at the membrane . The prior model for Btk activation involved only a PIP3-dependent membrane recruitment as the key step in switching on the kinase . The ability of IP6 to activate Btk points to an as yet unappreciated complexity to the activation mechanism . We have determined four crystal structures as part of this work . Two of these structures help explain how Btk is autoinhibited . These are the structures of the Btk SH3-SH2-kinase module ( determined at 2 . 6 Å resolution ) and of the PH-TH-kinase unit ( determined at 1 . 7 Å resolution ) . The other two structures are of the isolated PH-TH module bound to IP6 ( determined at 2 . 3 Å resolution ) and the isolated kinase domain with mutations in the activation loop ( determined at 1 . 6 Å ) . The details of the crystallographic experiments are given in Tables 1 , 2 . 10 . 7554/eLife . 06074 . 005Table 1 . Data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 005Src-like module of mouse Btk ( 217-659 ) PH-TH-kinase unit of bovine BtkBtk PH-TH module bound to IP6Btk kinase domain with mutations in the activation loopPDB ID4XI24Y934Y944Y95Data collectionKAu ( CN ) 2NativeNativeNative Wavelength ( Å ) 0 . 94741 . 0001 . 0001 . 000 Space groupP 31 2 1P 2 21 21P 1P 1 a , b , c ( Å ) 132 . 2 , 132 . 2 , 107 . 678 . 6 , 38 . 3 , 157 . 637 . 2 , 64 . 0 , 80 . 050 . 9 , 79 . 0 , 79 . 2 α , β , γ ( ° ) 90 . 0 , 90 . 0 , 120 . 090 . 0 , 90 . 0 , 90 . 082 . 0 , 88 . 8 , 89 . 890 . 7 , 89 . 9 , 90 . 0 Resolution ( Å ) 43 . 2–2 . 650–1 . 743 . 9–2 . 347 . 9–1 . 6 Rsym ( % ) 8 . 7 ( >100 ) 12 . 2 ( 79 . 7 ) 6 . 0 ( 53 . 8 ) 5 . 1 ( 71 . 7 ) I/σ ( I ) 10 . 6 ( 1 . 1 ) 17 . 0 ( 3 . 5 ) 10 . 8 ( 1 . 2 ) 13 . 2 ( 3 . 2 ) Completeness ( % ) 91 . 9 ( 69 . 6 ) 99 . 2 ( 98 . 5 ) 94 . 9 ( 71 . 8 ) 96 . 9 ( 94 . 7 ) Redundancy4 . 8 ( 3 . 5 ) 7 . 8 ( 7 . 4 ) 2 . 0 ( 1 . 7 ) 24 . 4 ( 3 . 1 ) Wilson B factor72 . 815 . 752 . 318 . 1Refinement Resolution43 . 2–2 . 650–1 . 746 . 5–2 . 347 . 9–1 . 6 Reflections57 , 643 ( 30 , 933 ) 53 , 35730 , 516157 , 562 Rfree reflections1525200020252009 Rwork/Rfree0 . 237/0 . 2520 . 167/0 . 1950 . 236/0 . 2540 . 155/0 . 187 No . atoms Protein3459697050538911 Ligands280216961 Water052539737 Average B factors Protein114 . 427 . 762 . 122 . 5 SolventN/A33 . 066 . 230 . 1 Root mean square deviation from ideality Bonds ( Å ) 0 . 0060 . 0050 . 0030 . 014 Angles ( ° ) 1 . 110 . 9770 . 7641 . 571 Ramachandran statistics Favored ( % ) 9198 . 5898 . 297 . 11 Disallowed ( % ) 2 . 40 . 00 . 00 . 18 MolProbity clash score9 . 62 . 85 . 162 . 66The CC1/2 values for the PH-TH-kinase dataset , IP6-bound PH-TH dataset and the kinase domain with mutations in the activation loop dataset are 99 . 9 ( 86 . 5 ) , 99 . 9 ( 55 . 4 ) and 99 . 9 ( 90 . 7 ) , respectively . 10 . 7554/eLife . 06074 . 006Table 2 . Data statistics for the Src-like module of BtkDOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 006Src-like module of mouse Btk ( 217-659 ) Data collectionKAu ( CN ) 2DMAAu2 ( NO3 ) 3native Wavelength ( Å ) 0 . 94740 . 94740 . 94740 . 9474 Space groupP31 2 1P31 2 1P31 2 1P31 2 1 a , b , c ( Å ) 132 . 2 , 132 . 2 , 107 . 6132 . 5 , 132 . 5 , 107 . 3131 . 9 , 131 . 9 , 107 . 6131 . 8 , 131 . 8 , 107 . 0 α , β , γ ( ° ) 90 . 0 , 90 . 0 , 120 . 090 . 0 , 90 . 0 , 120 . 090 . 0 , 90 . 0 , 120 . 090 . 0 , 90 . 0 , 120 . 0 Resolution ( Å ) 43 . 2–2 . 641 . 7–3 . 541 . 7–3 . 440–4 . 0 Rsym ( % ) 8 . 7 ( >100 ) 8 . 7 ( 32 . 7 ) 7 . 5 ( 32 . 5 ) 5 . 1 ( 71 . 7 ) I/σ ( I ) 10 . 6 ( 1 . 1 ) 9 . 7 ( 3 . 0 ) 12 . 1 ( 3 . 6 ) 18 . 4 ( 12 . 3 ) Completeness ( % ) 91 . 9 ( 69 . 6 ) 92 . 4 ( 94 . 2 ) 94 . 5 ( 95 . 9 ) 92 . 3 ( 100 ) Redundancy4 . 8 ( 3 . 5 ) 4 . 34 . 1N/A Wilson B factor72 . 858 . 870 . 867 . 3 We determined the structure of a construct of mouse Btk spanning residues 214 to 659 , which starts just before the SH3 domain and ends at the C terminus of full-length Btk . The structure resembles those described previously for autoinhibited , Src-like , non-receptor tyrosine kinases , with the SH2-kinase linker sandwiched between the SH3 domain and the small domain of the kinase ( the N lobe of the kinase ) ( Figure 1B ) . The crystal structure has two molecules of Btk in the asymmetric unit , intertwined as a domain-swapped dimer in which each SH2 domain has pieces from two molecules ( Figure 1—figure supplement 1A ) . The domain-swapped SH2 domain packs against the large domain of the kinase ( the C lobe of the kinase ) , but Btk lacks a Src-like C-terminal tail to stabilize this interaction . The sequence identity between Btk and c-Abl is 42% , 31% , and 46% in the SH3 , SH2 , and kinase domains , respectively; for Btk and c-Src , it is 44% , 29% , and 40% . Superposition of the C lobes of the Btk and c-Abl kinase domains gives an approximately 20° relative rotation of their SH3 domains; superposition of the SH3 domains , a relative rotation of about 15° for their SH2 domains . Similar results are obtained for a comparison to c-Src . Coupling between the SH3 and SH2 domains may therefore be somewhat different in these kinases ( Figure 1C ) . The structure of the SH3 domain of Btk is nearly identical to those determined previously using NMR ( the r . m . s . deviation for 53 Cα atoms is 0 . 9 Å ) ( Hansson et al . , 1998; Tzeng et al . , 2000 ) , except for the expected differences seen at the N and C termini . When compared with the other autoinhibited SH3-SH2-kinase structures ( c-Src , Hck , c-Abl ) , the conserved features of the SH3/linker interaction in Btk are the polyproline type-II helix conformation adopted by residues 383–387 in the SH2-kinase linker , including the proline at position 385 , and the overall hydrophobic character of the peptide-binding pocket on the SH3 domain ( Figure 1D ) . The tryptophan sidechain in the center of this pocket donates a hydrogen bond to a backbone carbonyl of the linker , as in most other SH3-ligand complexes . A cluster of hydrophobic sidechains extends from the binding pocket and the SH3 ‘RT loop’ , across the SH2-kinase linker segment , to the β2-β3 loop and strand β4 of the kinase ( Figure 1D ) . Binding of exogenous protein ligands to the Btk SH3 domain would disassemble the Src-like inactive conformation , as seen for the Src family kinase Hck ( Moarefi et al . , 1997 ) . Diffuse electron density for the SH2 domain at every stage of our structure determination shows that this part of the polypeptide chain occupies a range of positions and adopts multiple local conformations . The domain itself has a marginally stable internal structure , as shown by the domain swap , and it is probably imperfectly locked in place . The SH2 βA/αA loop lies between kinase-domain helices αB and αF . Density features for this loop are relatively sharp , and the rest of the domain may pivot somewhat about its contact with the C lobe . The domain swap occurs in the same location within the SH2 domain as a domain swap seen in a Grb2 SH2-domain crystal structure ( Schiering et al . , 2000 ) . The SH2 domain opens up at a position within the subsidiary β-sheet , so that helix B of one polypeptide chain packs against the main β-sheet of the other and vice versa ( Figure 1—figure supplement 1B ) . This exchange occurs in solution for Btk , as only the dimeric species crystallizes . The similar domain swap seen in two SH2 domains not closely related in their specificity or functions suggest that other SH2 domains may also have the same bipartite character . There is no evidence that a domain-swapped dimer ( or , indeed , any stable dimer ) is a functional state of intact Btk , and so it is likely that the assembled state involves a single polypeptide chain , rather than two inter-folded chains . The relatively high concentration of purified protein could have promoted dimer formation in our preparations . In the case of Grb2 , the structure of the intact protein , a monomer in solution , shows no domain swap , and loss of contacts between the deleted N- and C-terminal SH3 domains can account for destabilization of the central SH2 when it is prepared as an isolated domain ( Maignan et al . , 1995 ) . For Btk , the PH-TH module might stabilize the Src-like module . The unphosphorylated activation loop of Btk folds into the mouth of the catalytic cleft , blocking part of the substrate-binding site , and helix αC lies at the margin of the N lobe , preventing formation of a Glu-Lys salt bridge conserved in all catalytically active protein kinases ( Taylor and Kornev , 2011 ) ( Figure 1D ) . The activation loop of Btk is in a conformation seen in the structure of a Btk kinase domain-inhibitor complex ( PDB: 3OCS; [Di Paolo et al . , 2011] ) , and in autoinhibited c-Src ( the r . m . s . deviation from c-Src is 2 . 3 Å for the 24 α-carbons of the activation loop ) , but the cleft between the two lobes of the kinase is substantially more open in Btk than in c-Src . Helix αC is also displaced from the catalytic cleft , as in inactive c-Src or cyclin-dependent kinases . The segment of the activation loop that contacts the N lobe shifts with it , with respect to c-Src as a reference , by up to 3 Å , away from the C lobe . A network of polar interactions present in the autoinhibited forms of c-Src and Hck is also present in the Btk conformation in our crystals ( Figure 1—figure supplement 1C ) . Glu 445 , which would buttress Lys 430 in the active conformation , interacts instead with Arg 544 and Tyr 545 . Arg 544 stacks in turn on Arg 520 , clamping the latter against the phenyl ring of Phe 574 . Each of these arginines also appears to have an amino–aromatic interaction: Arg 544 with Phe 517 and Arg 520 with Phe 574 . In the active conformation of Btk , the same two arginines are expected ( based on the structure of the active Lck kinase domain , PDB: 3LCK; [Yamaguchi and Hendrickson , 1996] ) to engage in salt bridges with a phosphorylated Tyr 551 . Activation thus entails a shuffling of salt-bridge and polar hydrogen-bond partners , described in the case of Src kinases as an ‘electrostatic switch’ ( Ozkirimli and Post , 2006; Banavali and Roux , 2009 ) ( Figure 1—figure supplement 1C ) . In addition , the inactive conformation is stabilized by interactions made by the SH2-kinase linker . Leu 390 of the linker ( corresponding to Trp 260 in c-Src ) is sandwiched between residues Trp 421 and Tyr 461 in the kinase domain ( Figure 1—figure supplement 1D ) . The hydrophobic packing between Trp 421 and Leu 390 would not be possible in the active conformation of the kinase domain . The structure just described demonstrates that the SH3-SH2 domain in a Tec-family member can assemble into a regulatory clamp , just as it does in other Src-like kinases . In the crystal structures of c-Src , Hck , and c-Abl , a latch is present—either in the form of a phosphorylated C-terminal tail ( c-Src , Hck ) or a myristoyl group bound in a pocket on the C lobe of the kinase domain ( c-Abl ) . In both cases , the latch secures the SH2-domain/C lobe contact , precisely the interface that appears to be poorly formed in our Btk crystals . As we show below , the latch in Btk is provided by the PH-TH module . We expressed a ‘PH-TH-kinase’ construct of bovine Btk that contains the PH-TH module connected to the kinase domain by the 13-residue SH2-kinase linker . We replaced six residues in the activation loop of Btk with the corresponding ones in Itk , shown previously to stabilize the activation loop of the Btk kinase domain ( Joseph et al . , 2013 ) . This improved the poorly ordered crystals obtained with the unmutated form , leading to a structure of the PH-TH-kinase with a bound ATP-competitive inhibitor ( CGI1746 , [Di Paolo et al . , 2011] ) at 1 . 7 Å resolution ( Figure 2A ) . 10 . 7554/eLife . 06074 . 007Figure 2 . Crystal structure of the PH-TH-kinase construct of Btk . ( A ) Crystal structure of the PH-TH-kinase construct . The PH-TH module is connected to the kinase domain via a 13-residue linker from the SH2-kinase linker of Btk . The kinase domain is bound to the inhibitor CGI1746 ( Di Paolo et al . , 2011 ) . The canonical PIP3/IP4 binding pocket of the PH domain is located distal to the interface with the kinase , and the peripheral binding site for IP6 is indicated . ( B ) The interface between the PH-TH module and the kinase domain of Btk . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 00710 . 7554/eLife . 06074 . 008Figure 2—figure supplement 1 . Structure of the kinase domain of Btk and its interaction with the PH domain . ( A ) Crystal structure of the isolated Btk kinase domain with mutations in the activation loop . These mutations ( L542M , S543T , V555T , R562K , S564A , and P565S ) are based on a previous study ( Joseph et al . , 2013 ) , and they improved the quality of the crystals of the PH-TH-kinase construct . The structure of the mutant kinase is very similar to that of the wild-type Btk kinase domain ( PDB: 3OCS ) . ( B ) Comparison of the Btk PH-TH-kinase structure ( left ) and the Akt1 PH-kinase structure ( right ) . In Akt1 , The PH domain is docked between the N lobe and the C lobe of the kinase domain , and it blocks substrate access to the catalytic cleft . The lipid-binding site on the PH domain is buried at the PH-kinase interface , and binding of PIP3/IP4 breaks the PH-kinase interface and activates Akt1 ( Wu et al . , 2010; Calleja et al . , 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 008 The overall conformation of the kinase domain , as well as the specific conformation of the activation loop , is very similar in the two structures we have determined ( Figure 2—figure supplement 1A ) . We also determined the structure of the isolated Btk kinase domain with the six mutations described above and found by comparing it to wild-type Btk that the mutations do not introduce major perturbations in the structure ( Figure 2—figure supplement 1A ) . Details of expression , purification , crystallization , and structure determination are in the ‘Materials and methods’ . The PH-TH module contacts the kinase N lobe . The interface differs substantially from the one found in Akt1 , another PH-domain regulated kinase ( Wu et al . , 2010 ) ( Figure 2—figure supplement 1B ) . In Akt1 , the PH domain sits in front of the catalytic cleft of the kinase , blocking peptide substrate binding . In Btk , the PH-kinase interface does not directly occlude the active site . The interface combines a set of hydrophobic contacts with a hydrogen-bond network ( Figure 2B ) . In particular , Tyr 134 at the C terminus of helix α2 of the PH-TH packs into a groove between Trp 395 at the linker-kinase junction and helix αC in the kinase N lobe . We expect these interactions to stabilize the inactive conformation of the kinase , by hindering the shift in helix αC that accompanies kinase activation . Experiments confirming these expectations are described below , in the section on ‘Autoinhibition’ . We combined the two structures just described to build a model for full-length Btk ( Figure 3A ) . Aligning the kinase domains in the two structures gives some local overlap between helix α2 in the PH-TH module and the β3/β4 loop in the SH3 domain , but otherwise compatible positions for the regulatory domains ( Figure 3—figure supplement 1A ) . We therefore carried out molecular dynamics simulations of the two component structures and of the composite assembly , to determine whether energetically minor adjustments could resolve the remaining clashes . The trajectories ranged in length from 60 to 150 ns ( see ‘Materials and methods’ for computational details ) . 10 . 7554/eLife . 06074 . 009Figure 3 . A structural model for full-length Btk . ( A ) A composite model for full-length Btk . The PH-TH domain sits on top of the kinase domain and the SH3 domain , serving as a ‘latch’ that presumably stabilizes the Src-like module in the autoinhibited conformation . The 45-residue linker between the PH-TH module and the SH3 domain is represented by the dotted brown line . Two orthogonal views of an instantaneous structure at 360 ns from a molecular dynamics trajectory for the composite model are shown . ( B ) Inter-domain interactions in an instantaneous structure at 360 ns from a simulation of the composite model . The SH2/kinase interface and the PH-TH /SH3 interface are formed and remain stable during the molecular dynamics simulation . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 00910 . 7554/eLife . 06074 . 010Figure 3—source data 1 . Structures from several time points in the simulation of the composite model for full-length Btk . The structures are taken every 20 ns from a 360 ns molecule dynamic simulation of the composite model . See ‘Materials and methods’ for the details of simulations . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 01010 . 7554/eLife . 06074 . 011Figure 3—figure supplement 1 . Utilization of molecular dynamics in constructing a model for full-length Btk . ( A ) Overlay of the crystal structures of the Btk Src-like module and the PH-TH-kinase construct , using the kinase domain C lobe as the reference . Sidechain clashes ( circled ) are observed between helix α2 in the PH-TH module and the β3/β4 loop segment in the SH3 domain . ( B ) Fluctuations in the Btk Src-like module occurred during a 100 ns molecular dynamics simulation . An instantaneous structure ( t = 89 ns ) from the simulation is overlaid on the crystal structure of the Src-like module . ( C ) Fluctuations in the Btk PH-TH-kinase construct during a 100 ns molecular dynamics simulation . An instantaneous structure ( t = 12 ns ) from the simulation is overlaid on the crystal structure of the PH-TH-kinase construct using the kinase domain N lobe as the reference . The PH-TH module pivots ( black arrow ) about an apparent anchor point at the C terminus of helix α2 . ( D ) Steps used to generate a composite model for full-length Btk . An instantaneous structure ( t = 89 ns ) of the Src-like module and an instantaneous structure ( t = 12 ns ) of the PH-TH-kinase construct from the molecular dynamics trajectory are overlaid , using the kinase domain C-lobe as the reference . In contrast to a corresponding overlay of the crystal structures , there are no sidechain clashes between helix α2 in the PH-TH module and the loop in the SH3 domain . A model for full-length Btk is obtained by combining the PH-TH module with the Src-like module in the two instantaneous structures , and is subject to further molecular dynamics simulation . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 011 In simulations of the SH3-SH2-kinase module , the initial and final conformations of the SH3 and kinase domains had r . m . s . deviations in Cα positions of only ∼3 . 0 and ∼2 . 0 Å , respectively , after first aligning the C lobe of the kinase domain ( Figure 3—figure supplement 1B ) . The SH2 domain shifted by ∼6 Å , with a roughly 10° rotation that brought it closer to the C lobe of the kinase domain and closer to its position in the inactive forms of c-Src , Hck , and c-Abl . Some adjustment of the SH2 domain in these simulations was expected , as the domain swap in the crystals probably perturbed the domain to some extent , and the starting structure for the simulations was therefore itself an approximation . In simulations of the PH-TH-kinase construct , the interface with the kinase domain remained intact , but the PH-TH module wobbled back and forth through a rotation of about 20° ( Figure 3—figure supplement 1C ) . The averaged r . m . s . deviation between the crystal structure and structures from the end of the trajectories , aligned on the kinase domain , was ∼4 Å for the PH-TH module . The PH-TH module pivots about an apparent anchor point at the C terminus of helix α2 , at the center of the autoinhibitory contact ( Figure 3—figure supplement 1C ) . The fluctuations in the PH-TH orientation generate transient structures in which there would be essentially no clashes between the SH3 domain and the PH-TH module in the merged structures . We initiated multiple MD trajectories from such a composite model ( Figure 3A ) . The inhibitory interactions between each of the regulatory modules and the kinase domain were preserved , but various intra-domain rearrangements occurred during the course of these trajectories . The overall conformation of Btk at the end of the trajectories is very similar to the initial structure . The averaged r . m . s . deviations of the PH-TH module and the SH3 domain between the crystal structures and the end of trajectories , aligned on the Src-like module of the kinase , were ∼4 Å and ∼2 . 5 Å , respectively . The former is comparable to the value found in simulations of the PH-TH-kinase , but the latter is larger than in simulations of the Src-like module , probably because the SH3 domain reorients with respect to the kinase domain . The SH3 rotation results in a polar interface between SH3 and PH-TH , with good electrostatic complementarity , which preserves the interactions between the SH3 domain , the linker and the N lobe of the kinase ( Figure 4B ) . In these simulations , the r . m . s . deviations of the SH2 domain and the kinase N lobe are ∼2 . 6 Å and 2 . 0 Å , respectively , after 360 ns , similar to those of the Src-like module simulations . We emphasize , however , that restrictions in the time scale of the simulations allow only qualitative conclusions and do not give a high-resolution picture of the intact molecule . Structures from several time points in the simulation of the composite model are deposited on the eLife website ( Figure 3—source data 1 ) . 10 . 7554/eLife . 06074 . 012Figure 4 . Autoinhibition of Btk . ( A ) Activation of full-length bovine Btk ( residues 1 to 659 , 2 μM ) . Reactions are carried out in the presence of 10 mM Mg2+ , 150 mM NaCl , 1 mM ATP , 25 mM Tris-HCl pH 8 . 0 . The level of autophosphorylation is assayed by immunoblotting an SDS-PAGE gel with a non-specific , anti-phosphotyrosine antibody ( 4G10 , EMD Millipore ) ( upper panel ) . The amount of total protein loaded on the gel is measured by coomassie-blue staining . The kinase activity of Btk is assayed by a continuous kinase-coupled colorimetric assay , in the presence of 1 mM PLC-γ2 peptide substrate . See methods for detailed experimental procedures . ( B ) Comparison of the activation of the Btk Src-like module ( residues 217 to 659 ) , SH2-kinase ( residues 270 to 659 ) , and the kinase domain ( residues 394 to 659 ) . The SH2-kinase construct activates substantially faster than full-length Btk and the Src-like module of Btk . Activated full-length Btk degrades to a small extent over time , which results in some lower molecule-weight bands being detected on the western blot . ( C ) Activation of full-length Btk with mutations Y223A and Y268A . Tyr 223 and Tyr 268 are on the SH3/SH2-linker interface , and the two mutants activate faster than wild-type Btk . ( D ) Activation of full-length Btk with a double mutation ( R134E/Y133E ) . Arg 134 and Tyr 133 are located at the PH-TH/kinase interface . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 012 We probed the significance of the interactions seen in the crystal structures by studying how various mutations affect the rate of autophosphorylation and enzymatic activity . We avoided the heterogenous phosphorylation that accompanies expression in eukaryotic cells by using bovine Btk expressed in bacteria , which gives a pure , unphosphorylated product with good yield . Bovine Btk is 98 . 8% identical to human Btk in sequence , with only eight amino-acid differences over the entire protein . The mass of the bacterially expressed full-length Btk ( 76 , 379 Da ) , as determined by mass spectrometry , is consistent with the calculated molecular weight ( 76 , 381 . 2 Da ) . Incubation with ATP-Mg2+ initiates autophosphorylation , leading in turn to increased catalytic activity . We monitored activation in two ways . First , we monitored the phosphorylation by Btk of a peptide substrate derived from PLC-γ2 , using a continuous kinase-coupled colorimetric assay ( Figure 4A ) . Second , we followed accumulation of tyrosine-phosphorylated Btk by immunoblotting with a non-specific , anti-phosphotyrosine antibody ( 4G10 , EMD Millipore ) ( Figure 4A ) . The results of the two assays are in good agreement . The kinase domain of Btk has low catalytic activity and autophosphorylates very slowly , like the c-Abl kinase domain and unlike those of the Src family ( Figure 4B ) . For example , there is no detectable change over 90 min in the level of phosphorylation of the Btk kinase domain at 4 μM concentration . Inclusion of the SH2 domain and the SH2-kinase linker ( but not the SH3 domain or the PH-TH module ) increases Btk autophosphorylation substantially ( Figure 4B ) . We have not studied how the SH2 domain increases activity , but we note that in c-Abl , the SH2 domain docks onto the N lobe of the kinase domain and stabilizes the active conformation and that activation by the SH2 domain is also seen in Csk ( Sondhi and Cole , 1999 ) and c-Fes ( Nagar et al . , 2006; Filippakopoulos et al . , 2008 ) . The apparent linear array of domains detected by small-angle x-ray scattering from a partially phosphorylated form of full-length Btk might represent the activated rather than the inactive form , in a conformation similar to that of activated c-Abl ( Márquez et al . , 2003; Nagar et al . , 2006 ) . The autoactivation rate of the complete Src-like module of Btk is lower than that of the SH2-kinase module ( Figure 4B ) , as expected from the joint clamping effect of the SH3 and SH2 domains . Based on contacts seen in the crystal structure of the Src-like module of Btk , we introduced ( separately ) two mutations , Y223A and Y268A , into full-length Btk . These SH3-domain residues pack against Pro 385 in the SH2-kinase linker , and their phosphorylation ( or mutation to alanine ) would destabilize the autoinhibited conformation ( Figure 1C ) . The autoactivation rates of both mutants are indeed about sixfold greater than that of the unmutated enzyme and comparable to that of the SH2-kinase module ( the rate of activation is estimated from the slope of activity vs time , and is only a rough measure because of non-linearity; Figure 4C ) . An analogous interaction regulates c-Abl activation: phosphorylation of Tyr 89 , in the SH3-kinase interface , is necessary for full activation ( Brasher and Van Etten , 2000 ) . Mutations expected to disrupt the PH-TH-kinase interface likewise increase the autoactivation rate of full-length Btk , consistent with our inferences from examining the interface between the PH-TH and the kinase N lobe ( Figure 4D ) . Removal of the entire PH-TH module , while retaining the Src-like regulatory components , increases the rate of autoactivation about threefold , as estimated from the slopes of the activity curves in Figure 4 . We studied the effect on Btk activity of PIP3 in small unilamellar vesicles ( SUVs ) . SUVs containing 75% DOPC , 20% DOPS , and 5% PIP3 produce a rapid and substantial activation of Btk ( Figure 5A ) . At a Btk concentration of 2 μM , robust phosphorylation can be detected within the first 5 min of reaction in the presence of these vesicles . The extent of activation depends on the concentration of PIP3 in the vesicles , presumably because increasing the PIP3 fraction increases the surface density of Btk on the lipid vesicles and thereby increases the rate of trans-autophosphorylation ( Figure 5A ) . An estimate of the surface density of PIP3 on these vesicles suggests that the local concentration of Btk on the vesicle surface is in the millimolar range , that is , 100-fold greater than in bulk solution ( see ‘Materials and methods’ for detailed calculations ) . 10 . 7554/eLife . 06074 . 013Figure 5 . Activation of Btk . ( A ) Activation of full-length Btk on membranes . Left panel: Btk ( 1 μM ) autophosphorylates rapidly on lipid vesicles ( total lipid concentration , 1 mM ) containing 75% DOPC , 20% DOPS and 5% PIP3 . Right panel: end-point autophosphorylation assay of Btk in the presence of lipid vesicles . The molar concentration of DOPC is kept at 75% , and the lipid fractions of PIP3 is 0% , 2% , 5% , and 10% , with the remaining lipid fraction filled with DOPS . The measurements are performed after 2 min incubation with lipid vesicles in the presence of ATP/Mg2+ . ( B ) End-point autophosphorylation assays of Btk in solution in the presence of inositol phosphates . Upper panel: Btk ( 2 μM ) autophosphorylates rapidly in the presence of 100 μM IP6 in solution , with the results after the first 5 min of incubation shown here . Essentially no phosphorylation is detected in Btk samples incubated with IP3 , IP4 or I ( 1 , 3 , 4 , 5 , 6 ) P5 , and very low phosphorylation is detected in Btk samples incubated with I ( 1 , 2 , 3 , 5 , 6 ) IP5 and I ( 2 , 3 , 4 , 5 , 6 ) IP5 in the same experiment . Lower panel: the assay is repeated with other inositol phosphates at a higher protein concentration ( 4 μM ) and longer incubation time ( 10 min ) , and the gel was exposed for a longer time . There is a small amount of degraded Btk in the sample , as judged by the coomassie staining , leading to multiple bands in the western blot shown here and in panel C . ( C ) Activation of full-length Btk ( 2 μM ) in the presence of IP6 ( 100 μM ) . ( D ) Activation of full-length Btk ( 2 μM ) with increasing concentrations of IP6 . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 01310 . 7554/eLife . 06074 . 014Figure 5—figure supplement 1 . Activation of various deletion constructs of Btk . ( A ) End-point autophosphorylation assay for full-length Btk , the Src-like module , the SH2-kinase construct and the kinase domain in the presence/absence of IP6 . The measurements are carried out after incubating proteins ( 2 μM ) with and without IP6 ( 100 μM ) for 2 min in the presence of ATP/Mg2+ . ( B ) End-point autophosphorylation assay for Btk PH-TH-kinase constructs of various linker lengths in the presence/absence of IP6 ( 100 μM ) . The measurements are carried out after incubating proteins ( 2 μM ) with and without IP6 ( 100 μM ) for 2 min at the presence of ATP/Mg2+ . ( C ) Activation of the PH-TH-kinase construct ( 2 μM ) with a 13-residue linker in the presence and absence of IP6 ( 100 μM ) . ( D ) Activation of full-length Btk ( 2 μM ) with mutations R133E/Y134E in the presence and absence of IP6 ( 100 μM ) . Arg 133 and Tyr 134 are located at the PH-TH/kinase interface . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 014 We next studied the effect of inositol phosphates , the soluble headgroups of various phosphatidylinositol phosphate lipids , on Btk activity . The purpose of these studies was to see whether interactions with the soluble headgroups might activate Btk independent of membrane localization , as is the case with Akt ( Calleja et al . , 2009; Wu et al . , 2010 ) . The soluble headgroup of PIP3 is IP4 , and so this was included in our study , as were several other inositol phosphates ( Figure 5B ) . IP4 itself has no effect on the rate of Btk activation in solution , demonstrating that the role of PIP3 is to recruit Btk to the membrane without necessarily triggering an allosteric effect . We tested the effect of 10 other inositol phosphates on the activity of Btk ( Figure 5B ) . Quite unexpectedly , we discovered that inositol hexakisphosphate ( IP6 ) promotes Btk autophosphorylation strongly at 2 μM protein concentration , substantially enhancing Btk catalytic activity within the first 5 min after addition ( Figure 5C ) . We also saw a milder activation with certain inositol pentakisphosphates ( IP5 ) , but these required higher protein concentration and longer times for appreciable autophosphorylation than did IP6 ( Figure 5B ) . The EC50 value for IP6 activation is around 20 μM ( Figure 5D ) , in the range of the physiological concentration of IP6 in cells , which is 10 μM–80 μM ( Irvine and Schell , 2001; Shears , 2001 ) . In the presence of a saturating concentration of IP6 ( 100 μM ) , full-length Btk is activated within the first 5 min of reaction time , in contrast to the 30 min and 120 min that are required for the activation of the SH2-kinase module and full-length Btk , respectively , in the absence of IP6 or PIP3-containing vesicles ( Figure 5A , C ) . Once fully activated , the catalytic rates of various Btk constructs are the same with or without IP6 addition , suggesting that the effect of IP6 is to change the rate of autophosphorylation , and that it does not affect the catalytic machinery directly ( Figure 5C ) . By comparing constructs , we determined which domains in Btk are responsible for the IP6 response . Activation by IP6 required only the PH-TH module and not the SH3 and SH2 domains ( Figure 5—figure supplement 1A , B ) . A PH-TH-kinase construct that contained the PH-TH-SH3 linker and the SH2-kinase linker ( with a total linker length of 45 residues ) had the same response to IP6 as did full-length Btk ( Figure 5—figure supplement 1B ) . PH-TH-kinase constructs with shorter linkers also responded to IP6 . The shortest linker necessary for activation contained only 13 residues from the SH2-kinase linker; we used this construct to determine the crystal structure described earlier ( Figure 5—figure supplement 1C ) . The catalytic activity of the fully activated PH-TH-kinase construct was about threefold lower than that of full-length Btk . This lower activity can be explained by the absence of the SH2 domain . IP6 has no effect on activation of the isolated kinase domain of Btk , suggesting that it is not a direct allosteric activator of the enzyme . The protein kinase CK2 is activated by IP6 , which binds to the kinase domain and dislodges an inhibitory interaction between CK2 and the protein Nopp140 ( Lee et al . , 2013 ) . It seems unlikely that IP6 works by displacing an inhibitory interaction between the kinase and the PH-TH module , because the Src-like module ( lacking the PH-TH module ) activates only slightly faster than full-length Btk in the absence of IP6 . Nevertheless , in order to test this possibility , we introduced two mutations ( Y133E and R134E ) at the PH-TH-kinase interface that are expected to disrupt the inhibitory docking of the PH-TH module on the kinase domain . Full-length Btk bearing these two mutations retains activation by IP6 , showing that IP6 does not act by dislodging the interactions seen in the crystal structure ( Figure 5—figure supplement 1D ) . The structure of the PH-TH-kinase construct does not provide direct clues about how IP6 might activate Btk . The IP4-binding site on the PH-TH module ( the ‘canonical binding site’ ) is located ∼15 Å from the kinase-interacting surface of the PH-TH module ( Figure 2A ) . The structure of the PH-TH-kinase construct , in which no inositol phosphates are bound , resembles closely that of the liganded PH domain ( Baraldi et al . , 1999 ) in the vicinity of the PH-kinase interface , consistent with the failure of IP4 to activate Btk . We used isothermal titration calorimetry ( ITC ) to study the interaction between IP6 and various constructs of Btk ( Figure 6A ) . The ITC measurements show that among the four domains in Btk , only the PH-TH module interacts with IP6 . At 150 mM NaCl concentration , the titration curve for IP6 binding to the PH-TH module can be fit by a one-site model for binding , with a dissociation constant , Kd , of 250 nM ( Table 3 ) . We introduced two mutations , R28C and N24E , at the canonical binding site in the PH-TH module . Arg 28 and Asn 24 have been shown previously to be critical for IP4 binding ( Hyvönen and Saraste , 1997; Baraldi et al . , 1999 ) . IP4 binds to the PH-TH module with a Kd value of 30 nM , and the R28C/N24E mutation in the PH-TH module reduces IP4 binding to an undetectable level ( Table 3 ) . The ITC signal of IP6 binding to the R28C/N24E mutant is reduced greatly but not completely abolished , and the residual signal corresponds to a dissociation constant between 5 μM and 10 μM . 10 . 7554/eLife . 06074 . 015Figure 6 . Binding of IP4 and IP6 to the Btk PH-TH module . ( A ) Representative isothermal titration calorimetry data for the Btk PH-TH module and its R28C/N24D variant binding to IP4 or IP6 . The protein concentration is 20 μM and that of the inositol phosphates goes up to 300 μM . Experiments are performed at 20°C . Titration curves are shown with the baseline-corrected raw data . The parameters from fitting a one-site binding model are listed in Table 3 . ( B ) Activation of full-length Btk N24D/R28C mutant ( 2 μM ) in the presence and absence of IP6 ( 100 μM ) . Asn 24 and Arg 28 are critical residues for PIP3/IP4 binding in the canonical lipid-binding pocket of the PH-TH module . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 01510 . 7554/eLife . 06074 . 016Table 3 . Thermodynamic parameters for Btk PH-TH module binding to IP6 and IP4DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 016ProteinLigandNKa ( ×106 ) Kd ( nM ) ΔH ( Kcal/mol ) ΔS ( cal/mol/deg ) Wild-type PH-THIP61 . 1 ± 0 . 14 . 2 ± 0 . 3238 ± 32−1 . 1 ± 0 . 226 . 7Wild-type PHTHIP40 . 9 ± 0 . 139 ± 1126 ± 64 . 4 ± 0 . 149 . 5PH-TH R28C/D24NIP60 . 7 ± 0 . 20 . 21 ± 0 . 034760 ± 700−1 . 7 ± 0 . 617 . 5The integrated heat from representative isothermal titration calorimetry experiments was fit with a one-binding site model . See methods for the details of data analysis . Although the residual binding of IP6 to the IP4-binding deficient mutant is relatively weak , the observed affinity is consistent with the EC50 value of 20 μM for IP6 activation , as determined in our activation assays ( Figure 5C ) . We observed no significant activation when the IP6 concentration was lower than 5 μM , a level still high enough to saturate the canonical binding site . We also found that the R28C/N24E mutant that is deficient in IP4 binding remains highly responsive to IP6 stimulation , with only a slight reduction in the activation kinetics as compared to the wild-type protein ( Figure 6B ) . These observations imply that the canonical binding site of the PH-TH module may not play a major role in the IP6-mediated activation of Btk . We determined a crystal structure of the Btk PH-TH module bound to IP6 at 2 . 3 Å ( Figure 7 ) . There are four PH-TH modules in the asymmetric unit of the crystal , which form two dimers , similar to those seen in other structures of the Btk PH-TH module ( Figure 7D ) ( Hyvönen and Saraste , 1997; Baraldi et al . , 1999 ) . The only structural changes in the IP6-bound PH-TH module , relative to the apo- or IP4-bound PH-TH module , are in the loop regions ( Figure 7—figure supplement 1B ) . 10 . 7554/eLife . 06074 . 018Figure 7 . Crystal structure of the PH-TH module bound to IP6 . ( A ) Structure of the PH-TH module bound to two IP6 molecules . One IP6 molecule is in the canonical lipid-binding site and the other IP6 molecule is in the peripheral binding site formed by strands β3 and β4 . The β1/β2 loop ( dotted lines ) is disordered in this structure . ( B ) IP6 coordination in the canonical binding site and in the peripheral binding site . ( C ) Activation of the Btk R52S/K49S mutant ( 2 μM ) in the presence and absence of IP6 ( 100 μM ) . Arg 52 and Lys 49 are two residues interacting with IP6 in the peripheral binding site . ( D ) A dimer assembly of the Btk PH-TH module in the crystal structure of the IP6-bound PH-TH module . The canonical binding site and the peripheral binding site are shown in blue and magenta circle , respectively . We refer to this dimer as the ‘Saraste dimer’ ( Hyvönen and Saraste , 1997 ) . ( E ) An electron density map , using σA-weighted 2m\|Fo\| − D\|Fc\| coefficients ( Read , 1986 ) for IP6 at the peripheral binding site , contoured at 1 . 5 σ . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 01810 . 7554/eLife . 06074 . 019Figure 7—figure supplement 1 . Aspects of the binding of IP6 to Btk . ( A ) An open-ended chain of PH-TH dimers in the crystal lattice , mediated by IP6 at the pheripheral site . IP6 is bound at the peripheral site in such a way that only one IP6 molecule occupies the two sites at each interface . ( B ) Structure superposition of the IP6-bound PH-TH module with apo- and the IP4-bound PH-TH module . The β1/β2 loop , which coordinates IP4 binding , adopts a different conformation in the apo-structure , and is partially disordered in the IP6-bound structure . ( C ) IP6 coordination in the peripheral binding site by a crystal symmetry mate of the PH-TH module in the crystal structure of IP6-bound PH-TH module . IP6 and the residues involved in coordination are shown in sticks . The hydrogen bonds are shown in dotted lines . This is a detailed view of the interfacial sites shown in the schematic diagram in Panel A . ( D ) Isothermal titration calorimetry data for Btk PH-TH mutant R52E/K49E/R28C/N24D titrated with IP6 . Arg 52 and Lys 49 are critical residues coordinating IP6 binding in the peripheral site . Arg 28 and Asn 24 are critical residues coordinating IP6 binding in the canonical site . Note that IP6 does not bind to this mutant form of the PH-TH module . ( E ) Activation of full-length Btk , Btk R52S/K49S mutant , and Btk R28C/N24D mutant in the presence of IP6 ( 100 μM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 019 The four PH-TH modules in the asymmetric unit have six bound IP6 molecules , each with clearly resolved electron density . Four of the IP6 molecules are in the canonical binding sites on each module . At those sites , residues Asn 24 , Tyr 39 , Arg 28 , and Lys 12 coordinate phosphate groups on the 2 , 3 , and 4 positions of the myo-inositol ring; these are also the key residues involved in IP4 binding ( Baraldi et al . , 1999 ) ( Figure 7B ) . The loop between strands β1 and β2 , which is important for IP4 binding , is largely disordered in our structure ( Figure 7A ) . There are fewer interactions at the canonical site between the PH-TH module and IP6 than seen for IP4 , consistent with the relative affinities of the two ligands ( Figure 6A ) . The two other IP6 molecules in the crystallographic asymmetric unit are bound at a peripheral site . This site is on the surface formed by an elongated β hairpin involving strands β3 and β4 of each PH-TH module . This site has not been seen in any other PH domain ( Lemmon , 2008 ) ; it differs from the two well-characterized lipid-binding sites in the PH domains of Pleckstrin ( Ferguson et al . , 1995 ) and β-spectrin ( Hyvönen et al . , 1995 ) . As in the canonical binding site , interactions between IP6 and the peripheral site are predominantly electrostatistic ( Figure 7B ) . Unlike the canonical site , which has a well-defined groove within which the inositol phosphates are bound , the peripheral site is relatively flat ( Figure 7B ) . A feature of the crystallographic lattice is that IP6 bound at the peripheral site is sandwiched between two PH-TH modules , interacting with the peripheral sites on both ( Figure 7—figure supplement 1A , C ) . The majority of interactions are with one subunit . Residues Arg 49 , Lys 52 , Lys 36 , and Tyr 40 at the peripheral binding site of one subunit ( Figure 7—figure supplement 1C ) coordinate phosphate groups at the 1 , 2 , 3 , and 4 positions of the myo-inositol ring . A subset of these residues , Lys 52 and Lys 36 , on the other PH-TH module interact from the opposite side of the myo-inositol ring with the phosphate group at position 3 ( Figure 7—figure supplement 1C ) . The bridging interactions provided by IP6 generate an open-ended chain of PH-TH dimers in the crystal lattice ( Figure 7—figure supplement 1A ) . Based on the limited interactions across this crystallographic interface , we do not believe that this open-ended chain is relevant to the function of Btk . We mutated Arg 49 and Lys 52 , in the peripheral binding site of the PH-TH module , to serine in full-length Btk . This Btk variant autophosphorylates slowly , as does wild-type Btk , and it has very little response to IP6 ( Figure 7C ) . In ITC experiments , we saw no detectable binding when we added IP6 to a PH-TH module with mutations at both the canonical and the peripheral sites , further confirming that the residual binding of IP6 to the IP4-binding deficient mutant is at the peripheral binding site ( Figure 7—figure supplement 1E ) . The disposition of the IP6 molecule can be determined unambiguously from the electron density because the phosphate group at the 2 position juts out from the plane of the myo-inositol ring ( Figure 7E ) . IP6 at the peripheral binding site interacts with the primary PH-TH module to which it is bound through the phosphates at positions 1 , 2 , 3 , and 4 . None of the inositol phosphates used in our experiments , other than IP6 , can fully satisfy this criterion , which may underlie the specificity of the PH-TH module for IP6 . Two isomers of IP5 , I ( 2 , 3 , 4 , 5 , 6 ) P5 and I ( 1 , 2 , 3 , 5 , 6 ) P5 , have a mildly activating effect on Btk; they are more effective than I ( 1 , 3 , 4 , 5 , 6 ) P5 . These comparisons indicate an important role in Btk activation for the phosphate at position 2 ( Figure 5B ) . In the cell , the only known enzyme that can add a phosphate to the hydroxyl group at position 2 of a myo-inositol ring is inositol 1 , 3 , 4 , 5 , 6-pentakisphosphate 2-kinase ( York et al . , 1999 ) , which produces IP6 . As a consequence , IP6 , and inositol pyrophosphates , such as IP7 and IP8 , may be the only cellular phosphates that can interact with the peripheral site in the Btk PH-TH module . Strong activation of Btk by concentration on PIP3-containing membranes suggests that IP6 might promote dimerization or oligomerization of Btk in solution . Neither multi-angle light scattering nor size exclusion chromatography gave convincing evidence for IP6-induced association , at Btk concentrations up to 30 μM . Nevertheless , we found it striking that in every crystal structure of the Btk PH-TH module , as well as in our structure of the PH-TH-kinase construct , the module forms very similar crystallographic dimers ( Figure 7D ) ( PDB entry: 1BTK [Hyvönen and Saraste , 1997]; PDB entry:1B55 and 1BWN [Baraldi et al . , 1999]; see also PDB entry: 2Z0P ) . This dimer was first identified and discussed by Marko Hyvönen and the late Matti Saraste; we refer to it as the ‘Saraste dimer’ . The Saraste dimer is formed by symmetric interactions between helix α1 and strands β3 and β4 of the PH-TH module ( Figure 7A ) . The largely hydrophobic dimer interface has a buried surface area of ∼1000 Å2 on each molecule . Phe 44 and Tyr 42 on strand β3 , Ile 9 on strand β1 and IIe 95 on helix α1 create a hydrophobic patch in one subunit of the dimer that packs tightly against the corresponding patch on the other subunit ( Figure 8A ) . 10 . 7554/eLife . 06074 . 020Figure 8 . The Saraste dimer of the PH-TH module is critical for Btk activation by IP6 . ( A ) Molecular details of the Saraste dimer interface in the crystal structure of the IP6-bound PH-TH module . ( B ) End-point autophosphorylation assays for wild-type Btk , Btk L29R/I9R mutant , Btk I94R/I95R mutant , and Btk Y42R/F44R mutant in the presence and absence of IP6 ( 100 μM ) . The measurements were made after incubating the proteins with and without IP6 in the presence of ATP/Mg2+ for 5 min . Residues Leu 29 , Ile 9 , IIe 94 , IIe 95 Tyr 42 and Phe 44 are located in the Saraste dimer interface , as shown in panel A . Activation of the Btk Y42R/F44R mutant is shown in the lower panel . ( C ) Activation of the Btk-Abl fusion construct ( 1 μM ) and that of a variant of this fusion protein in which the Btk PH-TH module is mutated ( Y42R/F44R ) . Measurements were made with 1 μM protein in the presence and absence of IP6 ( 100 μM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 020 To test whether the Saraste dimer interface is important for IP6-mediated Btk activation , we made three mutants in which we replaced hydrophobic residues at the Saraste dimer interface with arginines and compared their autophosphorylation rates and their response to IP6 with those of wild-type Btk . All three sets of mutant proteins showed little or no response to IP6 ( Figure 8B ) . These mutations do not affect the autophosphorylation rate of Btk in the absence of IP6 . We conclude that the Saraste dimer interface promotes Btk association in the presence of IP6 . The activation of Class I Histone Deacetylases ( HDACs ) requires the formation of a multi-subunit co-repressor complex , in which Ins ( 1 , 4 , 5 , 6 ) P4 serves as an intermolecular glue that cements the complex together ( Watson et al . , 2012; Millard et al . , 2013 ) . The mechanism of IP6-induced transient Btk dimerization is fundamentally different , in that the peripheral IP6 binding site is just outside the Saraste dimer interface , and IP6 does not participate in any interactions between the two Btk subunits in that dimer . If the effect of IP6 arises from its ability to promote dimerization of the PH-TH module , fusion of a Btk PH-TH module with the catalytic domain of another kinase that autophosphorylates slowly should impart activation by IP6 . A good example of such a kinase is c-Abl . Indeed , in chronic myelogenous leukemia , a chromosomal translocation results in fusion of a coiled-coil oligomerization domain with c-Abl , leading to hyper-activation of the BCR-Abl fusion protein by autophosphorylation . Several other fusion variants involving c-Abl , in which different oligomerization domains are present , are associated with hematological malignancies ( De Braekeleer et al . , 2011 ) . We generated a PH-TH-Abl fusion construct in which the PH-TH module of Btk is fused to the SH2-kinase module of c-Abl . We compared autophosphorylation rates of the PH-TH-Abl fusion protein with and without IP6 at a saturating concentration ( 100 μM ) . The fusion protein at a concentration of 1 μM autophosphorylated slowly in the absence of IP6 , while the IP6-bound protein became heavily phosphorylated within the first 30 s , the dead-time of our assay ( Figure 8C ) . We replaced Phe 44 and Tyr 42 , at the Saraste dimer interface , with arginine in the fusion protein . The autophosphorylation rate of the PH-TH-Abl mutant in the presence of IP6 was much lower than that of the IP6-bound wild-type protein , further evidence for the importance of the Saraste dimer interface in IP6-mediated activation ( Figure 8C ) . The Btk and c-Abl kinase domains are 46% identical in sequence , but most of the surface residues are different . We infer that , except for its enzymatic activity , the interaction properties of the Btk kinase domain are not important for the IP6 response . Because IP6 activates both kinases when they are linked to the Btk PH-TH module , we propose that IP6 promotes transphosphorylation of the attached kinase by inducing transient dimerization . Our preliminary studies indicate that the IP6-induced dimerization of the Btk PH-TH module is very weak , with a dissociation constant that is likely to be greater than 100 μM . It might appear counter-intuitive that IP6 can markedly accelerate activation at 2 μM Btk concentration , well below the dissociation constant . A simple calculation shows that even a modest increase in a small dimer population can lead to a sharp change in the rate of activation because the reaction is autocatalytic . The initial production of phosphorylated Btk drives the further production of this species at an accelerating rate ( Figure 9—figure supplement 1B ) . Our failure to detect stable dimers of the Btk PH-TH module or of full-length Btk in the presence of IP6 suggests that the monomeric PH-TH module has a conformation different from the one seen in crystal structures , which all show the same dimer contact . The high protein concentration in the crystal lattice ( ∼30 mM ) presumably drives the formation of the Saraste dimer , which is otherwise not favored . Molecular dynamics simulations suggest that the conformation of the PH-TH module seen in the Saraste dimer is unstable when the protein is monomeric . We initiated molecular dynamics trajectories using a monomeric PH-TH module from a crystal structure which contained neither IP6 nor IP4 . Within the first 30 ns of simulation , the β3-β4 hairpin bent towards the core of the PH-TH module ( Figure 9A and Figure 9—figure supplement 1A ) , and Phe 44 , at the heart of the Saraste dimer interface , changed its conformation and packed against the other three interfacial residues , Tyr 42 , Ile 9 , and IIe 95 ( Figure 9A ) . This conformational change closed the dimer interface , burying the interfacial residues . 10 . 7554/eLife . 06074 . 021Figure 9 . Possible allosteric and electrostatic effects of binding IP6 . ( A ) Fluctuations of the PH-TH module during a molecular dynamic simulation . An instantaneous structure ( t = 100 ns ) from a 100 ns simulation of the PH-TH module shows a conformational change that closes the Saraste dimer interface ( middle panel ) . Replacing these hydrophobic residues ( IIe 9 , Tyr 42 , Phe 44 and Ile 95 ) with alanine prevents the dimer interface from closing in the simulations ( lower panel ) . Although the β3/β4 loop becomes more dynamic , it remains open ( see Figure 9–figure supplement 1 ) . Binding of IP6 may shift the equilibrium between open and closed conformations . ( B ) Activation of the Btk R52E/K49E mutant ( 2 μM ) . Arg 52 and Lys 49 are two residues in the peripheral binding site . Mutation of these two residues to glutamate reduces the local charge at the peripheral site by a net four units , and results in Btk activation . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 02110 . 7554/eLife . 06074 . 022Figure 9—figure supplement 1 . Flexibility in the PH domain of Btk and simulation of the kinetics of Btk autophosphorylation . ( A ) Fluctuations in the β3/β4 loop of the wild-type PH-TH module and its I9A/Y42A/F44A/I95A variant . Each trajectory was sampled every 1 ns . The instantaneous structures are aligned using the crystal structure as the reference . The r . m . s . deviation was calculated for the backbone of residues 36 through 57 of the PH-TH domain . ( B ) Simulations of the kinetics of Btk autophosphorylation . We consider a very simple two-step Michaelis–Menten model for autophosphorylation , in which two Btk molecules first form an encounter complex , followed by phosphorylation of one of the two molecules in the complex . We assume that the rate of complex formation increases if IP6 is bound , and that the catalytic rate constant for phosphorylation of the other increases if one of the two Btk molecules is phosphorylated . This model is described by the following reaction scheme: ( 1 ) Btk+Btk⇌k−1k1Btk·Btk , ( 2 ) Btk·Btk→k2pBtk+Btk , ( 3 ) pBtk+Btk⇌k−3k3pBtk·pBtk , ( 4 ) pBtk·Btk→k4pBtk+pBtk . The on-rate for a diffusion-limited encounter between two proteins is in the range of 108 to 107 M−1s−1 . We assume that in the absence of IP6 the on-rate for Btk dimer formation is much slower than this ( k1 = k3 = 104 M−1s−1 ) and that IP6 binding increases the on-rate by 10-fold ( k1 = k3 = 105 M−1s−1 ) . The dissociation rate constant is chosen so that the dissociation constant is 2 mM and 200 μM in the absence and presence of IP6 , respective ( k-1 = k-3 = 20 s−1 ) . The catalytic constant ( k4 ) for activated Btk is set to 1 . 0 s−1 , comparable to the values determined for other tyrosine kinases , such as Src and ZAP70 ( Mukherjee et al . , 2013 ) . We assume that the catalytic constant is 10-fold smaller for unphosphorylated Btk ( k2 = 0 . 1 s−1 ) . To compare with our experimental data , the initial concentration of Btk was set at 1 μM . The integration of the differential kinetic equations was done using BerkeleyMadonna ( http://www . berkeleymadonna . com/ ) . As shown in the diagram , the autophosphorylation kinetics generated by this simple reaction scheme is consistent with the experimentally observed kinetics . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 022 The hydrophobicity of the dimer interface drove the observed conformational change . Molecular dynamics simulations initiated from crystal structures but with the four interfacial residues substituted by alanine reduced the extent of conformational change in this area ( Figure 9A ) . The peripheral binding site for IP6 is adjacent to the β3-β4 hairpin that folds in and closes the dimer interface in the simulation . Thus , one effect of IP6 binding may be to stabilize the open form of the hairpin . It also appears that IP6 might act by altering the electrostatic potential of the PH-TH module . We examined the surface electrostatic potential of the Btk PH-TH module by viewing graphical displays of solutions to the Poisson-Boltzmann equation ( Dolinsky et al . , 2004 , 2007 ) . The PH-TH domain carries seven net positive charges; it is also electrostatically polarized , with the positively charged residues concentrated on the broad , contiguous surface composed of the β3-β4 hairpin , the IP4 binding loop and the N-terminal region of the loop connecting strand β5 and helix α2 . In the Saraste dimer , positive electrostatic field lines calculated based on the monomer run into each other at the dimer interface , suggesting that electrostatic forces oppose dimer formation . We introduced two charge-reversal mutations at the peripheral IP6-binding site ( K49E and R52E ) , to reduce the positive charge on the protein by four units . The autophosphorylation rate of full-length Btk with the K49E/R52E mutations was substantially higher than that of the wild-type protein , with pronounced phosphorylation observed in the first 2 min of the reaction ( Figure 9B ) . We suggest that there are two barriers to the dimerization of the PH-TH module of Btk . First , the dimer interface of the PH domain is probably closed in the monomeric protein . Second , an electrostatic repulsion between PH domains further impedes dimerization . IP6 binding might stabilize the open form of the PH-TH module allosterically , while also alleviating the electrostatic repulsion . Our studies on Btk have revealed the structural basis for its autoinhibition . One key finding is that the Src-like module of Btk adopts a compact and autoinhibited form that resembles the autoinhibited structures of the Src kinases and c-Abl , in which the SH3 domain binds to the SH2-kinase linker and holds the kinase domain in an inactive conformation . Our structure of the PH-TH-kinase construct of Btk shows that the PH-TH module stablizes the inactive conformation of the kinase domain , as well as the assembled conformation of the Src-like module of Btk . A second , and quite unexpected , finding is that IP6 can activate Btk strongly by binding to a newly identified site on the PH domain , probably by promoting transient dimerization of the PH-TH module . Because IP6 synthesis requires the precursor molecule IP3 , which is itself produced as a result of Btk activation , a transient elevation of IP6 may create a positive feedback loop for Btk activation , either on the membrane or in the cytoplasm . Formation of the Saraste dimer interface is compatible with the simultaneous engagement of PIP3 by both PH domains in the dimer , raising the possibility that IP6 can stimulate dimerization of membrane-tethered Btk ( Figure 10 ) . IP6 binding to the peripheral binding site may also enhance PIP3 affinity in the canonical binding site and sensitize Btk to PIP3 levels . The linker connecting the PH-TH unit to the SH3 domain promotes dimerization of Btk by interacting with the SH3 domain of a second molecule , providing an additional mechanism for Btk dimerization at high local concentration ( Laederach et al . , 2002 ) . High levels of IP6 may down-regulate the amount of Btk on membranes by competing for PIP3 binding in the canonical binding site , as proposed for Akt1 regulation by IP7 ( Chakraborty et al . , 2010 ) . 10 . 7554/eLife . 06074 . 023Figure 10 . Models of Btk autoinhibition and activation . ( A ) The PH-TH module stabilizes the assembled conformation of the Src-like module of Btk in its inactive conformation . The activation of Btk involves formation of a transient dimer through the PH-TH module , which promotes trans-autophosphorylation and activates Btk . The stimulation of dimerization could occur at the membrane or in solution , as in our experiments . ( B ) IP6 may stimulate membrane-tethered Btk . Formation of the Saraste dimer is compatible with the simultaneous engagement of PIP3 by both PH domains in the dimer , and is compatible with the binding of IP6 at the peripheral binding site . This raises the possibility that IP6 can stimulate dimerization of the membrane-tethered Btk . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 02310 . 7554/eLife . 06074 . 024Figure 10—figure supplement 1 . Sequence alignment of Btk and Itk . The alignment was performed using ClustalW2 ( Larkin et al . , 2007 ) and visualized by Espript3 ( Robert and Gouet , 2014 ) . The alignment shows that the adoption of the compact autoinhibited conformation by the Src-like module , seen in our crystal structures , is likely to be conserved in Itk . In contrast , residues at the Saraste dimer interface and at the peripheral IP6 binding site in Btk are not conserved in Itk . DOI: http://dx . doi . org/10 . 7554/eLife . 06074 . 024 The Src-like modules of Btk and Itk have 55% sequence identity , with all the residues critical for maintaining the assembled and autoinhibited Src-like conformation of Btk conserved in Itk ( Figure 10—figure supplement 1 ) . The Src-like module of Itk is therefore likely also to adopt the assembled Src-like conformation when inactive . The roles of the PH-TH module in regulating Btk activity are not likely to be conserved in Itk . The four residues at the peripheral binding site that coordinate IP6 in Btk are not conserved in Itk , unlike those that coordinate PIP3/IP4 at the canonical lipid-binding site . This lack of conservation also extends to the key residues at the inhibitory interface between the PH-TH module and the kinase domain in Btk and to residues at the Saraste dimer interface ( Figure 10—figure supplement 1 ) . A different soluble inositol phosphate , IP4 , promotes PIP3 binding by the Itk PH-TH module; this activity is important for Itk regulation in T-cells ( Huang et al . , 2007 ) . Thus , regulation by soluble inositol phosphates may be a general feature of Tec family kinases , but the specific effector and the molecular mechanism of the kinase response may be different in each case . The regulation of tyrosine kinases by soluble second messengers has not received much attention in the past . Our discovery that IP6 activates Btk , along with the known effect of IP4 on Itk , points to the importance of further studies on these and soluble effectors . DNA encoding bovine Btk ( residues 1 to 659 ) was cloned into pET-28 expression plasmid yielding N-terminally 6-histidine-tagged SUMO-fusion protein . The hexahistidine-SUMO tag was cleavable by Ulp-1 SUMO-specific protease . For protein expression , the plasmid was transformed into BL21 DE3* ( Novagen , Australia ) cells that also contain the expression plasmids for GroES/GroEL and phosphatase YopH ( Seeliger et al . , 2005 ) , and then grown in Terrific Broth ( TB ) media supplemented with 50 mg/l kanamycin , 50 mg/l streptomycin , and 50 mg/l chloramphenicol at 37°C . At a cell density corresponding to an absorbance of 1 . 5 at 600 nm , the temperature was reduced to 18°C , and protein production was induced by mixing the bacterial culture with equal volume of 4°C TB media supplemented with 1 mM IPTG , 100 μM ZnCl2 , 50 mg/l Kanamycin , 50 mg/l streptomycin , and 50 mg/l chloramphenicol . After 16 hr , cells were harvested by centrifugation , resuspended in Ni-NTA buffer A ( 500 mM NaCl , 20 mM imidazole , 25 mM Tris pH 8 . 5 , 5% glycerol , and 1 mM DTT ) , and flash-frozen in liquid nitrogen . Cells were then thawed on ice and supplemented with a mixture of protease inhibitors ( aprotinin 2 μg/ml , benzamidine 15 μg/ml , leupeptin 2 μg/ml , PMSF 1 mM , pepstatin A , 1 μg/ml ) . After cell lysis by French press and removal of cell debris by centrifugation , clear lysates were loaded onto a HisTrap Ni-NTA column ( GE Healthcare , UK ) , washed with 100 ml Ni-NTA buffer A . Proteins were eluted with a single step of Ni-NTA buffer A supplemented with 250 mM imidazole . Proteins were buffer exchanged into storage buffer ( 25 mM Tris-HCl , pH 8 . 0 , 400 mM NaCl ) by FastFlow desalting column ( GE Healthcare ) , and affinity tags were removed by incubation with Ulp-1 protease at 4°C overnight . Cleaved proteins were collected in the flow-through during Ni-NTA affinity chromatography and were subjected to size-exclusion chromatography on a Superdex 200 column ( GE Healthcare ) equilibrated in storage buffer . Proteins were concentrated on an AmiconUltra centrifugal filter device ( 50 kDa cutoff; Millipore ) to a final concentration of ∼400 μM . Protein aliquots were frozen in liquid nitrogen and stored at −80°C . The Src-like module ( residues 217 to 659 ) , SH2-kinase module ( residues 270 to 659 ) , and the kinase domain ( residues 394 to 659 ) of bovine Btk , various bovine Btk PH-TH-kinase constructs and the PH-TH-Abl fusion construct ( BTK residues 1 to 216 fused to human c-Abl residues 140 to 518 ) were all expressed and purified using the same protocol . The PH-TH module of bovine Btk ( residues 1 to 172 ) was prepared in the same way as described above expect that the co-expression of GroEL/GroES and YopH is not necessary . Btk samples were diluted to 4 μM using the buffer that contains 150 mM NaCl , 25 mM Tris , pH 7 . 5 , and 5% glycerol . For experiments involving lipid vesicles and inositol phosphates , additional 500 μM lipid and 200 μM inositol phosphates are supplemented into the dilution buffer , respectively , followed by a 10-min incubation at room temperature . The diluted samples were mixed with equal volume of the reaction buffer that contains 150 mM NaCl , 20 mM MgCl2 , 2 mM ATP , 25 mM Tris , pH 7 . 5 , 2 mM sodium vandate , and 5% glycerol and incubated at room temperature . The final concentration of Btk samples is 2 μM . At each time point , the samples were either frozen in liquid nitrogen and stored at −80°C for the coupled kinase assays , or mixed with equal volume of quench buffer that contains 2X SDS-PAGE buffer supplemented with 100 mM EGTA , followed by incubating at 95°C in a heat block for 15 min for western blot assays . A continuous pyruvate-kinase coupled assay was performed to measure the kinase activity of the proteins as described ( Barker et al . , 1995 ) , with minor modifications . The ATP concentration was kept at 1 mM in all the assays . The buffer used contains 150 mM NaCl , 10 mM MgCl2 , 25 mM Tris , pH 7 . 5 , 2 mM sodium vandate . The protein concentrations were kept at 1 μM for all Btk constructs . The peptide sequence is ERDINSLYDVSRMYVDPSEIN , which was derived from residues 746 to 766 of PLC-γ2 . The Km value for phosphorylated full-length Btk for this peptide is 925 ± 117 μM , and the peptide concentration was kept at 1 mM in all experiments . The reactions were performed in a 96-well plate and were monitored by SpectraMax ( Molecular Devices , Sunnyvale CA ) . The ATP turn-over rate ( v0: min−1 ) was calculated by fitting the data acquired from first 150 s using a linear regression model . The levels of autophosphorylation of purified Btk samples were monitored using non-specific anti-phosphortyrosine antibody 4G10 ( EMDMillipore ) . The total amount of Btk was monitored using coomassie-blue stained SDS-PAGE . For each set of experiments , 10 μl of each Btk sample was loaded on two 12% SDS-PAGE gels . The two gels were run at 250 V , 400 mA for 35 min at room temperature . One gel was transferred to 50 ml fixing buffer containing 50% ethanol and 10% acetic acid and was heated for 30 s in a microwave . The gel is then transferred to 50 ml staining buffer that contains 5% ethanol , 7 . 5% acetic acid , and 0 . 00025% coomassie brilliant blue R-250 and was heated for another 30 s in a microwave . The gel was then left to cool at room temperature in the staining buffer on a rocker for 1 hr and was then imaged using ImageLab ( Bio-Rad , Hercules CA ) . The other gel was transferred into western blot transfer buffer that contains 25 mM Tris at pH 7 . 4 , 192 mM glycine , and 20% methanol and was incubated at room temperature for 15 min . Protein samples were then transferred to PVDF membranes from the gel using the semi-dry transfer device TRANS-BLOT ( Bio-Rad ) . The membranes were then blocked for 1 hr in TBST buffer ( 20 mM Tris pH 7 . 5 , 150 mM NaCl , 0 . 1% Tween-20 ) supplemented with 5% dry milk , and incubated with primary antibody 4G10 ( 1:2000 dilution ) at 4°C overnight . The membranes were then washed three times in TBST buffer , and incubated with HRP-linked anti-mouse IgG antibody ( 1:5000 dilution ) in TBST buffer at room temperature for 1 hr . The membranes were then washed three times in TBST buffer , incubated with 0 . 5 ml WesternBright Quantum ( Advansta , Menlo Park CA ) for 1 min , and were imaged using ImageLab ( Bio-Rad ) . The western blots were reproduced at least three times and the most representative ones were chose to present . Three lipids , 18:1 DOPS , 18:1 ( Δ9-Cis ) ( DOPC ) , and 18:1 PI ( 3 , 4 , 5 ) P3 ( Avanti , Alabaster AL ) were dissolved in chloroform and stored at −20°C . For liposome preparations , the three lipids are mixed in a test tube at the desired molar ratio ( 10:75:5 ) ; chloroform was evaporated under a nitrogen atmosphere for 15 min; the samples were transferred to a vacuum desiccator and dried overnight at room temperature . Dry films were hydrated in buffer containing 25 mM Tris-HCl ( pH 7 . 4 ) and 100 mM NaCl to a concentration of 10 mg/ml . The hydrated liposomes are heterogeneous in size and the diameters of the liposomes vary from ∼50 nm to ≥ 1 μm . To prepare small unilamellar vesicles , hydrated liposomes were subject to seven freeze–thaw cycles using liquid nitrogen , followed by extrusion through 100-nm filters ( Avestin , Canada ) . The liposomes prepared using this method have diameters around 100 nm , as measured using negative staining electron microscopy ( data not shown ) . The isothermal titration calorimetry experiments were performed using MicroCal-autoITC 200 ( GE Healthcare ) . Various Btk samples are diluted in the binding buffer containing 150 mM NaCl , 25 mM Tris , pH 7 . 5 , and 5% glycerol to a final concentration of 20 μM . IP4 and IP6 were diluted using the same buffer to a final concentration of 300 μM . Each set of ITC experiments included three samples: 420 μl of Btk protein sample in the cell , 150 μl of IP4 or IP6 at a concentration 300 μM in the syringe , and 500 μl of binding buffer . All samples are stored at 4°C before the titration experiments . The ITC experiments were performed at 20°C . An initial injection of 0 . 5 μl was excluded from data analysis , followed by 14 injections of 3 μl each , separated by 180 s with a filter period of 5 s . The protein solution was stirred at 500 rpm over the course of titration . Titration curves were fit with a one-site binding model . The three fitting parameters are stoichiometry , N , association constant , Ka and binding enthalpy , ΔH . The binding entropy is then calculated using formula: ΔS = ( ΔH + RTlnKa ) /T . Thermodynamic parameters for various Btk constructs binding to IP4 and IP6 are listed in Table 1 . Molecular dynamics trajectories were generated using the Gromacs 4 . 6 . 2 package ( Pronk et al . , 2013 ) using the ff99SB-ILDN force field ( Lindorff-Larsen et al . , 2010 ) . All simulations were in water , using the TIP3P water model; appropriate counterions ( Na+ and Cl− ) were added to neutralize the net charges . After initial energy minimization , the systems were subjected to 100 ps of constant number , volume and temperature ( NVT ) equilibration , during which the system was heated to 300 K . This was followed by a short equilibration at constant number , pressure , and temperature ( NPT , 100 ps ) . Finally , the production simulations were performed under NPT conditions , with v-rescale thermostats in Gromacs 4 . 6 . 2 , respectively , in the absence of positional restraints . Periodic boundary conditions were imposed , and particle-mesh Ewald summations were used for long-range electrostatics and the van der Waals cut-off is set at 10 Å . A time step of 2 fs was employed and the structures were stored every 2 ps . His 143 , Cys 154 , Cys 155 , and Cys 165 , which coordinate Zn2+ in the PH domain were set to be deprotonated . No additional positional restrains are applied to the Zn2+ ion , and its sp3 tetrahedral coordination remains intact during all MD simulations . The visual molecular dynamics ( VMD ) analysis toolkit ( Humphrey et al . , 1996 ) was used to make some of the root mean square deviation ( RMSD ) measurements , others were done using PyMol . The active conformation of the Btk kinase domain was modeled based on that of the Lck kinase domain ( PDB: 3LCK ) ( Sicheri et al . , 1997 ) , using Modeller ( Eswar et al . , 2006 ) . The sequence of the kinase domains of Lck and Btk were aligned by ClustalW ( Larkin et al . , 2007 ) . We chose the model with the lowest Discrete Optimized Protein Energy ( DOPE: −0 . 84 ) as representative of the active conformation of Btk . We estimated the effective concentration ( C ) of Btk on a lipid vesicle by the following equation:C=NV×Na . N is the number of Btk molecules on a lipid vesicle , Na is the Avogadro number , and V is the effective volume in which Btk moves on a lipid N0=AA0 vesicle . An estimate for V is given by:V=A×L . Here , A is the surface area of a lipid vesicle and L is the length of a Btk molecule , ∼ 100 Å , as measured from the composite model of full-length Btk . Assuming all the PIP3 molecules ( molar fraction , f ) are bound to Btk , then , N=f×N0 , where N0 is the total number of lipids on a vesicle . This is given by , N0=AA0 , in which A0 is the surface area of a lipid molecule . We assume all lipids have the same surface area for their headgroups , which is roughly 0 . 6 nm2 ( Gulik-Krzywicki et al . , 1967 ) . Therefore C is given by:C=NV×N0=f×AA0A×L×Na=fA0×L×Na . In our experiments , f = 0 . 05 , A0 = ∼60 Å2 , Na = 6 . 03 × 1023 mol−1 , and L = ∼100 Å , giving a value of 12 mM for C .	The human immune system is our defense against a variety of intruders , including bacteria , viruses , and other microbes , and provides our body with long-lasting protection against these invaders . B-cells—one type of immune cell—produce proteins called antibodies that can recognize microbes from previous invasions . This allows other immune cells to find the intruders and destroy them more efficiently . People suffering from a genetic disease called X-linked agammaglobulinemia have fewer fully mature B-cells than most people and are extremely susceptible to bacterial infections . The disease is caused by mutations in the gene encoding an enzyme called Bruton's tyrosine kinase ( Btk ) , which is involved in B-cell development . This enzyme is usually kept inactive until the B-cells are needed , but the molecular mechanism of this restraint remains to be discovered . Wang et al . have used X-ray crystallography and biochemistry to produce a three-dimensional atomic model of the Btk enzyme in its inactive form . The model and functional studies reveal that when the Btk enzyme is inactive , it folds up into a compact shape . Within the enzyme , a ‘PH-TH module’ and two other domains work together to stabilize this folded-up form . Although earlier studies indicated that the Btk enzyme becomes activated only when it is recruited to cell membranes , Wang et al . found that a small molecule called IP6 ( short for inositol hexakisphosphate ) can activate the Btk enzyme even in the absence of a membrane . This molecule recognizes a previously unnoticed binding site on the PH-TH module , and causes the PH-TH module in the inactive Btk enzyme to bind briefly to the PH-TH modules in other Btk enzymes . This brief binding event is an important step in the process that activates the enzyme . IP6 might also be involved in recruiting the Btk enzyme to the cell membrane in B cells , but this proposal needs to be tested experimentally . The details of the BTK activation mechanism reported by Wang et al . could prove useful to researchers developing new treatments for immunodeficiency diseases .	[ "Abstract", "Introduction", "Results", "and", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	Autoinhibition of Bruton's tyrosine kinase (Btk) and activation by soluble inositol hexakisphosphate
AAA+ unfoldases are thought to unfold substrate through the central pore of their hexameric structures , but how this process occurs is not known . VAT , the Thermoplasma acidophilum homologue of eukaryotic CDC48/p97 , works in conjunction with the proteasome to degrade misfolded or damaged proteins . We show that in the presence of ATP , VAT with its regulatory N-terminal domains removed unfolds other VAT complexes as substrate . We captured images of this transient process by electron cryomicroscopy ( cryo-EM ) to reveal the structure of the substrate-bound intermediate . Substrate binding breaks the six-fold symmetry of the complex , allowing five of the six VAT subunits to constrict into a tight helix that grips an ~80 Å stretch of unfolded protein . The structure suggests a processive hand-over-hand unfolding mechanism , where each VAT subunit releases the substrate in turn before re-engaging further along the target protein , thereby unfolding it . Electron microscopy density maps are deposited in the Electron Microscopy Databank under accession codes EMD-8658 and EMD-8659 . Atomic models are deposited in the Protein Databank under accession codes 5VC7 and 5VCA . An expression vector for the T . acidophilum VAT gene with the N-terminal 181 residues deleted ( ΔN-VAT ) ( Huang et al . , 2016 ) was prepared using a pProEx expression plasmid ( Invitrogen , Carlsbad , CA ) . The construct included an N-terminal His6-tag and a tobacco etch virus ( TEV ) cleavage site between the His6-tag and the protein sequence . Expression and purification of ΔN-VAT were performed following the protocol described previously for full-length VAT ( Huang et al . , 2016 ) . Walker B ( catalytically inactivating ) mutations ( Werbeck et al . , 2008 ) in VAT were made in a pET-28b vector that encoded codon-optimized VAT with an N-terminal His6-tag and a tobacco etch virus ( TEV ) cleavage site between the His6-tag and the ΔN-VAT sequence . The mutant protein was purified in the same way as wild type VAT and ΔN-VAT ( Huang et al . , 2016 ) . To produce mixed complexes of wild type and catalytically inactivated VAT , different ratios of wild type and mutant protein were mixed and diluted to 15 μM ( protomer ) in unfolding buffer ( 50 mM HEPES pH 7 . 5 , 200 mM NaCl , 6 M guanidine hydrochloride ) , and then refolded into hexamers by fast dilution at a ratio of 1:20 ( v:v ) into a refolding buffer ( 50 mM HEPES pH 7 . 5 , 200 mM NaCl , 0 . 5 M arginine ) . After concentrating with a centrifuged concentrating device ( Amicon , Millipore , Etobicoke , Canada ) , refolded VAT mixtures were subject to size exclusion chromatography and the hexamer fractions were collected . Little or no aggregate or monomeric VAT was detected after refolding . The ADP- and ATPγS-bound forms of ΔN-VAT were prepared by incubation overnight at 25°C with apyrase ( New England Biolabs , Whitby , Canada; 0 . 1 U/mg of ΔN-VAT ) . Apyrase was then removed by size exclusion chromatography and nucleotide was added to 5 mM . Purified ΔN-VAT at ~20 mg/mL in buffer ( 50 mM HEPES pH 7 . 5 , 100 mM NaCl ) was incubated with 5 mM ATPγS ( 90% pure , Sigma-Aldrich ) for 5 min at room temperature before preparing cryo-EM grids . Immediately before grid freezing 0 . 05% ( w/v ) IGEPAL CA-630 ( Sigma-Aldrich ) was added to the protein solution to increase the number of particles adopting side views on the grid . Sample ( 2 . 5 μL ) was applied to nanofabricated holey gold grids ( Marr et al . , 2014; Meyerson et al . , 2014; Russo and Passmore , 2014 ) , with a hole size of ~1 μm and blotted using a modified FEI Vitribot for 5 s before plunge freezing in a liquid ethane/propane mixture ( ratio of ~3:2 ) held at liquid nitrogen temperature ( Tivol et al . , 2008 ) . Micrographs were acquired as movies with an FEI Tecnai F20 electron microscope operating at 200 kV and equipped with a Gatan K2 Summit direct detector device camera . Movies , consisting of 30 frames at two frames/s , were collected with defocuses ranging from 1 . 7 to 2 . 9 μm . Movies were obtained in counting mode with an exposure of 5 electrons/pixel/s , and a total exposure of 35 electrons/Å2 . Whole frame alignment of movies was performed with alignframes_lmbfgs ( Rubinstein and Brubaker , 2015 ) and the resulting averages of frames were used for contrast transfer function ( CTF ) determination with CTFFIND4 ( Rohou and Grigorieff , 2015 ) and automated particle image selection with RELION ( Scheres , 2015 ) . 851 particle images were selected manually and used to generate two 2D class averages that served as templates for automated particle selection in RELION . Automated particle selection was followed by manual de-selection of non-protein features and protein from fibrils ( Figure 1d ) . Particle coordinates were used to extract particle images in 256 × 256 pixel boxes from the unaligned movies , while performing individual particle movement correction and exposure weighting with alignparts_lmbfgs ( Rubinstein and Brubaker , 2015 ) . Magnification and CTF parameters were corrected for a previously measured 2% magnification anisotropy ( Zhao et al . , 2015b ) resulting in a calibrated pixel size of 1 . 45 Å . A total of 171 , 381 candidate particle images were subjected to two rounds of 2D classification , ignoring the CTF until the first peak , in RELION ( Scheres et al . , 2007 ) . Visual inspection and selection of 2D classes showing hexameric ΔN-VAT yielded 93 , 469 particle images , which were then transferred to the program cryoSPARC ( Punjani et al . , 2017 ) for ab initio 3D classification and refinement . Reference-free 3D classification runs using the stochastic gradient descent algorithm were performed multiple times with between two and six classes , before selecting two classes that yielded the highest quality maps . The best map for the stacked-ring conformation came from reference-free classification with two classes , while the best map for the substrate-engaged conformation came from reference-free classification with three classes . Only particle images that were confidently assigned to a class ( cryoSPARC class probability ≥0 . 9 ) were used in refinement . With this threshold , gold-standard refinement of 75 , 205 particle images from the six-fold symmetric class , applying C6 symmetry , yielded a map at a resolution of 3 . 9 Å as measured by Fourier shell correlation ( FSC ) . With the class probability threshold of 0 . 9 , gold-standard refinement of 13 , 238 particle images from the substrate-engaged class with no symmetry applied yielded a map at a resolution of 4 . 8 Å by FSC . FSC curves were calculated with masks applied to remove solvent regions from the background of the maps . Masks are prepared automatically in cryoSPARC by binarizing the map , extending the edges of the resulting shape by a set distance , and tapering the mask to 0 over a second fixed distance with a cosine edge . Similar to RELION ( Scheres , 2012 ) , cryoSPARC measures and corrects for the effects of masking by randomization of high-resolution phases as described previously ( Chen et al . , 2013 ) . The mask correction process introduces a dip in the corrected FSC at the resolution where randomization begins in the calculation . The final tight masking conditions were determined automatically by cryoSPARC by adjusting the extension and fading distances of the mask to optimize the reported resolution without introducing correlation from the mask . Most of the density for loops , β-sheets and α-helices in the 3 . 9 Å resolution map were of sufficient quality to allow model building . A homology model of VAT that had been previously flexibly fit into a lower-resolution map ( Huang et al . , 2016 ) was rigidly fit into the map as individual domains ( NBD1 and NBD2 ) with UCSF Chimera ( Goddard et al . , 2007 ) . Final models were built with successive rounds of manual model building in Coot ( Emsley et al . , 2010 ) and real space refinement in Phenix ( Afonine et al . , 2013 ) , and gave an EMringer score of 1 . 17 , which is slightly better than the typical score of 1 . 0 for a map at this resolution . For the substrate-bound complex , the resolution was insufficient to determine side chain orientations , and a refined structure from the symmetric class was fit into the density by molecular dynamics flexible fitting ( MDFF [Trabuco et al . , 2008] ) . Density for each protomer was segmented with UCSF Chimera ( Goddard et al . , 2007; Pintilie et al . , 2010 ) . Only backbone atoms were subject to the fitting force during simulation , and optimization of geometry and rotamers was performed with Phenix ( Afonine et al . , 2013 ) . The auto-degradation of ΔN-VAT was measured following the protocol of Sauer and coworkers ( Barthelme et al . , 2014 ) . Hexameric ΔN-VAT ( 0 . 3 μM ) was mixed with an ATP-regenerating system ( 20 U·mL-1 pyruvate kinase; 15 mM phosphoenolpyruvate ) in the presence or absence of 10 mM ATP , with or without 0 . 9 μM T . acidophilum 20S proteasome . The reactions were carried out at 45°C with a total reaction volume of 50 μL ( 50 mM HEPES pH 7 . 5 , 200 mM NaCl , 120 mM MgCl2 ) . Samples were withdrawn from the reaction at different times , quenched by heating in SDS-loading buffer , and subjected to SDS-PAGE . Intensities of the ΔN-VAT bands were analyzed with Image Lab 4 . 0 . 1 . Steady state ATP hydrolysis was measured with an enzyme-coupled assay described previously ( Nørby , 1988 ) . The reaction was carried out at a VAT concentration of 2 . 5 μM ( protomer ) , and included an ATP regeneration system that contained 2 . 5 mM phosphoenolpyruvate , 0 . 2 mM NADH , 50 μg/mL pyruvate kinase , 50 μg/mL lactate dehydrogenase , 3 mM ATP , 200 mM NaCl , 120 mM MgCl2 , and 50 mM HEPES ( pH 7 . 5 ) . VAT complexes were prepared as mixtures of wild type and inactivated subunits ( Walker B mutations ) where Glu291 ( NBD1 ) and Glu568 ( NBD2 ) were substituted with Gln , followed by the unfolding/refolding protocol described above . Hydrolysis of ATP was monitored in a UV-visible light spectrometer at 340 nm , 25°C , in the presence or absence of 30 µM GFP substrate with an 11 residue ssRA tag ( Barthelme et al . , 2014 ) . All measurements were done in triplicate . Similar results were obtained in the absence of substrate . The expected activity of VAT complexes prepared by mixing different ratios of wild type and mutant protomers can be calculated as described previously ( Werbeck et al . , 2008 ) . We define r = [VATE291Q , E568Q]/[VAT]T , which can be calculated from the concentration of mutant VAT protomers that are mixed with wild type , where VATE291Q , E568Q is the VAT protomer bearing the E291Q/E568Q mutations and [VAT]T = [VATE291Q , E568Q]+[VATwild type] . We assume that the unfolding/refolding process leads to equal probabilities of insertion of either a mutant or wild type protomer into a VAT hexamer , so that r is the probability of any protomer in the hexamer being a mutant . Under this assumption the fraction of VAT molecules with k mutant protomers is given by ( 1 ) P ( k ) = ( 6k ) rk ( 1−r ) 6−k where ( nk ) =n ! k ! ( n−k ) ! , with the concentration of functional ATP hydrolysis sites in the protomer given by 2 × ( 6-k ) ( two nucleotide sites per VAT ) . Assuming further that ATPase activity in the functional sites is not affected by the incorporation of inactive protomers until m are added , in which case all hydrolysis stops , one can write the fractional activity of a solution of VAT complexes as a function of r , as ( 2 ) A ( r ) =16∑j=0m−1 ( 6−j ) ( 6j ) rj ( 1−r ) 6−j where A ( r ) is activity normalized relative to the expected hydrolysis rate for fully wild type VAT hexamers at an identical concentration . Our structural model predicts that a single defective protomer is enough to eliminate ATPase activity in the hexamer . In this case m = 1 and A ( r ) = ( 1−r ) 6 . In contrast , in the case where there is no coupling between active and inactive protomers , A ( r ) =16∑j=06 ( 6−j ) ( 6j ) rj ( 1−r ) 6−j=∑j=06 ( 6j ) rj ( 1−r ) 6−j−16∑j=06j ( 6j ) rj ( 1−r ) 6−j ( 3 ) =1−r Figure 4c plots measured A ( r ) versus r , as obtained experimentally , showing that ATP hydrolysis rates decrease rapidly with r . Although the dependence of A on r is slightly less steep than expected for a model where a single defective protomer completely eliminates ATP hydrolysis ( see text ) the experimental data points decrease more steeply than predicted for m = 2 ( two defective protomers are required to eliminate hydrolysis ) . Unfolding activity was assayed by monitoring the loss of fluorescence of a substrate consisting of green fluorescent protein ( GFP ) with an 11-residue ssRA extension at its carboxy terminus ( Weber-Ban et al . , 1999 ) . Briefly , VAT at a concentration of 5 μM of hexamer was mixed with 1 μM GFPssRA and 1 . 6 μM GroEL ( D87K ) , which acts as a trap for unfolded GFPssRA , keeping it in an unfolded state . Reactions were carried out in the presence of 100 mM ATP , 50 mM HEPES ( pH 7 . 5 ) , 200 mM NaCl , and 120 mM MgCl2 . Fluorescence of GFPssRA was monitored in a fluorescence plate reader ( SpectraMax M5e ) with an excitation wavelength of 480 nm and emission wavelength of 520 nm . Initial slopes of the reaction were used to calculate the unfolding rate of GFPssRA . VAT complexes were prepared as mixtures of wild type and inactivated subunits ( Walker B mutations ) as described for the ATPase assay above .	Proteins perform many important jobs inside cells . Each protein is made of a chain of building blocks called amino acids that fold together to form precise shapes . Old , damaged or poorly constructed proteins tend to be harmful to cells so it is important that they are recycled . Ring-shaped enzymes called AAA+ unfoldases help to recycle these proteins by unfolding their amino acid chains . However , since these enzymes only interact with their target proteins for a short time it has been difficult to find out what happens when the target protein binds to the enzyme . AAA+ unfoldases use chemical energy provided by a molecule called ATP to drive protein unfolding . Ripstein et al . used a technique called electron cryomicroscopy to study a AAA+ unfoldase from a microbe called Thermoplasma acidophilum . The enzymes were supplied with a molecule called ATPγS , which is similar to ATP but releases its energy more slowly . Ripstein et al . used this slowed down reaction , as well as new computer algorithms to analyse the microscopy images , to examine the structure of the target bound to the enzyme . The experiments show that proteins move through the centre of the enzyme’s ring as they unfold . At any time the enzyme is attached to the amino acid chain of the target protein in several places , which allows it to pull the protein through the ring in a “hand-over-hand” motion . Energy from each ATP molecule drives the enzyme to release one contact with the amino acid chain and form a new contact slightly further along the chain . It remains unclear how AAA+ unfoldases identify and bind to the proteins they unfold and how they work alongside other recycling proteins inside cells . AAA+ unfoldases have an essential role in the biology of cells , and contribute to cancer and several other human diseases . Therefore , understanding how these proteins work may aid the development of new drugs to treat these diseases .	[ "Abstract", "Introduction", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2017	Structure of a AAA+ unfoldase in the process of unfolding substrate
For intracellular pathogens , residence in a vacuole provides a shelter against cytosolic host defense to the cost of limited access to nutrients . The human pathogen Chlamydia trachomatis grows in a glycogen-rich vacuole . How this large polymer accumulates there is unknown . We reveal that host glycogen stores shift to the vacuole through two pathways: bulk uptake from the cytoplasmic pool , and de novo synthesis . We provide evidence that bacterial glycogen metabolism enzymes are secreted into the vacuole lumen through type 3 secretion . Our data bring strong support to the following scenario: bacteria co-opt the host transporter SLC35D2 to import UDP-glucose into the vacuole , where it serves as substrate for de novo glycogen synthesis , through a remarkable adaptation of the bacterial glycogen synthase . Based on these findings we propose that parasitophorous vacuoles not only offer protection but also provide a microorganism-controlled metabolically active compartment essential for redirecting host resources to the pathogens . Many microorganisms develop inside eukaryotic cells , either free in the cytosol or enclosed in a vacuole ( Creasey and Isberg , 2014; Fredlund and Enninga , 2014 ) . Each microorganism adapts its intracellular metabolism to the nutrient supply of the host ( Abu Kwaik , 2015; Eisenreich et al . , 2010 ) . One recognized advantage of a vacuole is that it provides a shelter against cytosolic host defense ( Kumar and Valdivia , 2009 ) , to the cost of limited access to cytosolic nutrients . Acquisition of nutrients through this barrier is a crucial feature of host-microbe interaction . Chlamydiae are Gram-negative obligate intracellular bacteria found as symbionts and pathogens in a wide range of eukaryotes , including protists , invertebrates and vertebrates ( Horn , 2008 ) . The developmental cycle of Chlamydiae involves two morphologically distinct forms . Infectious particles , called elementary bodies ( EBs ) , are small and adapted to extracellular survival . After invasion of the host cell , they establish a parasitophorous vacuole called an inclusion , and convert within the first hours into larger organisms with higher metabolic activity , called reticulate bodies ( RBs ) . RBs replicate several times within the inclusion , until they differentiate back into EBs , in a non synchronous manner ( AbdelRahman and Belland , 2005 ) . The human adapted strain C . trachomatis is the leading cause of infectious blindness ( Taylor et al . , 2014 ) as well as of sexually transmitted infections caused by bacteria . Infections of the urogenital mucosae often stay asymptomatic causing irreparable damage leading to ectopic pregnancies or tubal factor infertility ( Brunham and Rey-Ladino , 2005 ) . C . trachomatis displays a genome reduced to around one million base pairs and therefore highly relies on the host with regard to several essential metabolic pathways , such as nucleotide or amino acid biosynthesis ( Stephens et al . , 1998 ) . Lipid droplets and peroxisomes have been observed in the inclusion lumen , indicating that this compartment is able to engulf large particles to meet bacterial needs ( Boncompain et al . , 2014; Kumar et al . , 2006 ) . C . trachomatis , and the closely related mouse and hamster pathogen C . muridarum , are unique amongst Chlamydiae for their ability to accumulate glucose ( Glc ) under the form of glycogen in the inclusion lumen ( Gordon and Quan , 1965 ) . Glc deprivation leads to a complete loss of infectivity ( Harper et al . , 2000; Iliffe-Lee and McClarty , 2000 ) . The C . trachomatis genome encodes for all the enzymes necessary for a functional glycogenesis and glycogenolysis ( Stephens et al . , 1998 ) ( Figure 1A ) . Although they do not accumulate glycogen in the inclusion , other Chlamydia species also have a complete set of those enzymes , a surprising observation given that this pathway is absent in most intracellular bacteria ( Henrissat et al . , 2002 ) . Intriguingly , pathogenic Chlamydiae lack a hexokinase , the enzyme phosphorylating Glc , and therefore rely on the import of phosphorylated sugars for glycolysis or glycogenesis . During bacterial glycogenesis , Glc-1-phosphate ( Glc1P ) serves as a substrate for the enzyme ADP-Glc pyrophosphorylase ( GlgC ) giving rise to ADP-Glc , which in turn is the building block for a linear chain of α1 , 4-linked Glc molecules . Branching of chains through α1 , 6-glucosidic bond is introduced by the branching enzyme GlgB , giving rise to glycogen . The glycogen phosphorylase GlgP , the debranching enzyme GlgX and the amylomaltase MalQ are involved in the degradation of the glycogen particle to Glc1P ( Colleoni et al . , 1999; Seibold and Eikmanns , 2007; Wilson et al . , 2010 ) . Finally , the very large majority of circulating C . trachomatis strains contain a 7 . 5 kb plasmid . Loss of this plasmid is associated with decreased GlgA expression and impaired glycogen accumulation ( Carlson et al . , 2008 ) . 10 . 7554/eLife . 12552 . 003Figure 1 . Glycogen accumulation in C . trachomatis inclusions . ( A ) Glycogen metabolism in bacteria . In green: glycogen synthesis . In blue: glycogen degradation . Glc1P is the substrate of GlgC for ADP-Glc synthesis . GlgA ( glycogen synthase ) produces linear glycogen chains via α-1 , 4 glycosidic bonds , and the branching enzyme ( GlgB ) introduces branches through α-1 , 6 linkages . Glycogen depolymerization in Glc1P is the result of the activity of GlgP ( glycogen phosphorylase ) , GlgX ( debranching enzyme ) and MalQ ( 4-α-glucanotransferase ) . The phosphoglucomutase MrsA converts Glc1P to Glc6P . The arrows point to the site within the polysaccharide that is subjected to enzymatic activity . Genes for all these enzymes are present in C . trachomatis . ( B ) HeLa cells were infected for 30 hr with C . trachomatis . Glycogen was visualized by TEM after PATAg stain . Grey arrows point at glycogen . Green arrows point at examples of EBs and red arrows at RBs . The picture on the bottom shows an enlargement of the boxed region containing three EBs . Scale bars: 1 µm ( top ) , 200 nm ( bottom ) . ( C , D ) Cells were Glc-deprived 48 hr prior to infection . 10 mg/ml Glc were added 24 hpi and cells were ( C ) fixed immediately or ( D ) 4 hr after Glc administration . Note that no glycogen is detectable in the bacteria while it is highly abundant in the inclusion lumen . Scale bar: 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 00310 . 7554/eLife . 12552 . 004Figure 1—figure supplement 1 . Relocation of glycogen stores to the inclusion during infection . Cells were ( A ) non-infected , or infected with C . trachomatis for ( B ) 24 hr or ( C ) 48 hr , fixed in PFA and processed for PAS stain . ( D ) Enlargement of the boxed region in ( B ) Note that glycogen particles ( white arrowheads ) are still detected in cells infected for 24 hr but not later , while inclusion glycogen content strongly increases with infection time . Black arrowheads point to examples of inclusions . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 00410 . 7554/eLife . 12552 . 005Figure 1—figure supplement 2 . Luminal and cytoplasmic glycogen differs in size . Cells were infected for 30 hr with C . trachomatis . The picture on the right shows an enlargement of the boxed region . Glycogen is visualized by TEM after PATAg stain . Glycogen deposits in the inclusion lumen ( arrowheads ) are on average of bigger size than in the host cell cytoplasm ( arrows ) . Scale bar: 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 00510 . 7554/eLife . 12552 . 006Figure 1—figure supplement 3 . Kinetics of glycogen accumulation . HeLa cells were infected with C . trachomatis for 8 , 16 , 20 , 24 or 48 hr . Glycogen is visualized by TEM after PATAg stain . White arrows point to inclusions . Glycogen is first detected in the inclusion lumen at 20 hpi , and increases strongly after . Scale bar: 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 006 It is widely believed that the abundance of polysaccharides in the inclusion reflects the accumulation of glycogen in the bacteria themselves . While luminal , extrabacterial , glycogen has been observed ( Chiappino et al . , 1995 ) , its origin has not been investigated . Here we show that luminal glycogen is in fact derived from bulk import of host glycogen and also de novo intraluminal synthesis , through the action of secreted bacterial enzymes . We reconstruct Glc metabolism in infected cells and demonstrate the ability for a microbe to convert its vacuole lumen into a compartment for regulated metabolic activity . To determine whether glycogen accumulated in the host cytosol , in the inclusion , or both , we labeled polysaccharides using periodic-acid-Schiff ( PAS ) staining at different times of infection . Large glycogen particles were detected in most non-infected cells ( Figure 1—figure supplement 1 ) . Twenty-four hours post infection ( hpi ) , glycogen was still detected in the cytoplasm of some of the infected cells , and the inclusions only showed weak PAS staining . However , 48 hpi no glycogen particle was detected in the cytoplasm of infected cells , while inclusions heavily stained with PAS , indicating that the global increase in glycogen content is accompanied by a shift in its original cytosolic localization , in favor of the bacterial inclusion . We used transmission electron microscopy ( TEM ) to determine more precisely its subcellular localization and detected the polysaccharide in two locations: in the inclusion lumen , and within EBs ( Figure 1B ) . Intraluminal glycogen deposits are physically larger than in the cytoplasm ( Figure 1—figure supplement 2 ) . We did not observe glycogen in RBs , in contrast to an earlier report ( Chiappino et al . , 1995 ) . In that publication , the presence of glycogen in the inclusion lumen was interpreted as the result of glycogen release from lysed bacteria . Considering the abundance of glycogen in the inclusion lumen relative to its amount in bacteria we considered this hypothesis unlikely . We tested it by depriving the cells of Glc for 48 hr before infecting them . Under these conditions , the inclusions contained no glycogen 24 hpi ( Figure 1C ) . Restoring Glc availability for 4 hr was sufficient to trigger the accumulation of glycogen in the inclusion lumen , but not in the bacteria ( Figure 1D ) . This experiment demonstrates that glycogen in the inclusion lumen does not result from the release of bacterial stores . The kinetics of glycogen appearance in the inclusion was carefully examined next by TEM . Luminal glycogen first appeared between 16 and 20 hpi ( Figure 1—figure supplement 3 ) , and was abundant 24 hpi . Thus , luminal glycogen deposit took place at a time when RBs largely predominate over EBs . This observation suggests that , while glycogen accumulation in the inclusion is most obvious at later stages of infection , when EBs accumulate , the process is initiated by RBs . Our conclusion raised an obvious question: how could a large polymer appear in the inclusion lumen ? Two mechanisms are conceivable: bulk translocation of host glycogen , or transport of monomeric substrates ( such as nucleotide-sugars or hexose phosphates ) across the inclusion membrane for de novo luminal polymerization . Gys1 , the host glycogen synthase , produces linear chains of Glc and is tightly bound to glycogen ( Luck , 1961; Stapleton et al . , 2010 ) . Gys1 staining with specific antibodies gave a patchy pattern in the cytoplasm , as expected since it mirrors the distribution of glycogen . In addition , we observed that the enzyme accumulated within the inclusion lumen . Staining was specific since it disappeared when Gys1 expression was knocked down using siRNA prior to infection ( Figure 2A ) . Observation of cytoplasmic markers in the inclusion can result from post-fixation artifact ( Kokes and Valdivia , 2015 ) . When we deprived the cells of Glc for two days before infection , and thus decreased cytoplasmic glycogen , Gys1 staining in the inclusion lumen was considerably reduced , while remaining abundant in the cytoplasm . This observation shows that Gys1 appearance in the inclusion lumen requires the presence of host glycogen , and is likely not the result of a post-fixation artifact . ( Figure 2—figure supplement1 ) . Altogether these data favored the scenario of bulk import of host glycogen and glycogen-bound Gys1 into the inclusion . TEM observations also came in support of this scenario , since we could observe one or more glycogen-filled vesicular structures in the inclusion lumen of about 20% of the glycogen-positive inclusions we examined ( Figure 2B ) . Since in some cases several membranes were observed around luminal glycogen , we hypothesized that cytoplasmic glycogen might be entrapped by an autophagy-dependent process in the host cytoplasm , before delivery to the inclusion lumen . To test this hypothesis we used an Atg5-/- mutant mouse embryonic fibroblast ( MEF ) cell line , which is deficient in autophagy ( Kuma et al . , 2004 ) . Inclusions of Atg5-/- MEFs still harboured glycogen , Gys1 and glycogen-filled vesicles , demonstrating that the pathway of bulk host glycogen uptake is independent of autophagy ( Figure 2—figure supplement 2 ) . 10 . 7554/eLife . 12552 . 007Figure 2 . Bulk import of cytoplasmic glycogen contributes to the accumulation of glycogen in the inclusion . ( A ) Gys1 is imported into the inclusion lumen . Cells were treated with siRNA control or against Gys1 prior to infection for 30 hr . DNA was stained in blue , Gys1 in green and the inclusion membrane in red ( anti-CT813 ) . The white arrowhead points to intraluminal Gys1 , see also xz ( top ) and yz ( right ) projections . Scale bar: 10 µm . ( B ) Examples of TEM images of glycogen-filled vesicles ( arrows ) in inclusions 30 hpi . Glycogen is visualized through PATAg staining . Scale bar: 500 nm . ( C , D ) Cells were treated with either siRNA control or siRNA against Gys1 48 hr prior to infection . ( C ) Coomassie staining of whole cell lysates as loading control ( left ) and immunoblot with an anti-Gys1 antibody ( right ) . The arrowhead points to Gys1 ( MW = 85 kDa ) . ( D ) PAS on non-infected cells fixed 48 hr after siRNA treatment ( top ) and on siRNA treated cells infected for 48 hr ( bottom ) . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 00710 . 7554/eLife . 12552 . 008Figure 2—figure supplement 1 . Luminal location of Gys1 strongly decreases in glucose-deprived cells . HeLa cells were Glc-deprived ( A ) or not ( B , same medium supplemented with 4 . 5 mg/ml Glc ) for 48 hr , prior to infection with C . trachomatis . Gys1 was stained in green and the inclusion membrane in red ( anti-CT813 ) . The white arrowheads point to intraluminal Gys1 . Scale bar: 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 00810 . 7554/eLife . 12552 . 009Figure 2—figure supplement 2 . Host glycogen imported in bulk from the host is not of autophagic origin . Wild-type ( WT ) or Atg5-/- mouse embryonic fibroblasts ( MEFs ) were infected for 30 hr with C . trachomatis . ( A ) PAS staining . Glycogen accumulation was identical in both cell lines . Scale bar: 10 µm . ( B ) DNA is stained in blue , Gys1 in green and the inclusion membrane in red ( anti-CT813 ) . While less abundant than in HeLa cells , Gys1 ( white arrows ) was detected inside inclusions of both cell lines . Scale bar: 10 µm . ( C ) Glycogen is visualized by TEM after PATAg stain . Glycogen-filled vesicles were observed in inclusions of both cell lines . Scale bar: 500 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 009 If bulk glycogen import is the only mechanism at work , depleting host glycogen stores should dramatically reduce intraluminal glycogen content . Cells were transfected with siRNA against Gys1 two days before infection . Silencing was efficient ( Figure 2C ) and there was little glycogen left in the host cytoplasm at the time of infection ( Figure 2D , top ) . However , knocking down Gys1 expression did not abolish glycogen accumulation in the inclusions ( Figure 2D , bottom ) . We measured the pixel intensity of the PAS staining in inclusions 48 hpi , and used inclusion staining of cells infected in the absence of Glc to determine the background level , which we subtracted . Inclusions of cells treated with siRNA against Gys1 contained 82% ( s . e . m . 4 . 9 ) of the glycogen measured in cells treated with control siRNA . This experiment strongly suggested that bulk import of host glycogen was at least partially responsible for intraluminal glycogen accumulation . However , it did not appear to be the main contributor and indicated that de novo glycogen synthesis in the inclusion lumen also took place . Synthesis of glycogen , and possibly its degradation into Glc monomers amenable to bacterial uptake , implies that glycogen synthesis and degradation enzymes are present in the inclusion lumen . While Gys1 import in the inclusion lumen might contribute , it cannot account for the glycogen accumulation observed in Gys1 depleted cells . We measured the expression profile of several proteins involved in glucose metabolism , together with control genes . For most proteins , RNA levels increased between 8 and 24 hpi , mirroring the known increase in bacterial metabolic activity in this time window ( Figure 3—figure supplement 1 ) . glgA stood out , as its transcription was delayed until 16 hpi . glgA expression thus correlates very well with the timing of glycogen appearance in the inclusion lumen , between 16 and 20 hpi , implicating that the enzyme might be involved in luminal glycogen synthesis . Indeed , secretion of GlgA into the inclusion lumen , as well as into the host cytoplasm , has recently been demonstrated ( Lu et al . , 2013 ) . In addition , a GlgB mutant strain shows massive precipitation of glycogen in the inclusion ( Nguyen and Valdivia , 2012 ) , due to accumulation of unbranched glycogen , implicating that GlgB normally functions in the inclusion lumen . These data strongly support the hypothesis that bacterial enzymes are at the origin of glycogen synthesis in the inclusion lumen . To determine if bacterial enzymes also take over glycogen depolymerization we obtained a specific antibody against the chlamydial debranching enzyme GlgX . We controlled the specificity of the staining by western blot and by immunofluorescence ( Figure 3—figure supplement 2 ) . In cells infected for 24 hpi or 48 hpi , GlgX was found in the inclusion lumen , with mostly no overlap with bacteria , demonstrating secretion within the inclusion lumen ( Figure 3A ) . Interestingly , GlgX was also detected on the inclusion membrane 24 hpi , but not 48 hpi ( Figure 3A and Figure 3—figure supplement 3 ) . Altogether , these data show that bacterial enzymes for glycogen synthesis and depolymerisation are present in the inclusion lumen , where they likely control glycogen polymerization/depolymerization . 10 . 7554/eLife . 12552 . 010Figure 3 . Bacterial glycogen metabolism enzymes are secreted in the inclusion lumen . ( A ) HeLa cells were infected for 24 hr or 48 hr with GFP expressing L2 ( GFP-Chlam , green ) and stained with an anti-GlgX antibody ( red ) and Hoechst ( blue ) . Insets to the right show enlargements of the boxed areas . Scale bar: 5 µm . ( B ) Heterologous test of secretion: the N-terminal 20 amino acids of the indicated proteins were fused to the reporter Cya , and constructs were transformed into the S . flexneri strains ipaB ( T3S+ ) and mxiD ( T3S- ) . Liquid cultures were fractionated into pellet ( P ) and supernatant ( S ) . All chimeras except GlgC/cya were detected in the supernatant in T3S competent bacteria and not in T3S defective bacteria , indicative of a functional T3S signal . The endogenous T3S substrate of S . flexneri , IpaD , serves as a positive control , while the non-secreted cAMP receptor protein ( CRP ) controls for non-specific leakage into the supernatant . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 01010 . 7554/eLife . 12552 . 011Figure 3—figure supplement 1 . Expression profiles of genes related to glycogen metabolism . qRT-PCR of selected genes related to glycogen metabolism was performed . Values were plotted against standard curves and cDNA was normalized with the overall chlamydial genomic DNA ( gDNA ) quantified in parallel samples ( cDNA/gDNA ) . An early chlamydial gene , euo , and two late genes , hctA and omcB , served as controls . Values were obtained from three independent experiments , each performed in triplicates . Error bars depict the standard deviation ( n=3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 01110 . 7554/eLife . 12552 . 012Figure 3—figure supplement 2 . Specificity of the anti-GlgX antibody . ( A ) Western Blot of HeLa cells non-infected ( nI ) or infected ( I ) with LGV for 24 hr . The anti-GlgX antibody detected a band of the expected molecular weight ( 73 kDa ) , only present in the infected sample . The lower band observed does not correspond to a bacterial antigen since it is also present in non-infected cells . ( B ) Cells were infected for 48 hr and processed for immunofluorescence as described in Figure 3 , except that the anti-GlgX antibody was preincubated with either unrelated control peptides , or with the peptides the antibody was raised against . DNA is in blue , bacteria in green and GlgX in red . The GlgX staining disappears upon preincubation with the GlgX peptides , demonstrating its specificity . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 01210 . 7554/eLife . 12552 . 013Figure 3—figure supplement 3 . GlgX accumulates at the inclusion membrane . HeLa cells were infected for 24 hr with LGV , fixed with 3% PFA and stained with the anti-GlgX antibody ( green ) , an antibody against the inclusion protein CT813 ( red ) and Hoechst ( blue ) . Insets to the right show enlargements of the boxed areas . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 013 How are these bacterial enzymes transported to the inclusion lumen ? None of them contain a signal peptide for transit to the periplasm and further transport by the type 2 or 5 secretion pathway ( SignalP 4 . 1 ) . Since type 3 secretion is the prominent pathway for protein secretion in Chlamydia ( Betts et al . , 2009 ) we looked for the presence of a type 3 secretion ( T3S ) signal ( Galan et al . , 2014 ) in the amino terminal domain of all glycogen metabolism enzymes . The N-termini of the proteins of interest were fused to a reporter ( calmodulin-dependent adenylate cyclase of Bordetella pertussis Cya ) , and the secretion of the resulting chimera was evaluated in Shigella flexneri , in an assay previously validated ( Subtil et al . , 2005; Subtil et al . , 2001 ) . Five out of the six chimera tested , GlgA/Cya , GlgB/Cya , GlgX/Cya , GlgP/Cya and MalQ/Cya , were secreted in the supernatant of cultures when transformed in a S . flexneri mutant with constitutive T3S ( ipaB strain ) , and not when transformed in a mutant deficient for T3S ( mxiD strain ) . The endogenous T3S substrate of S . flexneri , IpaD , was also secreted only in the ipaB background ( positive control ) , while cAMP receptor protein ( CRP ) , a non-secreted protein , was found exclusively in the bacterial pellet , excluding the possibility of non-specific leakage into the supernatant . The sixth chimera tested , GlgC/Cya , was not detected in the culture supernatant , suggesting that GlgC , which converts Glc1P into ADP-Glc , is not a substrate of T3S ( Figure 3B ) . Collectively , these data strongly support the second scenario we proposed: chlamydial enzymes for glycogen metabolism are secreted by a T3S dependent mechanism into the inclusion lumen , and account for the glycogen accumulation observed even in the absence of host glycogen . Secretion of the chlamydial glycogen degradation enzymes , GlgX , GlgP and MalQ , is consistent with a decrease in luminal glycogen content observed at late infection stages ( Gordon and Quan , 1965 ) , likely to feed EBs monomeric sugars ( Omsland et al . , 2012 ) . We investigated what form of Glc the bacteria are able to take up by incubating purified EBs with radioactively labeled [C14]-Glc , [C14]-Glc1P or [C14]-Glc6P , in absence or presence of a 50-fold excess of non-radioactive Glc , Glc1P or Glc6P . Only [C14]-Glc6P was taken up by bacteria , and only Glc6P could compete it out ( Figure 4A ) , demonstrating that the uptake was saturable , as expected for a transporter . There is only one annotated hexose phosphate transporter in C . trachomatis ( UhpC ) , which is likely responsible for Glc6P uptake , as the homologous protein in C . pneumoniae showed such activity ( Schwoppe et al . , 2002 ) . Therefore there is a discrepancy between the glycogen degradation product , Glc1P , and the substrate of import into the bacteria , Glc6P . To solve this conundrum we hypothesized that Glc1P might be converted into Glc6P in the inclusion lumen . C . trachomatis encodes for a phosphoglucomutase ( MrsA ) , an enzyme allowing the interconversion between Glc1P and Glc6P . Using the heterologous test of secretion we found that C . trachomatis MrsA contains a functional T3S signal ( Figure 4B ) . This suggests that Glc1P might be converted into Glc6P by MrsA inside the inclusion lumen , followed by uptake through UhpC . 10 . 7554/eLife . 12552 . 014Figure 4 . C . trachomatis takes up Glc6P and secretes phosphoglucomutase ( MrsA ) . ( A ) Purified EBs were incubated for 2 hr with [14C]-Glc , [14C]-Glc6P or [14C]-Glc1P in absence or presence of a 50-fold excess of non-radioactive Glc , Glc6P or Glc1P . Bacteria were subsequently washed and pelleted and radioactivity measured . Background ( Bg ) was measured after 1 min of incubation instead of 2 hr . CPM: counts per minute . Results show mean and S . D . ( n=3 ) . ( B ) Heterologous test of secretion was performed on MrsA as in Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 014 Our data strongly support the hypothesis that chlamydial glycogen synthase ( GlgA ) initiates de novo glycogen synthesis in the inclusion lumen . This raises the question of its substrate specificity . A sharp distinction between prokaryote and eukaryote glycogen synthases is that , almost without exception , bacterial glycogen synthases function on ADP-Glc while eukaryotic glycogen synthases use UDP-Glc as substrate . We reasoned that GlgA secretion into the host cytoplasm , which contains no ADP-Glc , might point to an unusual substrate specificity for this enzyme . Transfection of cells with flag-tagged chlamydial GlgA led to massive glycogen accumulation , proving that indeed C . trachomatis GlgA is functional on UDP-Glc ( Figure 5A ) . Interestingly , when transfected cells were subsequently infected , an increase in glycogen accumulation in the inclusion lumen was observed , and Flag-GlgA was detected in the bacterial compartment ( Figure 5—figure supplement 1 ) . High intraluminal glycogen content upon ectopic GlgA expression was also observed when cells were infected with a strain devoid of the natural plasmid of C . trachomatis ( Figure 5—figure supplement 1 ) , and Flag-GlgA was observed in the inclusion lumen also in that case ( not shown ) . Plasmid-loss is associated with decreased GlgA expression and impaired glycogen accumulation ( Carlson et al . , 2008 ) . In non-transfected cells the polysaccharide was still detected in EBs of the plasmid-less strain and , to a highly reduced level , in the inclusion lumen ( Figure 5—figure supplement 2 ) . Glycogen recovery upon GlgA transfection indicates that the low level of expression of GlgA in the plasmid-less strain accounts for the defect in glycogen storage . It is quite remarkable that a protein expressed by the host can compensate for a bacterial deficiency . 10 . 7554/eLife . 12552 . 015Figure 5 . Host UDP-Glc is the substrate for intraluminal glycogen synthesis . ( A ) PAS staining was performed 24 hr after transfection with chlamydial Flag-GlgA . ( B ) Zymogram analysis . Serial dilutions of lysates of E . coli lacking endogenous glgA and transformed with chlamydial glgA ( C . tr . ) or E . coli glgA ( E . coli ) were separated by native polyacrylamide electrophoresis and incubated in either 1 . 2 mM ADP-Glc , UDP-Glc or buffer only ( control ) . Glycogen production was visualized by iodine staining . ( C , D ) Cells were treated with either siRNA control or siRNA against UGP2 48 hr prior to infection . ( C ) Coomassie staining of whole cell lysates as loading control ( left ) and immunoblot with an anti-UGP2 antibody ( right ) . The arrowhead points to UGP2 ( MW = 56 kDa ) . ( D ) PAS on siRNA treated cells at time of infection ( top ) and at 48 hpi ( bottom ) . Knocking down UGP2 expression results in the decrease of luminal glycogen accumulation . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 01510 . 7554/eLife . 12552 . 016Figure 5—figure supplement 1 . Flag-GlgA is imported into the inclusion lumen and enhances luminal glycogen accumulation . ( A ) HeLa cells were transfected ( top right ) or not ( top left ) with Flag-GlgA before infection , and fixed 24 hpi . PAS staining revealed an increase of intraluminal glycogen ( arrowheads ) in Flag-GlgA expressing cells . C . trachomatis has a 8 kb plasmid , and its loss is associated with decreased GlgA expression and impaired glycogen accumulation ( Carlson et al . , 2008 ) ( bottom left ) . Remarkably , transfection of Flag-GlgA restored luminal glycogen accumulation in cells infected with the plasmid-less strain LGV 25667R ( bottom right ) . ( B ) Immunofluorescence on Flag-GlgA transfected cells infected with the wild-type LGV strain . DNA is stained in blue , Flag-GlgA in green and the inclusion membrane in red ( anti-Cap1 ) . Flag-GlgA is abundant in the host cytoplasm and the inclusion lumen ( see also xz ( top ) and yz ( right ) projections ) . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 01610 . 7554/eLife . 12552 . 017Figure 5—figure supplement 2 . The plasmid-less strain accumulates glycogen in EBs and - to a lesser extent than the wild-type strain - in the inclusion lumen . Cells were infected for 40 hr with the plasmid-less strain LGV 25667R . Glycogen is visualized by TEM after PATAg stain and is found in EBs ( black arrow heads ) and free in the inclusion lumen ( white arrow heads ) . Scale bar: 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 017 A zymogram analysis was performed to further compare chlamydial GlgA activity towards either UDP-Glc or ADP-Glc . Briefly , serial dilutions of lysates of E . coli lacking their endogenous glgA and transformed with either E . coli glgA or C . trachomatis glgA were separated on non-denaturing polyacrylamide gels that contained rabbit glycogen . The gels were subsequently incubated in buffer containing either UDP-Glc or ADP-Glc , and glycogen production was visualized by iodine staining . While E . coli GlgA showed activity exclusively upon incubation with ADP-Glc , as expected , the chlamydial GlgA showed activity with both substrates ( Figure 5B ) . We next asked which Glc derivative might be translocated across the inclusion membrane for de novo glycogen synthesis . UDP-Glc stood as the best candidate: it is a substrate for GlgA , and its import would relieve the bacteria from the cost of nucleotide-sugar synthesis . In addition , GlgC ( converting Glc1P into ADP-Glc ) was the only glycogen metabolism enzyme that failed the Shigella T3S test , suggesting that this enzyme may not be present within the inclusion lumen . To test the requirement for UDP-Glc for luminal glycogen synthesis we silenced the expression of the host UDP-Glc pyrophosphorylase UGP2 ( UGP2 catalyzes the conversion of Glc1P into UDP-Glc ) by siRNA for two days before infection . We measured the pixel intensity of the PAS staining in inclusions as described above . Glycogen accumulation within the inclusion lumen decreased to 57% ( s . e . m . = 5 . 8 ) in cells treated with a siRNA against UGP2 compared to cells treated with control siRNA ( Figure 5C , D ) . These data strongly argue for UDP-Glc being the substrate for sugar import into the inclusion . Note that while UGP2 silencing also impacted host glycogen stores , as expected , it did so to a lesser extent than Gys1 silencing ( Figure 2D ) . Thus the greater impact of UGP2 silencing on luminal glycogen stores cannot be explained by its indirect effect on bulk glycogen import . Instead it very likely reflects the requirement for UDP-Glc for de novo luminal glycogen synthesis . We have demonstrated that de novo glycogen synthesis takes place in the inclusion lumen , initiated by chlamydial glycogen synthase GlgA , using UDP-Glc as substrate . This implies that UDP-Glc is translocated from the host cytoplasm into the inclusion lumen . Among the large family of annotated human sugar transporters , only SLC35D2 was experimentally shown to be able to transport UDP-Glc ( Suda et al . , 2004 ) . In cells infected with C . trachomatis , a construct of the transporter bearing a C-terminal HA-tag was enriched around the inclusion ( Figure 6A ) . By TEM , we detected the presence of the transporter on the inclusion membrane ( Figure 6B ) . Staining was specific since we observed 0 . 96 gold particles per μm of inclusion membrane examined , against 0 . 12 gold particles per μm in non-transfected cells . The ability for SLC35D2 to reach the inclusion membrane indicates that it might play a role in UDP-Glc import . To test this hypothesis we silenced SLC35D2 expression using siRNA prior to infection . In these conditions , glycogen accumulation in the inclusion decreased to 43% ( s . e . m . = 1 . 7 ) ( Figure 6C , D ) , thus to a level similar to that when UGP2 had been silenced ( Figure 5C , D ) . These data strongly suggest that C . trachomatis co-opts the host nucleotide-sugar transporter to favor UDP-Glc import into the inclusion lumen . Finally we wondered if by simultaneously impairing bulk glycogen import and UDP-Glc translocation we could further decrease luminal glycogen content . We depleted Gys1 and SLC35D2 for 48 hr before infecting the cells , and applied PAS staining 48 hr after infection . Additional Gys1 depletion did not decrease luminal glycogen content further than what SLC35D2 depletion did , confirming that in C . trachomatis infected cells bulk import of cytoplasmic glycogen plays only a minor part in intraluminal glycogen accumulation ( Figure 6—figure supplement 1A ) . We measured the consequences on progeny 30 hpi to minimize the effect of SLC35D2 depletion on cell viability that we observed in some experiments . Decrease in progeny was never more than two-fold , indicating that even when both Gys1 and SLC35D2 were depleted simultaneously , there was no severe impact on bacterial development ( Figure 6—figure supplement 1B ) . 10 . 7554/eLife . 12552 . 018Figure 6 . SLC35D2 imports UDP-Glc into the inclusion lumen . ( A , B ) Prior to infection cells were transfected with SLC35D2-HA , and fixed 24 hpi . ( A ) Labelling of the HA tag ( green ) , inclusion membrane marker Cap1 ( red ) and DNA ( blue ) show recruitment of SLC35D2 to the inclusion periphery . Scale bar: 10 µm . ( B ) Duplicate samples were processed for TEM . Staining with anti-HA antibodies , followed with gold-coupled secondary antibodies , showed that SLC35D2-HA is detected on the inclusion membrane ( arrows ) . Scale bar: 250 nm . ( C ) Cells were treated with siRNA control ( ctrl ) or siRNA SLC35D2 48 hr and 4 hr prior to infection . Samples were taken before infection ( 0 hpi ) and 48 hpi , and RT-PCR was performed with primers specific to SLC35D2 . ( D ) PAS staining of siRNA ctrl or siRNA SLC35D2 treated cells 48 hpi . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 01810 . 7554/eLife . 12552 . 019Figure 6—figure supplement 1 . Depletion of Gys1 does not further decrease luminal glycogen content in cells depleted for SLC35D2 . Cells were treated with control siRNA , or siRNA against SLC35D2 , or SLC35D2 and Gys1 48 hr and 4 hr prior to infection . ( A ) Cells were fixed 48 hr after infection and PAS staining was applied . Mean intensity of glycogen staining in inclusions is expressed as a percentage of the mean value measured in cells treated with control siRNA . The median of three independent experiments is shown . ( B ) Reinfection assays assessing the effect of the indicated knockdowns on infectious progeny formation 30 hpi . Results are expressed as the percentage of progeny measured in cells treated with control siRNA . The median of three independent experiments is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 019 With the long-term goal of understanding the requirements for glycogen metabolism enzymes in a mouse model of infection we screened a mutant library of Chlamydia muridarum for strains with defects in glycogen accumulation . Most of the mutants obtained have a mutation in one of the enzymes for glycogen metabolism ( Giebel et al , in preparation ) . We selected two clones , P2B10 and P3B4 , which carry mutations in conserved residues of GlgA ( H131Y for P2B10 and G386E for P3B4 ) , suggesting that they may have a strong , if not complete , loss of function ( Figure 7—figure supplement 1 ) . Glycogen accumulation in P2B10 and P3B4 inclusions 48 hpi was reduced to 40 and 51% respectively of its level in the parental strain ( Figure 7A ) . Impact on infectivity was assessed by collecting the bacteria 48 hpi , and re-infecting fresh HeLa cells . Progeny was reduced 7-fold and 3-fold in P2B10 and P3B4 respectively , demonstrating that a defect in GlgA results in an impaired infectivity in vitro ( Figure 7B ) . However , it is impossible to discriminate if the deficiency is related to the defect in luminal storage of glycogen , or in glycogen synthesis in the bacteria , or both . We hypothesized that import of host Gys1 may partially compensate for the deficiency in secreted GlgA . In that case , depletion of Gys1 might affect luminal glycogen stores , and possibly decrease infectivity even further . In the parental strain , depletion of Gys1 reduced glycogen accumulation by 29% , similar to C . trachomatis reduction ( Figure 7A ) . In the mutants , depletion of Gys1 also decreased glycogen accumulation , demonstrating that part of the residual glycogen observed in the mutants is due to bulk glycogen import . The fact that glycogen stores were not fully depleted suggests that the mutants we selected show reduced GlgA activity rather than complete loss of function . Interestingly , decreasing luminal glycogen further by silencing Gys1 expression content did not have a higher impact on the progeny of the mutants than it had on the parental strain , rather less so ( Figure 7B ) . In particular , Gys1 depletion did not decrease progeny of the P2B10 mutant . This observation suggests that in this mutant , the defect in progeny is mostly due to a deficiency in bacterial glycogen synthesis rather than in luminal glycogen storage . 10 . 7554/eLife . 12552 . 020Figure 7 . Mutations in GlgA result in defect in glycogen storage capacity and loss of infectivity . Cells were treated with either siRNA control or siRNA against Gys1 48 hr and 4 hr prior to infection with glgA mutants P2B10 , P3B4 or with the parental wild-type strain . ( A ) PAS staining was performed 48 hpi . Pictures show representative fields for each strain . Mean intensity of glycogen staining in inclusions was quantified and expressed as a percentage of the mean value measured in cells treated with control siRNA and infected with the parental strain . The median of three independent experiments is shown . Mutations in GlgA resulted in reduced glycogen stores , which were decreased further upon depletion of Gys1 . Scale bar: 10 µm . ( B ) Progeny was determined 48 hpi in a reinfection assay . Results are expressed as the percentage of progeny measured in cells treated with control siRNA and infected with the parental strain . The median of three independent experiments is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 02010 . 7554/eLife . 12552 . 021Figure 7—figure supplement 1 . Alignment of GlgA sequences of different bacteria . Alignments were obtained using Clustal W . An asterisk indicates positions which have a single , fully conserved residue . A colon points to conservation between groups of strongly similar properties and the period indicates conservation between groups of weakly similar properties . The amino acids mutated in the strains used in this report are highlighted in grey ( H131Y for P2B10 and G386E for P3B4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 021 Altogether , these experiments confirm that glycogen accumulation in the inclusion lumen results from the activity of both host and bacterial glycogen synthases , the latter making the largest contribution . In addition , they demonstrate that impaired GlgA activity has a significant impact on bacterial infectivity . This work shows that two independent processes contribute to glycogen accumulation in the C . trachomatis inclusion lumen: bulk uptake from the host cytoplasm and de novo synthesis ( Figure 8 ) . It is a unique example of a bacterium utilizing the compartmentalization of eukaryotic cells , to the extent that energy stores are radically shifted toward the bacterium and made inaccessible to the host . 10 . 7554/eLife . 12552 . 022Figure 8 . Glc flux in C . trachomatis infected cells . Early during the infectious cycle ( left ) the inclusion contains mostly RBs . This developmental form does not accumulate glycogen and uses ATP rather than Glc6P ( Omsland et al . , 2012 ) . SLC35D2 , and possibly other transporters , are recruited to the inclusion membrane and UDP-Glc is translocated into the inclusion lumen . The activity of chlamydial glycogen metabolism enzymes , secreted by RBs into the inclusion lumen , leads to the onset of luminal glycogen synthesis between 16 and 20 hpi . In addition , host glycogen is imported into the inclusion lumen through invagination of the inclusion membrane . In culture cells de novo synthesis predominates over bulk glycogen import . Later on ( right ) , RBs start converting into EBs , which rely on Glc6P as energy source . EBs obtain Glc6P via the degradation of luminal glycogen into Glc1P , subsequently converted to Glc6P by the phosphoglucomutase ( MrsA ) and imported by UhpC . During RB to EB conversion T3S is turned off , allowing for intrabacterial activity of the glycogen metabolism enzymes , and glycogen accumulation in the bacteria . DOI: http://dx . doi . org/10 . 7554/eLife . 12552 . 022 Intraluminal glycogen accumulation had previously been thought to be due to bacterial lysis and to the release of chlamydial glycogen into the surrounding environment ( Chiappino et al . , 1995 ) . In the present work we observed that glycogen appears first in the inclusion lumen and only later in EBs . Thus , bacterial lysis is not the source of luminal glycogen . We identified two mechanisms by which glycogen accumulates in the inclusion lumen . First , the host enzyme Gys1 , which is known to bind to glycogen ( Roach et al . , 2012 ) , is translocated into the vacuole , arguing for uptake of host glycogen in bulk . The observation of glycogen-filled vesicles in the inclusion lumen also strongly argues for this scenario . Finally , depleting cytoplasmic glycogen by silencing Gys1 partially reduced luminal glycogen content in C . trachomatis and C . muridarum . Translocation of cytoplasmic glycogen confirms the ability for the inclusion to take up large particles from the host cytoplasm , already illustrated with the uptake of lipid droplets and peroxisomes ( Boncompain et al . , 2014; Kumar et al . , 2006 ) . This likely occurs through invagination of the inclusion membrane , but the underlying mechanisms remains to be investigated . While multi-layered glycogen-filled vesicles were observed , they still occurred in Atg5 deficient cells , implicating that they do not originate from an Atg5-dependent autophagic process . The observation that depletion of host glycogen ( through silencing of the host glycogen synthase Gys1 ) reduces luminal glycogen stores by no more than 20% demonstrates that bulk uptake is a minor pathway for glycogen accumulation in the inclusion lumen . It remains to be determined if this bulk import through a vesicular mechanism is glycogen-specific , or a general pathway used for the acquisition of a wide range of host components . We identified the second , and main , mechanism for luminal glycogen accumulation as de novo synthesis , through the action of chlamydial enzymes . Using a heterologous secretion assay , we have identified T3S signals in all but one ( GlgC ) glycogen metabolism enzymes of C . trachomatis . We had previously demonstrated the robustness of this assay , with an estimated 5% of false positives ( Subtil et al . , 2005 ) . Secretion of GlgA in the inclusion lumen and in the host cytoplasm was reported recently ( Lu et al . , 2013 ) . GlgA mutants of C . muridarum accumulated less glycogen in the inclusion lumen than their parental strain , demonstrating further that GlgA was indeed present and active in the inclusion lumen . Evidence for the secretion of the branching enzyme GlgB was brought by the observation that a GlgB mutant strain shows precipitation of unbranched glycogen in the inclusion ( Nguyen and Valdivia , 2012 ) . Finally in this paper , we show the secretion of GlgX using specific antibodies . Secretion of GlgX , and of other enzymes involved in glycogen metabolism ( except GlgA ) , was not detected by an earlier study using mouse antisera ( Lu et al . , 2013 ) . This study focused primarily on the detection of bacterial proteins in the host cytoplasm . Detection of the intraluminal pool , not overlapping with bacteria , might have been hindered by the signal coming from the pool of intra-bacterial enzymes . Finally , GlgX localizes not only to the inclusion lumen , but seems to be enriched at the inclusion membrane at 24 hpi . Interestingly , GlgP localization at the inclusion membrane was also reported ( Saka et al . , 2011 ) , indicating that the two enzymes might work in concert , degrading host glycogen in the vicinity of the inclusion membrane . Thus , altogether these data indicate that C . trachomatis uses T3S to transform the lumen of the inclusion into a glycogen storage compartment . So far , T3S is mostly described as a mechanism for protein translocation across a eukaryotic membrane , either the plasma membrane or the membrane of a parasitophorous vacuole ( Galan et al . , 2014 ) . The ability for chlamydial glycogen enzymes to reach the inclusion lumen is intriguing . Loose membrane-like structures can frequently be seen in the inclusion lumen , and might trigger the secretion of some effectors into the inclusion lumen , especially from bacteria not in contact with the inclusion membrane . Also , it has been demonstrated , in Yersinia pseudotuberculosis , that T3S substrates can translocate first on the surface of the bacteria , without previous membrane contact ( Akopyan et al . , 2011 ) . Alternatively , it may be that glycogen enzymes are first translocated into the cytoplasm , and transported back to the inclusion lumen directly or indirectly associated to glycogen , like Gys1 . Clearly , more work is needed to understand how T3S is regulated in Chlamydia , and what determines substrate translocation across the inclusion membrane . Our transcription analysis , in agreement with published data ( Belland et al . , 2003 ) , indicates that glgA transcription most likely controls the onset of glycogen synthesis , as its initiation coincides with the onset of intraluminal glycogen accumulation ( between 16 and 20 hpi ) . For all the other glycogen enzymes , transcription correlated with the known increase in bacterial metabolic activity between 8 and 24 hpi . The only other exception is glgB , which is one of the few early genes . It is not clear at this stage why the protein is made long before GlgA is present , when GlgA lies upstream of GlgB in the glycogen synthesis pathway . glgB is not in an operon , and its promoter region lies within a coding sequence , limiting the possibility of mutations in this region ( Albrecht et al . , 2011 ) . Early expression of glgB might thus be a remnant of a different regulation of expression of the glycogen metabolism enzymes during the evolution of Chlamydiales . To test the need for bacterial glycogen synthase activity in vitro we obtained two independent mutants in conserved residues of GlgA . Because of the mutagenesis strategy used , the mutants carried additional mutations ( the library has an average of 16 mutations/genome ) . However , their similar phenotypes in respect to glycogen storage and infectivity strongly support the hypothesis that those are due to the GlgA mutation and not to additional genetic defects . These mutants demonstrated that defects in GlgA activity result in a loss of infectivity in vitro . We have not yet been able to obtain a stable strain knocked-out for glgA expression , but this might be due to technical difficulties rather than essentiality of the gene , and the question as to whether glycogen synthesis is required for chlamydial development remains open . The very few C . trachomatis isolates that do not have the plasmid accumulate only minor amounts of glycogen compared to the wild-type strain , and GlgA expression is strongly reduced in these strains ( Carlson et al . , 2008 ) . When we ectopically expressed Flag-GlgA in cells infected with either the wild-type or the plasmid-less strain , we observed increased intraluminal glycogen accumulation for both strains . These data show that different levels of GlgA expression between the two strains account for their difference in terms of glycogen accumulation . When Flag-GlgA import into the inclusion took place , likely together with bulk glycogen uptake , the low level of endogenous GlgA in the plasmid-less inclusions was complemented , allowing for high glycogen accumulation . Presumably , in the plasmid-less strain , bulk import of host glycogen occurs normally , but remains below detection . The fact that this strain has no infectivity defect in vitro ( Carlson et al . , 2008 ) indicates that high luminal glycogen accumulation is not needed for chlamydial growth in vitro . Still , the observation that Gys1 depletion decreased progeny by about 30% in the wild-type C . muridarum strain suggests that bacterial development is somewhat sensitive to the amount of glycogen in the inclusion lumen , maybe at early times of infection , before glgA is highly expressed and glycogen accumulation peaks . Alternatively , disruption of the glycogen storage capacity in the host might have indirect effects on nutrient balance , and account for the small decrease in infectivity . In vitro polymerization assays using E . coli expressing C . trachomatis GlgA , as well as transfection experiments in a eukaryotic background where only UDP-Glc is available , proved that C . trachomatis GlgA can use UDP-Glc as substrate . This was unexpected , because eubacterial glycogen synthases normally use ADP-Glc . Consistent with this finding , we observed a strong decrease in intraluminal glycogen accumulation when we knocked down the human enzyme responsible for the generation of UDP-Glc ( UGP2 ) . Decrease of host glycogen levels ( and thus the reduction in the bulk import pathway ) could not account for this result , as the Gys1 knockdown was a lot more efficient in depleting glycogen in the host cytoplasm than the UGP2 knockdown . This experiment points to UDP-Glc as being the substrate imported into the inclusion lumen . If , as we had initially hypothesized , Glc6P or Glc1P were that substrate , a UGP2 knockdown should not produce any difference in intraluminal glycogen accumulation because both Glc6P and Glc1P would remain available , as they lie upstream of UDP-Glc generation . Energetically speaking , it is beneficial to import UDP-Glc rather than Glc6P or Glc1P , as it relieves the bacteria from the costly reaction of transferring a nucleotide to the sugar molecule . It also fits perfectly with the absence of a T3S signal in GlgC , suggesting that this enzyme , which makes ADP-Glc out of Glc1P , remains restricted to the bacteria and only serves bacterial glycogen production . Secretion of GlgC into the inclusion lumen would be superfluous , with GlgA being able to produce unbranched glycogen out of UDP-Glc . The identification of UDP-Glc as the sugar imported into the inclusion lumen was further comforted by our observation that knocking down the UDP-Glc transporter SLC35D2 led to a significant reduction of intraluminal glycogen staining . In addition , we observed that SLC35D2-HA accumulates at the periphery of the inclusion . Our data thus strongly indicate that SLC35D2 is at least partially responsible for the import of UDP-Glc . The fact that the reduction was only partial , and slightly less pronounced than after UGP2 silencing , might be due to residual SLC35D2 expression . In addition , other Golgi- or ER-based transporters from the SLC35-family of nucleotide-sugar transporters might transport UDP-Glc and be recruited to the inclusion membrane , accounting for the residual glycogen accumulation in the SLC35D2 knockdown . We demonstrated that de novo glycogen synthesis takes place in the inclusion lumen , triggered by the presence of GlgA and GlgB . Glycogen storage in the inclusion lumen would only be of benefit for the bacteria if they were subsequently degraded into monomers amenable to bacterial uptake . The early observation that glycogen decreases at very late infection times , is consistent with a late consumption of the stores ( Gordon and Quan , 1965 ) . Similarly to the glycogen synthesizing enzymes , the degrading enzymes GlgP , GlgX and MalQ possess T3S signals , and we demonstrated GlgX secretion using specific antibodies . We therefore hypothesize that these enzymes are active in the inclusion lumen , and generate free Glc1P . We could clearly demonstrate that Chlamydiae are not able to take up Glc nor Glc1P , but exclusively Glc6P . This is in agreement with data obtained on the homologous protein in C . pneumoniae , which transports Glc6P and not Glc1P ( Schwoppe et al . , 2002 ) . This apparent contradiction can be explained by the fact that the chlamydial phosphoglucomutase MrsA ( interconverting Glc1P and Glc6P ) also possesses a T3S signal , and is thus most likely secreted . We have attempted to demonstrate MrsA secretion using newly developed tools to transform C . trachomatis ( Wang et al . , 2011 ) . The C-terminal tagged MrsA was not expressed in transformed bacteria , neither with the endogenous promoter nor with a tetracyclin-inducible promoter ( not shown ) . Thus , while we propose that intraluminal glycogen is degraded into Glc1P and is converted to Glc6P in the inclusion lumen , the involvement of bacterial MrsA in this conversion remains to be confirmed . Altogether , our work reveals the origin of glycogen in the inclusion lumen and brings to light the complexity of the Glc flux in C . trachomatis infected cells ( Figure 8 , see legend for details ) . At a first glance , this complexity seems counter intuitive: why not import Glc6P directly from the host cell cytoplasm into the inclusion lumen and then directly into the bacteria ? UDP-Glc import is initiated very early on , at a time when most bacteria are RBs , which , importantly , use ATP as an energy source , while EBs use Glc6P ( Omsland et al . , 2012 ) . Uptake of Glc6P , before EBs appear , would dangerously increase the osmotic pressure in the inclusion . Stocking Glc in the shape of the osmotically inert glycogen brings an elegant solution . Why do C . trachomatis , amongst all the Chlamydia species , accumulate these high amounts of glycogen within their parasitophorous vacuole ? Clearly , as discussed earlier , glycogen accumulation is not necessary for growth in vitro . Plasmid-less strains , that do not accumulate glycogen , were highly attenuated in a genital model of mouse infection and in ocular infection of non-human primates ( Carlson et al . , 2008; Russell et al . , 2011 ) . To date it is not known whether the lack of glycogen accumulation plays a role in this reduction in infectivity , or other traits associated with the absence of plasmid . Surprisingly , it was recently shown that the plasmid-less strain induced similar level of infection and pathologies as the wild-type strain in a model of genital infection in non-human primates ( Qu et al . , 2015 ) . Thus , while , the ability for virtually all C . trachomatis clinical isolates to accumulate glycogen speaks for a selective advantage of doing so , we can only speculate on what this advantage might be . Many bacterial strains of the microflora of the female genital tract metabolize Glc as carbon source . Whether Glc can become limiting , especially during bacterial vaginosis , where the risk of becoming infected with C . trachomatis is elevated , is not known . It is an attractive hypothesis to explain why the ability for C . trachomatis to store Glc could come to an advantage in such specific conditions , which the macaque model would not reveal . One independent advantage of re-routing the host energy stores towards the inclusion lumen is that it might overall 'weaken' the host cell in its fight against the intruder . Indeed , many autonomous host cell defense mechanism rely on processes that require energy ( including host protein synthesis or phosphorylation cascades ) and that might be compromised in cells enduring sustained hijacking of its energy stores . This work demonstrates that a microbe can deplete the host from its energy stores , making them available only to themselves . We propose that such sequestration of host molecular complexes , or even organelles , inside the parasitophorous vacuole , could be broadly used by other intracellular parasites , to ensure nutrient access and/or to disrupt defensive signalling pathways . HeLa cells ( ATCC ) , wild-type and Atg5-/- mouse embryonic fibroblasts ( MEF ) ( generous gift from N . Mizushima , Tokyo Medical and Dental University ) were cultured in Dulbecco's modified Eagle's medium with Glutamax ( DMEM , Invitrogen , Carlsbad , CA ) , supplemented with 10% ( v/v ) fetal bovine serum ( FBS ) . Cells were routinely checked for absence of mycoplasma contamination by PCR . For experiments with medium containing different Glc concentrations , DMEM without Glc ( DMEM , Invitrogen ) was used and complemented with 5 mM sodium pyruvate ( Sigma-Aldrich ) , 10% fetal bovine serum ( FBS ) and the indicated Glc concentration ( Merck ) . C . trachomatis LGV serovar L2 strain 434 ( ATCC ) , the plasmid-less strain LGV L2 25667R ( Bowie , 1990 ) or GFP-expressing L2 ( L2IncDGFP ) ( Vromman et al . , 2014 ) were purified on density gradients as previously described ( Scidmore , 2005 ) . Chlamydia muridarum was a kind gift from Dr H Caldwell ( Rocky Mountain Laboratories , NIAID , Hamilton , MT ) . McCoy mouse fibroblast cells ( ATCC ) , and Vero monkey kidney cells ( ATCC ) were maintained and propagated as detailed in ( Rajaram et al . , 2015 ) . The GlgB mutant was kindly provided by Dr R Valdivia ( Nguyen and Valdivia , 2012 ) . The ipaB and mxiD strains are derivates of M90T , the virulent wild-type strain of Shigella flexneri , in which the respective genes ( ipaB and mxiD ) have been inactivated ( Allaoui et al . , 1993 ) . The Escherichia coli strain DH5α was used for cloning purposes and plasmid amplification . Both S . flexneri and E . coli strains were grown in Luria-Bertani medium with ampicillin at 0 . 1 mg/ml . HeLa cells were grown in wells , infected with C . trachomatis LGV serovar L2 strain 434 at an MOI of 0 . 1 and carefully trypsinized at the indicated time points . The cells were then washed with PBS once and fixed with 0 . 1 M cacodylate and 2 . 5% glutaraldehyde at room temperature for at least 30 min . PATAg staining was performed as described elsewhere ( Thiéry , 1967 ) . Briefly , thin sections were incubated in 1% periodic acid for 25 min and then washed several times in water , followed by an incubation step in 0 . 2% thiocarbohydrazide in 20% acetic acid for 45 min . Several washing steps in a graded acetic acid series to water were carried out and the thin sections were stained with 1% silver proteinate for 30 min . Samples were observed within a week after preparation . Images were obtained using a Tecnai T12 transmission electron microscope ( FEI ) at 100 kV and captured using an Eagle CCD camera ( FEI ) . Transfected HeLa cells were fixed for 15 min at room temperature in 8% paraformaldehyde ( PFA ) /0 . 2% glutaraldehyde in PHEM buffer ( 60 mM Pipes , 25 mM Hepes , 10 mM EGTA and 2 mM MgCl2 , pH 7 . 2 ) . Cells were carefully scraped off , centrifuged and fixation was allowed to continue for 45 min at room temperature with fresh fixative buffer before storage at 4°C . Then cells were subsequently washed three times with PHEM buffer and embedded in 12% gelatine in PHEM buffer at 37°C . Samples were centrifuged for 2 min at 20000 g and gelatine was solidified on ice for 20 min . Cubes ( 1 mm3 ) were excised and impregnated overnight with 2 . 3 M sucrose at 4°C ( Tokuyasu , 1973 ) . Ultrathin sections ( 50–60 nm ) were cut with a cryoultramicrotome ( UC6/FC6; Leica Microsystems , Vienna , Austria ) . Those sections were collected with 0 . 575 M sucrose/0 . 05% ( w/v ) methylcellulose in PHEM buffer and dipped on to formvar/carbon-coated Ni grids . Immunolabelling was performed with mounted sections which were washed three times for 15 min with 25 mM NH4Cl in PBS . The sections were blocked in 5% milk powder in PBS and subsequently incubated with a rat anti–HA antibody ( clone 3F10 Roche Diagnostics ) diluted at 1/50 in PBS-milk powder 5% for 30 min . The labelled sections were washed 6 times for 3 min in PBS-milk powder 0 . 5% and incubated for 30 min with a rabbit anti-rat IgG ( Dako Denmark A/S , Glostrup , Denmark ) diluted 1/200 in PBS-milk powder 0 . 5% . Sections were washed 6 times for 3 min with PBS milk-powder 0 . 5% and then incubated for 30 min with Protein-A gold-10 nm , diluted at 1/50 in PBS/0 . 5% BSA . The sections were then rinsed briefly 3 times and then 7 times for 3 min with PBS . The labelled sections were fixed for 5 min in PBS/1% glutaraldehyde and subsequently washed 8 times with distilled water . The sections were then washed twice with 0 . 4% aqueous uranyl acetate/1 . 8% ( w/v ) methylcellulose on ice and the incubation was continued with fresh solutions for 5 min on ice . Sections were observed under a Tecnai T12 transmission electron microscope ( FEI company ) at 120 kV . Images were taken with an Eagle CCD camera . HeLa cells grown on coverslips were infected with C . trachomatis LGV serovar L2 strain 434 with an MOI < 1 ( unless specified differently ) at 37°C and fixed in 4% PFA in PBS for 30 min at room temperature ( except when staining with anti-CT813 was intended , which required fixation in 2% PFA ) . Cells were blocked and permeabilized in 0 . 05% saponin and 0 . 1% bovine serum albumin ( BSA ) in PBS for 10 min before being subjected to antibody staining . The antibody against inclusion protein CT813 was kindly provided by Dr . G . Zhong ( San Antonio , Texas ) . Polyclonal anti-Cap1 antibodies were obtained by immunization of New Zealand white rabbits with purified recombinant Cap1 deleted of its last 167 amino acids and fused to a N-terminal His tag ( Agro-Bio ) . The polyclonal anti-GlgX antibody was equally purchased from Agro-Bio , and was directed against two peptides ( KHNEENGEYNRDGTSANC and HEDFDWEGDTPLHLPKEC ) . To investigate its specificity anti-GlgX was preabsorbed with either the two GlgX peptides or control peptides . For this , 1 µg/ml antibody was incubated with 20 µg/ml of peptides for 15 min at room temperature prior to immunostaining of cells . The monoclonal rat anti-HA antibody was purchased from Roche Diagnostics . Goat secondary antibodies were conjugated to Alexa488 ( Molecular Probes ) , or to Cy3 or Cy5 ( Amersham Biosciences ) . For periodic-acid-Schiff ( PAS ) stain cells grown on coverslips were fixed in 4% PFA/PBS for 30 min at room temperature and staining was performed as described ( Schaart et al . , 2004 ) . Briefly , cells were incubated in 1% periodic acid ( Sigma ) for 5 min . Thereafter coverslips were put in tap water for 1 min , quickly rinsed in mQ-H20 and then applied to Schiff reagent ( Sigma ) for 15 min at room temperature . Afterwards the coverslips were rinsed again in mQ-H20 , incubated in tap water for 10 min followed by an incubation step in PBS for 5 min . Periodic acid oxidizes the vicinal diols in sugars such as glycogen to aldehydes , which now in turn react with the Schiff reagent to give a purple/magenta colour . Images were acquired on an Axio observer Z1 microscope equipped with an ApoTome module ( Zeiss , Germany ) and a 63× Apochromat lens unless specified . Images were taken with a Coolsnap HQ camera ( Photometrics , Tucson , AZ ) using the software Axiovision . Gradient purified Chlamydia EBs were incubated in an axenic medium ( 5 mM KH2PO4 , 10 mM Na2HPO4 , 109 . 6 mM K-gluconate , 8 mM KCl , 1 mM MgCl2 ) ( Omsland et al . , 2012 ) supplemented with 0 . 2 mM α-D-[14C ( U ) ]-Glc 1-Phosphate , α-D-[14C ( U ) ]-Glc 6-Phosphate or D-[14C ( U ) ]-Glc ( Perkin Elmer ) ( 0 . 1 µCi per sample ) . In some samples 10 mM of the indicated cold monosaccharide was added in a competition assay . After two hours of incubation at 37°C the bacteria were pelleted ( 15000 g for 5 min ) and washed twice in 50 mM K2HPO4/KH2PO4 , 100 mM KCl and 150 mM NaCl . Radioactivity associated to the bacterial pellet and to the supernatant was measured by a scintillation counter . Genomic DNA from C . trachomatis LGV serovar L2 strain 434 was prepared from bacteria using the RapidPrep Micro Genomic DNA isolation kit ( Amersham Pharmacia Biotech ) . The first 20 to 30 codons of different chlamydial genes including about 20 nucleotides upstream from the translation start sites were amplified by PCR using primers listed in Supplementary file 1 and cloned into the pUC19cya vector as described ( Subtil et al . , 2001 ) . attB-containing primers ( Supplementary file 1 , Gateway , Life technologies ) were used to amplify and clone the C . trachomatis L2 glgA gene into a destination vector derived from the mammalian expression vector pCiNeo , providing an amino-terminal 3xflag tag , and into pDEST15 ( Gateway ) . siRNA transfections were performed with Lipofectamine RNAiMAX ( Life technologies ) according to the manufacturer's protocol . A mixture of 2 to 4 siRNA sequences ( Supplementary file 2 ) , with a final concentration of 10 nM each , was used and transfection was done twice , 48 hr and 4 hr prior to infection . siRNA efficiency was determined by immunoblot or RT-PCR ( see respective sections for more detail ) . Cells were transfected with plasmid DNA 24 hr after seeding using jetPRIME transfection kit ( Polyplus transfection ) according to the manufacturer's protocol . The constructs used were Flag-GlgA and SLC35D2-HA ( pMKIT-neo-hRel8-cHA , kindly provided by Nobuhiro Ishida , Chiba Institute of Science , Japan ) ( Ishida et al . , 2005 ) . Cells were treated with the indicated siRNA 48 hr and 4 hr prior to infection , which was performed at an MOI < 0 . 3 . Thirty or 48 hpi the cells were washed in PBS , detached and lysed with 1 mm glass beads . Fresh HeLa cells were inoculated with serial dilutions of the cell lysates and 24 hr later flow cytometry was used to determine the infection rate as described elsewhere ( Vromman et al . , 2014 ) . Briefly , cells were washed with PBS and gently trypsinized . After a washing step in PBS cells were fixed in PFA 2% in PBS and bacteria were stained using FITC-coupled anti-Chlamydia antibody ( Fitzgerald clone 502 #61R-1012 ) . Infection rates were measured using a Gallios flow cytometer ( Beckton Coulter ) . The cell image analysis software CellProfiler was used to quantify glycogen content in inclusions stained with PAS . Around 50 inclusions were manually encircled , and their size and total staining intensity was determined . PAS staining not linked to the presence of glycogen was estimated by doing the same procedure on inclusions in cells grown in the absence of Glc . The averaged value obtained was considered background value and was substracted . Glycogen content of inclusions of cells treated with control siRNA was considered 100% . Total RNA was isolated from 5 x 105 HeLa cells infected with C . trachomatis LGV serovar L2 after 1 , 3 , 8 , 16 , 24 hr or 40 hr of infection ( MOI of 10 for 1 , 3 , 8 hr and MOI of 0 . 1 for 16 , 24 , 40 hr ) with the RNeasy Mini Kit with DNase treatment ( Qiagen ) according to the manufacturer's protocol . RNA concentrations were determined by NanoDrop and the samples normalized to an equal RNA content . Reverse transcription ( RT ) was performed with SuperScript III Reverse Transcriptase ( Life Technologies ) and quantitative PCR ( qPCR ) undertaken with LightCycler 480 system using LightCycler 480 SYBR Green Master I ( Roche ) . In parallel , genomic DNA ( gDNA ) of each time point was purified with the DNeasy Blood and Tissue Kit ( Qiagen ) , and the amount of bacteria in the samples determined by qPCR using chlamydial primers . This was done to normalize the cDNA of the different samples . Primers are listed in Supplementary file 3 , their specificity was ensured through the analysis of melting curves . For the evaluation of siRNA SLC35D2 efficiency the steps until reverse transcription were the same , but the PCR was run with PrimeStar ( Clontech ) . Equal volumes were loaded on agarose gels and bands were revealed using UV-light visualizing ethidium bromide . Cell pellets were lysed with a lysis buffer ( 8 M urea , 30 mM Tris , 150 mM NaCl , 1% v/v sodium dodecyl-sulfate ) and proteins were subjected to sodium dodecyl-sulfate poly-acrylamide gel electrophoresis ( SDS-PAGE ) and transferred to a polyvinylidene difluoride ( PVDF ) membrane , which was blocked with 1× PBS containing 5% milk and 0 . 01% Tween-20 . The membranes were then immunoblotted with primary antibodies diluted in 1 x PBS and 0 . 01% Tween-20 . Primary antibodies used in the secretion assay were mouse anti-cya , rabbit anti-CRP and rabbit anti-IpaD and were generously given by Drs N Guiso , A Ullmann and C Parsot , respectively ( Institut Pasteur , Paris ) . Other antibodies used were rabbit anti-Gys1 ( Millipore #04–357 ) , rabbit anti-UGP2 ( GeneTex #GTX107967 ) , mouse anti-Flag M2 ( Sigma-Aldrich ) , goat anti-mouse IgG coupled to horseradish peroxidase ( HRP ) and goat anti-rabbit IgG-HRP ( GE Healthcare ) . Blots were developed using the Western Lightning Chemiluminescence Reagent ( GE Healthcare ) . Glycogen synthase genes ( glgA ) were amplified from genomic DNA of Escherichia coli K12 and Chlamydia trachomatis D/UW-3/CX and cloned into the expression vector pGEX ( GE Healthcare ) and pDEST15 , respectively . Precultures of wild-type and transformed E . coli were grown overnight in LB medium at 37°C , then transferred to fresh LB medium and grown until the optical density ( OD ) 600 nm reached 0 . 6 . Cultures were subsequently induced overnight with 0 . 5 mM of IPTG . Cells were harvested by centrifugation and disrupted by a French press at 1250 psi . The lysate was centrifuged at 16 , 000 g for 15 min at 4°C . Soluble proteins ( such as GlgA ) were found in the supernatant . Glycogen synthase activity was detected by zymogram analysis . The proteins in the supernatant were separated by non-denaturing polyacrylamide gel electrophoresis ( PAGE ) containing 0 . 6% rabbit glycogen ( Sigma-Aldrich ) . After electrophoresis , gels were incubated overnight at room temperature in glycogen synthase buffer ( 66 mM glycyl-glycine , 66 mM ( NH4 ) 2SO4 , 8 mM MgCl2 , 6 mM 2-mercaptoethanol , and 1 . 2 mM ADP-Glc or UDP-Glc ) . Glycogen synthase activity was then visualized as dark activity band after iodine staining . Analysis of secreted proteins was performed as described elsewhere ( Ball et al . , 2013; Subtil et al . , 2001 ) . Briefly , 1 ml of a 30°C overnight culture of Shigella flexneri ipaB or mxiD transformed with different Cya chimeras was inoculated in 30 ml of LB broth with 0 . 1 mg/ml ampicillin and incubated at 37°C for 4 hr . Bacteria were then harvested by centrifugation and the supernatant was filtered through a Millipore filter ( 0 . 2 μm ) . To precipitate the proteins 1/10 ( v/v ) of trichloroacetic acid was added to the supernatant and the precipitate as well as the bacterial pellet resuspended in sample buffer for analysis by SDS-PAGE and immunoblot . C . muridarum was mutagenized by treating infected Vero cells with 15 mg/ml ethyl methanesulfonate ( EMS ) for 1 hr , starting 16 hpi , resulting in an average of 16 mutations/genome with a range of 5–26 mutations/genome ( Rajaram et al . , 2015 ) . Mutagenized Chlamydia EBs were harvested at 26 hpi in sucrose phosphate glutamic acid buffer ( SPG ) using 3 mm glass beads , and aliquots were stored at -80°C . Mutant libraries were constructed as described previously ( Rajaram et al . , 2015 ) . Briefly , mutagenized C . muridarum isolates were plaque-cloned in McCoy cells and plaques were picked 10 days post infection . Isolated plaques were expanded in McCoy cells by infecting monolayers in 96-well plates with 30 µl of plaque lysates by centrifugation ( 1600 g for 1 hr at room temperature ) and rocking ( 30 min at 37°C ) . The library plates were stored at -80°C in SPG . Five µl of C . muridarum mutant isolates from the mutant library were used to infect fresh McCoy cell monolayers by centrifugation and rocking . At 28 hpi monolayers were fixed with 100% methanol for 10 min . Glycogen was assayed by staining with a 5% iodine solution ( 5% w/v iodine ( Sigma ) , 5% w/v potassium iodide ( Sigma ) , in a 50/50 solution of water and 100% ethanol ) for 10 min , followed by staining with a 2 . 5% iodine solution ( 2 . 5% w/v iodine , 2 . 5% w/v potassium iodide , in a 50/50 solution of water and 100% ethanol ) for 10 min . The iodine solution was removed and 100 µl PBS was added to each well . Plates were imaged using an EVOS® FL Auto cell imaging microscope . Glycogen deficient mutants were examined for mutations in known glycogen biosynthesis genes by Sanger sequencing or by whole genome sequencing as previously described ( Rajaram et al . , 2015 ) . P2B10 carried a C393T mutation in glgA , and P3B4 carried a G1157A mutation in the same gene .	Chlamydia trachomatis is the most common sexually transmitted bacteria that causes disease . Infections often do not produce any obvious symptoms , but can lead to infertility or other severe problems if left untreated . This microbe is also the leading cause of blindness by an infectious agent . The bacteria grow in the human body by infecting host cells . Inside these cells , the bacteria are found inside compartments known as inclusions , which protect them from the host’s defense responses and enable them to create a comfortable environment for themselves . However , this comes at a cost because the bacteria lose immediate access to the nutrients in the rest of the host cell . Thus , C . trachomatis has developed ways to import these nutrients into inclusions , and , more generally , to take the control of its interactions with the host cell . The inclusions built up by C . trachomatis contain a high amount of glycogen , a carbohydrate that generally acts as an energy storage molecule . Although this observation was made many decades ago , the molecular mechanism by which such a large molecule accumulates in the inclusion has not been clarified . Gehre et al . have now used a variety of cell biology techniques to address this question . The experiments show that there are two different pathways through which glycogen accumulates within the inclusion . Some glycogen is transported in bulk from the interior of the host cell into the inclusion . However , the bacteria also make new glycogen in the inclusion from a building block molecule called UDP-glucose . To do this , the bacteria recruit a host transport molecule to the membrane that surrounds the inclusion . This transport molecule brings UDP-glucose into the inclusion , where an enzyme called glycogen synthase – which is released by the bacteria – uses the UDP-glucose to make glycogen . The C . trachomatis glycogen synthase is unusual because most other bacteria can only make glycogen from another type of glucose . By using both pathways , C . trachomatis is able to trap most of the glycogen stores of the infected cell within the inclusion so that they are inaccessible to the host but ready for the bacteria to use . Previous work has shown that C . trachomatis is much better at accumulating glycogen than other Chlamydia bacteria are . Therefore , a future challenge will be to find out exactly how this helps C . trachomatis survive inside human cells .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "microbiology", "and", "infectious", "disease" ]	2016	Sequestration of host metabolism by an intracellular pathogen
In a presynaptic nerve terminal , synaptic strength is determined by the pool of readily releasable vesicles ( RRP ) and the probability of release ( P ) of each RRP vesicle . These parameters are controlled at the active zone and vary across synapses , but how such synapse specific control is achieved is not understood . ELKS proteins are enriched at vertebrate active zones and enhance P at inhibitory hippocampal synapses , but ELKS functions at excitatory synapses are not known . Studying conditional knockout mice for ELKS , we find that ELKS enhances the RRP at excitatory synapses without affecting P . Surprisingly , ELKS C-terminal sequences , which interact with RIM , are dispensable for RRP enhancement . Instead , the N-terminal ELKS coiled-coil domains that bind to Liprin-α and Bassoon are necessary to control RRP . Thus , ELKS removal has differential , synapse-specific effects on RRP and P , and our findings establish important roles for ELKS N-terminal domains in synaptic vesicle priming . Within a presynaptic nerve terminal , synaptic vesicle exocytosis is restricted to sites of neurotransmitter release called active zones . The active zone is a dense protein complex that is attached to the presynaptic plasma membrane and is exactly opposed to postsynaptic receptors ( Couteaux and Pecot-Dechavassine , 1970; Schoch and Gundelfinger , 2006; Südhof , 2012 ) . At the active zone , a small subset of the synaptic vesicles are primed in close proximity to presynaptic Ca2+ channels such that the incoming action potential leads to neurotransmitter release with minimal delay . The proteins of the active zone control the size of this pool of primed , readily releasable vesicles ( RRP ) and the release probability of those vesicles in response to an action potential ( Kaeser and Regehr , 2014; Alabi and Tsien , 2012 ) . Vesicular release probability , referred to in this paper as P , and RRP size together act to set synaptic strength , sometimes referred to as the synaptic probability of release ( Stevens , 2003; Zucker and Regehr , 2002 ) . It is well known that RRP size and vesicular release probability differ across synapses , contributing to the generation of unique release properties ( Abbott and Regehr , 2004 ) . In the hippocampus , for example , excitatory and inhibitory synapses have markedly different properties ( Kraushaar and Jonas , 2000; Salin et al . , 1996 ) . The underlying molecular mechanisms that control RRP and P are still only partially understood , and it is not known what components of the release machinery account for their synapse specific control . ELKS , RIM , Munc13 , RIM-binding protein ( RIM-BP ) , Bassoon/Piccolo , and Liprin-α proteins form a protein complex that defines the active zone ( Schoch and Gundelfinger , 2006; Südhof , 2012 ) . This protein complex includes many additional proteins that are not active zone specific ( Boyken et al . , 2013; Muller et al . , 2010; Schoch and Gundelfinger , 2006 ) . ELKS ( also called Erc , CAST , and Rab6IP2 ) was identified as an active zone protein through its interactions with RIM and named for its high content in the amino acids E , L , K , and S ( Nakata et al . , 1999; Wang et al . , 2002; Monier et al . , 2002; Ohtsuka et al . , 2002 ) . ELKS has known in vitro interactions with many active zone proteins and additional neuronal proteins ( Figure 1A ) . It contains a C-terminal region that binds to the PDZ domain of RIM ( Wang et al . , 2002; Ohtsuka et al . , 2002 ) and multiple coiled-coil stretches , which we have subdivided based on homology between the various vertebrate and invertebrate ELKS isoforms into four coiled-coil domains ( CCA-CCD , Figure 1A ) . The coiled-coil stretches bind in vitro to Liprin-α ( CCA-CCC , [Ko et al . , 2003] ) , Bassoon ( CCC , [Takao-Rikitsu et al . , 2004] ) , and β-subunits of Ca2+ channels ( CCD , [Kiyonaka et al . , 2012] ) . Vertebrate genomes contain two genes for ELKS , Erc1 and Erc2 ( Wang et al . , 2002 ) , whereas C . elegans expresses a single ELKS homolog ( Deken et al . , 2005 ) . D . melanogaster expresses a protein called Brp with homology to ELKS in the N-terminal but not the C-terminal half ( Wagh et al . , 2006; Kittel et al . , 2006; Monier et al . , 2002 ) . Vertebrate ELKS proteins are expressed as predominant , synaptic α-isoforms and shorter β-variants , which account for less than 5% of ELKS ( Kaeser et al . , 2009; Liu et al . , 2014 ) . In addition , ELKS C-terminal variants determine RIM-binding: the B-isoforms are prominently expressed in the brain and contain the RIM binding site , whereas A-isoforms are expressed outside the brain and lack RIM binding ( Wang et al . , 2002; Kaeser et al . , 2009 ) . 10 . 7554/eLife . 14862 . 003Figure 1 . ELKS1α and ELKS2α are co-expressed at excitatory synapses . ( A ) Schematic of ELKS protein structure . Arrows: transcriptional start sites of α- and β-ELKS , CCA-D: coiled-coil regions A - D ( ELKS1: CCA1MYG…SKI208 , CCB 209TIW…ENN358 , CCC 359MLR…EAT696 , CCD697LEA…EEE988; ELKS2: CCA1MYG…ARM204 , CCB205SVL…ENI362 , CCC363HLR…NIE656 , CCD657DDS…DEE917 , B: PDZ-binding sequence ( ELKS1: 989GIWA992 , ELKS2: 918GIWA921 ) of the ELKS-B C-terminal splice variant . Binding regions for interacting active zone proteins are indicated with black bars . ( B ) Sample images and quantification of ELKS1α ( left ) and ELKS2α ( right ) expression levels at excitatory and inhibitory synapses . VGAT or GAD2 ( red , inhibitory synapses ) and VGluT1 ( blue , excitatory synapses ) staining was used to define regions of interest ( ROIs ) , respectively ( control n = 4 independent cultures , cDKO n = 4 , 10 images were averaged per culture ) . All data are means ± SEM; p≤0 . 05 as determined by Student's t test . ( C ) Sample images ( top ) and correlation of expression levels of ELKS1α and ELKS2α ( bottom ) at excitatory ( left ) and inhibitory ( right ) synapses . Arrowheads indicate example puncta used to define ROIs . Data points represent the fluorescent intensity of ELKS1α within an ROI plotted against the ELKS2α signal in the same ROI . Within a single channel , individual puncta are normalized to the average intensity across all puncta ( excitatory synapses: 329 ROIs/30 images/3 independent cultures; inhibitory synapses: 250/30/3 ) . ρ: Spearman rank correlation between ELKS1α and ELKS2α . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 00310 . 7554/eLife . 14862 . 004Figure 1—figure supplement 1 . ELKS antibody specificity . ( A ) Western blot for testing specificity of ELKS2α ( 1029 , top ) and ELKS1α ( E-1 , middle ) antibodies against samples of HEK293T cells transfected with ELKS1αB or ELKS2αB cDNAs . The ELKS2α specific antibodies were raised in rabbits to a non-conserved sequence between ELKS1 and ELKS2 ( 109LSHTDVLSYTDQ120 ) , the E-1 antibody is commercially available . β-actin was used as a loading control . ( B ) Western blot testing reactivity of ELKS2α ( top ) and ELKS1α ( middle ) in cultured cDKO and control hippocampal neurons and whole brain homogenate . β-actin was used as a loading control . ( C ) ELKS2α antibodies were affinity purified using the ELKS2 peptide and characterized via immunostaining in cultured control and ELKS2α cKO neurons . ELKS2α cKO neurons were generated from ELKS2αfloxed mice ( Kaeser et al . , 2009 ) , and neurons were stained for ELKS2α ( 1029 ) , GAD2 , and VGluT1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 00410 . 7554/eLife . 14862 . 005Figure 1—figure supplement 2 . Frequency distributions of ELKS1α and ELKS2α at excitatory and inhibitory synapses . ( A ) Histogram displaying the frequency distribution of ELKS1α intensity within VGAT ( black bars , n = 194 ROIs/40 images/4 independent cultures ) or VGluT1 ( grey bars , n = 207/40/4 ) labeled puncta . ( B ) Histogram displaying the frequency distribution of ELKS2α intensity within GAD2 ( black bars , n = 170/40/4 ) or VGluT1 ( grey bars , n = 205/40/4 ) labeled puncta . The analysis shown in this figure uses the data presented in the Figure 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 005 The observation that ELKS binds to several active zone proteins has led to the hypothesis that ELKS scaffolds other active zone proteins , in particular RIM ( Takao-Rikitsu et al . , 2004; Ohtsuka , 2013; Ohtsuka et al . , 2002 ) . Invertebrate studies offer mixed support to this hypothesis . Loss of Brp disrupts the T-bar structures at the fly neuromuscular junction ( Kittel et al . , 2006 ) , but this function involves the C-terminal region of Brp ( Fouquet et al . , 2009 ) . In contrast , C . elegans ELKS is not required for recruitment of other active zone proteins ( Deken et al . , 2005 ) , but a gain of function mutation in syd-2 , the C . elegans Liprin-α homologue , requires ELKS for its synaptogenic activity ( Dai et al . , 2006 ) . Relatively little is known about the role and molecular mechanisms of ELKS in neurotransmitter release at vertebrate synapses . Previous studies showed that ELKS1α/2α boost Ca2+ influx at inhibitory synapses , whereas ELKS2α has a regulatory , non-essential function in RRP at these synapses ( Liu et al . , 2014; Kaeser et al . , 2009 ) . In contrast , at excitatory ribbon synapses of rod photoreceptors , ELKS2α/CAST may have a structural role to enhance synaptic transmission ( tom Dieck et al . , 2012 ) . These studies suggest the interesting possibility that ELKS may have differential roles in synaptic transmission between synapses . Thus far , ELKS functions have not been studied at small , excitatory synapses , the most abundant synapses in the vertebrate brain . Here , we establish that ELKS is prominently expressed at excitatory hippocampal synapses . In contrast to its roles in enhancing Ca2+ influx and release probability at inhibitory synapses , ELKS1α/2α control the size of the RRP at excitatory hippocampal synapses . These data establish that removal of ELKS has differential , synapse-specific effects on RRP size and P: boosting RRP at excitatory synapses but enhancing P at inhibitory synapses . Using structure-function rescue experiments , we then determine the sequences within ELKS required for RRP enhancement at excitatory synapses . Surprisingly , RIM-binding sequences are dispensable for this function , but CCA-CCC , which include binding sites for Liprin-α and Bassoon , control excitatory RRP size . Together , these data show that ELKS selectively controls RRP at excitatory synapses through its N-terminal protein interaction motifs . The distribution of individual ELKS proteins at excitatory and inhibitory synapses is not known . We generated ELKS2α specific antisera by immunization of rabbits ( Figure 1—figure supplement 1A–C ) and used this , in addition to an available ELKS1α specific antibody , to determine if either or both ELKS proteins are present at excitatory and inhibitory synapses . We employed immunostainings for ELKS1α and ELKS2α in cultured hippocampal neurons and analyzed their distribution using confocal microscopy . Both ELKS proteins were present at excitatory and inhibitory synapses and we observed higher intensity staining at excitatory synapses compared to inhibitory synapses for ELKS1α and ELKS2α ( Figure 1B ) . A recent proteomic study of release site composition found that overall differences between excitatory and inhibitory release sites are small ( Boyken et al . , 2013 ) . Interestingly , however , ELKS1 and ELKS2 were modestly enriched at docking sites for glutamatergic vesicles , consistent with our observation . We next examined the distribution of ELKS relative to one another to determine whether levels of individual ELKS proteins correlate positively or negatively at excitatory or inhibitory synapses . Synaptic markers for either excitatory or inhibitory synapses were used to define regions of interest ( ROI ) and co-stained for both ELKS1α and ELKS2α ( Figure 1C ) . We found a strong positive correlation between ELKS1α and ELKS2α at excitatory and inhibitory synapses and both ELKS proteins showed a single peak in the distribution of fluorescence intensity at each synapse ( Figure 1—figure supplement 2 ) . Together , these data reveal that ELKS1α and ELKS2α are enriched at excitatory synapses and they suggest that synapses rich in ELKS1α are rich in ELKS2α and vice versa . Excitatory transmission was not affected in single ELKS2α mutants ( Kaeser et al . , 2009 ) and excitatory synaptic transmission was not studied in ELKS1α/ELKS2α double mutants . We thus decided to determine the function of ELKS1α/2α at excitatory synapses in cultured hippocampal neurons employing mice in which the first coding exon of each gene , Erc1 and Erc2 , is flanked by loxP sites ( Liu et al . , 2014; Kaeser et al . , 2009 ) . At 3–5 days in vitro ( DIV ) , we infected the neurons with lentiviruses that express a cre-recombinase tagged with EGFP driven by a neuron-specific synapsin promoter ( Liu et al . , 2014 ) to generate ELKS1α/ELKS2α knockout neurons ( cDKO ) . In all experiments , control neurons were genetically identical with the exception that they were infected with lentiviruses that expressed an inactive , truncated cre protein . We only analyzed cultures in which no non-infected neurons could be detected . Using focal stimulation , we recorded action-potential evoked excitatory postsynaptic currents ( EPSCs ) and directly compared effects to inhibitory PSCs ( IPSCs ) . Similar to the effect on IPSCs ( Figure 2A ) , EPSCs were reduced to ~50% ( Figure 2B , C ) . At inhibitory synapses , removal of ELKS resulted in a 30% decrease in presynaptic Ca2+ influx which led to a reduction in P ( Liu et al . , 2014 ) and prolonged IPSC rise times ( Figure 2A ) . When we examined the EPSC rise times , there was no effect of ELKS1α/2α cDKO ( Figure 2B , C ) , suggesting that ELKS may operate differently at excitatory synapses . To determine whether the deficit in excitatory transmission was presynaptic , we recorded miniature EPSCs ( mEPSCs ) in the presence of TTX . mEPSC frequency was reduced by ~50% , but there was no change in mEPSC amplitude , rise time ( Figure 2D ) , or decay kinetics ( control τ = 5 . 727 ms , n = 16/3; cDKO τ = 6 . 321 ms , n = 15/3; p>0 . 05 ) . The number and size of excitatory synapses was also unchanged ( Figure 2—figure supplement 1 ) . Thus , ELKS1α/2α cDKO reduces release from excitatory presynaptic nerve terminals , but the mechanisms may be different from inhibitory synapses . 10 . 7554/eLife . 14862 . 006Figure 2 . ELKS1α/2α control neurotransmitter release at excitatory synapses . ( A–C ) Sample traces and quantification of IPSC ( A ) , AMPAR-EPSC ( B ) , and NMDAR-EPSC ( C ) amplitudes and rise times in control and ELKS1α/2α cDKO neurons . Bar graphs show quantification of the peak amplitude ( middle ) and quantification of the rise time from 20% to 80% of the peak amplitude ( right , A: control n = 18 cells/4 independent cultures , cDKO n = 18/4; B: control n = 18/4 , cDKO n = 16/4; C: control n = 15/3 , cDKO n = 15/3 ) . ( D ) Sample traces ( top ) and quantification ( bottom ) of mEPSC frequency , amplitude , and 20–80% rise time ( control n = 16/3 , cDKO n = 15/3 ) . Sample traces on the top left show 10 s of recording time . Sample traces on the top right are the overlayed averaged events from an individual cell in each condition normalized for amplitude . All data are means ± SEM; p≤0 . 05 , p≤0 . 01 , p≤0 . 001 as determined by Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 00610 . 7554/eLife . 14862 . 007Figure 2—figure supplement 1 . No change in synapse number in ELKS1α/2α cDKO cultures . ( A ) Sample images of excitatory ( green ) and inhibitory ( red ) synapses in control and cDKO neurons . ( B and C ) Quantification for excitatory ( B ) and inhibitory ( C ) synapses of fluorescence intensity , synapse number ( puncta/10 μm ) , and puncta size ( control n = 3 independent cultures , cDKO n = 3 , 10 images per culture were averaged per condition ) . All data are means ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 007 At inhibitory hippocampal synapses , ELKS boosts presynaptic Ca2+ influx to increase P and to accelerate the IPSC rise . To directly test the prediction that ELKS functions differently at excitatory synapses , we employed imaging to measure the presynaptic Ca2+ transient in response to a single action potential ( Figure 3A ) . Neurons were filled with the Ca2+ indicator Fluo-5F and a fixable Alexa-594 dye and Ca2+ influx into presynaptic boutons was imaged during a single action potential elicited by a brief current injection through the patch pipette . This method has revealed impaired Ca2+ influx at inhibitory ELKS deficient synapses ( Liu et al . , 2014 ) . After the experiment , cells were fixed and stained with GAD67 antibodies to exclude GAD67 positive inhibitory neurons from the analysis . This method reliably distinguished excitatory and inhibitory neurons in hippocampal cultures ( Figure 3—figure supplement 1 ) . We found that there was no change in peak action potential evoked Ca2+ influx into boutons of excitatory neurons ( Figure 3A ) . At inhibitory synapses , decreased Ca2+ influx in ELKS deficient neurons resulted in a reduction in P . Thus , our data suggest that initial P may not be affected at excitatory ELKS1α/2α cDKO synapses . Initial P is inversely correlated with paired-pulse ratios ( PPR ) . We measured PPRs of EPSC amplitudes by monitoring N-Methyl-D-aspartate receptor EPSCs ( NMDAR-EPSCs ) to circumvent the strong reverberant activity that is present in cultured networks when α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid ( AMPA ) receptors are not blocked ( Maximov et al . , 2007 ) . Consistent with the Ca2+ imaging data , but different from inhibitory synapses ( Liu et al . , 2014 ) , there were no changes in PPR in ELKS cDKO neurons across all tested interstimulus intervals ( Figure 3B ) . These experiments establish that defects in Ca2+ influx and P are unlikely to explain impaired neurotransmitter release at excitatory synapses of ELKS cDKO neurons . 10 . 7554/eLife . 14862 . 008Figure 3 . ELKS1α/2α do not control Ca2+ influx and release probability in excitatory nerve terminals . ( A ) Sample images ( top , imaged boutons are numbered and color coded ) , action potential and imaging traces ( middle ) and summary plots ( bottom ) of single action potential-induced Ca2+ transients imaged by Fluo-5F fluorescence in presynaptic boutons are shown , the inset shows the same plot for dendrites . Data are shown as mean ( line ) ± SEM ( shaded area ) and analyzed by two-way ANOVA: genotype , n . s . ; time , p<0 . 001; interaction , n . s . ( Boutons: control n = 60 boutons/6 cells /4 independent cultures , cDKO n = 80/8/4; dendrites: control n = 6 dendrites/6 cells/4 independent cultures , cDKO n = 8/8/4 ) . ( B ) Sample traces ( top ) and quantification of paired pulse ratios ( PPRs , bottom ) of evoked NMDAR-EPSCs . Example traces showing overlayed responses to pairs of stimuli at 50 , 100 , 500 , and 2500 ms interstimulus intervals . PPRs ( amplitude 2/amplitude 1 ) are plotted against the interstimulus interval . Significance as analyzed by two-way ANOVA: genotype , n . s . ; interstimulus interval , p<0 . 001; interaction , n . s . ( control n = 17 cells/3 independent cultures , cDKO n = 19/3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 00810 . 7554/eLife . 14862 . 009Figure 3—figure supplement 1 . Post-hoc identification of excitatory neurons after presynaptic Ca2+ imaging . ( A ) In each presynaptic Ca2+ imaging experiment , cultured neurons were stained post-hoc with GAD67 antibodies to label the soma and neurites of inhibitory neurons , and the imaged neurons were identified by the Alexa 594 fill ( red ) . The left panel shows an example of non-GABA neuron ( negative for GAD67 ) and the right panel a GABA neuron ( positive for GAD67 ) . ( B ) In an independent experiment , excitatory neurons were marked with a lentivirus expressing TdTomato ( arrowheads ) under a CaMKII promoter in addition to a lentivirus that labeled all neurons with nuclear EGFP ( merged image ) . Inhibitory neurons were identified by somatic and dendritic GAD67 staining ( arrow , GAD67 antibodies have a nuclear background staining in all neurons ) . Of the 114 neurons we analyzed , 80 were TdTomato positive ( excitatory ) , 23 were GAD67 positive ( inhibitory ) , and 11 were negative for both markers . Thus , 80 out of 91 of the GAD67 negative neurons were identified as excitatory neurons . The remaining neurons may be excitatory , inhibitory , or modulatory . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 009 We next tested whether the size of the RRP was changed at excitatory synapses . We stimulated release of the entire RRP using a hypertonic sucrose solution ( 500 mOsm ) and quantified RRP size by integrating the total charge transfer during the first ten seconds of the response ( Rosenmund and Stevens , 1996 ) . Although the hypertonic stimulus is non-physiological , this method has been used as a snapshot measurement of RRP in cultured neurons and has been insightful for genotype comparisons and for dissecting molecular mechanisms of RRP control ( Augustin et al . , 1999; Deng et al . , 2011; Neher , 2015; Rosenmund and Stevens , 1996 ) . At excitatory synapses , the RRP was reduced by ~40% ( Figure 4A ) in ELKS cDKO neurons , providing an explanation for a reduction in release in the absence of changes in P at excitatory ELKS1α/2α cDKO synapses . Short action potential trains ( 10 stimuli at 20 Hz ) , a more physiological stimulus , resulted in a 50% reduction in charge transfer during the stimulus train in ELKS1α/2α cDKO neurons compared to control neurons ( Figure 4B ) . The delayed charge transfer after stimulation ended was similarly reduced . These data are consistent with a reduced number of RRP vesicles available for release in response to a brief action potential train . 10 . 7554/eLife . 14862 . 010Figure 4 . ELKS1α/2α control RRP at excitatory synapses . ( A ) Sample traces showing AMPAR-EPSCs in response to superfusion with 500 mOsm sucrose ( left ) . The bar graph on the right shows the AMPAR-EPSC charge transfer , quantified as the area under the curve during the first ten seconds of the response ( control n = 30/3 , cDKO n = 29/3 ) . ( B ) Sample traces ( left ) of NMDAR-EPSCs during a short action potential train ( 10 stimuli at 20 Hz ) . The charge transfer during the train and in the two seconds immediately after the train is quantified separately on the right ( control n = 14/3 , cDKO n = 14/3 ) . All data are means ± SEM; p≤0 . 05 , p≤0 . 01 as determined by Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 010 One possible explanation for the phenotypic differences between inhibitory transmission as observed earlier ( Liu et al . , 2014 ) and excitatory transmission as described here is that removal of ELKS has effects that vary between cultures and over time , causing some cultures to have more pronounced effects on RRP size and others to exhibit changes in P . To control for this possibility , we conducted a series of experiments analyzing RRP size and P ( as measured by PPRs ) for inhibitory and excitatory transmission in the same cultures and matching sample size for all conditions . Both inhibitory and excitatory synapses had a reduction in action potential-evoked EPSC and IPSC amplitudes ( Figure 5—figure supplement 1 ) . At inhibitory synapses , ELKS1α/2α cDKO increased the PPR at low interstimulus intervals but had no significant effect on RRP size , consistent with a reduction of P ( Figure 5A , C ) and with our previous study ( Liu et al . , 2014 ) . In contrast , excitatory synapses in the same cultures showed no change in PPR but a reduction in RRP size ( Figure 5B , D ) . Notably , there was a small , non-significant trend towards a reduced RRP at inhibitory synapses in this experiment . To further rule out that there is a reproducible RRP reduction at inhibitory hippocampal synapses , we conducted a second experiment measuring RRP with a slightly different protocol as described in the methods section and used in ( Liu et al . , 2014 ) . Again , no change in RRP was detected ( Figure 5—figure supplement 2 ) . Thus , in three independent measurements ( Figure 5C , Figure 5—figure supplement 2 and Figure 7F in Liu et al . , 2014 ) , no reduction in the inhibitory RRP could be detected . These data confirm our previous experiments and establish synapse-specific roles for ELKS in neurotransmitter release: at excitatory synapses ELKS1α/2α primarily boost the RRP ( Figures 1–5 ) , whereas at inhibitory synapses ELKS1α/2α mainly control action potential induced Ca2+ influx to enhance P ( also see Figures 5 and 8 in Liu et al . , 2014 ) . 10 . 7554/eLife . 14862 . 011Figure 5 . Direct comparison of IPSC and EPSC phenotypes of ELKS1α/2α cDKO . ( A ) Example traces ( top ) showing overlayed IPSC responses to pairs of stimuli at 10 , 20 , 100 , and 500 ms interstimulus intervals . Paired-pulse ratios ( amplitude 2/amplitude 1 ) are plotted against the interstimulus interval ( 10 , 20 , 50 , 100 , 500 , and 2500 ms intervals ) . Significance as analyzed by two-way ANOVA: genotype , p<0 . 001; interstimulus interval , *p<0 . 001; interaction , n . s . Holm-Sidak post-hoc test: 10 ms , p<0 . 05; 20 ms , p<0 . 05 ( control n = 15 cells/3 independent cultures , cDKO n = 15/3 ) . ( B ) Example traces ( top ) showing overlayed EPSC responses to pairs of stimuli at 50 , 100 , 500 , and 2500 ms interstimulus intervals . Paired-pulse ratios ( amplitude 2/amplitude 1 ) are plotted against the interstimulus interval ( 50 , 100 , 500 , and 2500 ms intervals ) . Significance as analyzed by two-way ANOVA: genotype , n . s . ; interstimulus interval , p<0 . 001; interaction , n . s . ( control n = 15/3 , cDKO n = 15/3 ) . ( C ) Sample traces showing IPSCs in response to superfusion with 500 mOsm sucrose ( left ) and quantification ( right ) of IPSC charge transfer during the first ten seconds of the response ( control n = 15/3 , cDKO n = 15/3 ) . ( D ) Sample traces showing EPSCs in response to superfusion with 500 mOsm sucrose ( left ) and quantification ( right ) of EPSC charge transfer during the first ten seconds of the response ( control n = 15/3 , cDKO n = 15/3 ) . Data are means ± SEM; p≤0 . 01 as determined by Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 01110 . 7554/eLife . 14862 . 012Figure 5—figure supplement 1 . Direct comparison of IPSC and EPSC amplitudes of ELKS1α/2α cDKO . Quantification of IPSC ( left ) and NMDAR-EPSC ( right ) amplitudes in control and cDKO neurons from the same cultures ( control n = 15 cells /3 independent cultures , cDKO n = 15/3 ) . All data are means ± SEM; p≤ 0 . 05 , p≤0 . 01 as determined by Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 01210 . 7554/eLife . 14862 . 013Figure 5—figure supplement 2 . Inhibitory RRP size in ELKS1α/2α cDKO neurons . ( A ) Sample traces showing IPSCs in response to superfusion with 500 mOsm sucrose ( left ) and quantification ( right ) of IPSC charge transfer during the first ten seconds of the response ( control n = 9 cells/3 independent cultures , cDKO n = 9/3 ) . All data are means ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 013 Given the hypothesis that ELKS is a presynaptic scaffold , we set out to test the possibility that loss of ELKS1α/2α changes presynaptic levels of proteins involved in controlling RRP size at excitatory synapses . RIM and Munc13-1 were of particular interest , since both proteins localize to the active zone and knockouts of either protein have strong RRP impairments at excitatory hippocampal synapses ( Augustin et al . , 1999; Calakos et al . , 2004; Kaeser et al . , 2011 ) . Using confocal microscopy , we quantified the synaptic levels of the active zone proteins ELKS , RIM , Munc13-1 , Bassoon , and RIM-BP2 in cDKO and control neurons . With the exception of ELKS , no protein was significantly reduced at excitatory or inhibitory synapses ( Figure 6 ) . These data are consistent with normal electron microscopic appearance of ELKS1α/2α cDKO synapses ( Liu et al . , 2014 ) and indicate that ELKS is not essential to recruit the priming proteins RIM and Munc13-1 to the presynaptic nerve terminal . 10 . 7554/eLife . 14862 . 014Figure 6 . Active zone composition in ELKS1α/2α cDKO synapses . ( A ) Sample images of control and ELKS1α/2α cDKO neurons stained with antibodies against active zone proteins . Inhibitory synapses were marked with VGAT ( for ELKS , RIM and Bassoon ) or GAD2 ( for Munc13-1 and RIM-BP2 ) , excitatory synapses were marked with VGluT1 . ( B ) Quantification of active zone proteins within ROIs defined by excitatory ( left ) or inhibitory ( right ) synaptic markers ( RIM: control n = 4 independent cultures , cDKO n = 4; Munc13-1: control n = 4 , cDKO n = 4; Bassoon: control n = 3 , cDKO n = 3; RIM-BP2: control n = 3 , cDKO n = 3; in each culture , 10 images were averaged per culture and genotype ) . Data are means ± SEM; *p≤0 . 001 as determined by Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 014 Since ELKS1α/2α cDKO caused no detectable structural changes at the active zone , we instead turned to electrophysiological rescue experiments to determine how ELKS might control RRP size . ELKS binds directly to the PDZ domain of RIM through its four C-terminal residues and may have roles in active zone anchoring of RIM ( Lu et al . , 2005; Wang et al . , 2002; Ohtsuka et al . , 2002; Kaeser et al . , 2009 ) . Peptide injection experiments using the RIM-binding domain of ELKS support a role of the RIM-ELKS interactions in release ( Takao-Rikitsu et al . , 2004 ) . The CCD domain , which is adjacent to the RIM binding sequence , also binds to Ca2+ channel β4 subunits ( Kiyonaka et al . , 2012 ) . These interactions suggest that the C-terminus of ELKS may tether ELKS to the active zone to support its functions in release . We designed lentiviral rescue constructs in which we expressed either full length ELKS1αB or mutant ELKS1 that lacks either the entire C-terminal region including the PDZ binding motif ( △CCDB ) or just CCD ( △CCD ) ( Figure 7A ) . All three rescue constructs expressed at levels similar to wild type ELKS1α and localized to synapses ( Figure 7—figure supplement 1 ) , suggesting that the ELKS C-terminal regions are not necessary for ELKS localization . Surprisingly , ELKS1αB , ELKS1-△CCDB , and ELKS1-△CCD were sufficient to rescue excitatory RRP size in ELKS cDKO neurons ( Figure 7B ) . Thus , the ELKS C-terminal domains that bind to RIM and other presynaptic proteins are not necessary for its control of the RRP at excitatory synapses . Altogether , the C-terminal ELKS domains are unlikely to act as a central scaffolding hub at the active zone . 10 . 7554/eLife . 14862 . 015Figure 7 . C-terminal ELKS sequences do not support RRP in ELKS1α/2α cDKO neurons . ( A ) Schematic of ELKS1 rescue constructs; CCA-D: coiled-coil regions A-D , B: PDZ-binding motif; H: human influenza hemagglutinin ( HA ) tag , deleted sequences are illustrated as dashed lines . ( B ) Sample traces ( left ) and quantification ( right ) of the AMPAR-EPSC charge in response to hypertonic sucrose application , measured as area under the curve during the first ten seconds after the start of the stimulus ( control n = 26 cells/5 independent cultures , cDKO n = 27/5 , cDKO + ELKS1αB n = 21/5 , cDKO + ELKS1-△CCDB n = 23/5 , cDKO + ELKS1-△CCD n = 21/5 ) . All data are means ± SEM; p≤0 . 05 as determined by one-way ANOVA followed by Holm-Sidak multiple comparisons post-hoc test comparing each condition to cDKO . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 01510 . 7554/eLife . 14862 . 016Figure 7—figure supplement 1 . Expression and localization of ELKS1 C-terminal rescue constructs . ( A ) Sample images of control , cDKO , and cDKO + rescue neurons stained with antibodies against ELKS1α ( E-1 ) , the inhibitory synapse marker VGAT , and the excitatory synapse marker VGluT1 . ( B , C ) Quantification of ELKS1 fluorescent intensity within ROIs defined by VGluT1 ( B ) or VGAT ( C ) signals ( control n = 3 independent cultures , cDKO n = 3 , cDKO + ELKS1αB n = 3 , cDKO + ELKS1-△CCDB n = 3 , cDKO + ELKS1-△CCD n = 3 , 5–10 images per culture were averaged per condition ) . Data are shown as means ± SEM . ( D ) Representative western blot of control , cDKO , and cDKO + rescue neurons used for recording in Figure 7 . Samples were blotted using an ELKS1α specific antibody ( E-1 ) and β-actin was used as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 016 We next decided to take an unbiased approach and systematically tested all ELKS protein interaction sites covering the entire sequence of ELKS1αB in rescue experiments . We generated rescue constructs lacking N-terminal coiled-coils ( corresponding to ELKS1βB ) or the central coiled-coil region ( △CCC ) . ELKS1αB and ELKS1-△CCDB were used as positive rescue controls ( Figure 8A ) . All rescue constructs were successfully expressed , albeit at variable levels ( Figure 8—figure supplement 1A ) . Compellingly , neither ELKS1βB nor ELKS1-△CCC rescued the RRP in ELKS1α/2α cDKO neurons ( Figure 8B ) . The lack of rescue could be due to either a local function of the deleted coiled-coil regions or to a role for these regions in localizing ELKS to synapses . We distinguished between these possibilities by assessing the localization of each rescue construct using confocal microscopy . Since ELKS1βB lacks the antigen recognized by our ELKS1 antibody we stained for either the HA tag included in all rescue proteins ( Figure 8C and Figure 8—figure supplement 1B ) or for ELKS1α ( Figure 8—figure supplement 1C–E ) . Synaptic expression of the rescue constructs ranged from 50–200% of wild-type levels of ELKS1 . Importantly , all constructs localized to synapses . Thus vertebrate ELKS1 can be localized to synapses likely through multiple redundant interactions . While synaptic levels of rescue ELKS1αB were low , it rescued entirely . In contrast , ELKS1βB and ELKS1-△CCC both failed to rescue , but were expressed above ELKS1αB levels . Removal of ELKS reveals differential impairments of RRP and P at excitatory and inhibitory synapses , respectively ( Figure 5 ) . To test whether such differential effects are also reflected in the ELKS sequences that mediate rescue at each synapse , we tested whether ELKS1βB , an ELKS isoform that failed to rescue the excitatory RRP , was sufficient to restore inhibitory synaptic transmission ( Figure 8—figure supplement 2 ) . ELKS1βB was able to rescue action-potential evoked IPSC amplitudes , establishing that ELKS’ role in controlling P at inhibitory synapses does not require the N-terminal coiled-coil regions that control the excitatory RRP . Together , these experiments establish that the N-terminal coiled-coil sequences of ELKS1α are necessary for ELKS' role in enhancing the RRP at excitatory synapses , but not necessary for localizing ELKS to synapses . 10 . 7554/eLife . 14862 . 017Figure 8 . N-terminal coiled-coil domains of ELKS control RRP size at excitatory synapses . ( A ) Schematic of ELKS1 rescue constructs; CCA-D: coiled-coil regions A-D , B: PDZ-binding motif; H: human influenza hemagglutinin ( HA ) tag , black bar: antigen recognized by the ELKS1α antibody ( E-1 ) used in figure supplement 1C . Deleted sequences are illustrated as dashed lines , ( B ) sample traces ( left ) and quantification ( right ) of the AMPAR-EPSC charge in response to hypertonic sucrose application , measured as area under the curve during the first ten seconds after the start of the stimulus ( control n = 21 cells/4 independent cultures , cDKO n = 22/4 , cDKO + ELKS1αB n = 19/4 , cDKO + ELKS1βB n = 18/4 , cDKO + ELKS1-△CCC n = 20/4 , cDKO + ELKS1-△CCDB n = 21/4 ) . ( C ) Sample images ( left ) of control , cDKO , and cDKO + rescue neurons stained with antibodies against HA . Quantification ( right ) of HA fluorescent intensity within ROIs defined by VGluT1 ( control n = 3 independent cultures , cDKO n = 3 , cDKO + ELKS1αB n = 3 , cDKO + ELKS1βB n = 3 , cDKO + ELKS1-△CCC n = 3 , cDKO + ELKS1-△CCDB n = 3 , 5–10 images were averaged per culture and genotype ) . All data are means ± SEM; p≤0 . 01 , p≤0 . 001 as determined by one-way ANOVA followed by Holm-Sidak multiple comparisons post-hoc test comparing each condition to cDKO . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 01710 . 7554/eLife . 14862 . 018Figure 8—figure supplement 1 . Expression and localization of ELKS1 full length rescue constructs . ( A ) Representative western blot of control , cDKO , and cDKO + rescue neurons used for recording in Figure 8 . Samples were blotted using a pan-ELKS antibody ( P224 ) and β-actin was used as a loading control . ( B ) Quantification of HA staining fluorescent intensity within ROIs defined by VGAT signals ( control n = 3 independent cultures , cDKO n = 3 , cDKO + ELKS1αB n = 3 , cDKO + ELKS1βB n = 3 , cDKO + ELKS1-△CCc n = 3 , cDKO + ELKS1-△CCDB n = 3 , 5–10 images per culture were averaged per condition ) . Data are shown as means ± SEM . ( C ) Representative images of staining with ELKS1α antibodies ( E-1 ) in control , cDKO , and cDKO + rescue neurons . Note that ELKS1βB lacks the antigen for the ELKS1α antibody ( E-1 ) . ( D and E ) Quantification of ELKS1 fluorescent intensity within ROIs defined by VGluT1 ( D ) or VGAT ( E ) ( control n = 3 independent cultures , cDKO n = 3 , cDKO + ELKS1αB n = 3 , cDKO + ELKS1-△CCC n = 3 , cDKO + ELKS1-△CCDB n = 3 , 5–10 images were averaged per culture and genotype ) . All data are means ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 01810 . 7554/eLife . 14862 . 019Figure 8—figure supplement 2 . Rescue of action potential evoked IPSCs with ELKS1βB . Sample traces ( left ) and quantification ( right ) of the IPSC amplitude in control , cDKO , cDKO + ELKS1βB , and control + ELKS1βB ( control n = 11 cells/3 independent cultures , cDKO n = 11/3 , cDKO + ELKS1βB n = 10/3 , control + ELKS1βB n = 12/3 ) . All data are means ± SEM; p≤0 . 01 , *p≤0 . 001 as determined by one-way ANOVA followed by Holm-Sidak multiple comparisons post-hoc test comparing each condition to cDKO . Holm-Sidak multiple comparisons post-hoc tests comparing each condition to control did not reveal significant differences except for control vs . cDKO , p≤0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 14862 . 019 Our findings define a new functional role for ELKS in setting RRP size at excitatory hippocampal synapses . They extend beyond previous understanding of the role of ELKS in supporting presynaptic Ca2+ influx at inhibitory synapses ( Liu et al . , 2014 ) and demonstrate that this is not the primary function across all synapses . Most current molecular models of active zone function imply that active zones operate essentially identically across synapses despite the notion that the parameters they control , RRP and P , vary greatly between different types of synapses . Thus far , this assumption has proven principally true for genetic analyses of active zone protein function . For example , it is widely accepted that Munc13 is required at all synapses to prime vesicles and RIMs anchor and activate Munc13 to support priming ( Varoqueaux et al . , 2002; Andrews-Zwilling et al . , 2006; Deng et al . , 2011; Calakos et al . , 2004; Augustin et al . , 1999 ) . Our findings are a starting point for the dissection of synapse-specific architecture and function of the active zone . One possibility to explain synapse-specific roles is that ELKS1α and ELKS2α account for different functions , and that they are localized to specific subsets of synapses . Such differences have been observed for roles of Munc13-1 and bMunc13-2 in short-term plasticity ( Rosenmund et al . , 2002 ) . Our data make this possibility unlikely in the case of ELKS , because there is a strong positive correlation between ELKS1α and ELKS2α levels at all synapses . Furthermore , no ELKS isoform-specific protein interactions are described in the literature , and the interaction sequences are generally well conserved between ELKS1α and ELKS2α ( Monier et al . , 2002; Wang et al . , 2002 ) . Another possibility is that interaction partners of ELKS that mediate RRP control are distributed in a synapse specific fashion , and ELKS protein interactions at the active zone may be engaged differentially depending on the presence of specific interacting proteins . This model is supported by the observation that the sequence requirements for rescue of the excitatory RRP or the IPSC are different . ELKS interacts in vitro with many active zone proteins ( Ohtsuka et al . , 2002; Wang et al . , 2002; Schoch and Gundelfinger , 2006 ) . We find that ELKS N-terminal , but not C-terminal , sequences are required for RRP control . These N-terminal sequences have been shown to bind to Liprin-α via CCA-CCC ( Ko et al . , 2003 ) , to Bassoon via CCC ( Takao-Rikitsu et al . , 2004 ) , and they may mediate binding to Rab6 ( Monier et al . , 2002 ) . Because constructs including CCC but lacking CCA/B localize to synapses but do not to rescue , one possible explanation is that Liprin-α binding is required for ELKS function in controlling the excitatory RRP . Vertebrate genomes contain four genes encoding Liprin-α , but it is not known whether Liprin-α isoforms are differentially distributed to specific active zones . In fact , it is currently not clear for any of the vertebrate Liprin-α isoforms whether there is a tight association with the presynaptic active zone ( Zurner et al . , 2011; Spangler et al . , 2011 ) . Functional roles of vertebrate Liprin-α in synaptic transmission are not well understood , but a recent study supports presynaptic roles for Liprin-α2 at hippocampal synapses ( Spangler et al . , 2013 ) . In invertebrates , Liprin-α/syd-2 has synaptogenic activities , and effects of a gain of function mutation in syd-2 require ELKS ( Dai et al . , 2006; Patel et al . , 2006 ) , indicating important synaptic roles for Liprin-α/ELKS interactions . Thus , these previous studies are consistent with the hypothesis that ELKS controls RRP through Liprin-α . Nevertheless , it is possible that binding to Bassoon , Rab6 , or unknown proteins mediate the role of ELKS in enhancing RRP . A recent proteomic analysis of vesicle docking complexes revealed only modest differences in the composition of docking sites at glutamatergic compared to GABAergic synapses ( Boyken et al . , 2013 ) . Interestingly , however , Bassoon and ELKS were found to be enriched in glutamatergic synapses in this study , perhaps supporting a role for these proteins for promoting excitatory transmission in concert with one another . Furthermore , a recent study showed that Bassoon regulates release indirectly via RIM-BP through a specific Ca2+ channel subunit , which could result in a synapse-specific roles for Bassoon ( Davydova et al . , 2014 ) . In summary , these studies provide support for the hypothesis that ELKS may operate with Bassoon or Liprin-α to promote release in a synapse-specific fashion . Our studies also reveal that ELKS1α and ELKS2α are not required for active zone assembly and maintenance at hippocampal synapses , consistent with studies in C . elegans ( Deken et al . , 2005; Patel et al . , 2006 ) . At the D . melanogaster neuromuscular junction , however , Brp is essential for the formation of T-bars . The ELKS homology of Brp is limited to the N-terminal half of the protein ( Kittel et al . , 2006; Monier et al . , 2002 ) , making it possible that such strong scaffolding functions are specific to Brp and absent in ELKS . Alternatively , an ELKS scaffolding function may not be detected in our experiments due to redundant scaffolding activities in other proteins , for example β-ELKS or RIM ( Schoch et al . , 2002; Kaeser et al . , 2009; Liu et al . , 2014 ) . Several studies suggested that there is a good correlation between the number of docked vesicles and the size of the RRP at hippocampal synapses ( Schikorski and Stevens , 2001; Imig et al . , 2014 ) , but electron microscopic analysis of ELKS1α/2α cDKO synapses using conventional fixation has not revealed a deficit in vesicle docking ( Liu et al . , 2014 ) . Recent work has achieved better resolution in the analysis of docking by combining high-pressure freezing with EM tomography ( Imig et al . , 2014; Siksou et al . , 2007 ) . We can currently not rule out the possibility that ELKS1α/2α cDKO causes a subtle vesicle docking phenotype undetectable by the methods used in our previous work . An alternative hypothesis to a reduction of RRP at each excitatory synapse is the possibility that a subpopulation of excitatory synapses is very strongly affected by the removal of ELKS . Because the distribution of ELKS fluorescence intensity across excitatory synapses has a single peak , it is unlikely that ELKS only operates at a subset of excitatory synapses and that these synapses are silent in the absence of ELKS . However , additional work at clearly defined synapse populations with less heterogeneity than mixed hippocampal cultures will be necessary to address these questions . Finally , although ELKS1α/2α cDKO neurons do not reveal an RRP deficit as measured by hypertonic stimuli at inhibitory synapses ( Figure 5 and Liu et al . , 2014 ) , ELKS is known to regulate RRP at these synapses . Enhancing P by increasing extracellular Ca2+ cannot completely rescue synaptic transmission at inhibitory ELKS1α/2α cDKO synapses ( Liu et al . , 2014 ) , and removal of ELKS2α alone leads to an increase in the RRP at inhibitory synapses ( Kaeser et al . , 2009 ) . These experiments suggest that there may be molecular regulation of the RRP at inhibitory synapses through interplay between ELKS1 and ELKS2 , which are both present at inhibitory active zones . In the long-term , it will be important to understand how the synapse-specific molecular control of RRP and P contribute to circuit function . Human genetic experiments reveal that mutations in ERC1/ELKS1 may contribute to autism spectrum disorders ( Silva et al . , 2014 ) , and it is possible that the pathophysiology arises from synapse-specific misregulation of neurotransmitter release . All experiments using mice were performed according to institutional guidelines at Harvard University . Conditional double knockout ( cDKO ) mice that remove the ELKS1α/2α proteins were generated by crossing conditional knockout mice for the Erc1 ( [Liu et al . , 2014] RRID:IMSR_JAX:015830 ) and Erc2 ( [Kaeser et al . , 2009] RRID:IMSR_JAX:015831 ) genes . ELKS1α/2α cDKO mice were maintained as double homozygote line . ELKS2α specific antibodies were raised in rabbits using an ELKS2 peptide ( 109LSHTDVLSYTDQ120 ) . Peptides were synthesized and conjugated to keyhole lympet hemocynanin ( KLH ) via an N-terminal cysteine residue . Rabbits were inoculated at Cocalico Biologicals with KLH-conjugated ELKS2 peptides and given booster injections every two weeks , and bleeds were collected every three weeks . Sera were screened against protein samples harvested from cultured neurons , brain homogenate , and transfected HEK cells expressing either ELKS1αB or ELKS2αB . β-actin was used as a loading control . The serum with the highest immunoreactivity ( rabbit 1029 , bleed 5 ) against ELKS2 was affinity purified with the ELKS2 peptide coupled to an affinity column and used at 1:100 dilution . Primary mouse hippocampal cultures from newborn pups were generated as previously described ( Kaeser et al . , 2008 , 2011; Maximov et al . , 2007 ) . All lentiviruses were produced in HEK293T cells by Ca2+ phosphate transfection . Neurons were infected with viruses that express cre recombinase or an inactive cre truncation mutant under the human synapsin promoter ( Liu et al . , 2014 ) . Neuronal cultures were infected with 125–250 μl of HEK cell supernatant at 3–5 days in vitro ( DIV ) . Infection efficiency was monitored by an EGFP tag attached to nuclear cre , and only cultures in which no non-infected cells were found were used for experiments . Expression of rescue proteins was achieved with a second lentivirus driven in neurons by a human synapsin promoter and applied to the neurons at DIV 3 . Expression of rescue proteins was monitored by Western blotting and by immunostaining as described below . Neurons were fixed in 4% paraformaldehyde/phosphate-buffered saline , permeabilized in 0 . 1% Triton X-100/3% bovine serum albumin/phosphate-buffered saline , and incubated in primary antibodies overnight . The following primary antibodies were used: E-1 ( 1:500 , RRID:AB_10841908 ) , mouse monoclonal antibody , binds to ELKS1α isoforms ( ELKS1αB , ELKS1αA ) ; 1029 ( 1:100 ) , custom rabbit polyclonal antibody generated against ELKS2α ( 109LSHTDVLSYTDQ120 ) ; mouse anti-RIM ( 1:500 , RRID:AB_10611855 ) ; rabbit anti-Munc13-1 ( 1:5000; a gift from Dr . Nils Brose ) ; mouse anti-Bassoon ( 1:500 , RRID:AB_11181058 ) ; rabbit anti-RIM-BP2 ( 1:500 , SySy , #316103 ) ; rabbit anti-VGAT ( 1:500 , RRID:AB_887869 ) ; guinea pig anti-VGAT ( 1:500 , RRID:AB_887873 ) ; mouse anti-GAD2 ( 1:500 , RRID:AB_2107894 , also called GAD65 ) ; guinea pig anti-VGluT1 ( 1:500 , RRID:AB_887878 ) ; mouse anti-GAD1 ( 1:1000 , RRID:AB_2278725 , also called GAD67 ) . Secondary antibodies conjugated to Alexa Fluor 488 , 546 , or 633 were used for detection . Images were acquired on Olympus FV1000 or FV1200 confocal microscopes with 60x oil immersion objectives with 1 . 4 numerical aperture , the pinhole was set to one airy unit , and identical settings were applied to all samples within an experiment . Single confocal sections were analyzed in ImageJ software ( NIH ) . Background was subtracted using the rolling ball method with a radius of 2 μm . For quantification of synaptic protein levels , regions of interest ( ROIs ) were defined using VGAT , GAD2 , or VGluT1 puncta and the average intensity of the protein of interest ( in the 488 channel ) inside those ROIs was quantified . In Figure 1 , since differences in antibody affinity make raw fluorescence values between two different antibodies non-comparable , individual data points were calculated for each channel ( ELKS1α or ELKS2α ) by normalizing the fluorescence intensity within a single ROI to the average intensity across all ROIs . In Figure 6 , the intensity of the ELKS1α , RIM , Bassoon , and RIM-BP2 staining in cDKO neurons was normalized to the staining in control neurons . In all other figures where only ELKS1α staining is quantified , data are expressed in arbitrary units ( a . u . ) . When necessary , representative images were enhanced for brightness and contrast to facilitate visual inspection; all such changes were made after analysis and were made identically for all experimental conditions . All quantitative data are derived from ≥3 cultures; 5–10 fields of view were quantified per culture per genotype . For all image acquisition and analyses comparing two or more conditions , the experimenter was blind to the condition . Western blotting was performed according to standard protocols . After SDS-Page electrophoresis , gels were transferred onto nitrocellulose membranes and blocked in 10% ( w/v ) non-fat milk/5% ( v/v ) goat serum . Membranes were incubated with primary antibodies in 5% ( w/v ) non-fat milk/5% ( v/v ) goat serum for two hours at room temperature or overnight at 4°C . The following primary antibodies were used: mouse anti-ELKS1α ( 1:1000; E-1 , RRID: AB_10841908 ) , rabbit anti-ELKS2α ( 1:500; custom antibody 1029 ) , rabbit anti-ELKS1/2 ( 1:2000; P224 , gift of Dr . Thomas Südhof ) , mouse anti-β-actin ( 1:2000; RRID: AB_476692 ) . After washing , membranes were incubated with HRP-conjugated secondary antibodies in 5% ( w/v ) non-fat milk/5% ( v/v ) goat serum for one hour at room temperature , and chemiluminescence was used for detection after washing . Electrophysiological recordings in cultured hippocampal neurons were performed as described ( Kaeser et al . , 2008 , 2009 , 2011; Maximov et al . , 2007; Liu et al . , 2014 ) at DIV 15–19 . The extracellular solution contained ( in mM ) : 140 NaCl , 5 KCl , 2 CaCl2 , 2 MgCl2 , 10 HEPES-NaOH ( pH 7 . 4 ) , 10 Glucose ( ~310 mOsm ) . For evoked NMDAR excitatory postsynaptic currents ( EPSCs ) picrotoxin ( PTX , 50 μM ) and 6-Cyano-7-nitroquinoxaline-2 , 3-dione ( CNQX , 20 μM ) were added to the bath . For miniature EPSC recordings and RRP measurements tetrodotoxin ( TTX , 1 μM ) was added to block action potentials , in addition to PTX ( 50 μM ) for EPSCs or D- ( - ) -2-Amino-5-phosphonopentanoic acid ( APV , 50 μM ) and CNQX ( 20 μM ) for IPSCs . All recordings were performed in whole cell patch clamp configuration at room temperature . Glass pipettes for were pulled at 2–4 MΩ and filled with intracellular solutions containing ( in mM ) for EPSC recordings: 120 Cs-methanesulfonate , 10 EGTA , 2 MgCl2 , 10 HEPES-CsOH ( pH 7 . 4 ) , 4 Na2-ATP , 1 Na-GTP , 4 QX314-Cl ( ~300 mOsm ) and for IPSC recordings in Figure 1: 120 CsCl , 5 NaCl , 10 EGTA , 1 MgCl2 , 10 Sucrose , 10 Hepes-CsOH ( pH 7 . 4 ) , 4 Mg-ATP , 0 . 4 GTP , 4 QX314-Cl ( ~300 mOsm ) and Figure 5: 40 CsCl , 90 K-Gluconate , 1 . 8 NaCl , 1 . 7 MgCl2 , 3 . 5 KCl , 0 . 05 EGTA , 10 HEPES-CsOH ( pH 7 . 4 ) , 2 MgATP , 0 . 4 Na2-GTP , 10 Phosphocreatine , 4 QX314-Cl ( ~300 mOsm ) . Cells were held at at −70 mV for AMPAR-EPSC and IPSC recordings , at +40 mV for NMDAR-EPSC recordings and . Access resistance was monitored during recording and cells were discarded if access exceeded 15 MΩ or 20 MΩ during recording of evoked or spontaneous synaptic currents , respectively . Action potentials in presynaptic neurons were elicited with a bipolar focal stimulation electrode fabricated from nichrome wire . The RRP in Figure 4 and 5 was measured by application of 0 . 5 M sucrose in extracellular solution applied via a microinjector syringe pump for 10 s at a flow rate of 10 μL/min . For Figure 5—figure supplement 2 , the RRP was measured by focal application of 0 . 5 M sucrose with a picospritzer for 30 s in the presence of TTX ( 1 μM ) , as described in Liu et al . ( 2014 ) . Data were acquired with an Axon 700B Multiclamp amplifier and digitized with a Digidata 1440A digitizer . For action potential and sucrose-evoked responses , data were acquired at 5 kHz and low-pass filtered at 2 kHz . For miniature recordings data were acquired at 10 kHz . All data acquisition and analysis was done using pClamp10 . For all electrophysiological experiments , the experimenter was blind to the genotype throughout data acquisition and analysis . All Ca2+-imaging experiments were done in cultured hippocampal neurons infected with lentiviruses ( expressing active cre or inactive cre ) at DIV 5 . Presynaptic Ca2+ transients were examined at DIV15 - 18 in whole cell patch clamp configuration at room temperature . The extracellular solution contained ( in mM ) : 140 NaCl , 5 KCl , 2 CaCl2 , 2 MgCl2 , 10 Glucose , 0 . 05 APV , 0 . 02 CNQX , 0 . 05 PTX , 10 HEPES-NaOH ( pH 7 . 4 , ~310 mOsm ) . Glass pipettes were filled with intracellular solution containing ( in mM ) 140 K Gluconate , 0 . 1 EGTA , 2 MgCl2 , 4 Na2ATP , 1 NaGTP , 0 . 3 Fluo-5F , 0 . 03 Alexa Fluor 594 , 10 HEPES-KOH ( pH 7 . 4 , ~300 mOsm ) . Neurons were filled for 7 min and axons and dendrites were identified in the red channel . Presynaptic boutons were identified by their typical bead-like morphology . Neurons in which the distinction between axons and dendrites was unclear were discarded . 10 min after break-in , presynaptic Ca2+ transients were induced by a single action potential evoked via somatic current injection ( 5 ms , 800–1200 pA ) and monitored via Fluo-5F fluorescence . Images were acquired using an Olympus BX51 microscope with a 60x , 1 . 0 numerical aperture objective . Fluorescence signals were excited by a light-emitting diode at 470 nm , and were collected with a scientific complementary metal–oxide–semiconductor camera at 100 frames/s for 200 ms before and 1s after the action potential . After the experiment , coverslips were fixed using 4% paraformaldehyde/phosphate-buffered saline . Neurons were stained with GAD67 antibodies as described above and imaged neurons , identified post-hoc by their Alexa 594 filling , were identified as either inhibitory or excitatory neurons using confocal microscopy . In additional experiments ( Figure 3—figure supplement 1 ) , distinction between excitatory and inhibitory cultured neurons was further characterized by labeling excitatory neurons using a lentivirus expressing tandem dimer tomato ( TdTomato ) under a CamKII promoter ( pFCK ( 1 . 3 ) tdimer2W; gift from Pavel Osten; Addgene plasmid #27233 , [Dittgen et al . , 2004] ) . Ca2+ transients were quantified using ImageJ . For the analysis in boutons , ROIs were defined using pictures taken in the red channel , and 7–10 boutons were randomly selected from each neuron . Background was subtracted using the rolling ball method with a radius of 1 . 5 μm . After background subtraction , ( F−F0 ) /F0 was calculated ( F = average green emission in a bouton at a given time point , F0 = average fluorescent intensity in frames 0 to 20 before action potential induction ) . For dendritic measurements , a second order dendrite was selected from each neuron , and the fluorescence from a nearby empty region was referred to as background and subtracted from the ROI . For all Ca2+ imaging experiments , the experimenter was blind to the genotype throughout data acquisition and analysis . Statistical significance was set at p≤0 . 05 , p≤0 . 01 , and **p≤0 . 001 . Unless otherwise noted in the figure legends , all tests were performed using Student’s t tests to compare means . In cases where significance was determined by one-way ANOVA , Holm-Sidak multiple comparisons tests were used to compare all conditions against the cDKO condition to assess rescue . All experiments were done with using a minimum of three independent cultures and in each culture multiple cells ( typically 5–10 per culture and genotype ) or images ( typically 5–10 images per culture and genotype ) were analyzed .	Nerve cells in the brain communicate with one another at connections known as synapses: one nerve cell releases signaling molecules called neurotransmitters into the synapse , which are then sensed by the second cell . For the brain to work correctly , it is important that the nerve cells control when and how much neurotransmitter they release . Nerve cells package neurotransmitters into small packets called vesicles . These vesicles can be released at the so-called active zones of each synapse , though only a small subset of vesicles at a synapse are releasable . Many proteins at the active zone control the release of vesicles to influence how nerve cells communicate with each another . ELKS is one of the proteins found at the active zones of nerve cells that release either of the two most common neurotransmitters in the brain: glutamate and GABA . Held et al . have now found that the ELKS protein affects the release of these two neurotransmitters in different ways in the two types of nerve cells . The experiments showed that the number of releasable neurotransmitter-filled vesicles was lower in mouse nerve cells that release glutamate when the genes for the ELKS proteins were deleted in these cells . When the ELKS genes were deleted in the nerve cells that release GABA , the number of releasable vesicles remained the same , though the vesicles were less likely to be released . The fact that removing ELKS has different effects at these two types of synapses suggests that the active zone is not the same at all synapses . Furthermore , these results imply that ELKS is capable of fine-tuning the communication between nerve cells . Future experiments will address how glutamate- and GABA-releasing active zones differ at the molecular and structural levels . Ultimately , this will lead to a better understanding of how information is processed in the brain .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	ELKS controls the pool of readily releasable vesicles at excitatory synapses through its N-terminal coiled-coil domains
Neurotransmitter corelease is emerging as a common theme of central neuromodulatory systems . Though corelease of glutamate or GABA with acetylcholine has been reported within the cholinergic system , the full extent is unknown . To explore synaptic signaling of cholinergic forebrain neurons , we activated choline acetyltransferase expressing neurons using channelrhodopsin while recording post-synaptic currents ( PSCs ) in layer 1 interneurons . Surprisingly , we observed PSCs mediated by GABAA receptors in addition to nicotinic acetylcholine receptors . Based on PSC latency and pharmacological sensitivity , our results suggest monosynaptic release of both GABA and ACh . Anatomical analysis showed that forebrain cholinergic neurons express the GABA synthetic enzyme Gad2 and the vesicular GABA transporter ( Slc32a1 ) . We confirmed the direct release of GABA by knocking out Slc32a1 from cholinergic neurons . Our results identify GABA as an overlooked fast neurotransmitter utilized throughout the forebrain cholinergic system . GABA/ACh corelease may have major implications for modulation of cortical function by cholinergic neurons . For many years , neurons were thought to release only a single fast neurotransmitter ( Strata and Harvey , 1999 ) . This assumption led to classifying neuronal subtypes based on released neurotransmitter , a convention which helped predict a neuron's circuit function . However , many neuronal subtypes that release multiple fast neurotransmitters have now been described ( Hnasko and Edwards , 2012 ) . In some cases , the coreleased neurotransmitters have similar post-synaptic effects , such as inhibition mediated by GABA and glycine from spinal interneurons ( Jonas et al . , 1998 ) . In other instances , the effects of the two neurotransmitters may be different and synergistic . For example , coreleased GABA and glutamate are thought to control the balance of excitation and inhibition in the lateral habenula ( Root et al . , 2014; Shabel et al . , 2014 ) . In neuromodulatory systems , synaptic release of fast neurotransmitters along with slow neuromodulators has emerged as a common theme . In addition to the impact of dopamine , stimulation of dopaminergic terminals from the ventral tegmental area and substantia nigra compacta activates glutamate-mediated excitatory currents in the nucleus accumbens ( Stuber et al . , 2010; Tecuapetla et al . , 2010 ) and GABA-mediated inhibitory currents in the striatum ( Tritsch et al . , 2012 , 2014 ) . Likewise , serotonergic neurons of the dorsal raphe can trigger glutamate-mediated currents in post-synaptic neurons of the ventral tegmentum and nucleus accumbens which contributes to the signaling of reward ( Liu et al . , 2014 ) . Several cholinergic neuron populations also release multiple neurotransmitters . Retinal starburst amacrine cells ( SACs ) differentially release GABA and acetylcholine ( ACh ) based on the pattern of light stimulation ( Lee et al . , 2010 ) . In the central brain , selective activation of striatal cholinergic interneurons results in cholinergic and glutamatergic responses ( Gras et al . , 2008; Higley et al . , 2011; Nelson et al . , 2014 ) . Similarly , the cholinergic projection from habenula excites interpeduncular neurons through glutamate and ACh ( Ren et al . , 2011 ) . Some evidence suggests that the basal forebrain cholinergic ( BFC ) system , which provides the major source of ACh to cortex , may corelease GABA . The dorsal-most BFC neurons , which belong to the globus pallidus externus ( GP ) , express molecular markers for GABA synthesis and vesicular packaging and trigger GABAA receptor currents in GP and cortex when activated ( Tkatch et al . , 1998; Saunders et al . , 2015 ) . We therefore asked whether GABA corelease was a general feature of forebrain cholinergic neurons . To address this question , we selectively activated cholinergic fibers in the cortex , with the goal of identifying synaptic events triggered by endogenous release from forebrain cholinergic neurons . Recording from layer 1 interneurons , we observed not only the expected excitatory post-synaptic currents ( EPSCs ) mediated by nicotinic ACh receptors ( nAChRs ) , but an unexpected inhibitory post-synaptic current ( IPSC ) mediated by GABAA receptors . IPSCs insensitive to nAChR antagonists had onset latencies slightly faster than the nicotinic EPSCs ( nEPSCs ) , and could be directly evoked under pharmacological conditions in which action potentials were blocked , suggesting cholinergic neurons were directly releasing GABA in addition to ACh . Indeed , we found that cholinergic neurons throughout the forebrain commonly coexpressed the GABA synthetic enzyme GAD65 ( Gad2 ) , and the vesicular GABA transporter ( Slc32a1 ) , indicating that these neurons possess the necessary cell machinery for GABA transmission . Finally , we show that conditional deletion of Slc32a1 selectively in cholinergic neurons eliminates monosynaptic IPSCs while leaving nEPSCs intact , confirming the direct release of GABA from cholinergic terminals . These experiments suggest a previously overlooked capability of the cholinergic system to use GABA in synaptic signaling . To explore the effects that cholinergic neurons have on cortical circuitry , we used double transgenic mice to optogenetically activate neurons that express endogenous choline acetyltransferase ( Chat ) . These mice carried a knock-in allele linking Cre recombinase expression to the Chat locus through an internal-ribosome entry site ( Chat i-Cre ) as well as a Cre-activated channelrhodopsin-EYFP fusion allele ( Rosa26 lsl-ChR2-EYFP ) . Chat i-Cre; Rosa26 lsl-ChR2-EYFP mice expressed ChR2-EYFP throughout the forebrain , recapitulating known patterns of Chat expression in cortex ( Ctx ) , striatum , globus pallidus externus ( GP ) , and nucleus basalis ( NB , Figure 1A ) . To test whether ChR2+ cells express endogenous Chat , we performed ChAT immunohistochemistry on sections of Chat i-Cre; Rosa26 lsl-ChR2-EYFP mice ( Figure 1B ) . We focused on those Chat+ forebrain neurons positioned to innervate the cortex , including local Chat+ interneurons and the subcortical projections arising from the GP/NB . In both regions , ChR2+ neurons were immunopositive for ChAT , confirming our ability to selectively activate endogenous Chat+ inputs to cortex . 10 . 7554/eLife . 06412 . 003Figure 1 . Optogenetic stimulation of cortical Chat+ fibers evokes fast , monosynaptic GABAA receptor-mediated currents . ( A ) Low-magnification view of sagittal section from Chat i-Cre; Rosa26 lsl-ChR2-EYFP mouse forebrain . ChR2-EYFP ( green ) is expressed in the nucleus basalis ( NB ) , globus pallidus externus ( GP ) , striatum and cortex ( Ctx ) . ( B ) Higher-magnification view of ChR2-EYFP expression combined with ChAT immunostaining ( magenta ) in frontal cortex ( top ) and nucleus basalis ( bottom ) . Arrowheads indicate cells immunopositive for ChAT . ( C ) Example 2-photon stack from a layer 1 interneuron following whole-cell recording and dialysis with Alexa Fluor 594 . ( D ) High-magnification view of ChR2-EYFP+ somata and fibers in layers 1 and 2/3 of frontal cortex . NeuN immunostain ( magenta ) highlights the distribution of neuronal somata across layers . Arrowhead indicates an example layer 1 interneuron surrounded by ChR2-EYFP fibers . ( E ) Example PSCs from voltage-clamp recordings of three different layer 1 interneurons in response to blue light stimulation ( blue bar ) of ChR2+ cholinergic fibers . Neurons were voltage-clamped at −70 mV ( left ) to isolate EPSCs or at 0 mV ( middle and right ) to isolate IPSCs . PSCs recorded in the presence of glutamate receptor antagonists CPP and NBQX are shown in black and after bath application of nAChR antagonists ( MEC , MLA , and DHβE ) in red . ( F ) Example light-evoked IPSCs from a layer 1 interneuron in a Chat i-Cre; Rosa26 lsl-ChR2-EYFP mouse voltage-clamped at 0 mV in the presence of CPP and NBQX ( baseline ) and following subsequent bath application of ( from left to right ) nAChR antagonists , TTX , 4AP , and SR95531 . ( G ) Summary graph of IPSC peaks normalized to baseline ( n = 5 cells from 4 mice ) . Asterisk , condition vs baseline p < 0 . 05 , Mann–Whitney test ) . ( H ) Onset latencies for monosynaptic nEPSCs ( n = 41 cells ) , monosynaptic IPSCs ( n = 9 cells ) , and polysynaptic IPSCs ( n = 19 cells from 9 mice ) . Mean ( ±sem ) are shown in green . Asterisk , p < 0 . 05 , Mann–Whitney test . DOI: http://dx . doi . org/10 . 7554/eLife . 06412 . 003 To identify the synaptic signaling mechanisms engaged by activation of the cortical cholinergic system , we targeted layer 1 interneurons for whole-cell voltage-clamp recordings in acute brain slices . Layer 1 is strongly innervated by ChAT+ cells of the basal forebrain across species ( Mesulam , 1995; Mechawar et al . , 2000 ) including in Chat i-Cre; Rosa26 lsl-ChR2-EYFP mice , where ChR2-EYFP is expressed in a dense plexus ( Figure 1C , D ) . As expected , in a subset of interneurons ( n = 41 of 58 cells from 9 mice ) , we observed robust excitatory postsynaptic currents ( EPSCs ) at −70 mV in response to brief pulses of blue light ( 2–7 ms , Figure 1E , left ) . These EPSCs were not blocked by NBQX and CPP , ruling out a glutamatergic source , but were blocked by the nicotinic ACh receptor ( nAChR ) antagonists DHβE , MLA , and MEC , confirming their cholinergic identity ( nEPSCs ) . nEPSCs displayed a typical biphasic response , with a fast component , likely mediated by synaptic receptors containing the low-affinity α7 nAChR subunit , and a slow component , likely mediated by non-synaptic receptors expressing the high-affinity non-α7 subunits ( Bennett et al . , 2012 ) . In addition to the expected nEPSCs recorded at −70 mV , we also observed barrages of outward inhibitory postsynaptic currents ( IPSCs ) at 0 mV , indicative of signaling through GABA receptors ( n = 28 of 58 cells , Figure 1E , center ) . One possible explanation for these IPSCs could be ACh-mediated feed-forward activation of local interneurons , resulting in disynaptic release of GABA . Indeed , when nAChR antagonists were applied , the delayed outward IPSCs disappeared . However , in a subset of recorded cells IPSCs remained ( n = 9 of 58 cells , Figure 1F , right ) , suggesting that these PSCs were not dependent on nAChR signaling . To test if nAChR-resistant IPSCs are caused by direct release of GABA from cholinergic fibers , we bath applied the voltage-gated sodium channel antagonist TTX , which blocked light-evoked IPSCs ( Figure 1F , G ) . In the presence of TTX , IPSCs could be rescued by enhancing ChR2-mediated depolarization with the voltage-gated potassium channel blocker 4AP . Rescued IPSCs were subsequently blocked by the GABAA receptor antagonist SR95531 ( n = 5 cells from 4 mice ) . Moreover , nAChR-resistant ‘direct’ IPSCs had faster average onsets than both nEPSCs and nAChR-sensitive ‘indirect’ IPSCs ( nEPSCs , 4 . 0 ± 0 . 2 ms , n = 41 cells; direct IPSCs , 2 . 5 ± 0 . 2 , n = 9 cells; indirect IPSCs , 11 . 8 ± 2 . 3 , n = 19 cells from 9 mice , Figure 1H ) . These data suggest direct IPSCs are independent of nAChR signaling and mediated by GABAA receptors , consistent with monosynaptic release of GABA by cholinergic neurons . In support of this possibility , gene expression analyses have suggested some populations of Chat+ subcortical neurons contain the synthetic machinery for GABA ( Kosaka et al . , 1988; Tkatch et al . , 1998 ) . GABA corelease has been observed in other neuromodulatory systems , namely from dopaminergic neurons of the substantia nigra ( Tritsch et al . , 2012 ) . In those neurons , GABA is co-packaged with dopamine into vesicles by the transporter VMAT2 ( Slc18a2 ) , instead of by the typical vesicular transporter for GABA ( VGAT , encoded by Slc32a1 ) , which is necessary for packaging in most GABAergic neurons ( Wojcik et al . , 2006; Tong et al . , 2008; Kozorovitskiy et al . , 2012 ) . To assess whether cholinergic neurons could use VGAT to package GABA into vesicles , we tested for Slc32a1/ChAT co-expression throughout the brain , including in cortex , GP , NB , diagonal band of broca ( DBB ) /medial septum ( MS ) , and pedunculopontine nucleus ( PPN , Figure 2A , top ) . We labeled cells expressing endogenous Slc32a1 using double-transgenic mice that link Cre recombinase expression to the Slc32a1 locus ( Slc32a1i-Cre ) and carry a zsGreen Cre reporter allele ( Rosa26 lsl-zsGreen ) . Subsequent ChAT immunolabeling on sagittal and coronal sections from Slc32a1i-Cre; Rosa26 lsl-zsGreen mice was used to examine coexpression . Since zsGreen accumulates in somata and does not diffuse throughout the cytoplasm , this strategy allowed the clear identification of ChAT+ soma free from background fluorescence caused by Cre+ axons and dendrites . In cortex , GP , NB , and MS/DBB , nearly all of ChAT+ cell were Slc32a1+ ( zsGreen+/ChAT+ , from 3 mice: Ctx , 628/628; GP , 238/243; NB , 598/624; MS/DBB; 560/601 , Figure 2A , B ) . In contrast , ChAT+ cells of the PPN very rarely expressed Slc32a1 ( 3/157 , Figure 2B ) . These data suggest that in both cortical and subcortical forebrain regions , ChAT+ cells express the canonical molecular machinery to package GABA into vesicles . However , since ChAT+ neurons in the brainstem PPN do not express Slc32a1 , this GABAergic marker is not a ubiquitous feature of the central cholinergic system . 10 . 7554/eLife . 06412 . 004Figure 2 . Immunopositive ChAT cells in the forebrain express Slc32a1 . ( A ) Top , sagittal and coronal schematic views of a mouse brain showing cholinergic regions of interest . Red boxes indicate approximate locations for magnified regions below . Bottom , example single-plane image from a confocal stack from sections of a Slc32a1i-Cre; Rosa26 lsl-zsGreen mouse immunostained for ChAT ( magenta ) and reporting Cre expression ( green ) . Ctx , cortex; GP , globus pallidus externus; NB , nucleus basalis; MS/DBB , medial septum/diagonal band of broca; PPN , pedunculopontine nucleus . Insets show magnified view of individual neurons indicated by the white arrowhead . ( B ) Quantification of colocalization between cells expressing zsGreen Cre reporter and ChAT immunostain by brain region ( zsGreen+/ChAT+ , from 3 mice: Ctx , 628/628; GP , 238/243; NB , 598/624; MS/DBB , 560/601; PPN , 3/157 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06412 . 004 GABA is synthesized by one of two GABA synthetic enzymes , GAD65 or GAD67 , encoded by the genes Gad2 or Gad1 , respectively . GAD67 is expressed largely in cell bodies and is thought to be responsible for synthesizing GABA for general metabolic cell functions , whereas GAD65 expression is more prominent in axon terminals and is thought to mediate the majority of synthesis of synaptic GABA ( Soghomonian and Martin , 1998 ) . We examined if cholinergic neurons expressed GAD67 by immunostaining for ChAT in brain sections from knock-in mice where GFP replaces the first exon of the Gad1 gene ( Gad1GFP ) . We detected only minor overlap between neurons that stained positive for ChAT and those that expressed GFP in the NB , GP , MS/DBB , or PPN , except for cortical ChAT+ interneurons , where we observed significant overlap ( GFP+/ChAT + neurons: Ctx: 122/136; MS/DBB , 1/573; NB , 0/439; GP , 7/413; PPN , 12/246 , Figure 3—figure supplement 1 ) . This suggests that within the major subcortical cholinergic projections , GAD67-mediated GABA synthesis does not occur in cholinergic neurons . To test for coexpression of ChAT and GAD65 , we immunostained sections of knock-in mice where Cre recombinase was targeted to the endogenous Gad2 gene ( Gad2 i-Cre ) and visualized Cre expression with the zsGreen Cre reporter . In contrast to Gad1 , in cortex , GP , NB , and MS/DBB of Gad2 i-Cre; Rosa26 lsl-zsGreen mice , most ChAT+ neurons were Gad2+ ( zsGreen+/ChAT+ , from 4 mice: Ctx , 518/519; GP , 273/372; NB , 860/934; MS/DBB , 673/685 , Figure 3A , B ) . In the brainstem , however , few of the ChAT+ cells were Gad2+ ( PPN , 6/110 , Figure 3A , B ) . This ChAT co-expression pattern for Gad2 is similar to Slc32a1 , suggesting that forebrain cholinergic neurons possess the necessary cellular machinery to both synthesize and package synaptic GABA . 10 . 7554/eLife . 06412 . 005Figure 3 . Immunopositive ChAT cells in the forebrain express Gad2 . ( A ) Top , sagittal and coronal schematic views of a mouse brain showing cholinergic regions of interest . Red boxes indicate approximate locations for magnified regions below . Bottom , example single-plane image from a confocal stack from sections of a Gad2 i-Cre; Rosa26 lsl-zsGreen mouse immunostained for ChAT ( magenta ) and reporting Cre expression ( green ) . Ctx , cortex; GP , globus pallidus externus; NB , nucleus basalis; MS/DBB , medial septum/diagonal band of broca; PPN , pedunculopontine nucleus . Insets show magnified view of individual neurons indicated by the white arrowhead . ( B ) Quantification of colocalization between cells expressing zsGreen Cre reporter and ChAT immunostain by brain region ( zsGreen+/ChAT+ , from 4 mice: Ctx , 518/519; GP , 273/372; NB , 860/934;MS/DBB , 673/685; PPN , 6/110 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06412 . 00510 . 7554/eLife . 06412 . 006Figure 3—figure supplement 1 . Immunopositive ChAT cells of the cortex but not major subcortical projections express Gad1 . ( A ) Top , sagittal and coronal schematic views of a mouse brain showing cholinergic regions of interest . Red boxes indicate approximate locations for magnified regions below . Bottom , example single-plane image from a confocal stack from sections of a Gad1GFP ( green ) mouse immunostained for ChAT ( magenta ) . Ctx , cortex; GP , globus pallidus externus; NB , nucleus basalis; MS/DBB , medial septum/diagonal band of broca; PPN , pedunculopontine nucleus . Insets show magnified view of individual neurons indicated by the white arrowhead . ( B ) Quantification of colocalization between cells expressing GFP and ChAT immunostain by brain region ( GFP+/ChAT+ , from 4 mice: Ctx , 122/135; GP , 7/413; NB , 0/439;MS/DBB , 1/573; PPN , 12/246 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06412 . 006 Though the fast onset and persistence of GABAergic IPSCs in the presence of nAChR antagonists and TTX/4AP argues strongly for direct release of GABA from cholinergic terminals , we wished to confirm this finding using conditional genetics . To determine if GABA release is indeed monosynaptic , we took advantage of the observation that cholinergic neurons express Slc32a1 . If GABA is released directly from cholinergic neurons , then conditional knock-out of Slc32a1 selectively in these cells should abolish GABA release . We therefore bred triple transgenic mice which carried conditional ( floxed ) Slc32a1 alleles in addition to Chat i-Cre and Rosa26 lsl-ChR2-EYFP . These mice allowed us to compare the optogenetic stimulation of wild-type cholinergic neurons ( Slc32a1+/+ ) to those lacking Slc32a1 ( Slc32a1fl/fl ) in acute brain slices . In voltage-clamp recordings from layer 1 interneurons , we observed fast onset IPSCs in 31% of cells recorded in Chat i-Cre; Slc32a1+/+ mice ( n = 5 of 16 from 2 mice ) , but no direct IPSCs in Chat i-Cre; Slc32a1fl/fl mice ( n = 0 of 34 cells from 4 mice , Figure 4A , B ) . In contrast , the proportion and average peak amplitude of nAChR responses remains unchanged between Slc32a1+/+ and Slc32a1fl/fl mice ( Figure 4A , B ) . These data demonstrate that GABA but not ACh release from cholinergic neurons relies on cell autonomous Slc32a1 , ruling out a disynaptic mechanism . 10 . 7554/eLife . 06412 . 007Figure 4 . GABA release from cortical ChAT+ axons requires Slc32a1 . ( A ) Example light-evoked nEPSCs and IPSCs from four different layer 1 interneurons voltage-clamped at −70 or 0 mV from Chat i-Cre; Rosa26 lsl-ChR2-EYFP mice with wild-type cholinergic neurons ( Slc32a1+/+ ) or conditional Slc32a1 knock-out ( Slc32a1fl/fl ) . ( B ) Top , the number and proportion of layer 1 interneurons in which light-evoked nEPSCs or direct IPSCs were detected from Chat i-Cre; Rosa26 lsl-ChR2-EYFP mice with wild-type Slc32a1 alleles ( Slc32a1+/+ , from 2 mice ) or following conditional Slc32a1 knock-out ( Slc32a1fl/fl , from 4 mice ) . Bottom , PSC peaks for each condition . Means ( ±sem ) are shown in green . DOI: http://dx . doi . org/10 . 7554/eLife . 06412 . 007 Here , we provide evidence that the cortical cholinergic system is capable of GABAergic neurotransmission . In response to optogenetic stimulation of neurons expressing endogenous Chat , we observed PSCs mediated by both nAChRs and GABAA receptors in layer 1 interneurons . A subset of the evoked IPSCs appeared to be monosynaptic , based on latency and pharmacological analyses . In support of their GABAergic nature , the ChAT+ neurons which innervate cortex—local interneurons and subcortical projections arising from the GP/NB—express Gad2 and Slc32a1 , the canonical molecular machinery for GABA synthesis and vesicular packaging . Indeed , conditional knock-out of Slc32a1 selectively in cholinergic neurons eliminates light-evoked monosynaptic IPSCs . These genetic results confirm that cholinergic neurons release GABA directly . Cholinergic GABA release is likely to be a feature of most , but not all , central cholinergic neurons: the ChAT+ neurons of the MS and DBB—which innervate the hippocampus—also express Gad2 and Slc32a1 , whereas those of the midbrain PPN do not . The mode of neurotransmitter corelease can vary across neuron classes . In some instances , both neurotransmitters are released from the same synaptic vesicles . This is the case for GABA/glutamate corelease onto neurons of the lateral habenula , where individual miniature EPSCs can be observed with dual GABA/glutamate components ( Shabel et al . , 2014 ) . Copackaging in individual vesicles is also the case when the same vesicular transporter loads both neurotransmitters . For example , GABA is packaged in dopaminergic neurons by VMAT2 , which also packages dopamine ( Tritsch et al . , 2012 ) . Similarly , both GABA and glycine packaging in spinal interneurons rely on VGAT ( Wojcik et al . , 2006 ) . However , corelease from separate pools of synaptic vesicles also occurs . In retinal SACs , release of GABA and ACh can be functionally separated through patterned light stimulation or pharmacology . In Chat+ GP axons in cortex , GABA and ACh appear to be released from distinct vesicular pools which can be located within the same or neighboring pre-synaptic terminals ( Saunders et al . , 2015 ) . Though the precise mechanism by which each forebrain cholinergic population coreleases GABA and ACh remain unclear and further experiments are merited to explore how GABA/ACh cotransmission is regulated , the differences in the proportion of layer 1 interneurons showing ACh and GABA responses suggest some form of segregated release . The cholinergic system's function in promoting attention , alertness , and learning has classically been attributed to acetylcholine and its action as a diffuse volume transmitter , affecting cortical activity at relatively slow time scales . This model is supported by anatomical evidence showing widespread distribution of cholinergic fibers through all cortical layers with significant separation between the sites of release and ACh receptors ( Descarries and Mechawar , 2000; Mechawar et al . , 2000 ) . In addition , many in vitro pharmacological experiments have shown that ACh receptors can shape the signaling of other neurotransmitter systems , by altering properties of presynaptic release , synaptic plasticity , or the intrinsic excitability of targeted neurons ( Picciotto et al . , 2012 ) . However , more recent work has focused on the participation of ACh in rapid , wired neurotransmission , acting at tightly apposed synapses ( Sarter et al . , 2009; Poorthuis et al . , 2014; Sarter et al . , 2014 ) . Behaviorally relevant sensory cues can cause a fast , time-locked spike in ACh concentration , suggesting that ACh may mediate detection of that cue ( Parikh et al . , 2007 ) . In addition , fast onset currents mediated by nAChRs can be recorded in cortical interneurons following optogenetic activation of cholinergic fibers ( Arroyo et al . , 2012; Bennett et al . , 2012 ) . Our finding that cholinergic neurons also elicit fast-onset synaptic GABAA responses lends further support to the notion that the cholinergic system can rapidly affect cortical computations by acting at classical synapses . GABA corelease from cholinergic forebrain neurons may affect cortical function in several ways . First , at the circuit level , GABA release could act in a manner that reinforces the emerging concept that the cholinergic system disinhibits cortical firing . ACh release from basal forebrain neurons excites layer 1 and VIP+ interneurons , which in turn inhibit local interneurons that target principle neurons of the cortex ( Letzkus et al . , 2011; Pinto et al . , 2013; Fu et al . , 2014 ) . Depending on the timing and targeted cell types , GABA corelease could conceivably enhance this effect by inhibiting local interneurons , thereby promoting cortical activity . Second , nAChRs can regulate both pre and post-synaptic GABAergic signaling . For example , in hippocampal interneurons , post-synaptic nAChRs are present in inhibitory synapses ( Fabian-Fine et al . , 2001 ) and when activated , reduce GABAA receptor-mediated IPSCs in a Ca2+ and PKC-dependent manner ( Wanaverbecq et al . , 2007; Zhang and Berg , 2007 ) . Thus coreleased ACh and GABA could interact to modulate local synaptic signaling . Lastly , experiments poisoning or stimulating the basal forebrain cholinergic system have demonstrated that activity within this projection is necessary and sufficient for plasticity in sensory cortices ( Kilgard and Merzenich , 1998; Weinberger , 2004; Ramanathan et al . , 2009 ) . While ACh alone can induce functional changes in cortical circuits ( Metherate and Weinberger , 1990 ) , GABA may also contribute to synaptic rewiring in vivo . Addressing these questions experimentally will benefit from future work to clarify the basic synaptic anatomy and biochemical regulation of ACh/GABA corelease . Given the presence of GABA signaling machinery throughout the distinct forebrain cholinergic systems , corelease likely has a significant and fundamental effect on brain activity and cognition . Cre recombinase was targeted to specific cell types using knock-in mice to drive Cre expression under endogenous gene-specific regulatory elements using an internal ribosome entry site . Cre knock-in mice for choline acetyltransferase ( Chat ) ( Rossi et al . , 2011 ) and vesicular GABA transporter ( Slc32a1 ) ( Tong et al . , 2008 ) were provided by Brad Lowell ( Beth Israel Deaconess Medical Center ) and are available from the Jackson Labs ( Bar Harbor , ME; Chat i-Cre , stock #006410; Slc32a1i-Cre , stock #016962 ) . Gad2 i-Cre mice were purchased from Jackson Labs ( stock #010802 ) ( Taniguchi et al . , 2011 ) . Gad1GFP knock-in mice replace elements of Gad1 coding sequence with GFP ( Tamamaki et al . , 2003 ) . We did not distinguish between mice hetero or homozygous for transgenic alleles except where indicated . All experimental manipulations were performed in accordance with a protocol ( #03551 ) approved by the Harvard Standing Committee on Animal Care following guidelines described in the US National Institutes of Health Guide for the Care and Use of Laboratory Animals . Mice aged post-natal day 25–124 were deeply anesthetized with isoflurane and transcardially perfused with 4% paraformaldehyde ( PFA ) in 0 . 1 M sodium phosphate buffer ( 1× PBS ) . Brains were post-fixed for 1–3 days , washed in 1× PBS , and sectioned ( 40–50 μm ) coronally or sagittally using a Vibratome ( Leica Microsystems , Buffalo Grove , IL ) . For ChAT or NeuN immunohistochemistry , slices were incubated in a 1× PBS blocking solution containing 5% normal horse serum and 0 . 3% Triton X-100 for 1 hr at room temperature . Slices were then incubated overnight at 4°C in the same solution containing anti-choline acetyltransferase antibody ( AB144P; 1:100; Millipore , Billerica , MA ) or NeuN ( MAB377; 1:100; Millipore ) . The next morning , sections were washed three times for five minutes in 1× PBS and then incubated for 1 hr at room temperature in the blocking solution containing donkey anti-goat Alexa 647 or Alexa 594 ( for ChAT ) or anti-mouse Alexa 647 ( for NeuN ) ( 1:500; Molecular Probes , Eugene , OR ) . Slices were then mounted on slides ( Super Frost ) . After drying , slices were coverslipped with ProLong antifade mounting media containing DAPI ( Molecular Probes ) and imaged with an Olympus VS110 slide scanning microscope using the 10× objective . Confocal images ( 1–2 μm optical sections ) were acquired with an Olympus FV1000 laser scanning confocal microscope ( Harvard Neurobiology Imaging Facility ) through a 60× objective . Colabeling quantification was carried on images obtained from the Olympus VS110 slide scanning microscope using ImageJ . Acute brain slices were obtained from mice aged post-natal day 30–128 using standard techniques . Mice were anesthetized by isoflurane inhalation and perfused through the heart with ice-cold artificial cerebrospinal fluid ( ACSF ) containing ( in mM ) 125 NaCl , 2 . 5 KCl , 25 NaHCO3 , 2 CaCl2 , 1 MgCl2 , 1 . 25 NaH2PO4 , and 11 glucose ( ∼308 mOsm·kg−1 ) . Cerebral hemispheres were removed , placed in ice-cold choline-based cutting solution ( consisting of [in mM]: 110 choline chloride , 25 NaHCO3 , 2 . 5 KCl , 7 MgCl2 , 0 . 5 CaCl2 , 1 . 25 NaH2PO4 , 25 glucose , 11 . 6 ascorbic acid , and 3 . 1 pyruvic acid ) , blocked , and transferred into a slicing chamber containing ice-cold choline-based cutting solution . Sagittal slices ( 300–350 μm thick ) were cut with a Leica VT1000s vibratome and transferred to a holding chamber containing ACSF at 34°C for 30 min and then subsequently at room temperature . Both cutting solution and ACSF were constantly bubbled with 95% O2/5% CO2 . Individual slices were transferred to a recording chamber mounted on a custom built two-photon laser scanning microscope ( Olympus BX51WI ) equipped for whole-cell patch-clamp recordings and optogenetic stimulation . Slices were continuously superfused ( 3 . 5–4 . 5 ml·min−1 ) with ACSF warmed to 32–34°C through a feedback-controlled heater ( TC-324B; Warner Instruments ) . Cells were visualized through a water-immersion 60× objective using differential interference contrast ( DIC ) illumination . Epifluorescence illumination was used to identify those layer 1 interneurons surrounded by ChR2-EYFP processes . Patch pipettes ( 2–4 MΩ ) pulled from borosilicate glass ( G150F-3; Warner Instruments ) were filled with a Cs+-based low Cl–internal solution containing ( in mM ) 135 CsMeSO3 , 10 HEPES , 1 EGTA , 3 . 3 QX-314 ( Cl–salt ) , 4 Mg-ATP , 0 . 3 Na-GTP , 8 Na2-Phosphocreatine ( pH 7 . 3 adjusted with CsOH; 295 mOsm·kg−1 ) for voltage-clamp recordings . Series resistance ( <25 MΩ ) was measured with a 5-mV hyperpolarizing pulse in voltage-clamp and left uncompensated . Membrane potentials were corrected for a ∼7 mV liquid junction potential . In some cases after the recording was complete , cellular morphology was captured in a volume stack using 740 nm two-photon laser light ( Coherent ) . To activate ChR2 in acute slices from ChAT i-Cre; Rosa26 lsl-ChR2-EYFP mice , 473 nm laser light ( Optoengine ) was focused onto the back aperture of the 60× water immersion objective to produce collimated whole-field illumination . Square pulses of laser light were delivered every 20 s and power ( 2–7 ms; 4 . 4 mW·mm−2 ) was quantified for each stimulation by measuring light diverted to a focal plane calibrated photodiode through a low-pass dichroic filter . Following bath application of TTX and 4AP , in some cases , light power or duration was increased slightly to recover currents ( e . g . , changing the duration from 2 to 4 ms ) . Drugs ( all from Tocris , United Kingdom ) were applied via bath perfusion: SR95531 ( 10 μΜ ) , tetrodotoxin ( TTX; 1 μΜ ) , 4-aminopyridine ( 4AP; 500 μM ) , scopolamine ( 10 μΜ ) , 2 , 3-dihydroxy-6-nitro-7-sulfamoyl-benzo ( f ) quinoxaline ( NBQX; 10 μM ) , R , S-3- ( 2-carboxypiperazin-4-yl ) propyl-1-phosphonic acid ( CPP; 10 μM ) , N , 2 , 3 , 3-Tetramethylbicyclo[2 . 2 . 1]heptan-2-amine , ( MEC; 10 μM ) , [1α , 4 ( S ) , 6β , 14α , 16β]-20-Ethyl-1 , 6 , 14 , 16-tetramethoxy-4-[[[2- ( 3-methyl-2 , 5-dioxo-1-pyrrolidinyl ) benzoyl]oxy]methyl]aconitane-7 , 8-diol ( MLA; 0 . 1 μM ) , ( 2S , 13bS ) -2-Methoxy-2 , 3 , 5 , 6 , 8 , 9 , 10 , 13-octahydro-1H , 12H-benzo[i]pyrano[3 , 4-g]indolizin-12-one ( DHβE; 10 μM ) . CPP and NBQX were combined to make a cocktail of antagonists to target ionotropic glutamate receptors , while MEC , MLA , and DHβE were combined to make a cocktail to antagonize nicotinic receptors . Membrane currents and potentials were recorded using an Axoclamp 700B amplifier ( Molecular Devices , Sunnyvale , CA ) filtered at 3 kHz and digitized at 10 kHz using National Instruments acquisition boards and ScanImage ( available at: scanimage . org ) written in MATLAB ( Mathworks , Natick , MA ) . Electrophysiology and imaging data were analyzed offline using Igor Pro ( Wavemetrics , Lake Oswego , OR ) , ImageJ ( NIH , Bethesda , MD ) and GraphPad Prism ( GraphPad Software , La Jolla , CA ) . In figures , voltage-clamp traces represent the average waveform of 3–6 acquisitions . Peak current amplitudes were calculated by averaging over a 1 ms window around the peak . For pharmacological analyses , 3–7 consecutive acquisitions ( 20 s inter-stimulus interval ) were averaged following a 3-min wash-in period for NBQX and CPP or a 4-min wash-in period for MEC , MLA , and DHβE . For TTX and 4AP conditions , current averages were composed of the acquisitions following full block or first-recovery of ChR2 evoked currents , respectively . Data ( reported in text and figures as mean ± sem ) were compared statistically using the Mann–Whitney test . p values smaller than 0 . 05 were considered statistically significant .	Neurons communicate with one another at junctions called synapses . When an electrical signal arrives at the presynaptic cell , it triggers the release of molecules called neurotransmitters into the synapse . These molecules then bind to receptor proteins on the postsynaptic cell , starting a chain of events that leads to the regeneration of the electrical signal in the second cell . Broadly speaking , neurotransmitters are either excitatory , which means that they increase the electrical activity of the postsynaptic neurons , or they are inhibitory , meaning that they reduce postsynaptic activity . Initially , it was thought that neurons release only one type of neurotransmitter , but it is now known that this is not always the case . Many neurons within the spinal cord , for example , release two different inhibitory neurotransmitters , GABA and glycine , while some neurons in the midbrain release GABA and an excitatory neurotransmitter called glutamate . Saunders , Granger , and Sabatini now provide the first direct evidence that cholinergic neurons in different regions of the forebrain also release two neurotransmitters . Collectively known as the ‘forebrain cholinergic system’ , these cells are best known for producing the excitatory transmitter acetylcholine . However , Saunders et al . now show that this system also produces an enzyme that manufactures GABA , as well as a protein that pumps GABA into structures called vesicles , which are then released into the synapse . Although this is not concrete evidence for the release of GABA , Saunders et al . also show—with a technique called optogenetics , which involves the use of light to control neuronal activity—that some of the neurons in this system can trigger inhibitory responses in postsynaptic cells . Moreover , these responses can be blocked using drugs that occupy GABA receptors , or by using genetic techniques to delete the GABA-pumping protein from cholinergic neurons . Taken together , the results of these experiments strongly suggest that the cholinergic neurons throughout the forebrain—unlike , for example , the cholinergic neurons in the midbrain , the region of the brain that controls movement—possess the molecular machinery needed to produce and release GABA , in addition to acetylcholine . Given that the cholinergic system has a key role in cognition and is particularly susceptible to degeneration in Alzheimer's disease , the ability of these neurons to release GABA release could have widespread implications for the study and understanding of brain function .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "short", "report", "neuroscience" ]	2015	Corelease of acetylcholine and GABA from cholinergic forebrain neurons
Biological age measurements ( BAs ) assess aging-related physiological change and predict health risks among individuals of the same chronological age ( CA ) . Multiple BAs have been proposed and are well studied individually but not jointly . We included 845 individuals and 3973 repeated measurements from a Swedish population-based cohort and examined longitudinal trajectories , correlations , and mortality associations of nine BAs across 20 years follow-up . We found the longitudinal growth of functional BAs accelerated around age 70; average levels of BA curves differed by sex across the age span ( 50–90 years ) . All BAs were correlated to varying degrees; correlations were mostly explained by CA . Individually , all BAs except for telomere length were associated with mortality risk independently of CA . The largest effects were seen for methylation age estimators ( GrimAge ) and the frailty index ( FI ) . In joint models , two methylation age estimators ( Horvath and GrimAge ) and FI remained predictive , suggesting they are complementary in predicting mortality . Biological aging is a process characterized by progressive deteriorations occurring simultaneously at multiple levels , that is molecular , cellular , tissue- and organ-specific , in an organism ( Hayflick , 2002 ) . Due to its complex nature , quantifying this process has long been a challenge . Chronological age ( CA ) records the passage of time and serves as a convenient metric of aging; however , people of the same CA manifest diverse aging-related phenotypes . Hence , biological age measurements ( BAs ) , attempt to capture physiological changes along the aging course . Therefore , they can be used to assess health risks in individuals of the same age . Multiple BAs have been proposed , including omics-based clocks , clinical biomarkers , and assessments of functioning ( Jylhävä et al . , 2017; Xia et al . , 2017 ) . One of the potential BAs , telomere length , is the length of a nucleotide sequence at the chromosomal ends , and represents the capacity for cell division ( Mather et al . , 2011 ) . Epigenetic clocks comprise aging-related DNA methylation information from various genomic loci ( i . e . , CpG dinucleotides ) ( Hannum et al . , 2013; Horvath , 2013 ) . Another composite score derived from clinical measurements and blood biomarkers can be viewed as a physiological age measure ( Levine , 2013 ) . Furthermore , organismal functioning status could reflect the biological aging process from various aspects , such as cognitive performance , physical functioning , and overall frailty ( Reynolds et al . , 2005; Finkel et al . , 2019; Searle et al . , 2008 ) . Previous studies found low to moderate correlations among different types of BAs , for example telomeres , epigenetic clocks , clinical biomarkers , and frailty ( Belsky et al . , 2018; Kim et al . , 2017; Vetter et al . , 2019; Zhang et al . , 2018 ) . However , few studies comparing BAs have taken mortality risk into consideration , and none of them have included more than three types of BAs ( Kim et al . , 2017; Zhang et al . , 2018; Murabito et al . , 2018 ) . Thus , to grasp more of the complex aging process , different types of BAs should be studied jointly in the same cohort . In addition , to capture changes over time , as in aging , it is imperative to study the same individuals longitudinally , yet there is a lack of studies comparing longitudinal profiles of multiple BAs in the same population . Therefore , the present study aims to add to this knowledge gap in the field by examining longitudinal trajectories , correlations and associations to all-cause mortality using nine BAs in the Swedish Adoption/Twin Study of Aging ( SATSA ) ( Finkel and Pedersen , 2004 ) . Here , telomere length , four types of DNA methylation age estimators ( DNAmAges ) , physiological age , cognitive function , functional aging index ( FAI ) , and frailty index ( FI ) have been measured repeatedly across 20 years follow up . Nine BAs were included in the present study based on data availability in SATSA . Leukocyte telomere length was measured by quantitative polymerase chain reaction ( qPCR ) and presented as a relative telomere length ( Berglund et al . , 2016 ) . Genome-wide methylation levels in whole blood were assessed by Infinium HumanMethylation 450K BeadChip , and four types of DNAmAge ( Horvath , Hannum , PhenoAge , and GrimAge ) were then calculated using established algorithms based on the methylation levels of 353 , 71 , 513 , and 1030 CpG sites , respectively ( Hannum et al . , 2013; Horvath , 2013; Levine et al . , 2018; Lu et al . , 2019 ) . Physiological age was derived from a list of age-correlated biomarkers ( Pearson correlation coefficient with CA > 0 . 10 ) measured by blood tests and physical examinations ( see Materials and methods and Supplementary file 1B ) . Biomarkers were transformed into a calibrated BA value using principal component analysis and the Klemera and Doubal ( 2006 ) methods . Cognitive function measured performance in verbal ( crystallized ) ability , spatial ( fluid ) ability , memory , and perceptual speed using a cognitive testing battery . The four specific domains were combined into an overall score by principal component analysis ( Reynolds et al . , 2005 ) . FAI incorporated four functional aspects from self-reported information and physical examinations , that is sensory ( vision and hearing ) , pulmonary , strength ( grip strength ) , and movement/balance ( gait speed ) . The four indicators were standardized separately and summed to create a composite score ( Finkel et al . , 2019 ) . Lastly , a set of 42 self-reported deficits , covering a range of health domains , were taken into account in the development of the FI ( Jiang et al . , 2017 ) . ( Table 2 and Supplementary file 1B–C ) . For each BA , we calculated a corresponding BA residual by regressing out CA and sex effects as well as individual- and twin-pair related components from the BA using a linear mixed model ( see Materials and methods and Figure 1 ) . The BA residuals in the present study are commonly referred to as age acceleration in the DNAmAge-related literature . Of the 845 study participants , 59 . 5% were women , 42 . 2% attained above primary education , and 25 . 0% were current or former smokers; the average body mass index ( BMI ) was 25 . 7 kg/m2 , and mean chronological age was 63 . 6 years at baseline . On average , the levels of BAs were 0 . 73 for telomere length ( T/S ratio ) , 60 . 4 years for Horvath DNAmAge , 65 . 2 years for Hannum DNAmAge , 63 . 8 years for DNAmPhenoAge , 69 . 4 years for DNAmGrimAge , 64 . 7 years for physiological age , 51 . 5 for cognitive function , 48 . 3 for FAI , and 0 . 10 for FI . The baseline characteristics in individuals with complete measurements were similar to those in all individuals . ( Table 3 and Supplementary file 1D ) . On average , BAs were assessed between 2 . 5 times ( telomere length ) to 4 . 2 times ( FI ) in each participant with information on the respective BAs . ( Table 3 ) Longitudinal changes in BAs were modeled as functions of CA ( as a natural spline with three degrees of freedom ) and sex , with random effects introduced at the individual and twin-pair levels ( mixed models , see Materials and methods ) . Both individual-level BAs and population BA means over CA in men and women are presented in Figure 1 . We found that with increasing age , average telomere length and cognitive function declined , while all types of DNAmAges , physiological age , FAI , and FI increased . The profiles for the three functional BAs ( cognitive function , FAI , and FI ) show clear curvature , indicating an accelerated rate of change around the age of 70 , whereas the other BAs exhibit minimal or no curvature , corresponding to a constant change ( linear growth ) over the age span . In addition , we observed small to moderate sex differences in the mean levels of BAs . Women exhibited longer telomere length ( p=0 . 001 ) and lower DNAmAge ( p=0 . 013 for Horvath , p=0 . 001 for Hannum , and p<0 . 001 for GrimAge ) compared to men , but also worse functioning ( p<0 . 001 for FAI and p=0 . 001 for FI ) . ( Figure 1 ) . To examine the possible effect modification , we next introduced an interaction term between CA and sex in the mixed models . We found no evidence that the shape of the BA curves differed between men and women , except for in physiological age ( p<0 . 001 for sex interaction ) . However , the difference was mainly observed at the end of the CA spectrum ( i . e . , before the age of 50 and after the age of 85 ) . ( Figure 1—figure supplement 1 ) . BAs were broadly categorized into four groups according to the main biological structural levels where the BA measurements were implemented , that is telomere length , DNAmAges , physiological age , and functional BAs ( Panel A in Figure 2 ) . We estimated correlation coefficients between BAs and BA residuals and adjusted for repeated measurements using the rmcorr function in R ( Figure 2 and Supplementary file 1E ) ( Bakdash and Marusich , 2017 ) . Using all complete measurements , we estimated repeated measures correlation ( rmcorr , 22 ) between age and all nine BAs , which accounts for non-independence among observations and captures the common intra-individual association between age and each of the BAs in turn . Telomere length showed low correlations with both CA and BAs ( absolute correlation coefficients ≤ 0 . 16 ) ; the other BAs were correlated with CA and each other to varying degrees , with absolute correlation coefficients ranging from 0 . 24 to 0 . 87 . As indicated by the correlation matrices of BAs and BA residuals ( Panel B and C in Figure 2 ) , after regressing out CA from BAs , most of the original correlations were attenuated and moderate correlations remained only between Horvath and Hannum DNAmAges ( r = 0 . 35 ) , cognitive function and FAI ( r = −0 . 32 ) , and FAI and FI ( r = 0 . 31 ) . Correlation patterns of BAs and BA residuals in men and women were similar to those in the whole population . ( Data not shown ) . Next , we transformed correlation coefficients into Euclidean distances and then performed hierarchical cluster analysis . The same types of BAs , that is methylation BAs and functional BAs , tended to be more closely related . GrimAge and PhenoAge , however , were somewhat separated from the other two DNAmAges , especially using BA residuals . ( Figure 2—figure supplement 1 ) . To study the association between BAs and all-cause mortality , we used Cox regression models with attained age being time scale and accounted for left truncation and right censoring . To make the hazard ratios ( HR ) comparable , we rescaled all BAs and BA residuals by their standard deviation ( SD ) , so that HRs express changes in mortality risk due to a one-SD increment of the respective BA or BA residual . Fully adjusted models controlled for sex , educational attainment , smoking status , and BMI . All models were stratified by participants’ birth year ( categorized in 10 year intervals into five groups ) and implicitly adjusted for CA through the choice of attained age as the underlying time scale . When repeated measurements were available , only the first BA assessment was included in the survival analysis and referred to as the baseline BA . First , we looked at the effect of baseline BAs separately ( i . e . , one-BA models ) . During a median follow-up time of more than 15 years , we observed statistically significant BA-mortality associations for DNAmAges ( GrimAge , PhenoAge , and Horvath ) and functional BAs ( FI , FAI , and cognitive function ) , where a one-SD increase of BAcorresponded to a change in mortality risk by 39 , 26 , 17 , 32 , 27 , and 15% , respectively . Increased mortality risks were also found with changes in Hannum DNAmAges and physiological age , although the effects were generally weaker and came with a slightly higher level of imprecision , with HRs ( 95% CI ) being 1 . 17 ( 0 . 98 , 1 . 40 ) and 1 . 13 ( 0 . 97 , 1 . 31 ) . We found no evidence for telomere length as associated with mortality risk . ( Table 4 ) The corresponding HRs for the BA residuals were comparable to BAs ( Supplementary file 1F ) . Second , we explored the baseline BAs in relation to mortality risk by taking all BAs into consideration simultaneously in 288 individuals with complete measurements ( i . e . , nine-BA models ) . To avoid biased estimation due to collinearity ( as all nine BAs were correlated to varying extents ) , BA residuals instead of BAs were used . During a median follow-up time of 16 . 5 years , we observed that three BA residuals , Horvath DNAmAge , DNAmGrimAge , and FI , out of nine were associated with the risk of all-cause mortality independently of all other BAs . A one-SD increment of Horvath DNAmAge , DNAmGrimAge and FI residuals was associated with a 31% , 43% and , 58% higher risk of death , respectively , independent of the level of CA , all other BA residuals and common risk factors . The HRs for the other DNAmAges ( Hannum and PhenoAge ) , physiological age , and the other functional ages ( cognitive function and FAI ) were attenuated towards one after controlling for all BAs in the same model ( Table 5 ) . We replicated the first part of the survival analyses ( i . e . , one-BA models ) in subgroups classified by sex , age group , and smoking status , as well as in the sub-cohort of individuals with complete measurements . In the subgroup analyses , we observed evidence for potential effect modification . The associations of BAs with mortality risk were generally stronger in women ( except for Horvath DNAmAge , and physiological age ) , more pronounced in the younger individuals ( except for Horvath DNAmAge , physiological age and cognitive function ) , and a bit stronger in current or ex-smokers ( for Horvath DNAmAge and DNAmGrimAge ) compared to those in men , older individuals , and non-smokers , respectively . ( Figure 3 and Supplementary file 1G–I ) Furthermore , as with the results observed in the whole population , we found the same general pattern of BA associations with mortality in the sub-cohort with complete measurements . ( Supplementary file 1J ) We also adjusted for the presence of the previous diseases , including heart failure , stroke , diabetes , and cancer in the survival model and the observed results were largely unchanged . ( Supplementary file 1K ) . As biological aging does not proceed at a constant pace across time , monitoring longitudinal trajectories of BAs using repeated assessments to depict the characteristics of the aging course is of great importance . We observed an accelerated change in the aging pace in functional BAs , which are in parallel with previous studies focused on trajectories of cognition , physical function , and frailty ( Harris et al . , 2016; Stow et al . , 2018; Zaninotto et al . , 2018 ) . This result indicates that an aged population may expect deteriorations of their functional performance to speed up as their age approach 70 years . In particular , we observed that the aging acceleration of cognition ( cognitive function ) and physical function ( FAI ) seemed to take place before age 70 , while frailty ( FI ) went up at an increased rate after age 75 . This evidence suggests that accelerated cognitive and physical dysfunctions are likely to arise prior to frailty’s , and corresponding interventions and screenings could aim at people of different ages accordingly . In contrast to functional BAs , DNAmAges and physiological age accrued at a relatively constant rate . Due to their mode of construction , DNAmAges and physiological age were transformed into an age-calibrated scale in the unit of year . Thus , these BAs allow individuals to perceive an intuitive impression on their biological aging status by comparing BA with CA and have the potential to serve as a straightforward-to-use metric in the geriatric practice . However , in order to be clinically meaningful , they need to be validated on the individual level data as well . In addition , a drawback of this mode of construction is that it largely explains the linear shapes of the BA trajectories , making longitudinal BA characteristics less informative as the changes are constant . Another feature of physiological age is the strong correlation with CA; hence , it allows limited biology perception in prediction models on top of CA . This is again due to its mode of construction using a composite score of several biomarkers where each one , by definition , strongly correlates with CA . To improve construction models for BA , composite scores should balance the reliance on biological aspects and weight on CA . Sex differences in aging are observed at multiple biological levels ( Oksuzyan et al . , 2008; Ostan et al . , 2016 ) . Most notably , women tend to present lower mortality but higher morbidity rates , especially at advanced ages , compared to men of the same age . We found that , compared to men , women presented lower molecular ages ( telomere length and methylation age estimators ) but higher functional ages ( FAI and FI ) across the age spectrum , indicating different medical needs are likely to be encountered for men and women . Meanwhile , we observed no evidence for shape differences in the BA trajectories ( except for physiological age ) by sex . This suggests that aging proceeded at a similar rate in men and women since midlife onwards . For physiological age , the shape difference was mainly observed at the end of the age spectrum and could be influenced by methodological issues as the construction of physiological age was done differently in men and women . Putting our results in the perspective of previous studies , we confirm earlier findings in women , such as better telomere status and worse frailty ( Mu et al . , 2014; Bartley et al . , 2016 ) . One possible explanation for the sex difference in FI is that men and women tend to present different self-perception on aging-related performances due to social and cultural considerations when self-reported data are used ( Ostan et al . , 2016 ) . However , cognitive function and FAI incorporate objective measurements tested by trained nurses and both of them exhibited sex differences in the same direction . Hence , sex differences in the functional BA spectrum are not , or at least not exclusively , driven by social and cultural factors . Therefore , we suggest that sex differences should always be accounted for in studies of aging and , if possible , presenting stratified analyses should be encouraged . Aging is a complex process proceeding through interconnected biological mechanisms . BAs , in essence , are underlain by different sets of these mechanisms . Correlations of BAs could provide an idea of how one BA changes in accordance with another . We observed moderate to strong correlations between CA and all BAs , with the exception of telomere length , which was only weakly correlated with CA and the other BAs . These results are in line with previous research observing similar correlation pattern for 11 BAs , as well as low correlations between telomere length and other types of BA ( Belsky et al . , 2018 ) . As some degree of correlation between BAs is expected simply due to their relation with CA , we also investigated correlations between BA residuals , which by construction minimize correlation with CA . We found clear attenuation of the correlations between different types of BA residuals , that is methylation BAs and functional BAs as well as physiological age and telomere length . In other words , among individuals with the same CA , molecular BAs and functional BAs are only weakly associated . We interpret this as an indication that these categories largely reflect different aspects of the aging process . After regressing out the effect of CA , we still observed weak to moderate correlations between DNAmAges . These DNAmAges were initially trained to capture methylome differences associated with various aging-related phenotypes , namely CA for Horvath and Hannum , Phenotypic Age for PhenoAge , and mortality risk for GrimAge ( Hannum et al . , 2013; Horvath , 2013; Levine et al . , 2018; Lu et al . , 2019 ) . Employing correlated surrogates and measuring the same aging hallmark of epigenetic alteration in the development of these DNAmAges could partly explain the correlations observed among these residuals . Hierarchical clustering suggested GrimAge and PhenoAge were somewhat separated from the other two DNAmAges , which is expected as GrimAge and PhenoAge incorporate additional information on top of CA in the development of their BA algorithms . In addition , Horvath and GrimAge were the least correlated DNAmAge residuals , indicating that these two BAs captured uncorrelated methylome information to a larger degree than other DNAmAges did and might have the potential to inform aging-related outcomes independently ( as illustrated in the results of nine-BA survival analyses ) . Functional BA indicators included cognitive function , FAI , and FI , which quantify cognitive , physical , and frailty-related functioning . Intriguingly , the present analysis found comparable levels of correlation for these BAs as well as for their residuals , suggesting that most of these correlations were not simply driven by CA . Previous studies have also reported correlations between physical function , cognitive function , and health-related quality of life ( Kim , 2016 ) . As functional BAs measure complex phenotypes resulting from intricate biological mechanisms , further investigations are needed to disentangle the mechanisms underlying both the age-dependent and , in particular , the age-independent correlations . A majority of the included BAs are well-established aging indicators and were found to be associated with mortality in previous studies ( Hannum et al . , 2013; Levine , 2013; Shamliyan et al . , 2013; Wang et al . , 2018a ) . Our results were comparable to the current evidence . We first examined the baseline value of individual BAs in relation to the risk of death and found associations for all BAs except for telomere length , with DNAmGrimAge and FI demonstrating the strongest relationship . We next examined BA-mortality associations with all BAs included in the same model and observed significant evidence for Horvath DNAmAge , DNAmGrimAge , and FI , suggesting they were complementary in the prediction of mortality . In other words , Horvath DNAmAge , GrimAge and FI were likely to capture aging-related information that was both informative in the prediction of mortality risk and independent of other BAs . DNAmAges ( Hannum and PhenoAge ) and functional BAs ( cognitive function and FAI ) were likely to reflect some mortality-associated aging aspects that were shared and/or correlated with other BAs , as the corresponding HRs were attenuated to almost one when all BAs were included in the model . Few studies have focused on the comparison of the predictive aspect of BAs . Kim et al . ( 2017 ) found FI to outperform methylation age estimators in survival models by comparing only two types of BAs , frailty , and DNAmAge ( Horvath ) . Zhang et al . ( 2017 ) developed a methylation-based mortality score and observed a stronger mortality association than the FI-mortality relationship . Furthermore , Murabito and colleagues compared clinical age , inflammatory age , and DNAmAge ( Horvath and Hannum ) and concluded that they were complementary in predicting risk for mortality ( Murabito et al . , 2018 ) . Our study thus supports these findings , both in the way that mortality-oriented DNAmAges and FI were strongly associated with mortality , and in that certain types of BAs could reflect mortality risk independently . In particular , we add information by including other BAs , especially physiological age , a newly trained DNAmAge ( GrimAge ) and functional BAs . Physiological age is a composite biomarker of aging , developed from 10 blood and clinical markers . Levine et al . applied a similar method to develop a BA indicator and found it associated with mortality beyond CA in a large US population ( Levine , 2013; Klemera and Doubal , 2006 ) . We also observed that physiological age predicted mortality risk independently of age , albeit weaker and at a moderate significance level . Despite its weak mortality association , physiological age presents some benefits with respect to the physiological interpretation and the search for biological mechanism in future explorations , as the component biomarkers contributed to the development of physiological age were explicitly specified . DNAmGrimAge is derived via a two-step development method , in which methylation data were first used to predict a set of biomarkers ( plasma proteins and smoking pack-year ) , and the methylation-based biomarkers were then used in the prediction of mortality risk . As a result , the DNAmGrimAge is explicitly trained to be a mortality predictor; it also includes the largest number of genomic sites compared with the other three DNAmAges , and allows the methylation level of a single CpG site to contribute to mortality risk via different intermediate biomarkers , as the methylation-based biomarkers used somewhat overlapping CpG sites in the two-step training phase . Lu et al . ( 2019 ) reported the advantage of predicting mortality over other existing DNAmAges in a US population . These results point to the potential of utilizing methylation information to reflect complex aging phenotypes . The FI is developed using clinical information and is likely to capture an organismal perspective on aging that is closer to the clinical end-point of death compared to molecular-based BAs . In addition , the FI incorporates information in multiple health domains , in contrast to the other two functional BAs , cognitive function , and FAI , which exclusively measure cognitive or physical performance . The comprehensive nature of the FI could partly explain the strong and robust mortality association observed here and in previous studies ( Kojima et al . , 2018 ) . Functional BAs , especially FI , measure not only the biological aging process , but also social and psychological wellbeing . Consequently , disentangling the underlying biological mechanisms from the functional measures would be challenging , which impairs the ability of functional BAs to explain the innate biological aging process . Specifically , it is difficult to implement functional BA measurements in animal models , whereas molecular BAs can be easily investigated in animal studies . Additionally , functional BAs in humans tend to show less heterogeneity among the young , making them less ideal instruments to assess the aging process among young adults , compared to using molecular BAs . This is especially true when it comes to FAI as it only takes four functioning indicators into account . However , FI considers a wide range of health deficits ( 42 in our case ) and provides relatively precise scales to differentiate individuals with lower functioning scores . Furthermore , since biological , psychological , and social factors all contribute to the health status of human beings , the comprehensive nature of functional BAs also comes with important advantages for predicting aging-related outcomes , underlining the complementary nature of molecular and functional BAs . In the subgroup analysis , we found moderate effect modification from age and smoking status . A previous study in a large Swedish cohort from our research group also observed that the effect size of FI with mortality was attenuated with increasing age ( Li et al . , 2019 ) . DNAmGrimAge includes smoking ( pack-year ) associated CpG sites in its establishment ( Lu et al . , 2019 ) , which could partly explain a stronger association observed in smokers than non-smokers as it may capture smoking-based deterioration at the molecular level . In addition , the present results found a majority of the BA-mortality associations to be stronger in women compared to men , suggesting sex differences should be acknowledged in making use of BAs to inform clinically relevant decisions . It is noteworthy that in comparing the strength of relative risk of mortality related to each BA , we were not aiming to propose any superior BA , as mortality association alone is by no means the gold standard for a qualified BA in aging research . Evidence exists that some BAs have advantages in predicting other aging-related outcomes such as entry into care ( e . g . , FAI ) ( Finkel et al . , 2019 ) , which may be more relevant for developing policies and interventions . The present study has several strengths . First , we measured nine BAs in the same population from the molecular to the organismal level , which provided us an opportunity to examine BA characteristics from several molecular and functional aspects . Second , the BAs were assessed in a longitudinal manner to allow for descriptive and correlational analyses considering changes over time . Third , through the linkage to the Swedish national registers , we were able to keep track of study participants’ vital status for nearly 20 years , thus being able to predict the risk of death in relation to BA status prior to the occurrence of most aging-related outcomes . Several limitations should be taken into consideration when interpreting the present findings . First , only individuals with complete measurements were included in the nine-BA survival models . Complete observations may not be a representative subgroup of the entire study population and lead to selection bias . However , IPT occasions with incomplete BAs were mostly due to administrative reasons , such as methylation data were only assessed in five waves since IPT3 , and these factors are likely not related to subject-specific features . In addition , we replicated the one-BA survival models in individuals with complete measurements , and the results did not show major discrepancies . Second , the birth year of SATSA participants spanned a long period of time , from the year 1900 to 1948 , thus raising concerns for cohort effects . During this time , improved medical conditions or emergent public health events could affect both BA and mortality risk and cannot be controlled merely through the age scale in the present analysis . All survival models were stratified by participants’ birth year in 10 year intervals to in part alleviate bias due to the cohort effect . In summary , biological age is a construct built on molecular , cellular , and functional aspects of an individual´s health status and has the potential to change the way risk prediction is performed in the clinic . In the era of personalized medicine , individual health assessments are warranted and should rely on better methods than simply using the CA . In this study , we present the most comprehensive analysis of biological age in humans to date . We explain their correlations and longitudinal patterns . We highlight the sex-specific effects and point to the importance of providing sex-specific estimates in aging investigations and acknowledging sex-specific care in geriatric practices . Furthermore , we found BAs had the potential to provide mortality-relevant information independently of CA and independently of other types of BAs . Further studies are needed to investigate if the present findings could be extrapolated to a younger population , what the specific mechanisms underlying the BA correlations are , and possible modification effects on and between BAs . The Swedish Adoption/Twin Study of Aging ( SATSA ) was a population-based study consisting of reared apart and reared together twin pairs ( Finkel and Pedersen , 2004 ) . In-person testing ( IPT ) in SATSA was initiated in 1986 and nine complete waves of in-person testing were conducted through 2014 ( IPT1 to IPT10 , except for IPT four where only telephone interviews were performed ) , during which the information of a maximum of nine BAs , including telomere length , four DNAmAges ( Horvath’s , Hannum’s , PhenoAge , and GrimAge ) , physiological age , and three functional ages ( cognitive function , FAI , and FI ) were available . In total , 859 individuals participated in at least one IPT wave in SATSA , and 846 individuals who had at least one BAs assessed once since IPT1 to IPT 10 ( except for IPT4 ) and the information of vital status through linkage between the Swedish Twin Registry ( STR ) and the Swedish Population Register were included in the present study . The number of individuals and measurements by IPT and BAs were detailed in Table 1 and Supplementary file 1A . Sex , educational attainment , and baseline information of BMI , and smoking status were considered as covariates in the survival analyses . BMI was assessed through physical examination and other covariates were acquired through self-reported questionnaire data . Educational attainment was classified as primary education , lower secondary or vocational , upper secondary education , and tertiary education . BMI was derived as the body mass divided by the square of body height , expressed in units of kg/m2 . Smoking status was categorized as non-smokers , ex-smokers , and current smokers . Among the included individuals and measurements , 28 out of 846 individuals without education information were assigned into an unknown group , and another one and two individuals who missed the BMI and smoking status were imputed with the corresponding information collected from their nearest available IPTs . Since the participants in the present analysis were born over a long period of time ( 1900 to 1948 ) , we estimated the mortality associations with stratification on the calendar year of the participants’ birthday to avoid confounding from cohort effects . Individuals’ birth year was treated as a categorical variable with five groups in an interval of 10 years . All-cause mortality data , including vital status and dates of deaths , were obtained from linkages between the STR and Swedish Population Register through the personal identification number assigned to all residents . All-cause mortality data were updated on August 16 , 2018 . Characteristics of baseline measurements were presented as means ( standard deviations , SDs ) and proportions . Longitudinal changes of BAs were presented in plots with individual level measurements illustrated by dots , lines , and broken lines . Average trajectories of BAs were estimated through mixed linear regression models as described above , each of which included fixed effects of the intercept , CA , and sex , and a random effect of intercept at the individual and twin-pair level . P-values for sex effects were obtained from fixed effects within the mixed linear model . In addition , an interaction term between each BA and CA in natural spline was introduced to the model , and P-values for sex interaction were determined by the likelihood ratio test comparing the models with and without the interaction term . We estimated the repeated measures correlation coefficients between age and nine BAs using all complete measurements ( Bakdash and Marusich , 2017 ) . The correlation analyses were then replicated for CA and nine BA residuals to examine the correlations between the parts of BA residuals . We then transformed correlation coefficients to scaled squared Euclidean distances and performed hierarchical cluster analysis on BAs and BA residuals via Ward’s method . We used the Cox regression model to estimate the association between baseline BAs and the risk of all-cause mortality . When repeated measurements were available , only the first BA assessment was included and referred to as the baseline BA . In all models , we used attained age as the underlying time scale and the five groups of birth year as strata . Left truncation and right censoring were accounted for in the estimations; that is individuals entering into the cohort is conditional on their survival at baseline ages , and follow-up time may stop before death occurs . Follow-up time started from the first available measurement ( baseline measurement ) and ended at the date when they died or were censored on August 16 , 2018 . In addition , robust standard errors were introduced to adjust for relatedness within twin pairs and subjects . To achieve a direct comparison of BA effects in relation to mortality risk , we standardized all nine BAs to the mean of zero and SD of one across all available measurements and replicated the transformation in the BA residuals , such that the estimated hazard ratios ( HRs ) could be interpreted as the relative risk of death associated with a one-SD increase in the level of BA or BA residual . In the first part of the survival analyses , all models accounted for only one BA ( one-BA models ) . For those who had their BAs assessed for multiple times , the first available measurement was treated as the baseline measurement . For each BA , we estimated the association of BAs and BA residuals with mortality risk using two models , a univariate model and a multivariate model with sex , education , BMI , and smoking status as additional covariates . In the second part of the survival analyses , all nine BAs were accounted for altogether in the same survival model ( multi-BA models ) . Similarly , two models with or without adjustment for common risk factors were estimated . Only BA residuals were taken into consideration in the second part of survival analyses to avoid biased results caused by collinearity within BAs . P values were two-sided , and statistical significance was defined as p<0 . 05 . All analyses were conducted using Stata 15 . 1 and R 3 . 6 . 0 .	Everyone ages , but how aging affects health varies from person to person . This means that how old someone seems or feels does not always match the number of years they have been alive; in other words , someone’s “biological age” can often differ from their “chronological age” . Scientists are now looking at the physiological changes related to aging to better predict who is at the greatest risk of age-related health problems . Several measurements of biological age have been put forward to capture information about various age-related changes . For example , some measurements look at changes to DNA molecules , while others measure signs of frailty , or deterioration in cognitive or physical abilities . However , to date , most studies into measures of biological age have looked at them individually and less is known about how these physiological changes interact , which is likely to be important . Now , Li et al . have looked at data on nine different measures of biological age in a group of 845 Swedish adults , aged between 50 and 90 , that was collected several times over a follow-up period of about 20 years . The dataset also gave details of the individuals’ birth year , sex , height , weight , smoking status , and education . The year of death was also collected from national registers for all individual in the group who had since died . Li et al . found that all nine biological age measures could be used to explain the risk of individuals in the group dying during the follow-up period . In other words , when comparing individuals with the same chronological age in the group under study , the person with a higher biological age measure was more likely to die earlier . The analysis also revealed that biological aging appears to accelerate as individuals approach 70 years old , and that there are noticeable differences in the aging process between men and women . Lastly , when combining all nine biological age measures , some of them worked better than others . Measurements of methylation groups added to DNA ( known as DNA methylation age ) and frailty ( the frailty index ) led to improved predictions for an individual’s risk of death . Ultimately , if future studies confirm these results for measures from single individuals , DNA methylation and the frailty index may be used to help identify people who may benefit the most from interventions to prevent age-related health conditions .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "epidemiology", "and", "global", "health" ]	2020	Longitudinal trajectories, correlations and mortality associations of nine biological ages across 20-years follow-up
Enhancers constitute one of the major components of regulatory machinery of metazoans . Although several genome-wide studies have focused on finding and locating enhancers in the genomes , the fundamental principles governing their internal architecture and cis-regulatory grammar remain elusive . Here , we describe an extensive , quantitative perturbation analysis targeting the dorsal-ventral patterning gene regulatory network ( GRN ) controlled by Drosophila NF-κB homolog Dorsal . To understand transcription factor interactions on enhancers , we employed an ensemble of mathematical models , testing effects of cooperativity , repression , and factor potency . Models trained on the dataset correctly predict activity of evolutionarily divergent regulatory regions , providing insights into spatial relationships between repressor and activator binding sites . Importantly , the collective predictions of sets of models were effective at novel enhancer identification and characterization . Our study demonstrates how experimental dataset and modeling can be effectively combined to provide quantitative insights into cis-regulatory information on a genome-wide scale . Developmentally expressed genes in metazoans are regulated by diverse cis-regulatory elements , including distally-acting sequences termed enhancers ( Levine , 2010; Smith and Shilatifard , 2014; Heinz et al . , 2015 ) . Despite more than three decades of progress , surprisingly little is known about constraints on internal structural organization of binding sites ( 'grammar' ) within these elements ( Dickel et al . , 2013 ) . Some enhancers show little evolutionary variation , and permit no change in transcription factor binding sites without catastrophic effects on function ( Thanos and Maniatis , 1995; Kim and Maniatis , 1997 ) . Many developmental enhancers , however , demonstrate a more flexible deployment of binding sites , thus functionally conserved elements can exhibit a large degree of evolutionary variation ( Junion et al . , 2012 ) . Although high-throughput studies have dramatically increased our knowledge of genome-wide transcription factor occupancy and transcript expression , we have a limited ability to interpret the functional relevance of quantitative aspects of protein binding or DNA sequence variation . A quantitative understanding of the internal enhancer grammar of cis-regulatory elements will provide researchers with powerful tools to better understand the significance of genetic variation that is observed within and between species , critical for exploitation of burgeoning genomic resources . Thermodynamic models employ tools from statistical physics to model gene activity , providing a framework for understanding transcription factor interactions with specific DNA sequences to regulate gene expression ( Bintu et al . , 2005; Ay and Arnosti , 2011; Dresch and Drewell , 2012 ) Previous efforts at thermodynamic modeling in eukaryotic systems have demonstrated that diverse types of data can be fit , providing at least a qualitative level of prediction ( Zinzen et al . , 2006; Segal et al . , 2008; He et al . , 2010; Drewell et al . , 2014 ) . However , earlier studies relied on heterogeneous and low-resolution datasets , inherently limiting modeling effectiveness . In addition , few types of models were tested , reducing the chance that essential properties of transcription factor interactions will be captured . Here , we tested the hypothesis that an in-depth and quantitative analysis of key transcriptional regulators on an archetypal enhancer would reveal common transcriptional behaviors of these proteins for genome-wide analysis . To harness the potential of thermodynamic modeling approaches , we developed an in-depth enhancer perturbation analysis that takes advantage of the quantitative setting of the Drosophila blastoderm embryo . The rhomboid ( rho ) enhancer directs gene expression in the presumptive neuroectoderm under the control of the activator Dorsal , a homolog of NF-κB . The Twist activator and Snail repressor provide additional essential inputs ( Figure 1A ) . To extract fundamental information from this data set , we created and fit a comprehensive set of thermodynamic models designed to capture likely interactions between transcription factors as they interact with the enhancer . An extensive set of other coordinately-regulated Dorsal/Twist/Snail target genes was then used to assess the power of this modeling approach for interpretation of cis-regulatory variation . 10 . 7554/eLife . 08445 . 003Figure 1 . Experimental deep perturbation analysis of a neuroectoderm-specific enhancer . ( A ) rhomboid ( rho ) enhancer with footprinted Dorsal activator sites shown in green , Twist activator sites in yellow , and Snail repressor sites in red . ( B ) Drosophila embryo stained to measure rho reporter readout in the neuroectoderm with quantitative expression measured using confocal laser scanning microscopy . The embryo is oriented anterior up and dorsal to the right . rho-lacZ mRNA is shown in red , seven even-skipped mRNA stripes in green , and snail mRNA in the mesoderm ( left ) also in green . Average expression after normalization over embryos containing the wild-type enhancer in the region of interest ( white box ) is plotted in black at right; dashed red line indicates standard error; vertical dotted line indicating dorsal-most boundary of sna expression . ( C ) Effects on expression of removal of single activator sites – expression levels decrease by a similar extent when either Dorsal or Twist activator sites are removed . ( D ) Effects of double activator site mutation – expression levels are variably impacted depending on which pairs of sites are affected . ( E ) Effects on expression of removal of pairs of Snail repressor sites , leading to partial depression in ventral regions , or no derepression . For panels C–E , solid black lines indicate wild-type enhancer expression; solid blue lines indicate expression of the tested rho enhancer; dashed red lines indicate standard error; n = number of embryos imaged for the given construct shown ( in top right of each panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 00310 . 7554/eLife . 08445 . 004Figure 1—figure supplement 1 . Quantitative perturbation analysis of rho neuroectodermal enhancer . Averaged quantitative expression for each of the 38 variations of the rho enhancer that were tested; standard error of measurements indicated by gray lines surrounding mean values . Number of embryos imaged for each construct noted in corner ( n ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 004 rho is first expressed in two lateral stripes in the presumptive neurogenic ectoderm of the Drosophila embryo under cooperative activation by Dorsal and Twist ( Ip et al . , 1992; Hong et al . , 2008 ) . Expression is excluded from the mesoderm ( ventral region ) by Snail , a short-range repressor that interferes with activators located within ~100 bp of a Snail binding site ( Gray et al . , 1994 ) . We systematically mutated all activator and repressor binding sites in rho neurectodermal enhancer , removing Dorsal or Twist sites individually or in pairs to diminish activation , and removing multiple Snail sites to reduce repression ( Figure 1 , Figure 1—figure supplement 1 , and Supplementary file 1 ) . All 38 enhancers were cloned and integrated into the fly genome using a site-specific integration vector ( Bischof et al . , 2007 ) . We measured the transcriptional output using fluorescent in situ hybridization ( FISH ) and confocal laser scanning microscopy , and analyzed gene expression data using an image-processing pipeline ( Ay et al . , 2008 ) . Expression data from a total of 935 images - a minimum of ten embryos per construct - was normalized and combined to provide average expression patterns for each enhancer variant ( Figure 1B , Figure 1—figure supplement 1 , and Supplementary file 2 ) . Mutation of any single Dorsal or Twist activator binding site resulted in a measurable reduction of peak intensity and retraction of the rho stripe from the dorsal region , where activators Dorsal and Twist are present in limiting concentrations ( Liberman et al . , 2009; Rushlow et al . , 1989 ) . Strikingly , despite the differences in predicted binding affinities and relative positions of the motifs , the elimination of any site individually had similar quantitative effects , reducing gene expression to approximately 60% of the peak wild-type level ( Figure 1C , Figure 1—figure supplement 1 , and Supplementary file 2 ) . In contrast to this uniform picture , the impact of mutation of combinations of two Dorsal or Twist binding sites was highly variable , ranging from slightly lower expression to almost complete loss of activity ( Figure 1D , Figure 1—figure supplement 1 , and Supplementary file 2 ) . Overall , the double activator site mutagenesis revealed a complex picture of the contributions of activator sites to gene expression . We hypothesize that the variable effects of different pairwise mutations , as opposed to the rather similar effects of individual site loss , indicates that there are multiple and distinct thresholds for specific biochemical events occurring on the enhancer . In contrast to the perturbation of Dorsal and Twist elements , removal of Snail repressor binding sites revealed stark differences in the significance of individual motifs for overall activity . Mutation of all four Snail sites caused pervasive expression in the mesoderm , as expected , while constructs with a single intact Snail2 or Snail3 site showed substantial but not complete repression ( Figure 1—figure supplement 1 and Supplementary file 2 ) . Snail1 and Snail4 motifs were not nearly as effective at mediating repression , although these have similar binding affinities ( Figure 1E , Figure 1—figure supplement 1 , and Supplementary file 2 ) . Snail4 and Twist2 sites overlap , which may impair Snail binding and reduce repression efficiency . Direct examination of the data described above showed that the inter-relationships among activator and repressor sites were complex , and mutant phenotypes were not simply additive . Such complexity is a familiar facet of cis-element mutagenesis studies ( Swanson et al . , 2010; Evans et al . , 2012 ) . To extract the non-intuitive , quantitative information about transcription factor function and interactions from these results , we created an extensive set of quantitative models . Based on the likely importance of cooperativity ( Kazemian et al . , 2013 ) , we tested models that incorporated a variety of conceptions of distance-dependent homotypic and heterotypic cooperativity , as well as different distance-dependent 'quenching' or repressive interactions between repressors and activators ( Veitia , 2003; Wagner , 1999; Kulkarni and Arnosti , 2005; Arnosti et al . , 1996; Barolo and Levine , 1997 ) . We systematically tested continuous and step functions to determine possible distance relationships affecting interactions . In all , 15 formulations for cooperativity and eight formulations for quenching were employed ( Table 1 ) . Combining these two types of formulations resulted in 120 different models , which were trained on the quantitative expression data of the 38 enhancer variants . Parameters were estimated using CMA-ES , a global genetic algorithm , and overall performance was calculated from the fit to all constructs , using root mean square error ( RMSE ) as the objective function ( Hansen et al . , 2003 ) . Examination of model performance at three levels – global RMSE , construct-by-construct RMSE , and specific portions of the expression patterns - provided complementary insights into the nature of how the training data are fit , and the potential utility of the models on other enhancer sequences . 10 . 7554/eLife . 08445 . 005Table 1 . Parameters in each model . Number of parameters in cooperativity and quenching model combinations . 3 scaling factor parameters for Dorsal , Twist , and Snail are included in all models . Column 1 contains the nomenclature and number of parameters ( in parentheses ) for all 15 cooperativity models . Columns 2–9 in Row 2 contain nomenclature and number of parameters ( in parentheses ) for 8 quenching models . Each cooperativity and quenching scheme is described in the materials and methods section . The parameters being fitted for continuous functions are clearly laid out in this section . The following example illustrates the parameters in binned cooperativity and quenching functions . Model C14Q5: Parameters 1–3 are scaling factors for Dorsal , Twist , and Snail respectively . Parameters 4–15 reflect cooperativity ( separate parameters existing for each type of protein-protein interaction ) – parameters 4–6 represent Dorsal-Dorsal cooperativity ( at 1–60 bp , 61–120 bp , and 121+ bp between bound Dorsal proteins ) , likewise parameters 7–9 represent Twist-Twist cooperativity , parameters 10–12 represent Dorsal-Twist cooperativity , and parameters 13–15 represent Snail-Snail cooperativity . Paramters 16–23 reflect quenching – parameters 16–19 represent Snail quenching Dorsal ( at 1–25 bp , 26–50 bp , 51–75 bp , and 76+ bp between bound proteins ) , likewise parameters 20–23 represent Snail quenching Twist . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 005Quenching ModelCooperativity ModelQ1 – MSB ( 0 ) Q2 – Linear ( 2 ) Q3 – Logistic ( 2 ) Q4 – Gaussian ( 2 ) Q5 – Binned 4_25 ( 8 ) Q6 – Binned 4_35 ( 8 ) Q7 – Binned 4_45 ( 8 ) Q8 – Binned 10_10 ( 20 ) C1 – Linear ( 4 ) 799915151527C2 – Logistic ( 4 ) 799915151527C3 – Gaussian ( 4 ) 799915151527C4 – Binned 2_25 ( 4 ) 799915151527C5 – Binned 2_50 ( 4 ) 799915151527C6 – Binned 2_75 ( 4 ) 799915151527C7 – Binned 3_50 ( 6 ) 911111117171729C8 – Binned 3_60 ( 6 ) 911111117171729C9 – Binned 3_70 ( 6 ) 911111117171729C10 – Protein Binned 2_25 ( 8 ) 1113131319191931C11 – Protein Binned 2_50 ( 8 ) 1113131319191931C12 – Protein Binned 2_75 ( 8 ) 1113131319191931C13 – Protein Binned 3_50 ( 12 ) 1517171723232335C14 – Protein Binned 3_60 ( 12 ) 1517171723232335C15 – Protein Binned 3_70 ( 12 ) 1517171723232335 At a global level , the performance of the models was clearly co-dependent on both the cooperativity and repression formulations ( Figure 2A ) . Overall cooperativity formulation performance varied according to the type of quenching formulation; the best quenching scheme for one cooperativity scheme was not necessarily optimal for other cooperativity schemes . Such interactive effects are likely a reflection of parameter compensation . Models with more parameters tended to outperform those with the fewest , as expected , but it was notable that this was not a strict correlation; the models with the most parameters were not as effective as those with fewer . Additionally , there were measureable differences between models with identical numbers of parameters , suggesting that the different formulation of the schemes was interrogating aspects of enhancer grammar critical for the rho enhancer variants we were fitting . Best overall fits were observed using a model with cooperativity values parameterized in three 'bins' of 60 bp ( scheme C14 ) and quenching in four small 25 or 35 bp bins ( schemes Q5 and Q6 ) . 10 . 7554/eLife . 08445 . 006Figure 2 . Performance of 120 thermodynamic models . ( A ) Heat map representation of results of thermodynamic modeling of rho perturbation dataset; the performance of each model is represented by a rectangle . Models were globally fit using Root Mean Square Error ( RMSE ) , scale shown at right . Each model employs a distinct repression ( quenching ) scheme represented on horizontal axis , and a protein-protein cooperativity scheme on vertical axis . ( B ) Construct-by-construct performance ranking of 120 thermodynamic models ranked top to bottom . Vertical features are indicative of specific constructs that showed overall better or worse fit . ( C ) Examples of fitting of individual constructs by models ranking high ( C14Q5; 1st overall ) , medium ( C8Q7; 96th overall ) , or low ( C4Q4; 120th overall ) . Black points indicate measured average mRNA levels; red lines are model predictions . The high-ranking model correctly predicts both low expression ( repression ) in mesoderm and activation in neuroectoderm , medium model exhibits modest deviations in prediction of both repression and activation , and the low-ranked model exhibits particularly poor fits for repression activity . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 00610 . 7554/eLife . 08445 . 007Figure 2—figure supplement 1 . Cross validation of thermodynamic models . ( A ) Random five-fold cross validation was performed on 24 models as described in Materials and Methods to test for overfitting . Compared to fitting on the complete data set , most models showed only a modest decrease in performance in predicting left out sets ( left column , complete data set , right column , modeling on randomly selected partial data sets ) . In contrast , selective removal of specific data sets corresponding to one type of mutagenic perturbation ( all constructs with mutation to Snail sites , all constructs with mutations to Dorsal sites , etc . ) led to parameter sets that were generally significantly less able to fit left-out data ( central column ) . The comprehensive perturbation of the different types of motifs on the enhancers was thus an essential aspect of good model fitting . The low-performing model C4Q4 actually performed better during cross-validation , likely because the parameters determined from the original run do not represent a global minimum . ( B ) Construct-specific RMSE scores illustrate impact of specific portions of the data set on model performance . Here , predictions for 38 constructs are plotted for the 24 models that were fit on a data set lacking all Snail mutant constructs . The performance is specifically diminished for constructs 23–33 , which represent Snail mutant enhancers ( compare to Figure 2B ) . ( C ) Removal of bHLH site perturbation constructs changes little in overall fitting . ( D ) The removal of constructs with Dorsal site mutations is associated with poorer performance across a spectrum of models on constructs that represent Dorsal site mutants ( columns 8 , 11–12 , 15–16 ) . ( E ) Similarly , omission of mutant constructs with perturbed Dorsal and Twist sites degrades performance especially on constructs 4–5 , 9–10 , 14 , 17–21 that represent Dorsal+Twist mutations . ( F ) Removal of Twist alone mutants especially impacts fitting on construct 17 , the Twist1+2 mutant construct . Weak performance of C11Q4 and C2Q6 across the spectrum in B , C , and D represents suboptimal parameter estimation by the genetic algorithm specifically in these runs . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 007 To examine construct-specific performance , we plotted a heat map illustrating individual construct fits for all 120 models ( Figure 2B ) . The models , represented by rows , were ranked from best to worst based on global performance , and individual fits for each of the 38 rho enhancer variants were plotted in columns . The higher ranked models ( blue rows ) have generally lower RMSEs across all constructs , and groups of models with similar structures had similar patterns of better or worse performance on particular constructs . Some constructs were fit generally less well by most models . Those constructs containing the wild-type ensemble of activators ( columns 1 , 25 , 28 , 32 ) were among those that were less well fit than the bulk of constructs , from which we had removed activator sites; apparently the generally narrower neuroectodermal expression pattern of those constructs with some activator sites removed drove overall parameter fitting . Individual plots of the measured and predicted activity of individual enhancer modules show that some error arises from the models underestimating the activator potential specifically in more dorsal regions of the embryo , where activator concentration is limiting ( Figure 2C; e . g . construct 1: DMRW ) . The lower quality fit for these constructs does not represent a model failure; rather , these constructs are likely to be especially informative for activator function , and a modeling effort that entirely lacks these types of constructs would be more over-fit and less informative than the present one . Further examination of individual plots for specific genes provides additional insight into the nature of which features influence RMSE scores the most ( Figure 2C ) . A top-ranking model ( C14Q5 ) accurately captures high and low levels of Snail repression , deviating only in underestimating the expression of rho in regions most limiting for Dorsal and Twist . An intermediate-scoring model ( C8Q7 ) was partially successful in capturing general trends of activation with occasional overestimation of activity . This model misestimated Snail repression in some cases as well . The lowest performing model ( C4Q4 ) suffered from poor estimation in Dorsal/Twist activity , as well as a general absence of repression by Snail . Thermodynamic models calculate protein occupancy on an enhancer using descriptions of binding preferences distilled from position weight matrices ( PWMs ) , which can incorporate in vitro or in vivo protein-DNA interaction data . A number of studies have tested DNA-binding preferences of Dorsal , Twist , and Snail , therefore we tested the effects of different PWMs on model performance . Three similar but non-identical PWMs were tested for each factor , and all possible combinations of these PWMs were utilized in refitting the data set with a selection of 24 of the entire set of models , including most of the top performers . Overall ranks of model performance remained similar , although RMSEs were lowest using the PWM of Snail that was employed in the global analysis in Figure 2; this PWM predicted all four sites previously identified by in vitro footprinting ( Figure 3 ) . Thus , the biophysical information about protein-DNA interactions for the transcription factors in this system appears to be sufficiently complete to provide a robust platform for thermodynamic modeling . 10 . 7554/eLife . 08445 . 008Figure 3 . Effect of PWM settings on model performance . ( A ) Three versions of each binding site were considered as discussed in Materials and methods , drawing from previously reported in vitro or in vivo binding studies . ( B ) The correspondence between previously identified bindings sites from DNAseI footprinting studies ( colored squares ) and predicted binding sites using the diverse PWMs ( circles ) is superimposed on the structure of the rho enhancer . The size of the circles indicates the significance of the match using the particular PWM for that factor; Twist sites in green , Dorsal in blue , and Snail in red . ( C ) Heatmap showing global performance of 24 selected models using 27 different PWM combinations for Dorsal , Twist and Snail . Results were clustered by performance; columns of PWM settings were arranged with lowest average RMSE values to the left , then rows of models were arranged with lowest global RMSE towards the top . ( D ) PWMs used for different settings are denoted by ‘X’s; Snail3 PWM ( used in the rest of this study ) shows strong correlation with best model fitting . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 008 In addition to information about biophysical parameters of the system , model performance can be dramatically influenced by the nature of the perturbations studied . We used cross-validation , a statistical technique in which segments of the data set are left out during model fitting , to reveal possible overfitting , and determine the impact of certain portions of the data set on overall success in fitting . We performed cross-validation for the 24 models studied above both by leaving out specific parts of the dataset with common attributes ( e . g . all Dorsal or Snail site mutations ) as well as by randomly removing portions of the dataset . We used these two complementary approaches to not only test how much data is needed for the model fitting to be successful , but also to emphasize how important experimental design is in a study such as this one . Some studies rely on rely on saturation mutagenesis to achieve the same effect , but testing thousands of variants is impractical in this system , so a targeted survey was called for ( White et al . , 2013 ) . As anticipated , models were in general less able to fit constructs that were left out of the dataset , indicating the contribution that these particular enhancer constructs made to the fitting ( Figure 2—figure supplement 1 ) . Despite similar performance on the entire dataset , individual models showed differing levels of sensitivity to changes in the scope of the data set used for model fitting . For instance , RMSE values for C13Q5 were modestly increased by either systematic or random leave-sets-out treatment , while C11Q4 was more dramatically impacted . In general , removal of entire classes of constructs had a more profound impact than removal of randomly selected constructs . To better understand how specific constructs contribute to model performance , we analyzed how leaving out certain sets of constructs affects model predictions on each individual construct , including those used for fitting and those left out ( Figure 2—figure supplement 1 ) . The most profound effects were seen with the omission of the Snail repressor site constructs , where almost all models had striking increases in RMSE for all Snail constructs . Evidently , the elimination of activator sites alone is not sufficient to provide insights into how Snail affects the enhancer’s expression . Collectively , the scope of mutations assayed , targeting recognized Twist , Dorsal , and Snail sites , was sufficient to provide the required information to the modeling effort . The different effects of activator and repressor site mutations illustrate the importance of perturbation of each of these elements to fully explore the functional terrain of the enhancer; for a small number of constructs , a random perturbation of cis-regulatory sequences may not uncover the most informative changes related to transcriptional relationships ( Patwardhan et al . , 2012; Sharon et al . , 2012; Smith et al . , 2013 ) . All thermodynamic models used here characterized transcription factor activity by estimating parameters for activation and repression potential , as well as the distance-dependence of cooperativity and repression . Several broad trends emerged from analysis of parameter values across many model types , revealing possible biological implications for the modeled enhancer ( Figure 4 , Figure 4—figure supplement 1 , Supplementary file 3 ) . First , overall model performance was improved by specifying separate parameters for activator and repressor cooperativity ( Figure 2A; C1-9 vs . C10-C15 ) . In these latter models , repressor-repressor cooperativities were consistently small , implying that different Snail sites did not show more than additive contributions , a finding that is consistent with the induction of localized chromatin compaction by this class of short-range repressor , which may interfere with simultaneous interactions by repressors ( Li and Arnosti , 2011 ) . Activator-activator cooperativities were in contrast uniformly high , and in general did not show sharp distance dependencies . Such interactions may represent indirect interactions among activators , and may reflect joint repulsion of nucleosomes , or cooperative attraction of coactivators important for engaging the transcriptional machinery . The loose cooperative interactions of activators would explain the need for clustering of these proteins’ binding sites without strong constraints on their exact positioning , a flexibility that is seen with many developmental enhancers ( Figure 4B; Arnosti and Kulkarni , 2005 ) . This long-distance cooperativity contrasts with the short-range interactions often included in thermodynamic descriptions of transcription factor cooperativity ( Segal et al . , 2008 ) . A second trend noted in many but not all models was the distance-dependence of transcriptional repression , a feature inferred from previous studies of the Knirps , Snail , and Giant short-range repressors ( Gray et al . , 1994; Arnosti et al . , 1996; Hewitt et al . , 1999 ) ; ( Figure 4—figure supplement 1B; Supplementary file 3 ) . Thus , the overall landscape of estimated parameters both reflects certain known aspects of repression , and sheds light on activator cooperativity , which may involve primarily indirect , not rigidly constrained interactions in this system . 10 . 7554/eLife . 08445 . 009Figure 4 . Estimated model parameters suggest biological properties of long-distance activator-activator cooperativity and weak repressor cooperativity . ( A ) Representative estimated cooperativity parameters for Dorsal and Twist transcriptional activators show a trend of high levels of cooperativity with little distance dependence , consistent with indirect , long-range effects i . Dorsal-Dorsal , ii . Dorsal-Twist , iii . Twist-Twist , and iv . Snail-Snail cooperativity values of model C13Q3 . Low values estimated for repressor cooperativity are consistent with weak interactions between Snail proteins . One of the five runs for fitting this model had significantly higher RMSE than the other four; the corresponding parameters are indicated by a gray line . Other runs produced similar RMSE and parameters ( magenta and blue ) . ( B ) Illustration of how global long-range cooperative interactions between activators permit relaxed constraints on binding site arrangement , while distance-dependent repressor positioning anchors Snail sites to activator sites . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 00910 . 7554/eLife . 08445 . 010Figure 4—figure supplement 1 . Global trends observed for parameter values and evidence of compensation . ( A ) Scaling factors ( related to estimated transcription factor 'potency' ) identified by 120 models for i . Dorsal , ii . Twist , and iii . Snail . Models ranking 1–34 had similar , lower scaling factors for Dorsal , compared to relatively higher values for models ranking 35–86 . The scaling factors for the lowest ranked models 87–120 varied widely . No such trends were observed for better or worse performing models with respect to Twist scaling factors; significantly , there are only two Twist sites on the minimal rho enhancer , and thus fewer perturbations to constrain these values . Snail scaling factor values were clustered for the top 40 models , and then showed greater variation for lower performing models . For each model , the scaling factor is shown for the parameter estimation run ( of five total ) with the lowest RMSE . ( B ) Parameters estimated for model C11Q8 for i . scaling factors , ii . homotypic and heterotypic cooperativity , and iii . quenching parameters representing action of Snail on Dorsal and Twist . Global parameter estimation was carried out five times , with overall RMSE shown iv . . Although global fits were very similar , the parameter values vary , and showed evidence of compensation . In particular , the scaling factor for Twist obtained in the first and second runs ( green triangle and blue square ) is higher than the values obtained in the other runs . This higher scaling factor parameter is associated with a markedly lower Twist-Twist homotypic cooperativity parameter , shown in position 3 ( the only interval for this model that is filled; dashed line in lane 4 signifies that there are no Twist sites at the spacing represented by this interval ) . The trade-off between overall activity and cooperativity is likely to account for much of the second order effects noted for many models ( Figure 4—figure supplement 2 ) . In iii . , Snail quenching parameters for action on Dorsal ( 1–10 ) and Twist ( 11–20 ) show generally lower values for more distant 'bins' , consistent with a short-range of action . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 01010 . 7554/eLife . 08445 . 011Figure 4—figure supplement 2 . Model sensitivity analysis . The 24 selected models used for PWM and cross validation were examined for first- and second-order relative sensitivity ( complete results in Figure 4—figure supplement 2; Figure 4—figure supplement 2—source data 1 ) . Here we show high-scoring C14Q5 model ( A ) , as well as lower-scoring C3Q4 and C2Q1 models ( B , C ) . As shown with these three examples , most parameters exhibited much higher second-order than first-order effects , indicating that compensatory relationships among parameters are predominant . Even with such compensation , certain types of parameters ( e . g . relatively high activator cooperativities ) are preferred ( also seen in Figure 4—figure supplement 1 ) . Model structures influenced the first- and second-order sensitivities; models using quenching schemes Q2-4 exhibited a significant degree of first order sensitivity for the Dorsal scaling factor ( B ) , while almost every other model combination showed a significant extent of first-order sensitivity for the Snail scaling factor ( A , C ) . Overall sensitivities , first- and second-order combined , were dominated in most models by those for the Snail scaling factor , consistent with the significant impact the particular Snail PWM had on modeling performance ( see Figure 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 01110 . 7554/eLife . 08445 . 012Figure 4—figure supplement 2—source data 1 . Sensitivity analysis . Sensitivity analysis for each model including first- and second-order sensitivity of parameters and plots of these values . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 01210 . 7554/eLife . 08445 . 013Figure 5 . Application of rho cis-regulatory 'grammar' to heterologous regulatory elements . Discovery of short-range repression from modeling of rho perturbation dataset . ( A ) High ranking model C13Q6 ( 3rd overall ) correctly predicts distance-dependent significance of Snail sites within enhancer element; specifically , the third but not fourth construct positioned ectopic Snail binding sites ( purple balls ) close enough to Dorsal 1 and Dorsal 4 motifs to repress the expression in ventral-most parts of the embryo . ( B ) lower ranking model C2Q7 ( 78th overall ) correctly predicts activation potential , but fails to detect distance effects impacting repression by Snail . ( C ) Shows cartoons of the qualitative expression found by Gray and colleagues and the reporter genes tested in embryos ( Gray et al . , 1994 ) . ( D ) Correctly predicted expression patterns of cis-regulatory elements i . D . melanogaster vnd enhancer , ii . D . melanogaster twi proximal element , iii . D . erecta rho enhancer , and iv . D . ananassae rho enhancer . X-axis denotes distance from ventral end of the embryo , and Y-axis denotes expression values . Predictions of model C14Q5 ( ranked 1st overall ) are shown as red lines; measured average transgene expression in black points . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 01310 . 7554/eLife . 08445 . 014Figure 6 . Thermodynamic models predict quantitative outputs of additional , non-rho enhancers in D . melanogaster and putative rho enhancers from related drosophilids . ( A , B ) neuroectodermal elements scored for repression by Snail and activation by Dorsal and Twist , respectively . ( C , D ) mesodermal elements scored for lack of repression by Snail and preferential activation in mesoderm , respectively . Lowest ranked models performed poorly for correct prediction of Snail activity on neuroectodermal enhancers , but were also not liable to make a false positive call on Snail activity for mesodermal elements ( A , C ) . The D . erecta rho element was correctly called by most models , while some sog , vn , vnd , and brk constructs were less well fit by most models , indicating that there are likely enhancer-specific features of these elements that are not sampled by rho variations . For mesodermal elements derived from snail and twist proximal regions , those models that correctly scored Snail activity as low were able to produce more 'mesoderm-specific' type patterns . No model scored consistently highly across all constructs , but many performed well in a complementary fashion , suggesting that individual models are not overfit in identical manners . Activity of regulatory elements was experimentally measured in transgenic D . melanogaster , and output compared with the predicted expression from 24 models thermodynamic models fit on D . melanogaster rho expression . Fitting for predicted vs . measured expression was assessed using quantitative measures for correct expression as described in Materials and Methods . Lower scores represent better fits . D . erecta , D . ananassae , D . mojavensis , D . grimshawi , and D . virilis putative rho elements were assayed ( sequences indicated in Supplementary file 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 01410 . 7554/eLife . 08445 . 015Figure 7 . A panel of models fit on rho enhancer predicts additional neuroectodermal enhancer elements . ( A ) Predicted expression activity of a genomic region ( 36 kb ) flanking the neurectodermal brk gene , using a panel of 24 models , including 21 higher and three lower performing models . Blue marks indicate predicted mesodermal pattern , red neuroectodermal . The transcribed area for brk is indicated by the black rectangle; green peaks indicate in vivo binding for Dorsal , Twist , and Snail; previously mapped embryonic enhancers are indicated by gray rectangles . Rows of plots below the gene show expression patterns of: first row – top model , second row – composite of all 24 model predictions , third row – average of all 24 model predictions in solid black with standard deviation from the mean shown above and below in dotted black lines . Note that the average plot is similar to a plot of a weighted average using RMSEs or AICs to weight each model’s output . ( B ) Similar scan of the even-skipped locus ( 32 kb ) , which is not regulated by Dorsal , and where few areas are consistently identified as potential Dorsal/Twist/Snail enhancers . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 01510 . 7554/eLife . 08445 . 016Figure 7—figure supplement 1 . A panel of models fit on rho enhancer predicts additional neuroectodermal enhancer elements around sog and twi , with lower scores around ftz control region . DOI: http://dx . doi . org/10 . 7554/eLife . 08445 . 016 As we noted in previous studies , estimated parameters from thermodynamic models show different levels of sensitivity due to the structure of models and of data ( Dresch et al . , 2010 ) . We tested the set of 24 models with synthetic data sets to determine sensitivities , and determined that most parameters showed extensive mutual dependence and compensation , evidenced by the low first-order and high second-order sensitivities ( Figure 4—figure supplements 1 and 2 ) . The observed convergence of cooperativity and quenching parameters for a number of models during our data fitting thus suggests that although first-order sensitivities are low , there is important information inherent in these values . The objective of our study was to develop a quantitative understanding of novel enhancer sequences , using our models in a predictive mode . To test their efficacy , we first asked whether the perturbation analysis and modeling of rho was capable of uncovering a key property of modular enhancers in the embryo , namely the short range of transcriptional repressors that prevents 'cross-talk' with other nearby elements ( Gray et al . , 1994; Courey and Jia , 2001; Payankaulam et al . , 2010 ) . Previous work showed that Snail protein is representative of this major functional class of repressors , in studies that involved moving Snail binding sites within transcriptional regulatory regions ( Gray et al . , 1994 ) . Our perturbation dataset did not involve moving Snail binding sites around , however , the top third of models in our test set were able to accurately capture the effects of loss of Snail sites from a 700 bp native rho element , as well as correctly characterize modified enhancers in which novel Snail sites were inserted into active as well as inactive locations ( Figure 5A–C ) . To determine whether modeling from rho can be broadly extended to other regulatory elements , we measured the quantitative output of other enhancers targeted by Dorsal and Twist , and compared their expression patterns with the predictions of a spectrum of our models . Heterologous elements from D . melanogaster and homologous rho enhancers from other Drosophila species were successfully modeled by the models trained on the rho dataset ( Figure 5D and Figure 6 ) . The previous assays tested the models on elements already known to act as Dorsal response elements and to bind Dorsal/Twist . To determine whether models would be able to identify and correctly quantitatively score enhancers embedded in general genomic sequences , we tested the panel of models on 30–40 kb regions flanking a number of developmental genes . The regions were tiled in 500 bp overlapping sequences and predictions generated with parameters from the rho dataset fitting . Analysis of the neuroectodermal brk and sog genes , as well as the more widely expressed twi gene showed that most intergenic sequences did not give rise to any appreciable predicted expression ( Figure 7 and Figure 7—figure supplement 1 ) . To help one visualize how well the models agreed ( or disagreed ) on particular regions , average predictions from all 24 models were also generated for each 500 bp window , and plotted along with the standard deviation above and below the mean . A strong consensus among the 24 models for activity was found in two regions flanking the brk gene; these regions include mapped embryonic enhancers ( primary and shadow ) that are known to bind Dorsal and Twist in vivo ( Figure 7A ) . Some regions were less uniformly predicted by the models , and in these cases there was less uniformity about the prediction for mesodermal or neuroectodermal expression , indicating that Snail activity was variably predicted . These predictions tended to agree well with genomic protein occupancy information , although there were some regions predicted as enhancers where there was no protein occupancy measured . In contrast , there were fewer predicted active elements in the region surrounding even-skipped ( eve ) locus , which is expressed in an anterior-posterior pattern independent of Dorsal/Twist . In addition , for intergenic regions flanking eve , there was greater variability among models in predicted expression levels , patterns of expression , and whether an element was even likely to be active ( Figure 7B ) . Similar results were noted for neuroectodermal and mesodermal-specific short gastrulation ( sog ) and twist ( twi ) genes respectively , as well as the non-target gene fushi tarazu ( ftz ) ( Figure 7—figure supplement 1 ) . The use of multiple models as a predictive 'jury' may help overcome the overfitting that is inherent in modeling , as the effects of such bias for a given model may be more or less prominent in different contexts . The ensemble approach of averaging the 24 model predictions provides a convenient metric for transcriptional potential predictions , yet the graphic display of the individual models also serves an important role , to highlight particular regions where there is a high degree of uniformity about the prediction ( e . g . primary and shadow brk enhancers ) and those in which the models diverge sharply ( eve proximal region ) . Bioinformatic sequence analysis and direct comparison of chromatin features have been used to identify cis regulatory elements in metazoan genomes , however the significance of patterns of binding sites within such elements remains obscure . Many models simply rely on the number of sites to infer some level of activity ( Kazemian et al . , 2010; Heintzman et al . , 2009; Bonn et al . , 2012 ) . Our study indicates that the arrangement and quality of transcription factor binding sites contains an essential common 'grammar' of cis-regulatory regions that is shared across distinct regulatory elements; this stands in contrast to recent studies suggesting that in some cases transcription factors may recognize their target sequences primarily through protein-protein rather than protein-DNA interactions ( Kulkarni and Arnosti , 2005; Junion et al . , 2012 ) . Our models were biologically constrained , leading to structural similarities between the different model formulations . Although this may contribute to the clustering of parameter values we have observed , given the overall quality of predictions across the genome , we believe these models and parameter values to be biologically informative . Predictions of enhancer activity from sequence alone remain a challenge , but our study clearly shows that with a suitable training set , there are identifiable aspects of enhancer composition that allow for functional insights of mesodermal and neuroectodermal genes of the Dorsal regulon . Models such as these can guide future studies that mine population- and species-level variation of genomic sequences ( Mackay , 2010 ) . Two features were important for the successful implementation of this thermodynamic modeling: development of a high quality perturbation dataset , which in our case involved several dozen carefully designed constructs whose transcriptional output was measured quantitatively , and exploration of a variety of model structures . As we determined when applying these models to genome discovery , using a panel of models provides a more robust platform for interpreting possible regulatory information; the particular parameter sets developed from fitting the rho perturbation data may represent a certain level of overfitting with respect to all potential Dorsal/Twist target enhancers . These errors are partially cancelled out when the predictions of several dozen models can be evaluated , providing in effect a type of 'weather report' showing a certain chance of regulation at each element . Here , we have also used an ensemble approach to combine these predictions , averaging model predictions; an approach similar to that which was implemented in a recent study by Samee et . al . ( Samee et al . , 2015 ) . However , in their work they use a single thermodynamic model formulation and average the predictions obtained from different parameter sets with similar fits to their perturbation data . In addition to recapitulating known facets of cis-regulatory grammar in Drosophila including the distance-dependent activity of repressors , this modeling approach identified critical quantitative features of activator and repressor interactions at enhancers . Specifically , for Dorsal and Twist , key regulators of dorsal-ventral polarity in the early embryo , our modeling predicts strong distance-independent homo- and heterotypic cooperativity between activators , but weak cooperativity between the Snail short-range repressors . These effects are likely linked to chromatin-based activities of these proteins ( Li and Arnosti , 2011 ) . Previous , simpler models suggest that direct protein-protein interactions within a range of ~50 bp are likely to dominate such functional interactions ( Segal et al . , 2008 ) . The extensive survey of model structure allowed us to reject assumptions this simplifying assumption . These predicted properties of transcription factor activity were robustly observed for differently structured models , making us confident that they are biologically relevant and not the result of overfitting , and the initial application of models to genomic sequences points to the future use of these models in genomic characterization of the Dorsal-Twist regulon . Our ability to gain insight into the transcriptional landscape of this network is a proof of concept demonstrating the utility of such a modeling approach for identifying regulatory relationships in other systems . Our modeling approach was focused on three main regulators of rho activity , and did not take into account additional factors affecting cis-regulatory elements , such as intrinsic chromatin occupancy , modification of transcription factor activity by signaling , or additional proteins involved in regulation of some of the enhancers . Nonetheless , future models will incorporate such additional layers of information in more complex treatments , such as including basal chromatin patterns to bias the accessibility of transcription factor sites ( Bryant et al . , 2008; Floer et al . , 2010 ) . Our study has addressed only a fraction of the diversity of factors present in the fly embryo; we deliberately focused on the early Dorsal regulon for the richness of quantitative resources available to it , including transcription factor concentrations , binding specificity of trans-acting factors , and genomic data on in vivo targets . With continuing advances in genomics and high-throughput technologies , quantitative modeling can extend our knowledge of other regulons , building on extensive descriptions of gene and protein expression , genomic protein occupancy , and genomic variation . A greater promise for modeling the quantitative grammar of genomes will be in 'personalized genomics' , in which investigators predict the effects of sequence variation in human populations , and their physiological relevance in development and disease ( Corradin and Scacheri , 2014; Newman and Young , 2010 ) . The 318-bp rhomboid neurectodermal enhancers were cloned into AgeI and FseI restriction sites of the pHonda1 pattB-based targeted integration vector ( Ip et al . , 1992; Sayal et al . , 2011 ) . Enhancers were assembled from 40–60 bp overlapping synthetic 5’-phosphorylated oligonucleotides with 10 bp overhangs , which were annealed , and then ligated into pHonda1 . The footprinted binding sites for Dorsal , Twist and Snail , as well as two predicted E-box motifs thought to be bound by bHLH factors , were mutated as follows using sequences previously shown to affect rho enhancer activity ( Ip et al . , 1992; Jiang et al . , 1991 ) : Dorsal1 - GGGAAAAACAC to TTTAAAAACAC Dorsal2 - CGGAATTTCCT to CGTCAGTTAAT Dorsal3 - GGGAAATTCCC to TCTAGATTATC Dorsal4 - GGGAAAGGCCA to AGGCCTGGTCA Twist1 - CGCATATGTT to ACGCGTTGTT* Twist2 - AGCACATGTT to ACGCGTTGTT Snail1 – CAACTTGCGG to CAGAGCTCGG Snail2 - CACCTTGCTG to CAGGAGCTTG* Snail3 – CCACTTGCGCT to CCGCCGGCGT* Snail4 – GCACATGTTT to GCATATGTTT bHLH1 - CATTTG to TGATTC* bHLH2 - CAAGTG to TAGCGA* ( *novel mutations developed for this study ) The bHLH1 and bHLH2 sites were mutated simultaneously . The mutations are predicted to reduce the binding score for each transcription factor to near background values . Additional wild-type enhancers for other genes were created by PCR amplification from genomic DNA or by assembly using oligonucleotides as indicated above . We used the Clusterdraw bioinformatics tool to identify putative rhomboid regulatory sequences in non-D . melanogaster genomes ( Small et al . , 1992; Papatsenko , 2007 ) . Supplementary file 1 contains details of rhomboid and other genes’ enhancer sequences and their nomenclature . All constructs were integrated into the same site on chromosome 2 ( chromosomal location 51D; Bloomington stock center’s stock #24483 ) . DNA microinjections were performed in-house and by Rainbow Transgenic Flies , Inc . Transgenic lines were made homozygous , and only embryos from homozygous fly lines were used for confocal microscopy . Embryos were collected and fixed as previously described ( Small et al . , 1992; Kosman et al . , 2004; Janssens et al . , 2005 ) . Immunofluorescent in situ hybridization was done essentially as previously described with some modifications ( Kosman et al . , 2004; Janssens et al . , 2005 ) . All washes were done in 1 ml volume . About 50 μl of fixed embryos stored at -20°C in methanol were briefly washed six times with 100% ethanol , followed by a wash in xylenes for 30 min , and lastly , six times again with 100% ethanol . The embryos were then washed four times with 50%methanol-50%phosphate buffer-0 . 1%-Tween 80 ( PBT; 137 mM NaCl , 4 . 3 mM Na2HPO4 , and 1 . 4 mM NaH2PO4 ) and then with PBT four times , each for 2 min with continuous rocking . Embryos were washed in ( 1:1 , v/v ratio ) PBT/ hybridization solution ( hybridization solution: 50% formamide , 5X SSC [0 . 75M NaCl and 75 mM Na-citrate] , 100 μg/mL sonicated salmon sperm DNA , 50 μg/mL heparin , and 0 . 1% Tween 80 ) for 10 min , and then briefly in hybridization solution for 2 min . New hybridization solution was added , and the tubes were placed for 1 hr in a water bath at 55°C . Previously titrated antisense RNA probes of digU-labeled lacZ and biotin-labeled eve and sna were heated in 65 μL hybridization solution at 80°C for 3 min and directly placed on ice for 1 min; hybridization solution was completely removed from the embryos , and the probes were added to the embryos in a final volume of 65 μL in each tube , and incubated at 55°C overnight . After incubation , 1 mL of 55°C hybridization solution was added to each tube; all tubes were rocked at room temperature for 1 min , hybridization solution was changed , and tubes were incubated for another 1 hr at 55°C , followed by four washes with hybridization solution for 15 min each at 55°C and with hybridization solution and PBT ( 1:1 , v/v ratio ) two times at room temperature for 15 min . Five more washes were done with PBT for 10 min with rocking at room temperature . The embryos were washed with a blocking solution consisting of a mixture of PBT and 10% casein in maleic acid buffer ( Western Blocking Reagent; Roche , Indianapolis , IN 11921673001 ) ( 4:1 , v/v ratio ) . 0 . 5 ml of a 4:1 v/v mixture of PBT and 10X blocking solution containing primary antibodies ( 3 μl of sheep anti-digoxigenin , ( Roche 11333089001 ) ; 1 μl of mouse anti-biotin ( Invitrogen 03–3700 ) was added , and the tubes were rocked at 4°C overnight . Embryos were washed four times each with PBT for 15 min at room temperature . 0 . 4 ml of mixture of PBT and 10% casein blocking reagent and PBT ( 4:1 v/v ) , containing 8 . 0 μl of each secondary antibody ( donkey anti-sheep Alexa 555 ( Invitrogen A-21436 ) for detection of lacZ mRNA and donkey anti-mouse Alexa 488 ( Invitrogen A-21202 ) for detection of eve and sna mRNA ) Secondary antibodies that had been pre-absorbed for at least 2 hr against fixed yw embryos in PBT and 0 . 4 μl of TOPRO-3 DNA dye ( Invitrogen , T3605 ) were also added to each vial . Tubes were covered with aluminum foil to protect them from light and incubated overnight with rocking at 4oC . Embryos were then washed with PBT four times at room temperature for 5 min . with rocking , and washed in glycerol-PBT ( 7:3 , v/v ratio ) for 2 hr until the embryos settled to the bottom of the tubes . The embryos were then resuspended in 0 . 4 mL glycerol-PBT ( 9:1 , v/v ratio ) and 0 . 2 mL of PermafluorTM mounting medium ( Fisher TA-030FM ) , mounted on labeled slides , and covered with large rectangular Corning cover slips ( No . 1 . 5; 24 X 50 mm ) . The slides were protected from light and stored flat at room temperature until the embryos were imaged . An Olympus Spectral FluoView FV1000 Confocal Laser Scanning Microscope ( Olympus , Center Valley , PA ) configured on an IX81 inverted microscope was used for capturing the confocal fluorescent images . For each scan of mounted embryos on a particular day of imaging , the same microscope settings for wild-type rho transgenic embryos were employed to all images to allow for direct comparison of intensities . The 488 nm argon laser was used for excitation of the Alexa 488; the 559 nm solid-state laser was used for excitation of the Alexa 555 , and the 635 nm solid-state laser was used for excitation of the TOPRO-3 . Emitted fluorescence was detected using a 500–545 nm band pass filter for detection of the Alexa 488 , a 570–625 nm band pass filter for detection of Alexa 555 , and a 655-755 nm band pass filter for detection of TOPRO-3 . The pinhole aperture was set to 1 . 0 Airy unit . PMT voltage was set at maximum for images obtained from embryos transgenic for the wild-type rhoNEE enhancer , avoiding saturation of signal intensities . Other constructs were imaged subsequently on a single day without changing any of the microscope settings . Embryos were imaged at a scan speed of 12 . 9 s/scan and a Kalman average of 2 . A total of 21–30 confocal images through the Z thickness were acquired for each embryo with a Z-step interval of 1 . 16 μm per step . CLSM image data was stored as three separate stacks and projections of images for each channel . The section dimensions were 333 mm in length and width , and 1 . 73 mm in depth . Fluorescence pixels were recorded as 12-bit images and stored as TIFF files . To control for fading of signal post-staining , rho constructs containing the wild-type ensemble of activators were stained in parallel and used to normalize overall signal intensity for each imaging day . Stained embryos were imaged within a week to minimize loss of signal . Differences in probe bleaching , laser intensity , gain settings of the CCD etc . all impact overall signal intensity . For the 348 control images captured over 53 imaging sessions , the range of average peak intensities , prior to any background subtraction or normalization , was 56 . 8 – 255 units ( only 3 were at 255 , saturation value ) . The mean was 138 . 0 , with a standard deviation of 36 . 8 . Thus , for the large majority of captured signals , the day-to-day differences in intensities were not very great , and normalization procedures were not changing values by large factors . The background signals from non-expressing portions of the embryos were 52 . 4 +/- 23 . 8 ( S . D . ) , thus considerably ( 2X ) below the signal; and in all cases the strong signals measured on any day were well above background measured on any day . All confocal microscopy images were processed in a six-step procedure involving binary image generation , rotation , resizing , background subtraction , normalization and intensity-value extraction . Binary image generation , rotation and resizing were done as described previously ( Myasnikova et al . , 2005; Ay et al . , 2008 ) . The area of interest for all embryos comprised a region spanning from 40–60% egg length on the anterior-posterior axis . Ten samples , uniformly spaced , were taken from this region , plotted together and averaged along the dorsal-ventral axis ( as is illustrated in Figure 1 ) . For background subtraction , analysis of background signals from non-transgenic , yw flies showed a parabolic-shape ( Bergman et al . , 2005; Myasnikova et al . , 2005 ) , therefore a quadratic function was fit to the region of the signal representing the dorsal ectoderm , where rho is not expressed , and subtracted from the raw fluorescent signal . To normalize signals , values from each image were normalized to the average peak ( >95% ) wild-type signal obtained during the same imaging session . This procedure allows for images to be compared for a single construct imaged on multiple days , as well as to compare intensity from one construct to another . The average intensity profiles , along with standard error calculations are given in Figure 1—figure supplement 1 and Supplementary file 2 . For model fitting , we discretized the continuous expression data . We did this by taking the data in the region from 0–40% of the DV axis ( approximately 102 pixels in each image ) , and averaging every 6 pixels to result in 17 data points corresponding to this region; hence a data point every 2 . 5% of egg height ( as is illustrated in Figures 2 and 5 ) . We excluded the dorsal ectoderm from our modeling efforts , as we wanted to focus our attention on the areas of the embryo with varying expression levels across mutant constructs . For the 59 constructs analyzed , a total of 935 embryo images were taken , with a minimum of 10 images per construct . Late stage 5 ( pre-gastrulation ) cellularizing embryos were used for analysis , and eve expression was used to select the embryos in a narrow age range . Embryos were also selected based on their rotation , so that the rhomboid lateral stripe was near the center of the image , with a sufficient number of pixels in the dorsal region of the embryo for background estimation . Because there are slight differences in the reported PWMs for Dorsal , Twist , and Snail , we considered information from a variety of sources . For Dorsal , PWMs were obtained from two sources: a PWM generated by MEME analysis ( with default settings ) of footprinted binding sites found in FlyReg ( Bergman et al . , 2005; Noyes et al . , 2008 ) , herein referred to as DL1 , and bacterial one-hybrid experiments ( Zinzen et al . , 2006; Noyes et al . , 2008; Ozdemir et al . , 2011 ) , referred to as DL2 . The two position probability matrices were then averaged , and the log values calculated from this averaged matrix were used to yield a third hybrid PWM for Dorsal , DL3 . For Twist , PWMs were used from two different SELEX experiments ( Zinzen et al . , 2006; Ozdemir et al . , 2011 ) , herein referred to as TW1 and TW2 respectively . Subsequently , averaging the two PWMs as described above for Dorsal then derived a third hybrid PWM , referred to as TW3 . For Snail , three different sources were used: SELEX data from BDTNP ( http://bdtnp . lbl . gov ) , SELEX data from a previously published study ( Zinzen et al . , 2006 ) , and a PWM generated by MEME analysis of footprinted binding sites found in FlyReg ( Bergman et al . , 2005; Bailey et al . , 2009 ) , herein referred to as SN1 , SN2 and SN3 respectively . For analysis of enhancer sequences , we used the MAST program from the MEME software suite to identify putative binding sites ( Zinzen et al . , 2006; Bailey et al . , 2009 ) . The thresholds used in thermodynamic modeling were evaluated by recovery of known footprinted binding sites , although for some settings not all PWMs were able to find all footprinted sites . P-values for binding sites used in Figure 2–5 were set at p=0 . 001 for all factors . PWMs used were: DL1 , TW3 and SN3 . Quantitative values for concentrations of Dorsal , Twist and Snail were obtained for early Drosophila embryo ( stage 5 ) from a previously published study ( Fakhouri et al . , 2010; Zinzen et al . , 2006 ) . The published data consisted of 1000 average concentrations for each protein uniformly distributed along the DV axis . Since we were only concerned with the portion of the embryo in the ventral region , we took the region from 0 – 40% of the DV axis and chose a subset of the 1000 data point ( 17 uniformly distributed data points corresponding to this region; hence a data point every 2 . 5% of egg height ) as our Dorsal , Twist , and Snail concentration gradients . The data used for modeling is given below: Dorsal: 0 . 85326 0 . 77516 0 . 68914 0 . 59981 0 . 51152 0 . 42792 0 . 35175 0 . 28472 0 . 22757 0 . 18021 0 . 14193 0 . 11165 0 . 08811 0 . 07001 0 . 05618 0 . 0456 0 . 03746 Twist: 0 . 93224 0 . 88219 0 . 81279 0 . 70658 0 . 54216 0 . 34085 0 . 17674 0 . 08318 0 . 03873 0 . 01842 0 . 00892 0 . 00433 0 . 00208 0 . 00097 0 . 00044 0 . 00019 0 . 00008 Snail: 0 . 985 0 . 976 0 . 967 0 . 957 0 . 902 0 . 441 0 . 043 0 . 005 0 . 001 0 0 0 0 0 0 0 0 The modeling approach implemented in this study is a thermodynamic-based modeling approach , similar to models used in previous studies ( Zinzen et al . , 2006; Segal et al . , 2008; Fakhouri et al 2010; He et al . , 2010; ) . These models are derived using the law of mass action and thermodynamic equilibrium assumptions . They take information regarding the number and arrangement of TF binding sites , as well as TF concentrations , and output predicted levels of gene expression . Here , we use thermodynamic models that assume RNA polymerase ( RNAP ) is recruited by bound TFs , and thus model transcriptional output as proportional to the probability of the enhancer being in an ‘active state’ . Other assumptions used by all models tested in this manuscript include: To test different hypotheses about biochemical mechanisms of transcription factor activity on enhancers , several different schemes involving transcription factor cooperativity and short-range repression were implemented in our modeling effort . To create models that considered the diverse cooperativity and repression ( referred to as quenching ) relationships we propose , all possible pair-wise combinations of the fifteen cooperativity and eight quenching approaches were considered , generating 120 different models . For short-range repression , we used three continuous functions ( Linear-Q2 , Logistic-Q3 and Gaussian Decay-Q4 ) to describe change in repressor activity the percentage of time that the repressor is actively repressing ( or quenching ) an adjacently bound activator , as a function of the distance , d , in base pairs , from the repressor binding site to the activator binding site . When implemented , a=1 and b>0 is a model parameter for quenching functions . For cooperativity functions , ‘a’ and ‘b’ are both model parameters . An alternative approach involved 'binning' distances between activators and repressors . We fit quenching parameters ( Q ) for each of the bins . We also used the non-monotonic 'quenching' function ( Q1 ) derived from our analysis of short-range repression by the Giant protein in synthetic enhancer constructs ( Hansen et al . , 2003; Fakhouri et al . , 2010; Suleimenov et al . , 2013 ) . The binned quenching schemes are described as follows . The distances between binding sites were calculated from the center of the binding sites . Because of minimal center-to-center distances between Snail and Twist or Dorsal , the actual minimal distance possible is 11 bp in the wild-type rho enhancer sequence . Scheme Q5: q1: 1–25 bp , q2: 26–50 bp , q3: 51–75 bp , q4: 76–100 bp Scheme Q6: q1: 1–35 bp , q2: 36–70 bp , q3: 71–105 , q4: 106–140 bp Scheme Q7: q1: 1–45 bp , q2: 46–90 bp , q3: 91–135 , q4: 136–180 bp Scheme Q8: q1: 1–10 bp , q2: 11–20 bp… q9: 81–90 bp , q10: 91–100 bp For cooperativity functions , we use the same functions as above ( 1–3 ) to describe the multiplicative effect of cooperative binding between two adjacently bound activators , as a function of the distance , d , in base pairs , between the activator binding sites . When implemented as cooperativity functions , a>0 and b>0 are both model parameters . We considered two different ways of estimating cooperativity between transcription factors: heterotypic ( between Dorsal and Twist ) and homotypic ( Dorsal-Dorsal , Twist-Twist , or Snail-Snail ) . We tested three different continuous functions ( Linear-C1 , Logistic-C2 and Gaussian Decay-C3 ) , which were parameterized with a single pair of parameters for all homotypic interactions , and separate values for Dorsal-Twist cooperativity . Additional models with 'binned' distances were also considered . For each of the binned schemes , we used a simpler form in which all homotypic interactions are parameterized with the same values , and a more complex form where each type of protein interaction for a given bin size receives distinct parameters . Each of these schemes therefore generates two model forms – binned and protein-binned respectively . Schemes C4 and C10: c1: 1–25 bp , c2: >25 bp Schemes C5 and C11: c1: 1–50 bp , c2: >50 bp Schemes C6 and C12: c1: 1–75 bp , c2: >75 bp Schemes C7 and C13: c1: 1–50 bp , c2: 51–100 bp , c3: >101 bp Schemes C8 and C14: c1: 1–60 bp , c2: 61–120 bp , c3: >121 bp Schemes C9 and C15: c1: 1–70 bp , c2: 71–140 bp , c3: >141 bp For a summary of parameters in each model , see Table 1 . A global parameter estimation strategy , CMA-ES ( Covariance Matrix Adaptation - Evolutionary Strategy ) was applied to estimate the parameters ( Hansen et al . , 2003; Segal et al . , 2008; Fakhouri et al . , 2010; Suleimenov et al . , 2013 ) . Root mean square error ( RMSE ) was used as a measure of performance of different cooperativity and quenching schemes , as described previously ( Matsumoto and Nishimura , 1998; Segal et al . , 2008; Fakhouri et al . , 2010 ) . This RMSE was calculated using the 17 discrete expression points , as described above , and corresponding data points coming from the model predictions . Note that we used these 17 points , taken from 0–40% of the DV axis , and RMSE which gives each point equal weight , because we wanted to focus on the region of the embryo in which varying expression levels were observed and use the same scoring method with no bias across all construct . Due to the stochastic nature of starting points and fixed maximum number of runs for CMA-ES , estimations were run five times , which was empirically found to be sufficient to produce similar , minimal RMSE values for at least three of the runs in at least 47% of cases . Since we tested a number of different model formulations , to evaluate the performance of an ensemble of models , we used multiple different approaches . In one approach ( results shown in Figure 7 and Figure 7—figure supplement 1 ) , we calculated the average expression profiles of a panel of models on 500 bp fragments from genomic sequences flanking a number of developmental genes . We also investigated ensemble approaches using weighted averages , including an average weighted by the model’s performance ( one minus the RMSE ) , and an average weighted by the model’s AIC ( Akaike Information Criteria ) , a fitness measure which penalizes for the number of model parameters . The results obtained from these three ensemble approaches were similar . Therefore , only the unweighted average expression profiles are shown in Figure 7 and and Figure 7—figure supplement 1 . All image processing was done using ImageJ and MATLAB . Binding site locations were determined using the MAST algorithm in the MEME suite ( Bailey et al . , 2009 ) . All thermodynamic modeling was done using code written in C++ and run on the HPCC ( High Performance Computational Cluster ) at Michigan State University . Scripts to run MAST , create input files and run multiple versions of the model were written in C++ and Python . The C++ source code for the thermodynamic models used is available at http://www . github . com/arnosti-lab/ThermoModel , the MEME suite is available at http://meme . ebi . edu . au/meme/doc/download . html , and the C-source code for the CMA–ES algorithm is available at http://www . lri . fr/~hansen/cmaes_inmatlab . html#cpp ( Bailey and Gribskov , 1998; Hansen et al . , 2003 ) . The 38 constructs to be fitted were separated into 5 randomized partitions of size eight ( three partitions ) and seven ( two partitions ) . The partitions were computer-generated using the Python random . shuffle method , which is based on the Mersenne Twister algorithm ( Zinzen et al . , 2006; Matsumoto and Nishimura , 1998; Dresch et al . , 2010 ) . This process was repeated five times to give five different partitioning schemes . All 38 constructs were then predicted using parameters from each run , and average RMSE of the constructs left out was considered . Sensitivity analysis was performed for the 24 selected models as previously described ( Zinzen et al . , 2006; Dresch et al . , 2010 ) . Uninformative parameters , i . e . , those with empty bins , were excluded from the analysis . First-order relative sensitivity denotes the sensitivity of model to changes in values of a particular parameter , while second-order sensitivity denotes the sensitivity of model to changes in values of a parameter in combination with other parameters . A parameter with high first-order relative sensitivity is likely to be informative on its own , whereas a parameter with high second-order relative sensitivity implies that the model may have high predictive power , but the parameter values are not informative on their own due to inter-parameter dependencies . Scoring of predictions from experimentally measured enhancers cloned into pHonda1 that were not included in the model fitting shown in Figure 6 . For neurectodermal enhancers , a four-point scheme was applied to score Snail repression as well as neurectodermal activation . Snail repression was measured at nucleus 4 . Snail repression was scored as: Neurectodermal activation was scored at the peak in a four-point scheme: For mesodermal enhancers , scoring was done on a five-point scale for activation and four-point scale for Snail activity . The activation score is given below: Snail repression scale is given below:	DNA contains regions known as genes , which may be “transcribed” to produce the RNA molecules that act as templates for building proteins and regulate cell activity . Proteins called transcription factors can bind to specific sequences of DNA to influence whether nearby genes are transcribed . For example , so-called enhancer regions of DNA contain several binding sites for transcription factors , and this binding activates gene transcription . Little is known about how the transcription factor binding sites are organized in enhancer regions , which makes it difficult to use DNA sequence information alone to predict the regulation of genes . A transcription factor called Dorsal controls the activity of a network of genes that plays a crucial role in the development of fruit fly embryos . Dorsal binds to the enhancer region of a gene called rhomboid , which has been well studied and is known to be a fairly typical example of an enhancer region . To understand the regulatory information encoded in the DNA sequences of enhancers , Sayal , Dresch et al . have now used a technique called perturbation analysis to investigate the interactions that are likely to occur between Dorsal and other transcription factors as they bind to the rhomboid enhancer . This technique involves systematically mutating the enhancer to remove different combinations of transcription factor binding sites and quantitatively investigating the effect this has on gene activity . A large set of mathematical models were then trained using this data and shown to correctly predict the activity of a range of other gene regulatory regions . The collective predictions of the models identified new enhancer regions and revealed details about how different types of transcription factor binding sites are arranged within enhancers . As we enter an era where the DNA sequences of entire human populations are increasingly accessible , we would like to know the functional significance of changes in gene regulatory regions . Sayal , Dresch et al . show that the regulatory properties of specific control proteins are accessible by employing quantitative experiments and mathematical models . Similar studies will be required to learn how mutations found across the genome may alter gene expression , leading to better diagnosis and treatment of disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology" ]	2016	Quantitative perturbation-based analysis of gene expression predicts enhancer activity in early Drosophila embryo
The colon hosts gut microbes and glucagon-like peptide 1 secreting cells , both of which influence glucose homeostasis . We tested whether colectomy is associated with development of type 2 diabetes . Using nationwide register data , we identified patients who had undergone total colectomy , partial colectomy , or proctectomy . For each colectomy patient , we selected 15 non-colectomy patients who had undergone other surgeries . Compared with non-colectomy patients , patients with total colectomy ( n = 3 , 793 ) had a hazard ratio ( HR ) of clinically recorded type 2 diabetes of 1 . 40 ( 95% confidence interval [CI] , 1 . 21 to 1 . 62; p<0 . 001 ) . Corresponding HRs after right hemicolectomy ( n = 10 , 989 ) , left hemicolectomy ( n = 2 , 513 ) , and sigmoidectomy ( n = 13 , 927 ) were 1 . 08 ( 95% CI , 0 . 99 to 1 . 19; p=0 . 10 ) , 1 . 41 ( 95% CI , 1 . 19 to 1 . 67; p<0 . 001 ) and 1 . 30 ( 95% CI , 1 . 21 to 1 . 40; p<0 . 001 ) , respectively . Although we were not able to adjust for several potential confounders , our findings suggest that the left colon may contribute to maintenance of glucose homeostasis . The gut acts as a key regulator of glucose homeostasis . It influences appetite , gastric emptying , intestinal motility , digestion , and absorption of nutrients , as well as insulin secretion from the pancreas through hormonal secretion and nervous signaling ( Holst et al . , 2016 ) . Whereas the role of the small intestine in glucose homeostasis has been well elucidated ( Holst and Madsbad , 2016 ) , the role of the colon is less clear . The growing interest in the gut microbiota has led to an increased focus on colonic pathophysiology since the vast majority of gut microbes resides in the colon . It has been shown that patients with type 2 diabetes harbor an altered fecal microbiota ( MetaHIT consortium et al . , 2015; Karlsson et al . , 2013; Qin et al . , 2012 ) , but it is unknown whether it is causally involved in the pathogenesis of type 2 diabetes ( Allin et al . , 2015; IMI-DIRECT consortium et al . , 2018; MetaHIT Consortium et al . , 2016 ) . Whereas some mechanistic studies in mice suggest that obesity and its related phenotypes can be transmitted via the fecal microbiota ( Bäckhed et al . , 2004; Ridaura et al . , 2013 ) , elimination of the gut microbiota by antibiotics in humans , does not seem to have any short-term effect on glucose regulation ( Mikkelsen et al . , 2015; Reijnders et al . , 2016; Vrieze et al . , 2014 ) . In addition to hosting the gut microbiota , the colon also serves as an endocrine organ . Glucagon-like peptide 1 ( GLP-1 ) secreted by L-cells in the small intestine augments insulin secretion from the pancreatic β-cells in response to a meal ( Holst et al . , 2016 ) . L-cells are , however , not only present in the small intestine but also in the colon , where the cell density increases from the proximal colon to the rectum ( Eissele et al . , 1992; Gunawardene et al . , 2011; Jorsal et al . , 2018 ) . Accordingly , it has been shown that biologically active GLP-1 is present in colonic tissue in concentrations , which are similar to those in the small intestine ( Deacon et al . , 1995 ) , but the physiological role of colonic GLP-1 remains elusive . In addition to GLP-1 , colonic L cells also secrete the appetite-reducing hormone peptide YY ( PYY ) ( Svane et al . , 2016 ) . Patients who have undergone resection of the entire colon or parts of it , here denoted colectomy , may serve as an evident opportunity to study the role of the colon in health and disease ( Jensen et al . , 2015 ) . Glucose regulation in patients with colectomy has been examined in a few previous studies , the largest of which included 21 patients , and although results are not clear-cut , they suggest that colectomy may influence the entero-insulinar axis ( Besterman et al . , 1982; Nauck et al . , 1996; Palnaes Hansen et al . , 1997 ) . On one hand , based on studies showing that mice raised in the absence of microorganisms have less body fat and are less insulin resistant , removal of the human gut microbiota by colectomy may be expected to result in a reduced risk of type 2 diabetes ( Bäckhed et al . , 2004 ) . On the other hand , removal of GLP-1 and PYY producing L-cells by colectomy may be expected to lead to an increased risk of type 2 diabetes . In the present study , we tested the hypothesis that total and partial colectomy is associated with an altered risk of type 2 diabetes . For this purpose , we used the nationwide Danish National Patient Register to compare the frequency of clinically recorded type 2 diabetes in patients who had a colectomy with patients who had other types of surgeries not involving the gastrointestinal tract . We identified 8946 individuals with total colectomy and 90 , 594 individuals with partial colectomy or proctectomy in the Danish National Patient Register . As explained in detail in Materials and methods , we based our main analyses on those who were alive and who did not have a diagnosis of diabetes recorded during the first 1000 days after the surgery . Patients with records of total colectomy at different dates or more than one partial colectomy , patients younger than 30 years , patients who died within the first 1000 days of follow-up , patients diagnosed with diabetes before surgery or within the first 1000 days after surgery , and patients who had a follow-up time shorter than 1000 days were excluded ( Figure 1 ) . Thus , we included 46 , 279 patients who had undergone total colectomy , partial colectomy , or proctectomy . For comparison , we identified a total of 694 , 110 age , sex , and year of surgery matched non-colectomy patients who had undergone other types of surgery not involving the gastrointestinal tract ( Table 1 ) . Mean ( SD ) follow-up time was 7 . 1 ( 5 . 2 ) , 5 . 6 ( 4 . 6 ) , 6 . 3 ( 4 . 7 ) , 5 . 7 ( 4 . 8 ) , 6 . 4 ( 4 . 9 ) , 5 . 8 ( 4 . 8 ) years for patients with total colectomy , right hemicolectomy , resection of the transverse colon , left hemicolectomy , sigmoidectomy , and proctectomy , respectively . The outcome was routinely clinically diagnosed type 2 diabetes ( hereafter denoted ‘diabetes’ ) as recorded in the Danish National Patient Register . Figure 2 shows the time-corresponding cumulative hazards ( double Nelson-Aalen plot ) of diabetes for colectomy patients vs . non-colectomy patients . The slopes of the curves equalize the hazard ratios ( a slope of 1 . 00 corresponds to a HR of 1 . 00 ) . The cumulative hazard of diabetes was greater among patients who had undergone total colectomy compared with non-colectomy patients . Correspondingly , the HR of diabetes was 1 . 40 ( 95% confidence interval [CI] , 1 . 21 to 1 . 62; p<0 . 001 ) ( Figure 3 ) . The analysis of the risk of diabetes after partial colectomy revealed that the highest cumulative hazard was observed among patients who had undergone left hemicolectomy or sigmoidectomy . Accordingly , the HR was 1 . 41 ( 95% CI , 1 . 19 to 1 . 67; p<0 . 001 ) for patients who had undergone left hemicolectomy and 1 . 30 ( 95% CI , 1 . 21 to 1 . 40; p<0 . 001 ) for patients who had undergone sigmoidectomy . Right hemicolectomy was not statistically significantly associated with increased risk of diabetes ( HR 1 . 08 [95% CI , 0 . 99 to 1 . 19; p=0 . 10] ) . Resection of the transverse colon was infrequent , but showed the same estimate , although with wider confidence interval; HR 1 . 08 ( 95% CI , 0 . 76 to 1 . 54; p=0 . 66 ) , whereas proctectomy was not associated with increased risk of diabetes; HR 0 . 98 ( 95% CI , 0 . 91 to 1 . 07; p=0 . 71 ) . We stratified patients with colectomy into a group of colorectal cancer patients and a group with other colorectal diseases ( Supplementary file 1 lists the most common diagnoses for the patients with other colorectal diseases ) . In both groups of patients , total colectomy was associated with increased risk of diabetes ( Table 2 ) . The HR of diabetes after total colectomy associated with colorectal cancer was 1 . 61 ( 95% CI , 1 . 22 to 2 . 11 , p<0 . 001 ) , whereas the HR for total colectomy associated with other colorectal disease was 1 . 34 ( 95% CI , 1 . 13 to 1 . 59 , p<0 . 001 ) . Corresponding HRs were 1 . 24 ( 95% CI , 1 . 01 to 1 . 53 , p=0 . 04 ) and 1 . 88 ( 95% CI , 1 . 41 to 2 . 50 , p<0 . 001 ) for left hemicolectomy , and 1 . 11 ( 95% CI , 0 . 997 to 1 . 24 , p=0 . 06 ) and 1 . 47 ( 95% CI , 1 . 34 to 1 . 62 , p<0 . 001 ) for sigmoidectomy . If we , instead of following the patients from 1000 days after surgery , followed the patients from the date of surgery , we found a somewhat higher HR of diabetes after total colectomy: HR = 1 . 72 ( 95% CI , 1 . 56 to 1 . 89 ) , which , however , should be interpreted with caution due to non-proportional hazards . In contrast , HRs of diabetes after left hemicolectomy and sigmoidectomy remained similar to results from the main analyses ( Supplementary file 2 ) . When follow-up time began 500 days after surgery or 1500 days after surgery we observed similar HRs of diabetes as compared to the main analyses ( Supplementary file 2 ) . In the present nationwide study , we observed an increased risk of diabetes in colectomy patients . The risk was highest among individuals who had the left part of the colon removed , whereas resection of the rectum was not associated with any change in risk of diabetes . Moreover , we observed an increased risk of diabetes irrespective of whether the patients had co-occurring colorectal cancer or other underlying diseases . A major strength of the present work is its nationwide nature allowing inclusion of many thousands of patients who have had colectomies and a large number of non-colectomy patients leading to relatively narrow confidence intervals of the hazard ratios observed . Moreover , the data allowed by study design to control for known confounding by sex and age as well as the possible confounding factors implicit in the time period and factors influencing the decision-making to refer to and conduct surgery in general . However , various limitations of our study must be considered in the interpretation of the results as indicating a causal relation of colectomy to later diabetes risk . We did not have information on several other potential confounders , precluding adjustment for them . This implies the possibility that the reported estimates may be positively biased . Colorectal cancer is associated with adiposity ( Ning et al . , 2010 ) , which is also a major risk factor for diabetes , wherefore adiposity may confound our results . However , based on previously published data ( Berentzen et al . , 2011 ) , it can be calculated that if differences in BMI were to fully account for our observed HR of 1 . 4 of diabetes , a BMI difference of 3 units ( equaling a difference of 9 kg for individuals 175 cm tall ) would be necessary . Another important potential confounder that we did not account for is medication . However , the direction of medication effects on risk of diabetes is likely ambiguous . For example , whereas corticosteroids used for treatment of inflammatory bowel disease have well-documented side-effects resulting in impaired glucose regulation , biologicals such as tumor necrosis factor-α antagonists may alleviate inflammation and therefore potentially improve glucose regulation ( Antohe et al . , 2012; Parmentier-Decrucq et al . , 2009; Solomon et al . , 2011; Tam et al . , 2007 ) . As the effects of medication on glucose regulation are ambiguous , and since we observed an enhanced risk of diabetes among patients with colorectal cancer and other colorectal diseases , it seems unlikely that our findings are due to differences in medication . Also , confounding by indication , that is , the indication for the colectomy — in this case the underlying disease — may bias our results . Hence , the underlying disease may , either through its effect on the normal colonic function or via other pathways , directly increase the risk of diabetes . Likely , the fact that we achieved similar results when we started the follow-up time 1000 days after the surgery in our main analyses or 1500 days after surgery in the supplemental analyses may speak against such confounding . Last , differences in the life style , for example in physical activity , smoking , dietary habits , including alcohol and coffee drinking , of the colectomy versus the non-colectomy groups may have played a role as confounders . Another potential limitation of our study is that the outcome was based on data from the Danish National Patient Register , which includes data only on hospital inpatient and outpatient contacts . This implies that patients with undiagnosed diabetes and with diabetes diagnosed outside hospitals , but not recorded later in the hospitals , were missed , and that the observed incidence is an underestimation . On the other hand , it is also possible that the recorded diagnosis of diabetes is wrong . However , as we defined the non-colectomy group as individuals who had undergone various other types of surgery , we presume that the likelihoods of being referred to a hospital or an outpatient clinic , and thus the likelihoods of being registered with a clinical diagnosis of diabetes , are the same in the colectomy and non-colectomy patients . Further , the outcome diabetes based on the diagnosis code E14 ( unspecified diabetes ) may be subject to misclassification regarding the type of diabetes . The data available does not allow validation of the recorded diagnosis , but to avoid misclassification of type 2 diabetes as type 1 diabetes , we restricted the study population to individuals above 30 years , since type 1 diabetes is much more frequently diagnosed before than after that age . Also , type 1 diabetes is infrequent compared to type 2 diabetes , reducing the risk of significant misclassification in this regard ( Diabetesforeningen , 2018 Facts about diabetes in Denmark https://diabetes . dk/diabetesforeningen/in-english/facts-about-diabetes-in-denmark . aspx ) . If the observed association between colectomy and excess incidence of diabetes is not simply explained by chance , bias , or confounding , it may represent a causal effect of colectomy on the diabetes risk . Indeed , our prospective design , including a 1000 days delay after surgery , supports causality since the outcome occurred after the exposure and the risk remained equally elevated throughout the follow-up periods . The finding of increased risk following only colectomy including the left part of the colon , but not the remainder part or the rectum , suggests a specific causal effect of that part rather than a general effect of surgery on the colon , which would more likely reflect confounding . Obviously , experimental proof of a direct causal effect of colectomy on diabetes risk in humans is impossible , and epidemiological substitutes , such as instrumental variable analyses , for example by Mendelian randomization , may not be feasible . However , assuming causality may inspire investigations of the possible biological mechanisms that may have induced the association between colectomy and diabetes risk . Thus , it may be speculated that our findings are explained by removal of GLP-1 producing cells or alterations of the colonic gut microbiota composition and function . Recent studies suggest that modifications in GLP-1 release are directly involved in both development of diabetes ( Færch et al . , 2015 ) and diabetes remission after gastric bypass ( Holst and Madsbad , 2016 ) . However , the physiological role of GLP-1 secreted from the colon is unclear . Interestingly , the L-cell density increases from the proximal to the distal colon ( Eissele et al . , 1992; Gunawardene et al . , 2011 ) , and accordingly , we observed the highest risk among individuals who had the left part of the colon removed . The bacterial load also increases along the length of the colon ( Donaldson et al . , 2016 ) , and colectomies involving the left part of the colon therefore likely lead to removal of a larger part of the colonic microbiota than colectomies involving the right part of the colon . Notably however , it has been shown that ileocolonic resection leads to an altered colonization of the neo-terminal ileum ( GETAID et al . , 2016 ) suggesting that removal of the colonic gut microbiota may be somewhat compensated for by over-colonization of the remaining parts of the gastrointestinal tract . Such bacterial overgrowth may well change the microbiota-mediated conversion of primary bile acids to secondary bile acids , which through their hormone-like actions impact glucose metabolism ( Ridlon et al . , 2014 ) . In our study , proctectomy was , with quite narrow upper confidence limits , not associated with an increased risk of diabetes . This may be explained by the fact that feces , and thus bacteria and other stimuli of L-cells , is only present in the rectum during the short time periods preceding defecation , thus hindering continuous signaling between the lumen of the rectum and extra-intestinal tissues . Last , the removal of the entire colon or a part of it might influence the exchanges between the gut epithelium and ingested food and fluids , especially due to an altered bacterial degradation of otherwise indigestible fibers to short-chain fatty acids , which could in turn have an impact upon diabetes development ( Allin et al . , 2015 ) . Among the previous studies that have examined glucose regulation in colectomy patients ( Besterman et al . , 1982; Nauck et al . , 1996; Palnaes Hansen et al . , 1997 ) , the largest included 10 patients with an ileo-anal reservoir and 11 patients with an ileostomy who underwent an oral glucose tolerance test ( Palnaes Hansen et al . , 1997 ) . Compared with 10 healthy individuals , colectomy individuals had higher peak levels of plasma glucose and higher peak levels and area under the curve of plasma insulin suggesting impaired glucose tolerance and hyperinsulinemia . However , the healthy controls had lower body weight compared with the patients , and no major differences in plasma GLP-1 levels were observed ( Palnaes Hansen et al . , 1997 ) . The prior evidence from these studies for an association between colectomy and impaired glucose regulation is thus conflicting . In a previous study also based on the Danish National Patient Register , we found no corresponding difference in cardiovascular disease risk between individuals who had undergone a total colectomy and individuals who had undergone other types of surgeries leaving the gastrointestinal tract intact ( Jensen et al . , 2015 ) . While we would expect that an increased risk of diabetes would associate with subsequent increased risk of cardiovascular diseases , this discrepancy may be explained by the limited follow-up time , not allowing presence of diabetes yet to be manifested in increased cardiovascular risk . In conclusion , we observed an increased risk of clinically recorded type 2 diabetes among patients who had undergone total and partial colectomy with the risk being elevated only among individuals who had the left part of the colon removed . The increased risk of diabetes was observed in patients with colorectal cancer as well as in patients with other colorectal diseases . As our findings are based solely on register data , clinical-physiological studies are warranted to establish whether colectomy per se is associated with metabolic alterations indicating an increased risk of type 2 diabetes , and to understand the role of the colon in glucose homeostasis . Such studies could include comparisons of plasma glucose , insulin , and incretin hormone levels during a meal tolerance test before and after different types of colectomy with careful control for likely confounding factors such as those discussed above . The study is based on data from the Danish National Patient Register , which covers hospital inpatient and outpatient contacts for Danish patients from 1994 and onwards ( Lynge et al . , 2011 ) . Based on the Nordic Classification of Surgical Procedures ( NCSP , 1996 – 2015 ) and the Danish Surgical Procedure and Treatment Classification version 3 ( DOTC , 1994 – 1995 ) , we identified patients who had undergone total colectomy ( with and without resection of rectum ) , partial colectomy , or resection of the rectum ( proctectomy ) . We grouped partial colectomies into right hemicolectomy , resection of the transverse colon , left hemicolectomy , and sigmoidectomy ( Supplementary file 3 ) . We considered patients who had total colectomy ( without resection of rectum ) followed by resection of rectum as having a proctocolectomy at the date of rectum resection . For patients who had partial and later total colectomy , we disregarded the preceding partial removals . The selection of non-colectomy patients from the register was carried out among patients who had undergone other types of surgery not involving the gastrointestinal tract . Thus , the non-colectomy patients were selected from three groups of patients , and we selected five non-colectomy patients per colectomy patient from each group: ( 1 ) patients who had undergone orthopedic surgery , ( 2 ) patients who had undergone abdominal surgery leaving the gastrointestinal tract intact , and ( 3 ) patients who had undergone other surgery , unrelated to the gastrointestinal tract ( Supplementary file 4 ) . Non-colectomy patients were matched to colectomy patients by sex , year of birth ( ±1 year ) , and date of surgery ( ±1 year ) . If there were more than five patients available within each of the three surgical groups , we randomly selected five patients . The same series of matching patients may have been used for selection or non-colectomy patients for more than one colectomy patient implying that the individual non-colectomy patient may have been selected more than once ( random sampling with replacement ) . The outcome , clinically recorded type 2 diabetes , was defined as the International Classification of Diseases , 10th Edition ( ICD-10 ) codes E11 and E14 . Dates of deaths were obtained from the Danish Civil Registration System . Individuals who had a diagnosis of diabetes recorded before surgery were excluded . The project has been reported to the Danish Data Protection Agency ( ref 2015-54-0939 ) and the data extraction has been approved and delivered by Statens Serum Institut ( ref FSEID-00001627 ) . Informed consent and assessment of the proposal in scientific ethical committees are not required for registry-based research in Denmark . We analyzed the data using the ‘surv package’ in R version 3 . 2 . 2 , and two-sided p-values<0 . 05 were considered statistically significant . The cumulative hazard of diabetes was estimated using the Nelson-Aalen estimator . Cox proportional hazards regression models were used to estimate hazard ratios ( HRs ) of diabetes . A linear relationship between the time-corresponding Nelson-Aalen estimators indicates fulfillment of the assumption of proportional hazards . We based our main analyses on those who were alive and who did not have a diagnosis of diabetes recorded during the first 1000 days after the surgery , for two reasons . First , as expected , colectomy patients had an increased hazard rate of death after surgery , which may reflect postsurgical morbidity possibly influencing diabetes risk for other reasons than effects of colectomy as such; the mortality among patients with total colectomy decreased during this period and stabilized from about 1000 days after surgery ( Supplementary file 5 ) . Second , diabetes may be associated with the diseases leading to colectomy , and may still be undiagnosed or not recorded at the time of surgery , but possibly so during the first years following surgery . The follow-up time for each participant ended at occurrence of diabetes , death , or at the end of the study , December 18th 2015 , whichever came first , and the underlying time scale was time from 1000 days after surgery . None of the participants were lost during follow-up . In supplementary analyses , we examined the risk of diabetes when follow-up time was started at the date of surgery , 500 days after surgery , and 1500 days after surgery . To optimize the efficiency and minimize bias ( Sjölander and Greenland , 2013 ) of the comparisons of the colectomy and non-colectomy patients , the Cox regression models included adjustment for the matching variables , sex , age and year of surgery . Separate models were fitted for total colectomy and for each type of partial colectomy and proctectomy and their respective non-colectomy patients . To elucidate whether the underlying disease , for which the colectomy was performed , influenced the subsequent risk of diabetes , the study population was stratified into two groups: a group including patients with co-occurrence of colorectal cancer ( ICD-10 codes C18-C20 ) at the date of colectomy and a group including the remaining patients ( Supplementary file 1 ) .	The large intestine helps people process the food that has not been digested and absorbed in the small intestine . It houses billions of bacteria that break down food , especially fibers . It also produces hormones that may help control blood sugar after eating . People with type 2 diabetes , who have difficulty controlling their blood sugar after meals , have a different mix of gut bacteria than people without the disease . But scientists do not know if this unusual mix of bacteria is a result of having diabetes or its treatment , or if it contributes to the condition . They also do not know if hormones released by the large intestine play a role in the disease . One way to study the role of the large intestine in diabetes is to look at patients who have had all or part of it removed . A few small studies with 21 patients or fewer have looked at how well patients are able to control their blood sugar after removal of all or part of their large intestine . But the results were confusing . Now , Jensen , Sørensen et al . show that patients who have surgery to remove all or the left part of their large intestine are more likely to develop diabetes . They looked at patient records from a national registry in Denmark to see how many patients developed diabetes during up to 18 years after surgery . In the analysis , they compared post-surgical diabetes diagnoses in 3 , 793 people who had their whole large intestine removed , 42 , 486 who had part of it removed , and 694 , 110 people who had surgery on another part of the body . People who had their rectum removed did not have a greater risk of developing diabetes than people having other surgeries . But the risk was higher among those who had all or the left part of the large intestine taken out . This suggests that the left part of the large intestine helps people control their blood sugar levels . More studies are needed to confirm that the large intestine plays a role in diabetes and to identify the underlying mechanisms . Understanding how the large intestine helps control blood sugar may help scientists develop new ways of treating or preventing type 2 diabetes .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "medicine" ]	2018	Increase in clinically recorded type 2 diabetes after colectomy
RirA is a global regulator of iron homeostasis in Rhizobium and related α-proteobacteria . In its [4Fe-4S] cluster-bound form it represses iron uptake by binding to IRO Box sequences upstream of RirA-regulated genes . Under low iron and/or aerobic conditions , [4Fe-4S] RirA undergoes cluster conversion/degradation to apo-RirA , which can no longer bind IRO Box sequences . Here , we apply time-resolved mass spectrometry and electron paramagnetic resonance spectroscopy to determine how the RirA cluster senses iron and O2 . The data indicate that the key iron-sensing step is the O2-independent , reversible dissociation of Fe2+ from [4Fe-4S]2+ to form [3Fe-4S]0 . The dissociation constant for this process was determined as Kd = ~3 µM , which is consistent with the sensing of ‘free’ iron in the cytoplasm . O2-sensing occurs through enhanced cluster degradation under aerobic conditions , via O2-mediated oxidation of the [3Fe-4S]0 intermediate to form [3Fe-4S]1+ . This work provides a detailed mechanistic/functional view of an iron-responsive regulator . Iron–sulphur clusters are ubiquitous protein cofactors that play essential roles across all of life in processes as diverse as respiration , photosynthesis , and DNA replication ( Beinert et al . , 1997; Johnson et al . , 2005 ) , with recent evidence that they may be even more abundant than initially thought ( Rouault , 2015 ) . Elucidating the precise nature of their roles is a major challenge , which is often complicated by the extreme reactivity of the cluster to O2 and other gases . However , this very sensitivity has been exploited through the evolution of iron–sulphur cluster-containing transcriptional regulators that enable cells to sense and respond to , for example , oxidative stress and changes in concentrations of metabolically important species such as O2 and iron ( Beinert and Kiley , 1999; Crack et al . , 2014a ) . Iron is an essential micronutrient for nearly all of life , but is also potentially extremely toxic due to its ability to catalyse , via redox cycling , the formation of damaging reactive oxygen species . Consequently , for life to flourish , not only must sufficient iron be obtained from the environment , but its precise form must be carefully controlled ( Andrews et al . , 2003 ) . A central part of this control is exerted through the regulation of iron uptake into the cell in response to intracellular iron levels . In many bacteria , including such taxonomically diverse model organisms as Escherichia coli and Bacillus subtilis , iron uptake is under the control of the global iron regulator Fur ( Ferric uptake regulator ) . Iron is sensed through the availability and binding of Fe2+ directly to the Fur protein , resulting in a conformation that can bind cis-acting ‘Fur boxes’ close to the promoters of genes encoding proteins that function in the iron-uptake machinery , causing ( usually ) repression of transcription ( Lee and Helmann , 2007; Pohl et al . , 2003 ) . The direct binding of Fe2+ as a co-repressor also occurs in DtxR ( Diphtheria toxin Repressor ) from Corynebacterium diphtheriae and related species . Although DtxR has no significant sequence similarity to Fur , it shares significant structural features , in addition to Fe2+-binding ( D'Aquino et al . , 2005; Ding et al . , 1996 ) . Rhizobium and other closely related rhizobial genera that induce nitrogen-fixing nodules on legumes , as well as the pathogens Bartonella , Brucella , and Agrobacterium , contain a very different global iron regulator called RirA ( Rhizobial iron regulator A ) ( Todd et al . , 2002; Wexler et al . , 2003; Yeoman et al . , 2004; Todd et al . , 2005; Todd et al . , 2006; Rudolph et al . , 2006; Chao et al . , 2005; Ngok-Ngam et al . , 2009; Viguier et al . , 2005; Bhubhanil et al . , 2014; Crespo-Rivas et al . , 2019 ) , which has no structural or sequence similarity whatsoever to Fur or DtxR . Rather , RirA belongs to the Rrf2 super-family of transcriptional regulators ( Keon et al . , 1997 ) that includes IscR ( regulator of iron–sulphur cluster biosynthesis ) ( Rajagopalan et al . , 2013; Schwartz et al . , 2001 ) and NsrR ( regulator of nitrosative stress response ) ( Crack et al . , 2012; Crack et al . , 2015; Volbeda et al . , 2017 ) , both of which are homodimeric and bind an iron–sulphur cluster . In Rhizobium leguminosarum , the symbiont of peas , beans and clovers , RirA regulates many genes involved in iron homeostasis , by binding to cis-acting ‘IRO boxes’ ( Yeoman et al . , 2004 ) . Recently , it was shown that R . leguminosarum RirA binds a [4Fe-4S] cluster and that this form of the protein binds to the IRO box sequence ( Pellicer Martinez et al . , 2017 ) . Although the structure of RirA is not yet available , a model based on the structure of dimeric [4Fe-4S] NsrR , the first reported for any cluster-bound Rrf2 regulator ( Volbeda et al . , 2017 ) , is shown in Figure 1 . Exposure of [4Fe-4S] RirA to low iron conditions and/or O2 resulted in cluster conversion to generate a [2Fe-2S] form , which binds to the IRO box with lower affinity than the [4Fe-4S] form . The [2Fe-2S] form was also unstable under low iron conditions , resulting in apo-RirA , which did not bind the IRO box sequence ( Pellicer Martinez et al . , 2017 ) . To understand how RirA functions as an iron sensor and how it enables the cell to integrate iron and O2 signals requires that the mechanisms by which it responds to these two signals are elucidated . Traditional approaches to studies of iron–sulphur cluster proteins cannot readily resolve intermediates of cluster conversion/decay . ESI-MS utilizing solution and ionisation conditions under which proteins remain folded enables accurate mass detection of intact proteins and protein complexes , and has been used extensively to study protein-protein interactions , interactions of proteins with drugs , nucleic acids , sugars and lipids , as well as protein structural changes ( Liko et al . , 2016; Hopper and Robinson , 2014; Rodenburg et al . , 2017 ) . It has also been used to study protein-cofactor interactions where the cofactor remains bound following ionisation , and amongst these , ESI-MS of metalloregulators ( Glauninger et al . , 2018 ) and iron–sulphur cluster proteins ( Johnson et al . , 2000 ) and their reactivity ( Crack and Le Brun , 2019 ) have also been shown to be valuable . Time-resolved mass spectrometry has been exploited for studying protein metal ion binding ( Ngu and Stillman , 2006 ) and heme transfer ( Tiedemann et al . , 2012 ) , and for structural studies using H/D exchange ( Konermann et al . , 2011 ) . Recently , time-resolved ESI-MS provided detailed mechanistic information of the O2-sensing reaction of the [4Fe-4S] cluster binding FNR regulator ( Crack et al . , 2017 ) . Here , we employed time-resolved ESI-MS under non-denaturing conditions , along with EPR spectroscopy , to elucidate the exact series of molecular events involved in the conversion of the RirA cluster in response to key environmental signals . The data reveal , in remarkable detail , the nature of intermediates formed and differences in cluster conversion under aerobic and anaerobic conditions . Importantly , we find that the key sensing step , the initiation of cluster conversion/degradation , is dependent on Fe2+ but is independent of O2 , and that the Kd for the fourth iron binding to the cluster is in the low micromolar range , supporting a primary role for RirA as an iron sensor in vivo . Nevertheless , O2 influences the overall rate of cluster decay through the oxidation of a key cluster intermediate and of cluster-derived sulphide . Together , the data provide a comprehensive mechanistic and functional view of an important iron-responsive regulator . Regular electrospray ionisation mass spectrometry ( ESI-MS ) in water/acetonitrile acidified with formic acid results in denatured proteins and the loss of non-covalently attached cofactors . However , the use of a volatile buffer at physiological pH is known to preserve the protein’s folded state and , with it , any bound cofactors ( Crack et al . , 2017 ) . Therefore , using ESI-MS under such non-denaturing conditions , we set out to observe in real time the decay of the [4Fe-4S] RirA cluster in response to low iron conditions and to determine the identity of any intermediary cluster conversion breakdown products . The feasibility of this approach was indicated by the previously reported ESI-MS of [4Fe-4S] RirA ( Pellicer Martinez et al . , 2017 ) , in which both dimeric and monomeric forms were observed . Ionisation of RirA in the MS experiment results in partial dissociation of the RirA dimer into monomers . Initially , conditions for the formation of the monomeric form were optimised because this enabled the unambiguous assignment of cluster species . Analysis of the RirA dimer follows in a later section . The initial spectrum , Figure 2A ( black line ) , in which the protein was held under non-denaturing anaerobic , iron-replete conditions , was very similar to that previously described for cluster-bound RirA ( Pellicer Martinez et al . , 2017 ) ; the major peak at 17 , 792 Da corresponds to the [4Fe-4S] form ( see Supplementary file 1 for predicted and observed masses ) , but a range of lower intensity cluster breakdown species was also present , which are most likely due to cluster damage sustained during the exchange of the protein into the volatile buffer necessary for ESI-MS studies . Indeed , these were observed to decay away while the [4Fe-4S] form remained stable ( Figure 2—figure supplement 1 ) , consistent with previous studies that demonstrated that the [4Fe-4S] form of RirA is entirely stable under anaerobic , iron-replete conditions ( Pellicer Martinez et al . , 2017 ) . We previously showed that various Fe2+ chelators , including EDTA and Chelex 100 , can simulate low iron conditions by efficiently competing for Fe2+ . The lack of dependence on the concentration or type of chelator used indicated that this competition occurs through two equilibria corresponding to the loss of iron from the cluster and subsequent binding by the chelator . Consistent with this , under low iron conditions generated by the anaerobic addition of 250 µM EDTA , major changes in the ESI-MS spectrum were observed after 30 min , Figure 2A . ESI-MS changes were then measured over this time period ( t = 0 is the spectrum prior to the addition of chelator , t = 30 is 30 min post addition ) . While the [4Fe-4S] peak was observed to decay away , peaks corresponding to protein-bound cluster fragments ( 17 , 586–17 , 762 Da ) initially increased in intensity . These included [4Fe-3S] , [3Fe-4S] , [3Fe-3S] , [3Fe-2S] , [3Fe-S] , [2Fe-2S] and [2Fe-S] forms ( see Figure 2B and Supplementary file 1 ) . These species then subsequently decayed away ( Figure 2C ) , with a peak at 17442 Da , due to apo-RirA , becoming the major feature of the spectrum . Peaks at +16 , 32 and 64 Da , due to oxygen or sulphur adducts of apo-RirA , were also observed , though at relatively low intensity . The full time course is shown as a 3D plot in Figure 2D . To assist with unambiguous assignment of cluster-bound forms of RirA , [4Fe-4S] RirA with cluster sulphide specifically labelled with 34S was generated via in vitro cluster synthesis ( cluster reconstitution ) using 34S-L-cysteine ( Crack et al . , 2019 ) . The major peak in the deconvoluted ESI-MS spectrum was at 17 , 800 Da , shifted +8 Da relative to the natural abundance ( predominantly 32S ) form of [4Fe-4S] RirA , Figure 3A and Supplementary file 1 . Exposure of the 34S form to EDTA under anaerobic conditions resulted in a set of peaks , due to intermediate species , observed at masses shifted relative to those of the 32S sample , Figure 3B–G . The peaks initially assigned to [3Fe-4S] , [4Fe-3S] , [3Fe-3S] , [3Fe-2S] , [3Fe-S] , [2Fe-2S] and [2Fe-S] species were shifted by +2 Da for each sulphide , confirming the identification of these intermediate/product forms . As expected , the apo-protein peak was not shifted ( Figure 3H ) . The apo-RirA adduct at +16 and +32 Da were also not shifted , suggesting that these are due , respectively , to one and two O adducts , while the peak at +64 Da was shifted by + 4 Da , consistent with a double sulphur adduct derived from cluster sulphide ( Figure 3—figure supplement 1 ) . The latter arises as a result of oxidation of cluster sulphide , generating S0 , which can be incorporated into Cys thiol side chains . The origin of oxygen adducts is less clear . Recent ESI-MS studies of the anaerobic nitrosylation of NsrR also revealed O adduct species ( Crack and Le Brun , 2019 ) , and we note that under certain conditions aqueous samples may generate oxygen in the ESI source ( Banerjee and Mazumdar , 2012 ) . It is well established that non-denaturing ESI-MS can be used to follow chemical processes in solution ( e . g . ligand binding to a protein or isotope exchange over time ) , yielding quantitative thermodynamic and kinetic information ( Hopper and Robinson , 2014; Ngu and Stillman , 2006; Crack et al . , 2017; Pacholarz et al . , 2012; Schermann et al . , 2005 ) . To determine the sequence of events in RirA [4Fe-4S] cluster degradation , abundances of the different cluster fragment species in Figure 3 were analysed as a function of time . Figure 4A shows plots of relative intensities due to [4Fe-4S] and [3Fe-4S] species as a function of time following the addition of EDTA , while Figure 4B–E show equivalent plots for [3Fe-3S] , [3Fe-2S] , [2Fe-2S] and apo-RirA species . The data show that [4Fe-3S] and [3Fe-4S] were the first intermediates to reach their maximum abundance , consistent with their formation early in the conversion process . The [3Fe-3S] species was the next intermediate to maximise , followed by [3Fe-2S] species . The last of the intermediates to maximise in abundance and the last to decay away was [2Fe-2S] RirA . This was previously shown to be a stable intermediate following the [4Fe-4S] to [2Fe-2S] conversion promoted by EDTA ( Pellicer Martinez et al . , 2017 ) . However , the buffer conditions used in the present study , and which are required for the ESI-MS experiment , appear to reduce the stability of the [2Fe-2S] form leading to observation of apo-protein ( Figure 4E ) , as recently observed following conversion of the O2 sensor [4Fe-4S] FNR to its [2Fe-2S] form ( Crack et al . , 2017 ) . The observed sequence of cluster intermediates provided an outline for the mechanism of cluster conversion/degradation , which was used as the basis for global analysis of multiple ( n = 4 ) mass spectrometric kinetic data sets . Iterative optimisation of the mechanism/global fit resulted in the reaction scheme shown in Figure 5 , with corresponding fits of the peak intensities due to the formation and/or decay of [4Fe-4S] , [4Fe-3S] , [3Fe-4S] , [3Fe-3S] , [3Fe-2S] and [2Fe-2S] RirA species ( see the solid lines in the plots of Figure 4 ) . Rate constants obtained from the fit to the reaction scheme are given in Table 1 . The mechanistic scheme indicates that loss of a single iron or sulphide ion appears to be an obligatory first step in the cluster conversion process . However , the rate constant describing the loss of a sulphide to generate [4Fe-3S] is extremely low , indicating that this reaction does not occur as a significant part of the ( chemical ) conversion mechanism and that the observation of a [4Fe-3S] species ( which is present at time zero following addition of EDTA ) results from damage to the cluster during ionisation . It is important to note that the extent of cluster damage is low and this does not affect the mechanistic picture of the cluster conversion/degradation reaction that emerges . Thus , loss of an initial iron , [4Fe-4S] to [3Fe-4S] , is the heavily favoured route for the initiation of cluster conversion . Another important feature of the global fit is that the first step , loss of Fe2+ from [4Fe-4S]2+ , is reversible . The overall rate of cluster conversion observed here is significantly higher than that previously reported from absorbance kinetic experiments ( Pellicer Martinez et al . , 2017 ) . To confirm that ESI-MS and absorbance spectroscopy report on the same process , cluster degradation was followed by monitoring A382 nm ( corresponding to the maximum absorbance of the [4Fe-4S] cluster ) as a function of time for a [4Fe-4S] RirA sample in volatile ammonium acetate buffer ( necessary for ESI-MS measurements ) rather than the HEPES buffer previously used ( Pellicer Martinez et al . , 2017 ) , see Figure 4F . The decay fitted well to a bi-exponential function , where the initial , more rapid phase must correspond to the loss of the [4Fe-4S] cluster ( in which it converts to a species that cannot be identified by its absorbance alone ) , followed by a slower phase that corresponds to further decay of the cluster . The rate constant for the initial phase , k = 0 . 34 min−1 , is in excellent agreement with the rate constant ( k = 0 . 30 min−1 ) for the conversion of [4Fe-4S] to [3Fe-4S] , derived from the ESI-MS kinetic data ( Table 1 ) . Though [4Fe-4S] and [3Fe-4S] clusters often have similar absorbance extinction coefficients , loss of Fe2+ from the RirA [4Fe-4S]2+ cluster would be expected to result in some change in the UV-visible absorbance spectrum , as observed here and for a similar process in aconitase ( Emptage et al . , 1983 ) . Thus , the enhanced rate of conversion reported here is not a consequence of the ESI-MS method , but rather reflects the different buffer conditions used here compared to those of the previous study ( Pellicer Martinez et al . , 2017 ) . The RirA cluster is clearly more labile in the ammonium acetate buffer , most likely because acetate is a weak iron chelator . The ESI-MS data show that the [3Fe-3S] RirA species was formed from [3Fe-4S] and [4Fe-3S] clusters with similar observed rate constants ( k = 0 . 090 and 0 . 087 min−1 , respectively ) . The temporal behaviour of the [3Fe-3S] intermediate , where it was observed to maximise at ~4 min before decaying away , is consistent with it being an intermediate in the [4Fe-4S] to [2Fe-2S] cluster conversion pathway . A novel [3Fe-3S] intermediate was recently observed by mass spectrometry during the O2-dependent [4Fe-4S] to [2Fe-2S] cluster conversion reaction of FNR ( Crack et al . , 2017 ) . The observation of a similar intermediate here suggests that it may be a more widespread feature of [4Fe-4S] cluster conversions . While the structure of this particular [3Fe-3S] intermediate was not established , that of a small molecule [3Fe-3S]3+ cluster was recently published , revealing a planar hexagonal arrangement of alternating iron and sulphide ions ( Lee et al . , 2016 ) . If the [3Fe-3S] intermediate of RirA has a similar planar structure ( Figure 5 ) , it would reveal a key feature of the [4Fe-4S] to [2Fe-2S] cluster conversion , in which the geometry of the coordinating ligands changes from tetrahedral ( for the [4Fe-4S] cluster ) to planar ( for the [2Fe-2S] cluster ) . [3Fe-3S] RirA degraded predominantly to form [2Fe-2S] ( k = 0 . 5 min−1 ) , before finally generating apo-RirA ( k = 0 . 07 min−1 ) . The mechanistic scheme indicates that [3Fe-3S] may also decay to [3Fe-2S] and then on to apo-RirA ( in part via [2Fe-2S] ) . It is recognised that additional intermediate species are likely to be involved in the degradation of [3Fe-2S]/[2Fe-2S] to apo-RirA . While a [2Fe-S] form was observed in mass spectra , its temporal behaviour could not be sensibly modelled , suggesting that it might , at least in part , arise from spontaneous re-assembly of cluster fragments as iron/sulphide ions are released during cluster conversion/degradation . Previous studies of [4Fe-4S] RirA showed that the cluster is sensitive to O2 , undergoing conversion in an apparently similar manner to that observed under low iron conditions ( Pellicer Martinez et al . , 2017 ) . To gain a much more detailed view of this aspect of the conversion process , [4Fe-4S] RirA was investigated using non-denaturing ESI-MS under low iron conditions in the presence of O2 ( 228 µM ) . Cluster conversion/breakdown species were observed to form and decay as a function of time , Figure 6 and Table 1 . Plots of the individual cluster intermediate species as a function of time , Figure 7 , revealed marked differences to those in the absence of O2 . We note that the presence of O adducts in experiments carried out under anaerobic conditions suggested that some O2 might be generated during the ESI-MS experiment; however , the differences observed between the anaerobic and aerobic experiments suggest that this is relatively minor . Indeed , the formation of several intermediates , including the [3Fe-3S] and [2Fe-2S] forms , occurred significantly more rapidly under aerobic conditions , consistent with previous observations that a combination of low iron and O2 enhanced the rate of cluster conversion/degradation ( Pellicer Martinez et al . , 2017 ) . Interestingly , the formation of sulphur adducts of apo-RirA , which were confirmed to be derived from cluster sulphide by 34S-dependent mass shifts ( Figure 6—figure supplement 1 ) , occurred to a much greater extent in the presence of O2 , consistent with O2 acting as oxidant for S2- ions released from the cluster ( Figure 6 ) . Low abundance peaks due to oxygen adducts ( the first at apo-RirA +16 Da ) were also observed ( Figure 6—figure supplement 1 ) . The temporal behaviour of the double sulphur adduct of apo-RirA is shown in Figure 7E . A global analysis of the aerobic kinetic ESI-MS data was achieved on the basis of the same mechanistic scheme employed for the analysis of anaerobic data ( Figure 5 ) - see the solid lines in the kinetic plots of Figure 7 . The rate constants obtained from the fit are consistent with the enhanced rates of several steps in the conversion mechanism in the presence of O2 ( Table 1 ) . As a control , absorbance spectroscopy was again used to monitor the decay of A382 nm as a function of time for an aerobic [4Fe-4S] RirA sample in ammonium acetate buffer ( Figure 7F ) . In contrast to the equivalent anaerobic experiment ( Figure 4F ) , the data were fitted well by a single exponential function with a rate constant , k = 0 . 34 min−1 , again in excellent agreement with that obtained for the initial , [4Fe-4S] to [3Fe-4S] step of the cluster conversion reaction ( Table 1 ) . Previous studies of RirA revealed that a paramagnetic species was formed in RirA samples under oxidative conditions ( Pellicer Martinez et al . , 2017 ) . Thus , EPR spectroscopy was employed here to monitor the formation of paramagnetic intermediates as a function of time following exposure to low iron conditions in the presence of O2 . To facilitate direct comparison of kinetic data , samples were prepared at the same concentration , in the same buffer and at the same temperature as used for the mass spectrometric studies . An S = ½ signal at g = 2 . 01 , similar to that previously reported for RirA ( Pellicer Martinez et al . , 2017 ) , was observed , which was assigned to the [3Fe-4S]1+ cluster intermediate ( Figure 8A ) . However , the detection of a [3Fe-3S] intermediate also raised the possibility that this species might be responsible for the EPR signal , as recently proposed for cluster conversion in [4Fe-4S] FNR ( Crack et al . , 2017 ) . So , to investigate the EPR-active species in more detail , the saturation properties of the g = 2 . 01 EPR signal were studied as a function of microwave power and temperature ( Figure 8—figure supplement 1 ) . Signal intensities as a function of temperature at six microwave power values were plotted ( Figure 8B and Figure 8—figure supplement 2 ) . The inclusion of a 1/T hyperbola ( Figure 8B ) reveals the narrow temperature interval in which the EPR signal follows the Curie law ( based on non-saturated and non-broadened lines originating from an undisturbed Boltzmann distribution of energy levels populations ) for each microwave power . At a very low microwave power , such as 3 . 2 µW , the EPR signal perfectly obeyed the Curie law in the explored temperature range of 4–15 K . As the power was increased , the temperature dependence in the 4–9 K region became lower , indicating power saturation at these temperatures for all powers ≥ 100 µW . At temperatures above ~20 K , all dependences deviated from the Curie hyperbola , which is an effect of line broadening . No microwave power was able to produce any detectable EPR absorbance at temperatures above 50 K . This is a particular characteristic of [3Fe-4S]1+ clusters; they exhibit the most sensitive dependence on increasing temperature , with signal disappearing at a lower temperature than for any other cluster type ( Svistunenko et al . , 2006 ) . Such microwave power saturation behaviour , as well as the g-value and EPR line shape are highly characteristic of well-characterised [3Fe-4S]1+ clusters ( Beinert and Thomson , 1983; Cammack , 1992 ) and so we conclude that , in the case of RirA , the EPR active species is the [3Fe-4S]1+ cluster . It is perhaps surprising that none of the other intermediates identified by ESI-MS are detected by EPR; it appears that such species are diamagnetic . To determine the kinetic properties of the RirA [3Fe-4S]1+ intermediate , EPR measurements were performed for samples prepared at a range of time points following exposure to aerobic low iron conditions , Figure 8A . Quantification of the g = 2 . 01 signal revealed that ~ 1 µM ( ~4% ) of the cluster was present in the [3Fe-4S]1+ form at time zero , prior to the introduction of chelator and O2 . The signal intensity increased with time up to ~4 . 5 µM ( ~18% of the original cluster concentration ) by 3 min before decaying away over the next 15 min ( Figure 8A and C ) . The [3Fe-4S]1+ cluster contains three Fe3+ ions , meaning that it is formed from a [4Fe-4S]2+ cluster through loss of a Fe2+ ion and oxidation of the remaining Fe2+ to Fe3+ . EPR measurements were also carried out for [4Fe-4S] RirA samples under identical conditions except that O2 was omitted ( Figure 8D ) . A signal similar to that observed in the aerobic experiment was detected , but at a significantly lower concentration . Beginning at ~4% of original cluster concentration , the signal increased only to ~6 . 5% after 5 min , then remained at that level up to ~15 min before returning to its pre-EDTA exposure concentration ( Figure 8C and D ) . We note that the [3Fe-4S]1+ signal does not entirely decay away under either aerobic or anaerobic conditions . One possibility is that this low intensity residual signal represents an ‘off pathway’ form of the cluster that is less reactive . While the presence of O2 enhances several of the steps of cluster conversion/degradation , importantly , this is not the case for the initial reversible step in which the cluster loses a Fe2+ ion to form the [3Fe-4S] intermediate: rate constants for the initial step ( principally loss of iron ) are similar under aerobic and anaerobic conditions . However , the oxidation state of the [3Fe-4S] cluster is affected by the presence of O2 . A comparison of the ESI-MS data for [3Fe-4S] from Figure 7B and the EPR intensity data for [3Fe-4S]1+ is shown in Figure 8C . The data are generally in good agreement for the aerobic experiment . There is a slight lag in the appearance of the EPR signal relative to the ESI-MS [3Fe-4S] peak , which could arise from the initial formation of the EPR silent [3Fe-4S]0 intermediate , followed by its oxidation to the [3Fe-4S]1+ form in the presence of O2 . In the anaerobic experiment , the observed intensity of the [3Fe-4S]1+ cluster is much lower , consistent with the major form being the EPR silent [3Fe-4S]0 form . In combination , the ESI-MS and EPR data show that a [3Fe-4S] cluster intermediate is formed in both the absence and presence of O2 ( at similar abundance as judged by ESI-MS ) , but it is predominantly present in different oxidation states , being largely in the reduced [3Fe-4S]0 form under anaerobic conditions and in the oxidised [3Fe-4S]1+ form under aerobic conditions . The insensitivity of the initial step to O2 revealed here by time-resolved ESI-MS is consistent with the previous observation that the rate of the [4Fe-4S] to [2Fe-2S] cluster conversion reaction is O2-independent , leading to the conclusion that the rate-limiting step of the reaction does not involve O2 ( Pellicer Martinez et al . , 2017 ) . While the overall rates of reaction in the ammonium acetate buffer used here for ESI-MS and EPR experiments are greater than those previously reported , the global analyses indicate that , when O2 is present , the initial reaction ( corresponding to the loss of a single iron ( as Fe2+ ) from the [4Fe-4S]2+ cluster ) is , or is close to , the rate-limiting step ( Table 1 ) . This accounts for why the absorbance decay at 382 nm follows a single exponential in the presence of O2 but a bi-exponential in the absence of O2: in the former condition , the initial and subsequent reaction ( [3Fe-4S] to [3Fe-3S] ) occur at similar rates , while under anaerobic conditions , the second step is significantly slower . The independence from O2 of the initial step is consistent with RirA functioning principally as an iron sensor . In considering the question of why O2 should significantly affect the [3Fe-4S] to [3Fe-3S] conversion , it is likely that the [3Fe-4S]1+ species resulting from O2-mediated oxidation of the initially formed [3Fe-4S]0 cluster is more labile , and differences in oxidation state of intermediate cluster irons could be an important stability factor throughout . Another factor that is likely to be of at least equal importance is the enhanced oxidation of cluster sulphide in the presence of O2 , which would be expected to accelerate the degradation of the [3Fe-4S] , [3Fe-3S] and [2Fe-2S] forms towards apo-RirA . A comparison of the rate constants in Table 1 suggests that both sulphide and iron oxidation processes are important for the overall enhanced rate of degradation towards apo-RirA in the presence of O2 . The data presented above point to the equilibrium [4Fe-4S]2+ ↔ [3Fe-4S]0 + Fe2+ , as being key for iron sensing . If this is the case , then the binding affinity of the fourth iron should reflect the likely ‘free’ iron concentration of the cell , to which it would need to respond . Levels of ‘free’ or chelatable iron have not been reported for R . leguminosarum , but have been measured for Escherichia coli , in which it was found to be ~10 µM ( Keyer and Imlay , 1996 ) , and there are no a priori reasons why those in other proteobacteria should be very different . Mag-fura-2 , which binds Fe2+ to form a 1:1 complex with a Kd of 2 . 05 µM , has been used previously to probe Fe2+-binding to proteins ( Rodrigues et al . , 2015 ) . A titration of [4Fe-4S] RirA with an increasing concentration of mag-fura-2 was carried out . Importantly , the concentration of Fe2+-mag-fura-2 formed at the end of the titration corresponded to ~1 iron per [4Fe-4S] cluster ( Figure 9 ) , confirming that the process being measured corresponded to the [4Fe-4S]2+ ↔ [3Fe-4S]0 + Fe2+ equilibrium , and that no further cluster ( conversion/degradation ) reaction occurred on the timescale of the measurement . Resulting spectra were corrected for contributions from the underlying FeS cluster absorbance and normalised , as described in Materials and methods , see Figure 9 . The data describe the increasing proportion of Fe2+ that is bound by the ligand and are fitted by a simple binding isotherm , yielding an apparent Kd that reflects the competition between mag-fura-2 and [3Fe-4S] RirA for Fe2+ , from which the dissociation constant for RirA was readily determined as Kd = 3 ± 0 . 2 µM . This value is entirely consistent with the [4Fe-4S]2+ ↔ [3Fe-4S]0 + Fe2+ equilibrium functioning as the iron-sensing reaction . Conditions for the detection of the RirA dimer by ESI-MS were optimised ( Pellicer Martinez et al . , 2017 ) and time-resolved ESI-MS was used to follow cluster degradation in the RirA dimer under anaerobic conditions , see Figure 10A . The major species prior to addition of iron chelator were the [4Fe-4S]/[4Fe-4S] and [3Fe-4S]/[4Fe-4S] forms . In the dimer , the presence of two clusters in many cases precludes unambiguous identification of the intermediates because the distribution of iron and sulphur across the two clusters is unknown . Furthermore , there are many more possible intermediates of the dimer as each of the two clusters undergoes conversion/degradation . Nevertheless , the temporal behaviour of tentatively assigned species could be fitted to a basic model of cluster conversion in dimeric RirA , see Figure 10—figure supplement 1 . Plots of relative abundance of [4Fe-4S]/[4Fe-4S] , [3Fe-4S]/[3Fe-4S] , and [2Fe-2S]/[2Fe-2S] forms and fits to the global model are shown in Figure 10B , with rate constants in Supplementary file 2 . Additional plots of [3Fe-4S]/[4Fe-4S] , [2Fe-2S]/[3Fe-4S] , [apo]/[2Fe-2S] and [apo]/[apo] are shown in Figure 10—figure supplement 2 . A plot comparing the time dependence of the EPR signal intensity due to [3Fe-4S]1+ forms under aerobic and anaerobic conditions with the ESI-MS abundance of the [3Fe-4S]/[3Fe-4S] species is shown in Figure 10—figure supplement 3 . As in the case of the monomer , the aerobic ESI-MS and EPR data are in good agreement , consistent with the susceptibility of the [3Fe-4S]0 cluster to oxidation . We note that [3Fe-3S]/[3Fe-4S] and [3Fe-3S]/[3Fe-3S] cluster intermediates were not clearly observed; signals at the predicted masses for these species were observed , but only at very low intensities , indicating that these forms are reactive intermediates during the dimer cluster conversion process and therefore do not accumulate . Importantly , the rate constant derived from the global fit for the initial reversible loss of Fe2+ is very similar to that obtained from analysis of the monomer state ( 0 . 31 compared to 0 . 30 min−1 , Table 1 and Supplementary file 2 ) . Finally , the extent of sulphur adduct formation was greater than observed in the monomer region under equivalent conditions . Thus , within the limits of what we are able to observe , the RirA dimer behaved similarly to the monomer , consistent with monomer species forming from dimers during the measurement . The global iron regulator RirA differs in many respects from the well characterised bacterial iron regulator Fur , and so a distinct model is needed to account for how it senses and responds to the intracellular iron status . Data from ESI-MS under non-denaturing conditions , including 34S isotope exchange , along with EPR spectroscopy , have provided a highly detailed view of the [4Fe-4S] RirA cluster degradation reactions in response to low iron and O2 , another key environmental signal . In particular , the work further illustrates the remarkable potential of ESI-MS to provide time-resolved information on metallo-cofactor reactivity ( that involves a change in mass ) , which is not available from other techniques . As shown in Figure 5 , the mechanism involves a complex series of events initiated by the loss of iron , resulting in a [3Fe-4S] intermediate that decays further to [3Fe-3S] , [3Fe-2S] and [2Fe-2S] species along the pathway to apo-RirA . Importantly , ESI-MS and absorbance data showed that the iron ( Fe2+ ) dissociation step occurs at a similar rate under both aerobic and anaerobic conditions , consistent with RirA functioning principally as an iron sensor . Thus , the data presented here enabled us to elucidate at a molecular level the changes that occur in the transition from Fe-replete to Fe-depleted conditions . We propose a model in which the RirA [4Fe-4S]2+ cluster continually undergoes Fe2+ dissociation , to form a [3Fe-4S]0 form , and , under iron replete conditions , re-association to re-form [4Fe-4S]2+ . When cellular iron becomes scarce , competition for the dissociated Fe2+ increases , and re-association is disfavoured . In order to effectively sense cytoplasmic iron levels , the affinity ( dissociation constant ) of the fourth cluster iron must be in the range of normal free ( chelatable ) iron levels . Although unknown for R . leguminosarum , cytoplasmic free iron has been measured for E . coli and found to be ~10 µM ( Keyer and Imlay , 1996 ) . Assuming that free iron levels are not very different in R . leguminosarum , the Kd of ~3 µM determined here for the [3Fe-4S]0 + Fe2+ ↔ [4Fe-4S]2+ equilibrium is entirely consistent with an iron-sensing function . Indeed , we note that Fur , the iron-sensing regulator of E . coli , binds Fe2+ with a Kd of ~1–10 µM ( Bagg and Neilands , 1987; Mills and Marletta , 2005 ) . Although done in vitro , the parameters that emerged are in keeping with its role as a Fe-responsive regulator in vivo . Here , as previously ( Pellicer Martinez et al . , 2017 ) , we have used the iron chelator EDTA to simulate low iron , via its ability to coordinate Fe2+ that has dissociated from the RirA [4Fe-4S] cluster . While EDTA is clearly not a physiological ligand , it nevertheless induces low iron conditions by providing a sink for iron that simulates conditions under which cluster degradation occurs . Conditions such as this must exist in the cytosol of cells growing under iron-limiting conditions , where a myriad of iron-requiring systems are in competition for available iron , and where a drop in ‘free’ iron levels results in loss of RirA-mediated transcriptional repression leading to up-regulation of iron uptake systems . Competitive loss of Fe2+ from [4Fe-4S]2+ RirA results in a [3Fe-4S]0 form that is unstable , undergoing degradation to an apo-form . This lability under low iron conditions is not a general property of [4Fe-4S] clusters , as previously demonstrated for the closely related [4Fe-4S] NsrR ( Pellicer Martinez et al . , 2017 ) . This raises the question of why the RirA cluster is so fragile . One possibility is that the cluster is bound to the protein only by the three conserved Cys residues , with the tetrahedral coordination of the non-Cys bound iron completed by a weakly interacting ligand such as water or hydroxide . Such an arrangement , which would enhance the lability of the site-differentiated iron of the [4Fe-4S] cluster , would be similar to that found in the mammalian iron sensor protein IRE-BP ( iron regulatory element-binding protein ) ( Dupuy et al . , 2006 ) , and it is tempting to speculate that this may be a general feature of iron-sensing iron–sulphur clusters . We also note that the key step of iron-sensing in RirA , the simple equilibrium [4Fe-4S]2+ ↔ [3Fe-4S]0 + Fe2+ , mirrors the iron-sensing step of the better known bacterial global iron regulators Fur and DtxR in which Fe2+ binds the apo-form of these proteins ( Lee and Helmann , 2007; Pohl et al . , 2003; D'Aquino et al . , 2005; Ding et al . , 1996 ) . However , the relative complexity of the RirA cofactor ( co-repressor ) enables it to also directly sense O2 . A key step for O2-sensing is the oxidation of [3Fe-4S]0 to [3Fe-4S]1+ . The [3Fe-4S]1+ intermediate was more reactive than the [3Fe-4S]0 form , and several other subsequently formed intermediates also showed higher reactivity , leading to an overall enhanced rate of degradation ( Figure 5 and Table 1 ) . The susceptibility of cluster iron and sulphide to oxidation in the presence of O2 clearly drives the increased rate of cluster degradation . Thus , the [3Fe-4S]0 cluster formed via the reversible dissociation of Fe2+ from [4Fe-4S]2+ RirA is susceptible to oxidation in the presence of O2 . We propose that , even under iron-replete conditions , in the presence of O2 some oxidation of the RirA [3Fe-4S]0 form occurs before it can re-associate with Fe2+ , resulting in cluster degradation . Understanding how iron and O2 signals are sensed and integrated is important for all organisms that depend on both iron and O2 . In the case of R . leguminosarum , RirA regulates not only iron uptake , but also iron–sulphur cluster biogenesis ( Todd et al . , 2006 ) . The RirA O2-sensing mechanism enables the cell to meet the requirement for iron–sulphur cluster biosynthesis under aerobic conditions ( Imlay , 2006 ) . Under low iron and in the presence of O2 , cluster degradation occurs more readily as a result of the combined effects of competition for iron and the enhanced rate of cluster degradation under oxidising conditions . The rate at which cluster conversion/degradation occurs is relatively slow , particularly compared to the rate at which the [4Fe-4S] ↔ [3Fe-4S] equilibrium is established . In some cases at least , sensing processes are relatively slow reactions . For example , the master regulator of the aerobic/anaerobic switch in many bacteria , [4Fe-4S] FNR , undergoes a similar [4Fe-4S] to [2Fe-2S] cluster conversion reaction over minutes ( Crack et al . , 2008; Crack et al . , 2007 ) , with in vivo transcriptional responses occurring over a similar timescale ( Partridge et al . , 2007 ) . Responses to changes in iron levels mediated by Fur also occur over several minutes ( Pi and Helmann , 2017 ) . Thus , the cellular response to change does not necessarily need to occur instantly , perhaps reflecting that environmental changes themselves may occur over a period of time . The similarities between the FNR cluster conversion mechanism ( Crack et al . , 2017 ) and that described here for RirA are remarkable in that the two proteins have no sequence or structural similarity beyond the cluster , and FNR does not sense iron levels; its cluster is stable in the presence of iron chelators ( Crack et al . , 2008 ) . The advances reported here are relevant to all species that utilise RirA as a regulator , including several important pathogens of plants and animals . This work also contributes important new information about the widespread Rrf2 family of iron–sulphur cluster-binding regulators , and the variable ways in which they employ a cluster to sense environment change . Finally , it opens up new questions about RirA function , including: the nature of the cluster coordination; whether the Kd for the [4Fe-4S]2+ ↔ [3Fe-4S]0 + Fe2+ equilibrium and the mechanism of cluster disassembly are affected by [4Fe-4S] RirA complex formation with DNA; which steps of the degradation pathway are key for loss of DNA-binding; and , whether disruption of one cluster is sufficient to modulate DNA-binding of the RirA dimer . Efforts to address these are currently underway . R . leguminosarum RirA was over-expressed in E . coli and purified as previously described ( Pellicer Martinez et al . , 2017 ) . In vitro cluster reconstitution to generate [4Fe-4S] RirA was carried out in the presence of NifS , as described previously ( Crack et al . , 2014b ) . Protein concentrations were determined using the method of Bradford ( Bio-Rad ) , with bovine serum albumin as the standard . Cluster concentrations were determined by iron and sulphide assays ( Crack et al . , 2006; Beinert , 1983 ) or by using an absorbance extinction coefficient at 383 nm for the RirA [4Fe-4S] cluster of 13 , 460 ± 250 M−1 cm−1 ( Pellicer Martinez et al . , 2017 ) . 34S labelled [4Fe-4S] RirA was generated by reconstitution using 34S-L-cysteine as previously described ( Crack et al . , 2019 ) . [4Fe-4S] RirA was exchanged into anaerobic 250 mM ammonium acetate , pH 7 . 3 , using a desalting column ( PD-10 , GE Healthcare ) in an anaerobic glove box ( O2 < 2 ppm ) and diluted to ~30 µM cluster using the same buffer . To simulate low iron conditions , the soluble high affinity iron chelator EDTA ( Fe2+-EDTA , logK = 14 . 3 , Fe3+-EDTA , logK = 25 . 1 ) ( Martell and Smith , 1974 ) was added at 250 µM ( final concentration ) and the cluster response was followed via spectroscopy or mass spectrometry . To investigate the sensitivity of [4Fe-4S] RirA to low iron conditions in the presence of O2 , anaerobic protein samples in 250 mM ammonium acetate , pH 7 . 3 , were rapidly diluted with an air-saturated buffer containing EDTA to give the desired final O2 and EDTA concentrations and the sample was infused directly into the mass spectrometer via a syringe pump thermostatted at 37°C , or incubated at 37°C prior to rapid freezing in liquid nitrogen for EPR measurements . For absorbance experiments , 500 µM glutathione ( GSH ) was added to the sample to help solubilise oxidised sulphur species that otherwise caused significant scattering of light . The dissociation constant ( Kd ) for Fe2+ binding to [3Fe-4S]0 RirA was determined using an Fe2+-binding competition assay employing the well-known divalent metal ligand mag-fura-2 ( Rodrigues et al . , 2015 ) . Mag-fura-2 forms a 1:1 complex with Fe2+ , resulting in a shift of absorbance maximum from 366 nm ( for metal-free mag-fura-2 ) to 325 nm ( Rodrigues et al . , 2015 ) . Mag-fura-2 was dissolved in ultra-pure water to give a 2 mM stock solution and stored at −80°C until needed . Titrations of ~7 µM [4Fe-4S] RirA in 25 mM HEPES , 50 mM NaCl , 750 mM KCl , pH 7 . 5 with mag-fura-2 were carried out under anaerobic conditions at room temperature . Optimisation experiments showed that absorbance changes due to Fe2+-binding to the Mag-fura-2 occurred quickly and so UV-visible spectra were recorded using a Jasco V500 spectrometer immediately following each addition of the chelator , such that the titration was complete within ~60 min . Overnight incubation of RirA with Mag-fura-2 led to cluster conversion/breakdown , and increased coordination of Fe2+ to the chelator . Thus , the dissociation of a single Fe2+ from the [4Fe-4S]2+ cluster occurred on a very different timescale to cluster conversion , permitting investigation of the [4Fe-4S] to [3Fe-4S] equilibrium . A potential difficulty of using this or any ligand that gives rise to absorbance in the UV-visible region is that its absorbance spectrum overlaps that of the RirA cluster . Furthermore , the absorbance due to the cluster changes upon loss of a single Fe2+ , and so a robust method to correct for these changes was required . Fortunately , the absorbance profiles of apo- and divalent metal-bound mag-fura-2 overlap , with an isosbestic point at 346 nm , providing a means to correct for underlying absorbance changes due to the cluster ( Rodrigues et al . , 2015; Simons , 1993 ) . Firstly , the spectrum due to initial [4Fe-4S] cluster was subtracted from each subsequent spectrum . Then the data were normalised so that the resulting nest of spectra reported on the changing proportions of Fe2+-bound and apo-mag-fura-2 . Finally , spectra were corrected for any changes in the subtracted cluster spectrum using a scaling factor that preserved the isosbestic point . Spectra of mag-fura-2 in apo and Fe2+ forms ( in the absence of RirA ) were used to assist with this . Changes due to the scaling factor were <0 . 07 absorbance units , within the range expected for the absorbance change during cluster conversion . From this , the percentage of maximum Fe2+-mag-fura-2 complex present ( where 100% equated to the concentration of [4Fe-4S] RirA , that is one Fe2+ per cluster ) was plotted as a function of free mag-fura-2 . This resulted in a plot with a hyperbolic form , which was fitted with an equation describing a simple binding isotherm using Origin8 ( OriginLabs ) . This yielded an apparent competition dissociation constant from which the Kd for RirA was determined using the expression:KdMF2 app=KdMF2 ( 1+RirAKdRirA ) Where Kd ( MF2 app ) is the determined apparent Kd from the competition binding assay , Kd ( MF2 ) is the Kd for mag-fura-2 binding to Fe2+ , ( previously determined as Kd = 2 . 05 µM; Rodrigues et al . , 2015 ) and Kd ( RirA ) is the Kd for [3Fe-4S] binding to Fe2+ ( Hulme and Trevethick , 2010 ) . The resulting Kd and standard error was determined from three equivalent titration experiments . [4Fe-4S] RirA samples were infused directly ( 0 . 3 mL/h ) into the ESI source of a Bruker micrOTOF-QIII mass spectrometer ( Bruker Daltonics , Coventry , UK ) operating in the positive ion mode , and calibrated using ESI-L Low Concentration Tuning Mix ( Agilent Technologies , San Diego , CA ) . Mass spectra ( m/z 500–1750 for RirA monomer; m/z 1 , 800–3 , 500 for RirA dimer ) were acquired for 5 min using Bruker oTOF Control software , with parameters as follows: dry gas flow 4 L/min , nebuliser gas pressure 0 . 8 Bar , dry gas 180°C , capillary voltage 2750 V , offset 500 V , ion energy 5 eV , collision RF 180 Vpp , collision cell energy 10 eV . Optimisation of experimental conditions for the transmission of dimeric species was achieved by increasing the capillary voltage to 4000 V and the collision RF to 600 Vpp ( Laganowsky et al . , 2013 ) . Processing and analysis of MS experimental data were carried out using Compass DataAnalysis version 4 . 1 ( Bruker Daltonik , Bremen , Germany ) . Neutral mass spectra were generated using the ESI Compass version 1 . 3 Maximum Entropy deconvolution algorithm over a mass range of 17 , 300–18 , 000 Da for the monomer and 34 , 850–35 , 810 Da for the dimer . For kinetic modelling , in order to clearly resolve overlapping peaks , multiple Gaussian functions were fitted to the experimental data using a least-squares regression function in Origin 8 ( Origin Lab ) ( Laganowsky et al . , 2013 ) . Exact masses are reported from peak centroids representing the isotope average neutral mass . For apo-proteins , these are derived from m/z spectra , for which peaks correspond to [M + nH]n+/n . For cluster-containing proteins , where the cluster contributes charge , peaks correspond to [M + ( Fe-S ) x+ + ( n-x ) H]n+/n , where M is the molecular mass of the protein , Fe-S is the mass of the particular iron–sulphur cluster of x+ charge , H is the mass of the proton and n is the total charge . In the expression , the x+ charge of the iron–sulphur cluster offsets the number of protons required to achieve the observed charge state ( n+ ) ( Johnson et al . , 2000 ) . Predicted masses are given as the isotope average of the neutral protein or protein complex , in which iron–sulphur cluster-binding is expected to be charge-compensated ( Crack et al . , 2017; Kay et al . , 2016 ) . Mass spectra are plotted as percentage relative abundances , where the most abundant species is arbitrarily set to 100% and all other species are reported relative to it . Time-resolved MS intensity data for global analysis was processed to generate relative abundance plots of ion counts for the relevant species as a fraction of the total ion count for all species . This permitted changes in relative abundance to be followed without distortions due to variations in ionisation efficiency that normally occur across a data collection run . Some variation in the starting spectrum was observed due to the presence of cluster breakdown products , which affected concentrations of intermediates during the cluster conversion/breakdown process; these variations are represented by error bars in the relative abundance plots of Figures 4 and 7 . Such plots were analysed globally using the program Dynafit 4 ( BioKin Ltd ) ( Kuzmic , 1996 ) , which employs nonlinear least-squares regression of kinetic data , based on multi-step mechanisms , from which fits of the experimental data were generated and rate constants estimated , as previously reported ( Crack et al . , 2017 ) . The presence of breakdown products in the starting spectrum was accounted for by allowing starting abundance to be offset from zero . Briefly , the kinetic model consisted of a series of sequential or branched reactions beginning with the dissociation of Fe2+ from [4Fe-4S]2+ to form [3Fe-4S]0 and the reverse reaction , the binding of Fe2+ to [3Fe-4S]0 to reform the [4Fe-4S]2+ cluster . With the exception of the latter , which is a second order process , all steps in the proposed mechanism are first order . All steps of the mechanism are shown in Figure 5 . Absorbance kinetic data at A386 nm were recorded via a fibre optic link , as previously described ( Crack et al . , 2007 ) . EPR measurements were made with an X-band Bruker EMX EPR spectrometer equipped with a helium flow cryostat ( Oxford Instruments ) . Unless stated otherwise , EPR spectra were measured at 10 K at the following instrumental settings: microwave frequency , 9 . 471 GHz; microwave power , 3 . 18 mW; modulation frequency , 100 kHz; modulation amplitude , 5 G; time constant , 82 ms; scan rate , 22 . 6 G/s; single scan per spectrum . Relative concentrations of the paramagnetic species were measured using the procedure of spectral subtraction with a variable coefficient ( Svistunenko et al . , 2006 ) and converted to absolute concentrations by comparing an EPR spectrum second integral to that of a 1 mM Cu ( II ) in 10 mM EDTA standard , at non-saturating values of the microwave power . The RirA EPR signal saturation was studied by taking EPR measurements at 13 values of microwave power , ranging from 0 . 2 µW to 200 mW , at eight temperature values , ranging from 4 K to 50 K . The EPR sample was equilibrated at every new temperature for at least 8 min before the power set of spectra measurements commenced .	Virtually all life forms require iron to survive , yet too much of the metal can be catastrophic . In healthy cells , many systems regulate this delicate balance . For instance , when a protein called RirA is active in Rhizobium bacteria , it can sense high levels of the metal and helps to shut down the production of proteins that bring in more iron . RirA contains a cluster of four iron and four sulphur atoms , which acts as a sensor for iron availability: however , exactly how this cluster structure detects the levels of the metal in a cell was previously unclear . Pellicer Martinez , Crack , Stewart et al . used a technique known as time-resolved mass spectrometry to examine the sensory response of the iron-sulphur cluster of RirA when different levels of iron were available . The results revealed a ‘loose’ iron atom in the cluster; when iron levels fall down , this atom is rapidly lost as it is scavenged for use in other essential processes . Without it , the cluster in RirA collapses and the protein becomes inactive . This prompts the cell to produce proteins that enable it to take up iron from its surroundings . Once iron levels are high , RirA can regain its cluster and is active again , stopping the production of proteins that bring in more iron . Iron-sulphur clusters are common in many proteins , and this work offers new insight into their various roles . It also highlights the potential to use time-resolved mass spectrometry to examine biological processes in depth .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology" ]	2019	Mechanisms of iron- and O2-sensing by the [4Fe-4S] cluster of the global iron regulator RirA
Nucleotide-sugar transporters ( NSTs ) are critical components of the cellular glycosylation machinery . They transport nucleotide-sugar conjugates into the Golgi lumen , where they are used for the glycosylation of proteins and lipids , and they then subsequently transport the nucleotide monophosphate byproduct back to the cytoplasm . Dysregulation of human NSTs causes several debilitating diseases , and NSTs are virulence factors for many pathogens . Here we present the first crystal structures of a mammalian NST , the mouse CMP-sialic acid transporter ( mCST ) , in complex with its physiological substrates CMP and CMP-sialic acid . Detailed visualization of extensive protein-substrate interactions explains the mechanisms governing substrate selectivity . Further structural analysis of mCST’s unique lumen-facing partially-occluded conformation , coupled with the characterization of substrate-induced quenching of mCST’s intrinsic tryptophan fluorescence , reveals the concerted conformational transitions that occur during substrate transport . These results provide a framework for understanding the effects of disease-causing mutations and the mechanisms of this diverse family of transporters . Nucleotide-sugar transporters ( NSTs ) are products of the solute carrier 35 ( SLC35 ) gene family in humans and are a critical part of the glycosylation machinery in all eukaryotes ( Hadley et al . , 2014; Ishida and Kawakita , 2004; Song , 2013 ) . They are responsible for transporting nucleotide sugars from the cytoplasm , where they are synthesized , into the Golgi lumen where they are then utilized by glycosyltransferases to glycosylate proteins and lipids ( Figure 1—figure supplement 1A ) ( Capasso and Hirschberg , 1984; Milla and Hirschberg , 1989; Tiralongo et al . , 2006; Waldman and Rudnick , 1990 ) . NSTs provide an additional essential role of transporting the nucleotide monophosphate ( NMP ) byproduct of the glycosyltransferase reaction back to the cytoplasm where it can be recycled . Since many glycosyltransferases are inhibited by NMPs , this antiport property of NSTs provides an additional layer of regulation of glycan synthesis ( Hirschberg et al . , 1998 ) . Many NSTs are not obligatory antiporters , but the presence of NMPs in the lumen accelerates nucleotide sugar uptake several-fold through a poorly-understood process termed ‘trans-stimulation’ ( Capasso and Hirschberg , 1984; Milla and Hirschberg , 1989; Tiralongo et al . , 2006; Waldman and Rudnick , 1990 ) . Glycosylation is the most common form of protein and lipid modification . It affects nearly every aspect of biology by modifying protein folding , stability , and functional properties ( Dwek et al . , 2002; Moremen et al . , 2012; Ohtsubo and Marth , 2006; Stanley , 2011 ) . Glycans are typically terminated by sialic acids , which is important for many cell adhesion processes but also serves as ligands for pathogen invasion ( Varki , 2008; Varki and Schauer , 2009 ) . Since NST activity controls the concentrations of nucleotide sugars in the Golgi lumen , disruption of NST activity can have many adverse physiological effects . There are a number of genetic diseases caused by mutations in human NST genes that have severe and debilitating phenotypes ( Jaeken and Matthijs , 2007; Song , 2013 ) . In addition , some NSTs are potently inhibited by antiviral nucleoside analogs , which may form the basis for some of the side effects associated with this class of drugs ( Chiaramonte et al . , 2001; Hall et al . , 1994 ) . On the other hand , NSTs also represent potential therapeutic targets , as NSTs are virulence factors in many types of parasites and fungi that require extensive cell-surface glycoconjugates for effective pathogenesis ( Caffaro et al . , 2013; Descoteaux et al . , 1995; Engel et al . , 2009; Hong et al . , 2000; Liu et al . , 2013; Ma et al . , 1997 ) . Additionally , altered glycosylation profiles of cell-surface proteins are a property of malignant cells , and studies have shown that pharmacological block of some NSTs can inhibit tumor metastasis ( Caffaro and Hirschberg , 2006; Esko and Bertozzi , 2009; Hadley et al . , 2014; Ohtsubo and Marth , 2006; Song , 2013; Stowell et al . , 2015; Wang et al . , 2016 ) . Mammals express seven NSTs , which collectively transport the eight nucleotide sugars used in Golgi glycan synthesis ( Figure 1—figure supplement 1B ) . Each NST is highly selective for either one or two nucleotide sugars and the corresponding NMP – a critical property that is required for the proper hierarchical assembly of glycans . However , little is known about the underlying structural basis for this specificity or the mechanisms of substrate antiport . Here , we report X-ray crystal structures of the mouse CMP-sialic acid ( CMP-Sia ) transporter ( mCST ) in complex with CMP and CMP-Sia at 2 . 6 Å and 2 . 8 Å resolution , respectively . Analysis of these structures , coupled with functional characterization of mCST , elucidates the fundamental principles of the structure-function relationship of NSTs . We identified mCST , which is 91% identical to human CST ( hCST; Figure 1—figure supplement 2 ) , as a suitable candidate for structural studies after screening a large panel of NST orthologs . One obstacle that we encountered when characterizing this construct was our discovery that commercial stocks of CMP-Sia contain approximately 10% CMP ( Figure 1—figure supplement 3A ) , which has a ~ 100 fold higher affinity towards mCST ( Figure 1A ) . In addition , CMP-Sia hydrolyzes in aqueous solutions to yield CMP and Sia ( Beau et al . , 1984; Horenstein and Bruner , 1996 ) ( Figure 1—figure supplement 3B–H ) . To our knowledge , these issues have not been acknowledged or addressed in the CST literature . Therefore , we developed a method where we treat CMP-Sia with a nonselective phosphatase , Antarctic phosphatase ( AnP ) , to convert all of the free CMP to cytidine ( Figure 1—figure supplement 3B ) , which does not bind mCST ( Figure 1A ) . For binding and transport experiments , the AnP is removed by ultrafiltration and we discuss in the Methods how we account for any CMP-Sia hydrolysis that occurs during these experiments . When mCST is purified in detergent , it retains saturatable substrate binding , with a Kd of 6 . 3 ± 1 . 2 µM and 482 ± 147 µM for CMP and CMP-Sia , respectively ( Figure 1A ) . Purified mCST also transports CMP-Sia when reconstituted into liposomes , with a Vmax of 6 . 5 ± 0 . 6 nmol/mol mCST/min and a Km of 58 . 1 ± 13 . 2 µM ( Figure 1B and C ) , which is within the range of reported Km’s for CMP-Sia transport for human or mouse CST expressed in Golgi ( Aoki et al . , 2003; Chiaramonte et al . , 2001; Milla and Hirschberg , 1989; Tiralongo et al . , 2006 ) . CMP-Sia transport activity is greatly reduced when CMP is not included inside the vesicles ( Figure 1B ) , which is likely the result of the lack of trans-stimulation . We initially obtained crystals of full-length mCST in complex with CMP using the hanging-drop vapor diffusion ( HDVD ) method , which diffracted anisotropically to a moderate resolution ( 3 . 4 × 3 . 4 × 4 . 6 Å ) . We calculated an experimental electron density map using multiple isomorphous replacement with anomalous scattering using Hg and Pt derivatives ( Figure 2—figure supplement 1A–B ) . A model based on this map showed that the entire 20-residue C-terminus was disordered . After screening a series of deletion constructs , we identified one lacking the last 15 residues ( termed mCST∆C ) that was functionally identical to the full-length protein ( Figure 1 , Table 1 ) and produced lipidic cubic phase ( LCP ) crystals that diffracted to 2 . 6 Å . We solved the structure of these crystals with molecular replacement using the model of full-length mCST as the search model , and built a detailed model of mCST that refined to R/R-free factors of 24 . 1/25 . 2% ( Figure 2 , Figure 2—figure supplement 1C–D , Table 2 ) . We grew LCP crystals of mCST∆C in complex with CMP-Sia , which diffracted to 2 . 8 Å , by leaving AnP in CMP-Sia stocks and also adding additional AnP to maintain very low concentrations of CMP during crystal growth ( Figure 1—figure supplement 3C ) . The structure was solved with molecular replacement using the mCST∆C-CMP structure as a search model , and we refined the resulting model to R/R-free factors of 25 . 6/27 . 9% ( Figure 2 , Figure 2—figure supplement 1E–H , Table 2 ) . The HDVD and LCP structures were overall very similar ( Figure 2—figure supplement 2A ) , despite having very different crystal lattices ( Figure 2—figure supplement 2B–E ) . This indicates that the crystal packing interactions had no impact on the overall conformation of mCST . We will only discuss the higher-resolution LCP structures in the rest of the text . The structure of mCST consists of a 10-TM bundle ( Figure 2A–2C ) . The general topology resembles that seen in other members of the larger drug/metabolite transporter ( DMT ) superfamily , which includes proteins with 4–10 TMs ( Jack et al . , 2001; Västermark et al . , 2011 ) . The 10-TM DMT proteins are thought to have arisen from an ancient helix addition to the 4-TM proteins ( typified by EmrE ) followed by a gene duplication event . Although the two halves of mCST lack significant sequence identity , the 10-TM bundle of mCST is arranged with a pseudo-two-fold inverted symmetry ( Figure 2—figure supplement 3 ) , similar to what is seen in other 10-TM DMT protein structures ( Lee et al . , 2017; Parker and Newstead , 2017; Tsuchiya et al . , 2016 ) . The N- and C-termini are on the same side of the Golgi membrane and are thought to face the cytoplasm based on previous experiments using HA-tag insertions ( Eckhardt et al . , 1999 ) . We observe several intra-membrane interactions between mCST molecules in our various crystal lattices ( Figure 2—figure supplement 2B–E ) . However , since we determined that hCST , which is biochemically identical to mCST , is a monomer when purified in detergent ( Figure 2—figure supplement 2F–G ) , these interactions do not stem from a stable pre-formed oligomer that survived detergent extraction and purification . There is some evidence that suggests that some NSTs may form oligomers in the Golgi membrane ( Gao and Dean , 2000; Hong et al . , 2000; Puglielli and Hirschberg , 1999; Puglielli et al . , 1999 ) ; however , no experiments have directly investigated the oligomeric status of CSTs . Further work will be needed to determine if the interactions we see in our crystal contacts are physiologically relevant . The overall conformation of the CMP- and CMP-Sia-bound mCST structures is generally similar with a few key differences that will be discussed later ( Figure 2B ) . CMP and CMP-Sia both bind a similarly-shaped cavity inside the transporter , which is roughly halfway between the cytoplasmic and lumenal sides of the protein ( Figure 2D and E ) . An elliptically-shaped constriction ( Figure 2F–2H ) divides this cavity into two distinct regions: the substrate-binding cavity and the lumen-facing cavity . The dimensions of this constriction ( ~6×4 Å ) can perhaps permit the passage of CMP , but it is probably too small to let the bulkier Sia moiety of CMP-Sia pass through . For this reason , we think that this conformation of mCST represents a lumen-facing partially-occluded state . This state has not yet been observed in other 10-TM DMT protein structures and will be discussed in more detail below . Most members of the SLC35 family transport UMP and UDP-sugars , only one transports GMP and GDP-fucose , and only CST transports CMP and CMP-Sia ( Figure 1—figure supplement 1 ) . CST is highly selective for CMP , as UMP has a 20-fold higher Ki for inhibiting CMP-Sia transport ( Chiaramonte et al . , 2001 ) , and UMP and GMP have a 250- and 850-fold lower binding affinity than CMP , respectively ( Figure 1A and Table 1 ) . Our structure starts to explain how this is achieved . In mCST , the various groups of CMP are extensively coordinated by at least fifteen residues ( Figure 3A ) . Nearly a third of these interact with the phosphate group . The importance of these protein-phosphate interactions in mediating substrate binding is highlighted by: 1 ) the fact that cytidine alone does not bind mCST ( Figure 1A ) , and 2 ) among all the residues that line the substrate-binding cavity , the only ones linked to known disease-causing mutations interact exclusively with the phosphate ( Figure 3—figure supplement 1 and Table 3 ) . The cytosine group is closely coordinated by residues Lys55 , Tyr214 , and Asn210 , which explains why mCST cannot easily accommodate the much larger guanine base of GMP . On the other hand , UMP is very similar to CMP; however , the structure suggests that there are two key differences that explain their differences in affinity . The first is that UMP has a carbonyl at the C-4 position of the uracil base instead of the amine that cytosine has ( Figure 1—figure supplement 1C ) . In mCST , this amine is hydrogen-bonded to the Oδ1 of Asn210 , with the orientation of Asn210’s side chain stabilized by a hydrogen bond between the Nδ2 of Asn210 and the hydroxyl of Ser261 ( Figure 3A ) . In this orientation , Asn210 will not be able to form a hydrogen bond with the C-4 carbonyl of uracil . The partial negative charge of uracil’s carbonyl would also be electrostatically repelled by the partial negative charge of the π system of Phe195 . The second difference between cytosine and uracil is that the nitrogen at position three is protonated in uracil . In mCST , the hydroxyl of Tyr214 is hydrogen-bonded to the N-3 of cytosine . If we assume that UMP would bind in a similar orientation as CMP , then the proton on the N-3 of uracil would be ~1 . 7 Å away from the hydroxyl oxygen of Tyr214 . Therefore , it may still be able to hydrogen bond with Tyr214; however , NH–O hydrogen bonds are typically weaker than OH–N bonds hydrogen . In addition , tyrosine hydroxyls are generally thought to be relatively weak hydrogen bond acceptors on account of the lone pair of the hydroxyl’s oxygen partially delocalizing within the adjacent aromatic ring ( McDonald and Thornton , 1994 ) . These hypotheses are further supported by comparing the sequence of mCST to the closely-related UDP-Gal/GalNAc transporter ( UGT , SLC35A2; 44% sequence identity ) and the UDP-GlcNAc transporter ( NGT , SLC35A3; 41% sequence identity ) ( Figure 1—figure supplement 2 ) . Nearly all of the residues of mCST that are involved in coordinating CMP are conserved in UGT and NGT , with two key exceptions . The equivalent residues of Tyr214 and Ser261 in mCST are a Gly and Ala , respectively , in both UGT and NGT . An Ala at the Ser261 position would allow the conserved Asn at position 210 to orient its Nδ2 towards the binding pocket to be able to hydrogen bond with the C-4 carbonyl of uracil . On the other hand , the effect of having a Gly at the Tyr214 position is unclear . Sequence conservation in this region is high and homology models of UGT and NGT suggest that this Tyr-Gly substitution will result in the formation of a cavity that would be large enough to accommodate at least a few waters . Future work will be needed to understand what role , if any , a Gly at this position plays in affecting substrate specificity in UGT or NGT . Although interactions between the protein and nucleobase are important for nucleotide selectivity , the additional extensive interactions between the protein and the rest of the nucleotide are apparently employed by some exogenous ligands to achieve high-affinity binding . This is seen with the antiviral nucleoside analog azidothymidine monophosphate ( AZTMP ) . This molecule inhibits CST’s CMP-Sia transport activity with a Ki similar to that of CMP ( Chiaramonte et al . , 2001; Hall et al . , 1994 ) , despite the fact that the thymine base of AZTMP is essentially a methylated uracil , which would not be expected to have a strong interaction with CST . Instead , docking AZTMP into mCST suggests that the azide modification of the ribose group likely forms additional compensating interactions with the protein ( Figure 3—figure supplement 2A–B ) , which could be exploited in future efforts to pharmacologically target NSTs . In the mCST-CMP-Sia structure , the Sia moiety of CMP-Sia occupies a large cavity that was present but vacant in the CMP-only structure ( Figure 2D and E ) whereas the CMP moiety of CMP-Sia binds in a similar but slightly different orientation ( Figure 3B ) . The Sia moiety interacts with several residues of mCST through the various functional groups on the sugar ring ( Figure 3C ) . The C-1 carboxyl and C-6 glycerol groups make several polar interactions , including an ionized hydrogen bond between the C-1 carboxyl and Lys124 . This residue interacts with the α-phosphate of CMP in the CMP-only structure , but is shifted 2 Å to the side to accommodate Sia’s C-1 carboxyl . Docking of UDP-sugars into homology models of UGT and NGT suggest that this conserved lysine in these transporters would interact with the β-phosphate of the UDP moiety ( Figure 3—figure supplement 2D–F ) . The C-4 hydroxyl and C-5 N-acetyl groups do not make any direct polar interactions with the protein but may interact with a number of residues through van der Waals forces . Cavities adjacent to the C-5 N-acetyl could likely accommodate one of the most common sialic acid derivatives where the C-5 substituent is an N-glycolyl instead of an N-acetyl , which are also referred to as Neu5Gc and Neu5Ac , respectively ( Figure 1—figure supplement 1C , Figure 3—figure supplement 2G–H ) ( Varki and Schauer , 2009 ) . Neu5Ac is converted to Neu5Gc by a cytoplasmic hydrolase that is found in most mammals except for humans . However , Neu5Gc is still found in human glycans , which is thought to arise primarily from dietary sources and has been associated with an increased risk for inflammation and carcinomas ( Samraj et al . , 2014; Varki , 2008; Varki and Schauer , 2009 ) . The protein sequence for the region around the N-acetyl is nearly identical between mCST and hCST , which explains how hCST can still transport CMP-Neu5Gc . Overall , the Sia moiety of CMP-Sia makes far fewer protein contacts than the CMP moiety . This difference is consistent with the observation that , unlike CMP , Sia alone does not bind mCST ( Figure 1A ) and does not competitively inhibit CMP-Sia transport ( Carey et al . , 1980; Perez and Hirschberg , 1987 ) . This suggests that mCST’s mechanism for discriminating among different nucleotides , as described above , is also the primary means for nucleotide-sugar selectivity . An additional component of selectivity may also arise from the large volume of the Sia-binding cavity , which may not be able to effectively coordinate the other smaller nucleotide-coupled sugars . On the other hand , UDP-sugar transporters must utilize the converse of these principles in order to discriminate among the six UDP-coupled sugars . This concept is supported by sequence comparison between mCST and UGT/NGT ( Figure 1—figure supplement 2 ) , which shows that the most significant differences among the Sia-binding-site residues are the exchange of several non-polar and small side chains in CST to longer , mostly polar side chains in UGT and NGT . Homology models of UGT and NGT suggest this would result in a smaller binding pocket , which would clash with the various groups of Sia , but would be able to make several key interactions with substrates that may be important in substrate selectivity ( Figure 3—figure supplement 2C–F ) . Compared to the CMP-bound structure , the N-terminal two-thirds of TM1 of the CMP-Sia-bound structure twists and moves toward the substrate binding pocket ( Figures 2B and 4A ) . This brings three residues from TM1 ( Met20 , Ala24 , and Tyr27 ) into close contact with the Sia moiety and is also associated with several other conformational changes in various regions of the protein . The first is that there is a ~ 0 . 5 Å displacement of the adjacent TM8 which is also associated with smaller but significant displacements of the other nearby TMs surrounding the CMP site ( Figure 4A ) . These movements cause the side chains surrounding the CMP site to generally move away from each other , creating a slightly larger and differently-shaped binding pocket ( Figure 4B–4E ) – the latter of which is reflected in how the CMP moiety of CMP-Sia binds in a different orientation than CMP alone ( Figure 3B ) . The second conformational difference associated with CMP-Sia binding is that TM1 is brought into closer contact with TM9 ( Figure 4F–G ) . This is associated with a mostly rigid-body reorientation of a three-helix domain comprised of TMs 5 , 9 , and 10 ( Figure 4H ) , which will be discussed in more detail in another section below . This increased interaction between TMs 1 and 9 is also associated with TM1 becoming generally more ordered . For example , we were able to see electron density for all of TM1 , starting at Asn7 , in the CMP-Sia-bound structure . However , we did not see strong density for residues 7–14 in the CMP-bound structure and therefore these residues were not modeled ( Figure 2B ) . We propose that these conformational differences may at least partially account for the ~100 fold difference in binding affinity between CMP and CMP-Sia ( Figure 1A ) . First of all , the protein distortions needed to accommodate the Sia moiety of CMP-Sia may be energetically costly . In particular , the increased interaction between TMs 1 and 9 and the stabilization of TM1 may lead to a reduction in the conformational entropy of mCST to a degree that is not adequately compensated by enthalpic gains . Similarly , the number of rotatable bonds in CMP-Sia is nearly three times that of CMP ( 18 versus 7 ) , so stabilization of the Sia moiety of CMP-Sia through interactions with residues on TMs 1 and 4 ( Figures 3C and 4A ) may also result in a greater entropic penalty to the binding free energy . Finally , the larger and differently-shaped CMP-binding pocket may weaken the interactions between the protein and the CMP moiety of CMP-Sia by negatively impacting the distance and geometry of the numerous non-covalent bonds between the protein and the CMP moiety of CMP-Sia . An additional component of the differential binding affinity could be related to the reduced charge of the phosphate group in CMP-Sia versus CMP . The highest pKa of the phosphate group of CMP is 6 . 3 , meaning that it will predominantly have a −2 charge at neutral pH . However , the phosphate group of CMP-Sia only has one titratable oxygen with a very low pKa , so it will only have a −1 charge ( although CMP-Sia’s overall charge is still −2 on account of the C-1 carboxyl group on the Sia moiety ) . Therefore , considering the important role that the phosphate group plays in mediating CMP binding to the protein , as described above , a reduction in the phosphate group’s charge state may manifest as a lower substrate binding affinity . In support of this , elegant studies on the effect of pH on UDP-GlcNAc transporter activity showed that monoanionic UMP is a poor substrate compared to dianionic UMP ( Waldman and Rudnick , 1990 ) . As mentioned above , we describe our structures of mCST as partially-occluded because the constriction that separates the substrate-binding site from the Golgi lumen is most likely too small to allow the free passage of a substrate . In order to understand how a substrate can freely exchange between the substrate-binding site and the Golgi lumen , it is first important to understand how the constriction is formed . This is illustrated in Figure 5A and D where it can be seen how the constriction is primarily formed by the lumenal end of TM9 being dissociated from the core of the protein . However , the lumenal ends of TMs 1 and 8 are still engaged with the core of the protein to form a lid over the substrate-binding pocket . Therefore , we propose that the lumenal ends of TMs 1 and 8 must also move away from the core of the protein to allow the substrate-binding cavity to be freely accessible from the lumen . Such a fully-open lumen-facing state may resemble the conformations seen in the structures of other DMT proteins that were determined in the absence of a substrate: the fungal GDP-mannose transporter , Vrg4 ( Parker and Newstead , 2017 ) , and the bacterial amino acid transporter , YddG ( Tsuchiya et al . , 2016 ) . As shown in Figure 5E and F , compared to mCST , both TMs 1 and 8 in Vrg4 and YddG are dissociated from the core of the protein to provide full accessibility to the substrate-binding site . The concept that mCST may transition between a substrate-free fully-open state and a substrate-bound partially-occluded state is supported by characterization of mCST’s intrinsic tryptophan fluorescence ( Figure 5G and H ) . We found that when CMP binds mCST , the intrinsic tryptophan fluorescence of mCST is significantly quenched . We further showed that this is due to only one of the four tryptophans in mCST , Trp207 , which is located on the cytoplasmic end of TM7 . Since tryptophan fluorescence is sensitive to the polarity of the local environment , these results are consistent with the region around this section of TM7 changing conformation upon substrate binding . Indeed , a comparison of the partially-occluded conformation of mCST with the fully-open conformations of Vrg4 and YddG show that the cytoplasmic ends of TMs 2 and 8 , which flank Trp207 , would be expected to undergo significant conformational rearrangements as the protein transitions from the fully-open to the partially-occluded state ( Figures 5B , C and I ) . mCST appears to be a nearly-obligatory antiporter – that is , very little CMP-Sia uptake is observed unless vesicles are pre-filled with CMP ( Figure 1B ) . This phenomenon , which has been previously described in the literature for CST and several other NSTs , has also been referred to as trans-stimulation or trans-acceleration ( Capasso and Hirschberg , 1984; Milla and Hirschberg , 1989; Tiralongo et al . , 2006; Waldman and Rudnick , 1990 ) . Further structural comparison of mCST with other DMT protein structures provides insight into how the presence of a substrate on the lumenal side of the protein is coupled to the rate of substrate uptake from the cytoplasm . The algal triose phosphate transporter GsTPT2 is another DMT protein that is also an obligatory antiporter . The structure of GsTPT2 was solved in complex with inorganic phosphate ( Pi ) and was observed to be in a fully-occluded state ( Lee et al . , 2017 ) . This structure is generally similar to the partially-occluded state of mCST , with the major difference being that the extracellular ( lumenal for mCST ) end of TM9 is associated with the core of the protein , thus blocking the substrate-binding cavity from having any access to either side of the membrane ( Figure 6A and B ) . The Pi molecule in the GsTPT2 structure is coordinated by Lys and Arg residues coming from the middle of TMs 4 and 9 ( Figure 6C and D ) . Based on this , the authors propose the following mechanism to describe the strict requirement of obligatory substrate antiport . Formation of the occluded state is driven by binding of the negatively-charged Pi substrate which allows close approximation of the positive charges of the Lys and Arg residues . However , in the absence of a substrate , electrostatic repulsion between these Lys and Arg residues will drive apart TMs 4 and 9 to stabilize a fully-open conformation and prevent the protein from undergoing a conformational transition to face the other side of the membrane . mCST also has lysines on TMs 4 and 9 – one on each – which are involved in coordinating the phosphate of CMP and CMP-Sia ( Figure 6D ) ; however , Lys272 is on the cytoplasmic end of TM9 . This part of TM9 is not expected to dramatically change conformation between the substrate-bound partially-occluded state and the substrate-free fully-open state , as described in the previous section ( Figures 5A–5F and 6C ) . Furthermore , if mCST were to adopt a fully-occluded conformation that resembles what is seen in GsTPT2 , then the cytoplasmic end of TM9 would essentially act as a mostly-stationary hinge during a transition to this state from either the fully-open or partially-occluded states ( Figure 6C and D ) . Therefore , we do not think that electrostatic repulsion between lysines on TMs 4 and 9 plays a major role in regulating the stability of different lumen-facing conformational states of mCST . We instead propose that , in the absence of a substrate , the partially-occluded state of mCST may be less stable than the fully-open state by virtue of the differential arrangement of TMs 1 , 8 , and 9 . TM1 in the partially-occluded state of mCST is mostly orthogonal to TMs 8 and 9 , whereas it is more parallel in the fully-open state of Vrg4 ( Figure 6E and F ) . This results in significantly less surface area being buried among TMs 1 , 8 , and 9 in the partially-occluded versus fully-open states ( 991 , 1037 , and 1531 Å2 for mCST-CMP , mCST-CMP-Sia , and Vrg4 respectively ) . The additional ~500 Å2 of buried surface area between TMs 1 , 8 , and 9 in the fully-open conformation may contribute to the stabilization of this state . This may present enough of a free-energy barrier that would impair the protein from undergoing the necessary structural rearrangements needed to face the cytoplasmic side . We further propose that substrate binding would lower this barrier by favoring the conversion to the partially-occluded state . This may be driven by a bound substrate interacting with residues on the core of the protein as well as with residues on TMs 1 and 8 . For instance , in addition to interactions with residues on TMs 2–4 , 6 , and 9 as shown in Figure 3 , CMP also interacts directly , or through structured waters , with residues Ala253 , Gly257 , Thr260 , and Ser261 on TM8 ( Figure 3A and Figure 2—figure supplement 1I and J ) ; and CMP-Sia interacts with these same residues on TM8 as well as with Met20 , Ala24 , and Tyr27 on TM1 ( Figures 3B , C and 4A ) . The combination of these protein-substrate interactions along with additional protein-protein interactions between the lumenal ends of TMs 2–4 and the lumenal ends of TMs 1 and 8 ( Figures 2C and 5D ) may favor the formation of the partially-occluded state by sufficiently offsetting the otherwise unfavorable loss of ~500 Å2 of buried surface area among TMs 1 , 8 , and 9 . The formation of a partially-occluded state may be a necessary intermediate state that allows the protein to transition to face the cytoplasmic side . For CST , it has been reported that lumenal CMP is 2–3 fold more effective at trans-stimulation of CMP-Sia uptake than an equivalent lumenal concentration of either UMP or CMP-Sia ( Chiaramonte et al . , 2001; Tiralongo et al . , 2006 ) . One possible reason for this may be the large difference in binding affinities between CMP and either UMP or CMP-Sia ( Figure 1A ) . In other words , if we assume that we are measuring binding affinities towards the lumen-facing partially-occluded state , then for a given sub-saturating concentration of a substrate in the lumen , CMP will have a greater fractional occupancy of the lumen-facing partially-occluded state . As mentioned previously , mCST has symmetry-related structural repeats ( Figure 2—figure supplement 3 ) – similar to what is seen in other DMT proteins ( Lee et al . , 2017; Parker and Newstead , 2017; Tsuchiya et al . , 2016 ) as well as in many other transporter families ( Drew and Boudker , 2016; Forrest , 2013 ) . Although there are exceptions , one commonly-observed alternating access mechanism that has been observed in these types of transporters is that symmetry-related conformational changes are involved in alternately exposing the protein to either side of the membrane ( Drew and Boudker , 2016; Forrest , 2013 ) . Symmetry-related exchange mechanisms have been proposed for other DMT proteins ( Lee et al . , 2017; Parker and Newstead , 2017; Tsuchiya et al . , 2016 ) and we think that mCST may operate in a similar fashion ( Figure 7 ) . As the protein transitions from a lumen-facing partially-occluded state to face the cytoplasmic side , it may first pass through a fully-occluded state that resembles the GsTPT2 structure . As shown in the structural comparison between mCST and GsTPT2 , this would not only involve the lumenal-end of TM9 associating with the core of the protein but also a rearrangement of TMs 5 and 10 ( Figure 6A and B ) . As discussed when comparing the structural differences between the CMP and CMP-Sia bound mCST structures , TMs 5 , 9 , and 10 do have a propensity to move independently of the rest of the protein ( Figure 4H ) . The role of these TMs in substrate transport is supported by the fact that mutations of residues at the interfaces of these TMs ( in CST and other human NST isoforms ) form the basis for several diseases ( Figure 3—figure supplement 1 and Table 3 ) . As the protein continues to transition , TM4 may move away from the protein core to form a cytoplasmic-facing partially-occluded state . Formation of a cytoplasmic-facing fully-open state would then involve the cytoplasmic ends of TMs 3 and 6 moving away from the protein core ( Figure 7 ) . After substrate exchange , it’s possible that similar conformational transitions would happen in the reverse order to ferry a nucleotide sugar to the Golgi lumen . Cytoplasmic concentrations of CMP and CMP-Sia have been estimated to be between 5–44 µM for CMP and 75–450 µM for CMP-Sia ( Chiaramonte et al . , 2001; Briles et al . , 1977; Hauschka , 1973; Nakajima et al . , 2010; Traut , 1994 ) . On the other hand , lumenal concentrations of NMPs are much higher . CMP concentrations in particular have been estimated to be in the range of 0 . 167–2 . 1 mM ( Fleischer , 1981; Waldman and Rudnick , 1990 ) . This concentration gradient of NMPs across the Golgi membrane accounts for the observation that CST , like many other NSTs , is a secondary active transporter and can concentrate nucleotide sugars inside the Golgi lumen ( Hirschberg et al . , 1998 ) . This is an important property of this system as many glycosyltransferases have Km’s for nucleotide sugars that can approach mM concentrations and sialyltransferase Km’s in particular range from 0 . 05 to 3 . 2 mM ( Gupta et al . , 2016 ) . The kinetics of substrate efflux from the Golgi through CST are not well defined; however , high lumenal CMP concentrations coupled with CMP’s relatively high binding affinity for the partially-occluded state will ensure efficient conversion of CST back to a cytoplasmic-facing state . Although reported values for the Ki of CMP inhibition of CMP-Sia uptake ( Chiaramonte et al . , 2001; Tiralongo et al . , 2000 ) are similar to reported Km values of CMP-Sia uptake ( Aoki et al . , 2003; Chiaramonte et al . , 2001; Milla and Hirschberg , 1989; Tiralongo et al . , 2006 ) , the approximately 10-fold higher cytoplasmic concentrations of CMP-Sia will ensure that CMP-Sia is transported into the Golgi lumen at a high fraction of Vmax . These structures of mCST represent the first high-resolution structures of an NST in complex with its physiological substrates . Analysis of the structures and homology models of related NSTs explains the elegant mechanisms through which NSTs discriminate between different nucleotides and nucleotide-sugar conjugates . Comparing mCST structures to other DMT protein structures provides insight into the overall transport mechanism and explains the mechanisms of obligatory antiport and trans-stimulation , which is further supported by characterization of the differential binding affinities of CMP and CMP-Sia as well as by identifying the tryptophan that senses CMP-induced conformational changes in mCST . These analyses explain how disease-causing mutations in human NST genes affect transport activity and provide a framework for pharmacologically targeting NSTs . Full-length mCST cDNA was synthesized ( Bio Basic ) and subcloned into the Pichia pastoris expression vector pPICZ with a C-terminal PreScission protease site , followed by green fluorescent protein ( GFP ) , and then a His10 tag . The mCST∆C construct was generated by modifying the full-length construct using site-directed mutagenesis to remove the last 15 residues ( 322-336 ) . Full-length and mutant mCST constructs were expressed in P . pastoris and purified as previously described ( Whorton and MacKinnon , 2011; Whorton and MacKinnon , 2013 ) with a few modifications . Milled cells were solubilized for 1 . 5 hr at 4°C in the following buffer: 50 mM HEPES pH 7 . 5 , 150 mM NaCl , 0 . 01 mg/ml deoxyribonuclease I , 0 . 7 μg/ml pepstatin , 1 μg/ml leupeptin , 1 μg/ml aprotinin , 1 mM benzamidine , 0 . 5 mM phenylmethylsulfonyl fluoride , and 2% ( w/v ) n-dodecyl-β-D-maltopyranoside ( DDM ) ( Anatrace , solgrade ) . The lysate was centrifuged at 35 , 000 g for 35 min at 4°C to pellet the unsolubilized material . The clarified supernatant was pooled and the pH was adjusted to 7 . 2 with 5 M NaOH , then added to Talon resin ( 0 . 175 ml/g of cells; Clontech ) pre-equilibrated in Buffer A ( 50 mM HEPES pH 7 . 5 , 150 mM NaCl , 0 . 1% ( w/v ) DDM ( solgrade ) ) and incubated at 4°C for 2 hr under gentle rotation . 5 mM imidazole was added during binding to the Talon resin . The resin was washed in batch with five column volumes ( cv ) of Buffer A with 5 mM imidazole by pelleting at 1250 g for 5 min and re-suspending in fresh buffer . Washed resin was loaded onto a column and further washed with five cv Buffer A + 20 mM imidazole , then two cv Buffer A + 40 mM imidazole at about 1 ml/min using a peristaltic pump . The column was then eluted with Buffer A + 300 mM imidazole . Peak fractions were pooled and 1 mM DTT , 1 mM EDTA was added . PreScission protease was added at 1 μg protease per 20 μg of protein to cut the C-terminal GFP tag overnight at 4°C . The cleaved protein was concentrated in a 50 K MWCO concentrator ( Millipore ) to run on a Superdex 200 gel filtration column ( GE Healthcare ) in Buffer B ( 25 mM HEPES pH 7 . 5 , 150 mM NaCl , 0 . 1% ( w/v ) DDM ( anagrade ) , 5 mM DTT , and 1 mM EDTA ) . We characterized the oligomeric state of purified human CST ( hCST ) as it was initially one of our most promising constructs but did not grow well-diffracting crystals . Compared to mCST , it has an identical migration time on a gel filtration column ( data not shown ) ; therefore , we expect that the results from the following experiments will also be applicable to mCST . The molecular weight of the protein component of the purified hCST-lipid-detergent complex , which has a theoretical molecular weight of 37 kDa , was determined with the ‘SEC-UV/LS/RI’ approach using the ‘three detector’ method ( Arakawa et al . , 1992; Folta-Stogniew , 2006; Hayashi et al . , 1989 ) . 250 μl of purified hCST at 0 . 8 mg/ml was loaded onto a Superdex 200 column equilibrated in Buffer B . The output from the column was then passed through UV absorbance , light scattering , and refractive index detectors . The chromatography and analysis of the data were performed at the Biophysics Resource of Keck Facility at Yale University . For the cross-linking reactions , either disuccinimidyl suberate ( DSS; Thermo Fisher Scientific ) or bis ( sulfosuccinimidyl ) suberate ( BS3; Thermo Fisher Scientific ) were added at the indicated concentrations to 1 mg/ml purified hCST in Buffer A . The reactions were incubated for 30 min at room temperature ( RT ) and then quenched by adding 1 M Tris-HCl pH 7 . 5 to a final concentration of 100 mM . Samples were then incubated for 15 min at RT before being run on a 12% SDS-PAGE gel . CMP-Sia ( Carbosynth ) purity was assessed by reverse phase HPLC using an established protocol ( Nakajima et al . , 2010 ) with some modifications . Briefly , an XSelect CSH C18 column ( 3 . 5 μm , 2 . 1 × 150 mm; Waters ) was first equilibrated for at least 30 min ( at 0 . 2 ml/min ) in 70% Buffer C ( 0 . 1 M KPO4 pH 6 . 5 , 8 mM tetrabutylammonium hydrogensulfate ( TBHS ) ) and 30% Buffer D ( 100% acetonitrile ) . A sample method involved first running 100% Buffer C for 10 . 4 min , then a 0–30% Buffer D linear gradient for 1 . 6 min , and then maintaining this mobile phase for 15 min before running a linear gradient back to 100% Buffer C over 1 min , and then maintaining this mobile phase for 6 min , all at 0 . 2 ml/min . For a series of experiments , we would first pre-condition the column by twice injecting 10 µl of Buffer C . We would then inject 10 μl of each CMP-Sia sample . The flow-through was passed through an absorbance detector set to a wavelength of 270 nm to maximize the signal for cytidine-containing compounds . Sialic acid eluted in the void volume and was not detected under these conditions . All samples were kept at 4°C until immediately before injection onto the column . All buffers and chromatographic steps were at RT . Antarctic phosphatase ( AnP , New England Biolabs ) was added to CMP-Sia samples to convert all free CMP to cytidine . This typically involved first concentrating the AnP while also reducing the glycerol concentration by diluting to 100 U/ml in AnP buffer ( 50 mM Bis-Tris-Propane-HCl , pH 6 . 0 , 0 . 1 mM ZnCl2 and 1 mM MgCl2; New England Biolabs ) and then concentrating to 10 , 000 U/ml with a 10 K MWCO concentrator ( Millipore ) . For co-crystallization experiments , CMP-Sia powder was dissolved with a 6000 U/ml AnP solution to a final concentration of 413 mM . Additional ZnCl2 and MgCl2 were added to a final concentration of 0 . 2 mM and 2 mM , respectively , and this mixture was then incubated for 28 hr at RT . After confirming the absence of CMP using the HPLC assay ( described above ) , this stock was then used immediately for crystallization experiments . For binding and transport assays , CMP-Sia powder was dissolved into a 10 , 000 U/ml AnP solution to a final concentration of 100 mM . Additional ZnCl2 and MgCl2 were added to a final concentration of 0 . 2 mM and 2 mM , respectively , and this mixture was then incubated for 8 hr at RT . The AnP was then removed by filtering this mixture through a 3 K MWCO concentrator ( Millipore ) , which retained the 70 kDa AnP . We determined that this step removed any detectible AnP activity from our stocks ( discussed below ) . The elimination of CMP from CMP-Sia stocks was confirmed by HPLC analysis ( described above ) before aliquots were frozen and stored at −80°C . The rate of CMP production by CMP-Sia hydrolysis in an AnP-treated CMP-Sia sample , after the AnP was removed , was determined in conditions that mimic our binding and transport assay conditions . This was done by incubating 10 mM AnP-treated CMP-Sia in either a buffer at pH 6 . 5 ( 0 . 1 M KPO4 pH 6 . 5 ) or a buffer at pH 7 . 5 ( 20 mM HEPES pH 7 . 5 and 0 . 15 M NaCl ) for various times at RT , as shown in Figure 1—figure supplement 3D–F . These samples were then kept on ice before diluting to 2 mM in Buffer C and then running on a C18 HPLC column ( at RT ) as described above . The fraction of each species present in a sample was determined by dividing the area of each peak ( at the following elution times: ~3 min for CMP; ~3 . 8 min for cytidine; and ~5 min for CMP-Sia ) by the total combined area of the CMP , cytidine , and CMP-Sia peaks . This method of species quantitation was used to account for small differences in injection volumes between runs . The slope of a linear regression fit of the fraction of each species as a function of time gave the rate of species production/degradation for a given pH . The fraction of CMP present at 0 min is due to a combination of CMP not removed during AnP treatment as well as CMP that is produced during the ~5 min that the sample spends at RT in the pH 6 . 5 mobile phase during the HPLC run before eluting . These two sources were differentiated by first calculating the amount of CMP produced during the HPLC run using the rate of CMP produced at pH 6 . 5 as determined above . The residual amount of CMP was taken to have originated from incomplete removal during AnP treatment . The results from this analysis are shown in Figure 1—figure supplement 3G–H . The starting CMP-Sia fraction of 88 . 68% was used to determine the actual concentration of CMP-Sia in any given sample of AnP-treated CMP-Sia . The relatively slow rate of CMP-Sia degradation ( −0 . 0049 %/min ) means that the concentration would only change by 0 . 15% in 30 min at RT , which was not considered significant enough for analyzing results from transport and binding experiments . On the other hand , while the rate that CMP is produced from CMP-Sia hydrolysis is nearly as slow ( 0 . 0051 %/min ) , this still results in a non-negligible amount of CMP . For transport assays , the relatively low concentrations and short incubation times limit the total CMP concentration to 48 nM at the end of the time course shown in Figure 1B and only 26 nM at the highest concentration of CMP-Sia used in the titration shown in Figure 1C . As these concentrations are at least 100-fold lower than the reported Ki for CMP inhibition of CMP-Sia transport ( Chiaramonte et al . , 2001 ) , these concentrations of CMP were not considered significant . However , the higher concentrations and long counting times for the SPA assay resulted in higher concentrations of CMP contamination , ranging from 22 nM to 65 µM . We discuss how this was taken into account in the SPA assay section below . Finally , the rate of cytidine production was determined to be essentially 0 %/min , considering the error on the measurement , which demonstrates that the filtration step to remove AnP , described above , is sufficient to eliminate all AnP activity . For hanging-drop vapor-diffusion crystallization ( HDVD ) experiments , purified mCST was first concentrated to 5 mg/ml using a 50 K MWCO concentrator ( Millipore ) and then CMP ( Sigma ) was added to 20 mM . This sample was then mixed 1:1 with 200 nl of the crystallization solution ( 25 . 6–26 . 8% PEG 400 , 0 . 1 M Tris pH 8 . 5 , and 0 . 1 M magnesium acetate ) and incubated at 20°C . Rectangular rod-shaped crystals typically appeared after about five days and would continue to grow for approximately eight more days . The crystals were cryoprotected by increasing the concentration of PEG400 in the well up to 32–35% and incubating for about 16–24 hr . The crystals were then harvested directly from the drop and flash-frozen in liquid N2 . For lipidic cubic phase ( LCP ) crystallization experiments , mCST∆C was first purified in Buffer B with 0 . 03% DDM , then concentrated to 20 mg/ml , and then either CMP was added to 20 mM , or AnP-treated CMP-Sia was added to 18 . 4 mM ( additional AnP was also added to bring the final concentration to 390 U/ml ) . Either of these samples was then mixed 2:3 with monoolein ( Nu-Chek Prep ) and then 70 nl of this material was deposited on a glass slide ( Molecular Dimensions ) . 600 nl of the crystallization solution ( 26 . 7–30% PEG 300 , 0 . 1 M MES pH 6 . 5 , and 0 . 1 M NaCl for CMP; 28 . 05% PEG 350 MME , 40 mM Tris pH 8 . 0 , and 40 mM NaCl for CMP-Sia ) with either 10 mM CMP or 8 mM CMP-Sia ( additional AnP was also added to bring the final concentration to 173 U/ml in 600 nl ) was then added . The drop was sealed with a glass slide on the top and incubated at 20°C . Long needle-shaped crystals typically appeared after 1–2 days and would grow for another 2–4 days . The crystals were then harvested directly from the drop and flash-frozen in liquid N2 . All diffraction data were processed with XDS ( Kabsch , 2010 ) and further analyzed using Pointless ( Evans , 2011 ) and Aimless ( Evans and Murshudov , 2013 ) . CCP4i ( Winn et al . , 2011 ) was used for project and job organization . Diffraction data for the HDVD mCST-CMP native crystals were collected at the Advanced Photon Source ( APS ) beamline 23ID-B and the Advanced Light Source ( ALS ) beamline 5 . 0 . 2 , using X-ray wavelengths of 1 . 03313 Å and 1 . 0 Å , respectively . Data from five different crystals were merged using XSCALE . The diffraction data were highly anisotropic , so they were truncated and scaled using the UCLA anisotropy server ( Strong et al . , 2006 ) . Data for Hg and Pt derivatives were collected at the APS 21-D and ALS 5 . 0 . 2 beamlines , respectively , using 1 . 0052 Å and 1 . 07234 Å wavelengths , respectively . These derivatives were prepared by adding ~0 . 5 µl of 10 mM solutions of either CH3HgCl or ( NH4 ) 2PtCl4 , prepared in a buffer that mimicked the mother liquor , to crystal drops and soaking for either 10 min or 24 hr , respectively before harvesting . Hg or Pt sites were located with HKL2Map ( Pape and Schneider , 2004 ) and the SHELX ( Sheldrick , 2008 ) suite of programs . Initial phases were determined using the MIRAS method in SHARP ( Bricogne et al . , 2003 ) . Phases were extended and improved using solvent flattening , histogram matching , and 2-fold non-crystallographic symmetry averaging using the program DM ( Cowtan , 1994 ) . An atomic model was built using Coot ( Emsley et al . , 2010 ) and improved through iterative cycles of refinement using Refmac ( Murshudov et al . , 2011 ) . The register for nearly all of the TMs could be determined either from the presence of a labeled cysteine or bulky side chains . Diffraction data for the LCP crystals were collected at the APS 23ID-B and 23ID-D beamlines using a wavelength of 1 . 0332 Å . Data from 26 different crystals were merged using XSCALE . The LCP mCST-CMP crystal structure was solved by molecular replacement using Phaser ( McCoy et al . , 2007 ) , using the HDVD mCST-CMP crystal structure model as a search model . Iterative cycles of model building in Coot and refinement using Refmac were used to add more than 40% additional atoms to the model , including CMP . The model includes residues 15–317 , but lacks residues 164–167 of the loop between TMs 5 and 6 . This model was also then refined against the HDVD mCST-CMP dataset , which yielded the model shown in Figure 2—figure supplement 2A–C . The statistics for this model are shown in Table 2 . The LCP mCST-CMP-Sia crystal structure was also solved by molecular replacement using Phaser , using the LCP mCST-CMP crystal structure model as a search model . Iterative cycles of model building in Coot and refinement using Refmac were used to build the more ordered N-terminus and to add CMP-Sia . The model includes residues 7–317 , but lacks residues 161–167 of the loop between TMs 5 and 6 . All models were validated using MolProbity ( Chen et al . , 2010 ) as implemented in Phenix ( Adams et al . , 2010 ) . The HDVD mCST-CMP , LCP mCST-CMP , and LCP mCST-CMP-Sia models had 98 . 3 , 99 . 3 , and 98 . 0% , respectively , of their residues in the preferred region of a Ramachandran plot and no outliers . Figures were prepared using PyMOL ( Schrodinger , 2015 ) . The purification protocol for mCST constructs used for transport assays was altered slightly to achieve higher purity . This involved dialyzing the PreScission protease digestion overnight at 4°C against Buffer A to remove the imidazole . The cleaved protein was then run over a 1 ml Talon column equilibrated in Buffer A . The flow through as well as two subsequent one cv washes were then concentrated and run on a Superdex 200 column as described above . Purified constructs were then incorporated into lipid vesicles comprised of yeast polar lipid ( YPL ) extract ( Avanti Polar Lipids ) . A typical reconstitution entailed first drying down 40 mg of YPL from a chloroform suspension using a stream of argon , followed by two washes with pentane and then placing the dried film of lipids in a vacuum desiccator overnight . The lipid film was re-suspended at 11 . 1 mg/ml in Buffer E ( 20 mM HEPES pH 7 . 5 and 0 . 1 M KCl ) and then sonicated to form small unilamellar vesicles . DDM was then added to a final concentration of 5 mM and incubated for 1 hr at RT . 200 μg purified mCST ( or an equivalent volume of the gel filtration buffer for protein-free vesicles ) , either with or without 300 μM CMP ( final concentration ) , was then added . Additional gel filtration buffer was added to bring the final lipid concentration to 10 mg/ml and the mixture was then incubated for 1 hr at 4°C . Proteoliposomes were formed by removing DDM with three successive additions of Bio-Beads ( SM-2; Bio-Rad ) at 100 mg/ml . For the first two additions , the Bio-Beads were incubated for 2 hr at 4°C under gentle rotation followed by a third incubation overnight . The liposomes were then aliquoted , flash frozen in liquid N2 , and stored at −80°C . We were typically able to incorporate about 60% of the starting amount of protein into the vesicles , which we determined by solubilizing small amount of vesicles with DDM and then running on a gel filtration column ( data not shown ) . For transport assays , vesicles were thawed and extruded 10 times through a 0 . 4 μm Whatman Nuclepore filter ( GE Healthcare ) . For a typical reaction , 45 µl of extruded vesicles was used . External CMP was removed by pelleting the vesicles by ultracentrifugation ( 194 , 800 g for 60 min ) , removing the supernatant , washing the pellet with Buffer E , and then resuspending the vesicles in 20 µl of cold Buffer E and kept on ice . Transport was initiated by bringing the volume of the vesicles to 50 µl using 30 µl of RT Buffer E containing the indicated concentration of CMP-Sia which contained 30–50 nM [3H]CMP-Sia ( 20 Ci/mmol; American Radiolabeled Chemicals ) . The mixture was then incubated at RT for the indicated time for time-course experiments or for 30 s to determine the initial velocity for substrate titration experiments . Transport was then stopped by adding 0 . 6 ml ice-cold Buffer E and storing on ice . The transport reactions were then filtered through 0 . 22 μm mixed cellulose ester membranes ( Millipore ) and washed with 3 × 2 ml ice-cold Buffer E . Counts from protein vesicles at 0 min and 4°C ( similar to background counts from reactions with protein-free vesicles ) were subtracted from the total counts from the protein-containing vesicles to determine specific counts . Transport data were fit to a Michaelis-Menten model to determine Km and Vmax . For scintillation proximity assays ( SPA ) , mCST constructs were purified as described above , except the GFP-His10 tag was not removed and DTT and EDTA were eliminated from Buffer B during the final size exclusion chromatography step . A typical binding experiment involved combining 2 μM purified protein , 0 . 4 mg Copper HIS-Tag PVT SPA beads ( PerkinElmer ) , 30 nM [3H]CMP ( 20 Ci/mmol; American Radiolabeled Chemicals ) and various concentrations of cold substrates in a final volume of 50 μl . Non-specific counts were determined by setting up identical reactions without protein . The protein was added to beads for 30 min on ice to allow the protein to bind the beads before the ligands were added . The reactions were prepared in 96-well , white clear-bottom Isoplates ( PerkinElmer ) and incubated for 5 min on ice followed by a minimum incubation of approximately 30 min at RT before being counted in a Wallac MircoBeta TriLux scintillation counter ( PerkinElmer ) using the ‘SPA-counting mode . ’ For the CMP titration , the relatively high affinity of CMP meant that we had to titrate CMP close to the concentration of mCST used in the assay , so the following homologous binding model that takes into account ligand depletion was used to determine the fraction of CMP-bound mCST ( fP ) : ( 1 ) fP= S+S′+K+P− ( S+S′+K+P ) 2−4SP2Pwhere S is the concentration of cold CMP that was titrated , S’ is the concentration of [3H]CMP , K is the Kd for CMP , and P is the concentration of mCST . K was determined by fitting the experimental data to this model . Ligand depletion only had a minor effect on the interpretation of our results as Kd’s that were determined with this model only differed by approximately 1 . 3-fold from IC50’s determined from a simple one-site binding model . For other titrations , the fraction of mCST bound to the indicated substrate ( fP ) was modeled using the following equation: ( 2 ) fP=S1S1+KS1 ( 1+S2KS2 ) where S1 and S2 are the concentrations of substrate 1 ( the titrated substrate ) and substrate 2 ( either just [3H]CMP or the combination of [3H]CMP and cold CMP generated from CMP-Sia hydrolysis , which is discussed below ) , respectively . KS1 and KS2 are the Kd’s of substrate 1 and 2 , respectively . KS2 is the Kd for CMP , which was determined using Equation 1 as described above . KS1 was determined by fitting the experimental data to this model . For CMP-Sia titrations , the final concentration of contaminating CMP at each assay point was determined by multiplying the rate of CMP production , as determined above , by the total time that elapsed between when the assay plate was placed at RT to when an individual well was counted . This elapsed time typically varied from between 30–60 min , as each well was counted for 1 min . While the concentration of CMP in some wells was relatively high , the Kd’s that we determined through this method only differed by approximately 2-fold from IC50’s determined from a simple one-site binding model where the generation of CMP from CMP-Sia hydrolysis is ignored . The effect of CMP on the intrinsic tryptophan fluorescence of wild-type and mutant mCST was measured using a Photon Technology International dual-monochromator fluorometer . Since cytidine’s absorption spectrum ( λmax=270 ) partially overlaps that of tryptophan’s ( λmax=280 ) , this leads to an apparent fluorescence quenching termed the inner-filter effect ( Lakowicz , 2006 ) . To account for this , we first minimized this effect by using a slightly higher excitation wavelength ( 300 nm ) that is not as well absorbed by cytosine . What little inner-filter effect remained could be measured and corrected for by performing the CMP titration in the presence of mCST denatured with sodium dodecyl sulfate ( SDS ) . A typical experiment involved putting 400 μl of 6 μM mCST into a quartz cuvette and then recording the emission spectrum from 315 to 500 nm . Subsequent spectra would then be recorded after adding 0 . 4–2 . 5 μl of a concentrated stock of CMP to achieve the desired concentration . An identical titration was performed in the presence of 2 . 5% SDS . A spectrum collected from a buffer-only sample was used for subtracting the background fluorescence from the spectra of protein samples . The λmax for the tryptophan fluorescence of mCST was around 328 nm and did not change as a function of CMP concentration . Peak height was determined by averaging the values from 321 to 335 nm . To correct for the inner-filter effect , a correction factor for each CMP concentration was first calculated by dividing the peak height of the SDS-treated sample without CMP by the peak heights of each SDS-treated sample with various concentrations of CMP . The peak heights for the native protein samples were then multiplied by this correction factor . Correction values were typically very small up to 700 μM CMP ( ranging between 1–1 . 03 ) and increased slightly to ~1 . 2 for 7 mM CMP . The corrected peak heights were then used to calculate fractional quenching as a function of CMP concentration . The quenching data were fit to a version of Equation 1 that lacked the S’ terms to determine the Kd’s for CMP . AZTMP was manually docked into the mCST-CMP structure . Structural homology models of UGT and NGT were generated using the SWISS-MODEL web server ( Waterhouse et al . , 2018 ) , with the mCST-CMP-Sia structure used as a template . UDP-sugars were docked into these models using AutoDock4 ( Morris et al . , 2009 ) with the flexible side chain covalent docking method , as previously described ( Bianco et al . , 2016 ) . To do this , UMP was first placed in the same pose that CMP is found in the mCST-CMP-Sia structure and it was set to act as part of the rigid component of the protein . This greatly reduced the computation required and is justified by the nearly-complete conservation of the residues that comprise the nucleotide binding site in mCST , UGT , and NGT . The sugar moiety of the UDP-sugar was then treated as a flexible residue . The residues that line the sugar binding site were also set to be flexible while the rest of the protein was kept rigid . The ligand poses shown in Figure 3—figure supplement 2D–F are the lowest-energy poses out of 1000 docking runs . Atomic coordinates and structure factors are available in the Protein Data Bank ( PDB ) with entries 6OH4 ( HDVD mCST-CMP ) , 6OH2 ( LCP mCST-CMP ) , and 6OH3 ( LCP mCST-CMP-Sia ) .	The cells in our body are tiny machines which , amongst other things , produce proteins . One of the production steps involves a compartment in the cell called the Golgi , where proteins are tagged and packaged before being sent to their final destination . In particular , sugars can be added onto an immature protein to help to fold it , stabilize it , and to affect how it works . Before sugars can be attached to a protein , they need to be ‘activated’ outside of the Golgi by attaching to a small molecule known as a nucleotide . Then , these ‘nucleotide-sugars’ are ferried across the Golgi membrane and inside the compartment by nucleotide-sugar transporters , or NSTs . Humans have seven different kinds of NSTs , each responsible for helping specific types of nucleotide-sugars cross the Golgi membrane . Changes in NSTs are linked to several human diseases , including certain types of epilepsy; these proteins are also important for dangerous microbes to be able to infect cells . Yet , scientists know very little about how the transporters recognize their cargo , and how they transport it . To shed light on these questions , Ahuja and Whorton set to uncover for the first time the 3D structure of a mammalian NST using a method known as X-ray crystallography . This revealed how nearly every component of this transporter is arranged when the protein is bound to two different molecules: a specific nucleotide , or a type of nucleotide-sugar . The results help to understand how changes in certain components of the NST can lead to a problem in the way the protein works . Ultimately , this knowledge may be useful to prevent diseases linked to faulty NSTs , or to stop microbes from using the transporters to their own advantage .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics" ]	2019	Structural basis for mammalian nucleotide sugar transport
Preterm infants that suffer cerebellar insults often develop motor disorders and cognitive difficulty . Excitatory granule cells , the most numerous neuron type in the brain , are especially vulnerable and likely instigate disease by impairing the function of their targets , the Purkinje cells . Here , we use regional genetic manipulations and in vivo electrophysiology to test whether excitatory neurons establish the firing properties of Purkinje cells during postnatal mouse development . We generated mutant mice that lack the majority of excitatory cerebellar neurons and tracked the structural and functional consequences on Purkinje cells . We reveal that Purkinje cells fail to acquire their typical morphology and connectivity , and that the concomitant transformation of Purkinje cell firing activity does not occur either . We also show that our mutant pups have impaired motor behaviors and vocal skills . These data argue that excitatory cerebellar neurons define the maturation time-window for postnatal Purkinje cell functions and refine cerebellar-dependent behaviors . Abnormal cerebellar development instigates motor diseases and neurodevelopmental disorders including ataxia , dystonia , tremor , and autism . These conditions have a high incidence in premature infants as well as in newborns with cerebellar hemorrhage ( Dijkshoorn et al . , 2020; Limperopoulos et al . , 2007; Steggerda et al . , 2009; Zayek et al . , 2012 ) . The affected children ultimately attain a smaller cerebellar size compared to children born full-term ( Limperopoulos et al . , 2005; Volpe , 2009 ) , and they have altered functional connectivity between the cerebellum and the neocortex ( Herzmann et al . , 2019; Hortensius et al . , 2018 ) . Observations from human clinical data indicate a strong correlation between abnormalities in cerebellar size and the subsequent incidence of cognitive disorders , suggesting that the late-term stage of cerebellar development encompasses a critical developmental time-window for establishing not only the motor but also the non-motor circuitry that is required for complex cerebellar functions and behavior ( van der Heijden et al . , 2021b; Volpe , 2009; Wang et al . , 2014 ) . The cerebellum undergoes a series of dynamic developmental changes during the third trimester and early postnatal stages , a period that corresponds to the first 2 postnatal weeks in mice ( Sathyanesan et al . , 2019 ) . At the level of gross morphology , the rapid proliferation of granule cell precursors and the integration of granule cells into the cerebellar circuit result in a fivefold enlargement of cerebellar size ( Chang et al . , 2000 ) , an increase in foliation complexity ( Corrales et al . , 2006 ) , and the formation of the tri-laminar cerebellar cortex ( Miyata et al . , 2010 ) . At the cellular level , the differentiation and arrival of mature granule cells into the circuit modifies the laminar localization and synaptic specification of cerebellar afferents that project to Purkinje cells ( van der Heijden and Sillitoe , 2021 ) . Despite an extensive literature describing the temporal anatomical changes in the cerebellum and studies postulating that perturbations to the cerebellum during this dynamic period may lead to cognitive , social , and motor deficits ( Volpe , 2009; Wang et al . , 2014 ) , there is still relatively little known about how the anatomical aspects of cerebellar development correspond to its functional maturation programs that establish animal behavior . Previous studies show that preterm birth in pigs , as well as postnatal hemorrhage and hypoxia ( which mimics cerebellar hemorrhage ) in mice , result in decreased granule cell numbers ( Iskusnykh et al . , 2018; Yoo et al . , 2014 ) and abnormal motor control ( Sathyanesan et al . , 2018; Yoo et al . , 2014 ) . Importantly , postnatal hypoxia causes impairments in the spontaneous firing properties of Purkinje cells in mice ( Sathyanesan et al . , 2018 ) , and impaired Purkinje cell firing properties are also observed in prematurely born baboons ( Barron and Kim , 2020 ) . It is not surprising that decreasing the number of granule cells results in abnormal Purkinje cell function , since each Purkinje cell may integrate inputs from up to two hundred fifty thousand excitatory granule cell synapses ( Huang et al . , 2014; Napper and Harvey , 1988 ) . Nevertheless , genetic silencing of granule cells causes only modest alterations to the baseline firing properties of Purkinje cells ( Galliano et al . , 2013 ) , perhaps because the predominant Purkinje cell action potential , called the simple spike , is spontaneously generated by the Purkinje cells ( Raman and Bean , 1999 ) . In line with this hypothesis , gross motor control is not affected in mice with impaired granule cell signaling , although they do have impaired motor learning ( Galliano et al . , 2013 ) . In contrast , some mouse models that lack the majority of granule cells ( agranular mice ) have such severe motor defects that these mice are named after their phenotype; for example scrambler , weaver , reeler , and staggerer ( Bolivar et al . , 1996; Falconer , 1951; Heuzé et al . , 1997; Le Marec et al . , 1997; Sheldon et al . , 1997 ) . The discordance between the functional outcomes in mutant mice with reduced granule cell numbers and mice lacking granule cell signaling raises the question whether granule cell neurogenesis , rather than granule cell synaptic signaling , contributes more to the maturation of Purkinje cell firing properties in vivo . To probe these cellular interactions , we manipulated the mouse cerebellum by genetically blocking neurogenesis of excitatory neurons . We used an En1Cre allele ( Wurst et al . , 1994 ) to delete the proneural gene , Atoh1 , from the embryonic hindbrain . Atoh1 is necessary for the development of excitatory cerebellar neurons ( Ben-Arie et al . , 1997; Rose et al . , 2009 ) and is not expressed in the inhibitory Purkinje cells . In this agranular cerebellar model , we test how Purkinje cells develop their structure , connectivity , and function using immunostaining , neural tract tracing , and in vivo electrophysiology performed in the second postnatal week . We also examined motor function and vocal skills in the mutant pups to test how the structural and functional changes in the cerebellum impact the expression of behaviors that are usually acquired during the first two weeks of life . To test the hypothesis that granule cell neurogenesis is essential for the functional development of Purkinje cells , we first needed to establish an appropriate agranular mouse model . In previously generated models , the mice lack granule cells due to spontaneously occurring mutations in genes with widespread expression patterns ( Falconer , 1951; van der Heijden and Sillitoe , 2021; Sheldon et al . , 1997 ) . Therefore , one could not differentiate cell-extrinsic from cell-intrinsic effects in Purkinje cells ( Dusart et al . , 2006; Gold et al . , 2007 ) . To impair granule cell neurogenesis in a manner independent of genes expressed in Purkinje cells , we made use of the distinct origins of Purkinje cells and granule cells to establish a new model of agranular mice ( Figure 1A–B; Hoshino et al . , 2005; Rose et al . , 2009 ) . Atoh1 is necessary for the development of excitatory cerebellar neurons , including granule cells ( Ben-Arie et al . , 1997 ) , which are the most populous cell type in the cerebellum and make up at least 99% of all neurons in the cerebellum ( Consalez et al . , 2020 ) . En1 is a homeodomain transcription factor that is expressed in the mesencephalon and rhombomere 1 by embryonic day 8 ( E8 ) , where it is required for the formation of the cerebellum ( Davis and Joyner , 1988; Wurst et al . , 1994 ) . The intersection of the Atoh1 and En1 lineages converges on excitatory neurons that originate from the rostral rhombic lip ( rRL ) ( Wang et al . , 2005 ) , but does not include inhibitory neurons like Purkinje cells that originate in the ventricular zone ( Hoshino et al . , 2005 ) . We demonstrate this selectivity by conditional expression of TdTomato , which is expressed by an intersectional reporter allele that is dependent on both Cre- ( En1 [Wurst et al . , 1994] ) and FlpO- ( Atoh1 [van der Heijden and Zoghbi , 2018] ) mediated excision of a stop-cassette ( Figure 1C ) . As expected , we observed TdTomato expression in the granule cell layer and molecular layer of the cerebellar cortex , where the axons ( parallel fibers ) of granule cells reside , but not in the Purkinje cell layer ( Figure 1C ) . Our rationale for choosing to impair granule cell neurogenesis by employing a conditional deletion of Atoh1 from the En1 domain was because Atoh1 null mice are neonatal lethal ( Ben-Arie et al . , 1997 ) . We generated conditional knockout mice by crossing mice that are heterozygous for the En1Cre knock-in allele ( Wurst et al . , 1994 ) and heterozygous for Atoh1 with mice that are homozygous for a LoxP-flanked Atoh1 allele ( Atoh1fl/fl ) ( Shroyer et al . , 2007 ) . From this cross , we obtain control mice with one of the following three genotypes: Atoh1fl/+ mice , Atoh1FlpO/fl mice , En1Cre/+;Atoh1fl/+ mice . The different control alleles were analyzed separately for the behavioral data and subsequently grouped together for anatomical and electrophysiological analyses due to the absence of any significant behavioral differences . In addition , the cross generates our experimental En1Cre/+;Atoh1fl/FlpO mice that we will hereafter refer to as En1Cre/+;Atoh1fl/- mice because the Atoh1FlpO is a functional null-allele ( van der Heijden and Zoghbi , 2018 ) . We previously showed that these En1Cre/+;Atoh1fl/- mice are informative for assessing the period of dynamic cerebellar development ( P7-P14 ) but the animals succumb at around weaning age ( P21 ) , likely due to a combination of impaired motor coordination and breathing abnormalities ( van der Heijden and Zoghbi , 2018 ) . Using these mice , we have now investigated how the En1-mediated deletion of Atoh1 influences the period of cerebellar growth that usually occurs in the second postnatal week . We found that cerebellar size rapidly expands between P7 and P14 in control mice ( Figure 1D , cerebellum is pseudo-colored in red ) but that no such size-expansion occurs in the En1Cre/+;Atoh1fl/- mice ( Figure 1E ) . When we overlay the outlines of the cerebellum , we furthermore observe that the size of the cerebellum in En1Cre/+;Atoh1fl/- mice is markedly smaller at both P7 and P14 compared to that of similar aged controls ( Figure 1F ) . In sagittal tissue sections cut through the hindbrain and then stained with cresyl violet to visualize cell nuclei , we also observe that P7 and P14 En1Cre/+;Atoh1fl/- cerebella lack a densely packed granule cell layer , that En1Cre/+;Atoh1fl/- cerebella do not form cerebellar lobules , and that there is no substantial increase in cerebellar size that takes place during the second postnatal week , which are morphological milestones typically observed in control mice ( Figure 1G–I ) . At P14 , sagittal cerebellar tissue sections from En1Cre/+;Atoh1fl/- mice measure just 11 . 3% ( ±1% , n=12 sets ) the size of paired control tissue sections and in addition , coronal-cut cerebellar tissue sections from En1Cre/+;Atoh1fl/- mice measure just 20 . 2% ( ±3% , n=6 sets ) the size of paired control tissue sections . Moreover , as expected based on the expression patterns of the two genes we used in our manipulation , the deletion of Atoh1 from just the En1 domain did not result in any gross morphological abnormalities in brain regions other than the cerebellum , including the olfactory bulb , hippocampus , striatum , thalamus , pons , vestibular nuclei , inferior olive , and neocortex ( Figure 1—figure supplement 1 ) . Previous studies on agranular animal models have reported a spectrum of morphological changes in the cerebellum , depending on the degree of granule cell loss; a range of cerebellar size differences compared to controls , thinner to completely absent granule cell layers , and fewer or complete lack of cerebellar foliation ( compared to controls , the cerebellum in X-irradiated rats has: smaller size , thinner granule cell layer [Armstrong and Hawkes , 2001; Zanjani et al . , 1992]; staggerer and weaver mice: smaller cerebellar size , thinner granule cell layer , less foliation [Armstrong and Hawkes , 2001; Zanjani et al . , 1992]; reeler and scrambler mice: smaller cerebellar size , thinner granule cell layer , no foliation [Goldowitz et al . , 1997]; Atoh1 chimera mice: smaller cerebellar size , thinner/no granule cell layer , less/no foliation [the phenotype is variable depending on the percentage of chimerism; Jensen et al . , 2004] ) . Our En1Cre/+;Atoh1fl/- mice encompass all these malformations , without the variability that was observed in Atoh1 chimera mice ( Jensen et al . , 2004 ) or the gross abnormalities observed outside of the cerebellum that was reported for the reeler and scrambler mice ( Goldowitz et al . , 1997 ) . Therefore , our En1Cre/+;Atoh1fl/- mice encompass all of the morphological changes observed in previous agranular models , although at the more extreme end of the severity spectrum , with little experimental variability and high regional specificity . Next , we used immunohistochemistry to investigate which of the main cellular components of the cerebellum is affected in En1Cre/+;Atoh1fl/- mice . Based on the intersectional strategy we employed in our studies , we would expect a reduction in excitatory lineage cells with minimal changes to inhibitory neurons , including Purkinje cells . We show that deletion of Atoh1 prevents granule cells from reaching differentiation by staining for GABARα6 , which marks mature granule cells ( Figure 1—figure supplement 1–2A ) . However , it remained unclear whether this change in signal was due to abnormal maturation or impaired granule cell neurogenesis . To address this question , we decided to employ a genetic strategy to label all En1;Atoh1 lineage neurons in control and conditional knockout mice with an intersectional TdTomato-reporter allele ( Figure 1—figure supplement 1–3A–B ) . In this strategy , control animals are En1Cre/+;Atoh1FlpO/+;Rosalsl-lfl-TdTomato/+ mice ( as in Figure 1C ) and conditional knockout mice are En1Cre/+;Atoh1FlpO/fl;Rosalsl-lfl-TdTomato/+ mice . Within the granule cell layer of control mice , the TdTomato+ signal saturates due to the dense packing of granule cell bodies . This leads to a signal density of 99 . 6% ( ± 0 . 1% , N=2 , n=10 ) within the granule cell layer of lobule IX in control mice ( En1Cre/+;Atoh1FlpO/+;Rosalsl-lfl-TdTomato/+ ) , compared to just 43 . 4% ( ± 3 . 5% , N=3 , n=15 ) in the ventral-caudal regions of cerebella in conditional knockout mice ( En1Cre/+;Atoh1FlpO/fl;Rosalsl-lfl-TdTomato/+ ) ( Figure 1—figure supplement 1–3D ) . Despite the mainly qualitative utility of these data , this reduced signal density does not provide direct numerical insight in the quantity of En1;Atoh1 lineage neurons that escape our genetic manipulation because it does not account for the origins of the signal saturation , the nearly 50-fold size difference between control and conditional knockout cerebella , or the difference in cellular compartments present in the area of interest ( only cell bodies and dendrites in control mice , versus cell bodies , dendrites , and axons in conditional knockout mice ) . Unfortunately , the signal density in the control animals prevented us from manually counting individual En1;Atoh1 lineage neurons in control mice , but we were able to count the total number of TdTomato+ , En1;Atoh1 lineage neurons in sagittal tissue sections from conditional knockout mice . We found an average of 925 ( ± 22 , N=3 , n=12 ) TdTomato+ cells in each tissue section ( Figure 1—figure supplement 1–3A–BC–D ) . We usually obtain ~70 sagittal sections from a single conditional knockout mouse brain ( N=5 ) , leading us to estimate that the total number of En1;Atoh1 lineage cells that escape the conditional knockout of Atoh1 from the En1 lineage to be in the range of 65 • 103 ( /thousand ) neurons per mutant mouse brain . The total number of cerebellar neurons ( 99% of which are granule cells ) in control mice has previously been estimated in the range of millions ( between 35 • 106 and 70 • 106 neurons [Consalez et al . , 2020; Herculano-Houzel et al . , 2006; Surchev et al . , 2007; Vogel et al . , 1989] ) . We therefore estimate an approximate 103 ( thousand ) -fold reduction in the total number of En1;Atoh1 lineage neurons in our mutant mice , or less than 1% of the En1;Atoh1 lineage neurons escape the genetic manipulation . Consistent with the substantially reduced number of En1;Atoh1 lineage neurons in the conditional knockout mice , we observe that En1Cre/+;Atoh1fl/- conditional knockout mice have a reduced pool of excitatory unipolar brush cells , which also derive from the En1;Atoh1 lineage ( Figure 1—figure supplement 1–4A ) , compared to the density of unipolar brush cells that are localized to lobule IX and X in control mice ( Figure 1—figure supplement 2B–D ) . When we counted the number Tbr2+unipolar brush cells in control and conditional knockout mice , we found an approximately eight-fold reduction in the number of unipolar brush cells ( Figure 1—figure supplement 4C–E ) . Interestingly , none of the Tbr2+ escaper unipolar brush cells in En1Cre/+;Atoh1FlpO/fl;Rosalsl-lfl-TdTomato/+ conditional knockout mice express the TdTomato reporter ( Figure 1—figure supplement 4A ) , suggesting that these cells may derive from a different sub-lineage . We also counted the number of excitatory cerebellar nuclei neurons , that derive from the En1;Atoh1 lineage ( Figure 1—figure supplement 4B ) and found an approximate fivefold reduction in their number ( Figure 1—figure supplement 4F–H ) . Unlike the escaper Tbr2+ unipolar brush cells , cerebellar nuclei neurons in the conditional knockout mice express the TdTomato reporter , but the NFH signal appears more fibrous . Taken together , we find a significant reduction in the total number of all types of excitatory neurons that are derived from the En1;Atoh1 lineage . We also observed an abnormal localization of GABARα6 expression in granule cells ( Figure 1—figure supplement 2A ) and NFH expression in excitatory nuclei neurons ( Figure 1—figure supplement 4F–G ) , which is an interesting molecular response to progressing this far into development without Atoh1 . To confirm that our genetic strategy did not prevent the development of inhibitory neurons , specifically Purkinje cells , we stained control cerebella using a combination of Calbindin , which marks Purkinje cells , and DAPI , which binds to DNA and ubiquitously marks cell nuclei . In sagittal and coronal tissue sections cut through the cerebella from control mice , the staining again highlights the increase in size between P7 and P14 , as well as the formation of perfectly delineated dorsal-ventral layering of cells and the establishment of the characteristic folding that defines the cerebellar lobules ( Figure 1J ) . These morphological features are not evident in the cerebellum of the mutant mice . In addition to the easily visualized difference in the pattern of rostra-caudal lobules and reduced cerebellar size , the mediolateral morphological divisions of the cerebellum are not apparent and lamination is almost non-existent in P7 and P14 En1Cre/+;Atoh1fl/- mutant mice ( Figure 1K ) . In En1Cre/+;Atoh1fl/- mice , we could identify Purkinje cells ( Calbindin+ cells , Figure 1K ) , confirming that conditional deletion of Atoh1 does not block Purkinje cell neurogenesis . However , the cellular organization of Purkinje cells in cerebella from En1Cre/+;Atoh1fl/- mice is markedly different from controls . We confirmed the stunted cerebellar size and lack of lobule complexity by comparing the outline of cerebella from control to En1Cre/+;Atoh1fl/- mice ( Figure 1L ) . In line with our conditional knockout that targets neurons in the excitatory lineage , we further found that the representation of different classes of inhibitory neurons , including molecular layer interneurons , basket cells , and Golgi cells , remains robust and clearly detected by cell-type-specific markers ( Figure 1—figure supplement 2E–H ) . These data indicate that there are simultaneously occurring alterations in cerebellar morphology and cytoarchitecture in the mutant mice , both owing to the principal defect , which is the elimination of excitatory neurons . One of the hallmarks of cerebellar circuit organization are the longitudinal zones , or modules , that have a specific pattern and function ( Apps et al . , 2018 ) . While Purkinje cells differentiate into distinct molecular subtypes shortly after birth ( Wassef et al . , 1990 ) , they initially group together into cellular clusters and only form the more precise and sharply delineated longitudinal zones as the cerebellum greatly expands in size and volume ( Dastjerdi et al . , 2012; Larouche and Hawkes , 2006 ) . Here , we investigated whether Purkinje cell molecular subtypes in En1Cre/+;Atoh1fl/- mice form distinct zones . As expected , at P14 , ZebrinII+ and PLCβ4+ expressing Purkinje cells in control mice are organized in alternating zones ( Figure 2A–G ) . In contrast , ZebrinII+ and PLCβ4+ expression patterns in the En1Cre/+;Atoh1fl/- cerebellum are organized into a series of clusters rather than sharp zones . The overall pattern of clusters in the mutants resembles the architecture that is typically observed in normal neonates ( Figure 2H–N; Fujita et al . , 2012; Sugihara and Fujita , 2013 ) . Some Purkinje cells , here marked with PLCβ4+ , over-migrate beyond the boundaries of the cerebellum and into the inferior colliculus that is immediately adjacent ( arrowheads Figure 2K ) . We also observed co-expression of ZebrinII+ and PLCβ4+ in the same clusters of cells in the En1Cre/+;Atoh1fl/- mice ( arrowheads Figure 2M–N ) . In normal mice , postnatal ZebrinII+ expression sweeps over the whole cerebellum becoming expressed in all Purkinje cells by around P10 , before it segments into zones by ~P15 ( Tano et al . , 1992 ) . The prevalent overlap in what are typically complementary markers at P14 in En1Cre/+;Atoh1fl/- mice supports the hypothesis of an altered temporal progression of cerebellar development that has affected Purkinje cell patterning . The zonal organization of Purkinje cells is also essential for the formation of patterned synaptic inputs . For example , spinal cord mossy fibers send their projections to the granular layer below ZebrinII- zones ( Sillitoe et al . , 2010 ) . During development , these projections directly target Purkinje cells ( Kalinovsky et al . , 2011; Sillitoe , 2016 ) , but with the integration of granule cells into the circuit , mossy fibers displace their synapses onto granule cells ( Arsénio Nunes et al . , 1988; Pan et al . , 2009 ) . To test how this sequence of events was impacted by the lack of granule cells in En1Cre/+;Atoh1fl/- mice , we injected the anterograde tracer WGA-Alexa 555 into the lower thoracic-upper lumbar spinal cord of P12 mice ( Gebre et al . , 2012; Lackey and Sillitoe , 2020 ) . We observed WGA-Alexa 555 labeled fibers and terminals at P14 after two days of tracer transport . In controls , mossy fibers project to the cerebellar cortex in a zonal pattern that reflects the topography of ZebrinII ( Figure 2O–P; Brochu et al . , 1990; Sillitoe and Hawkes , 2002 ) . As in the controls , we found that spinocerebellar mossy fibers projected mainly to ZebrinII- domains in the En1Cre/+;Atoh1fl/- mice . However , in controls the mossy fibers occupy the domains in the granular layer that are located directly below the Purkinje cell zones , whereas in the mutants the mossy fibers remain intermingled within Purkinje cell clusters ( Figure 2O–Q ) . These results confirm that excitatory cerebellar neurons are not essential for initiating the zonal topography of mossy fiber projections that arise from the spinal cord . However , the organization of mossy fibers into poorly defined clusters in the mutants indicates that integration of excitatory neurons into the circuit is important for the architectural rearrangement and the refinement of both the Purkinje cells and afferents into zones . Previous studies suggest that decreasing excitatory input alters Purkinje cell dendrite outgrowth ( Bradley and Berry , 1976; Pan et al . , 2009; Park et al . , 2019; Takeo et al . , 2021 ) , which results in abnormal dendrite morphology in agranular mice ( Sotelo and Dusart , 2009 ) . Therefore , we next investigated the cellular morphology of Purkinje cells in control and En1Cre/+;Atoh1fl/- mice by staining mouse brains using a modified Golgi-Cox method . We found that Purkinje cells in En1Cre/+;Atoh1fl/- mice had stunted and smaller dendritic arbors compared to the controls ( Figure 3A ) , and unlike in the control cerebella , neighboring Purkinje cells did not orient their arbors in the same direction ( Figure 3A ) . Sholl analysis revealed that the Purkinje cells in the mutants were smaller with less bifurcated dendritic branches ( Figure 3B–C ) . Furthermore , spine density was significantly lower in En1Cre/+;Atoh1fl/- mice compared to the control mice ( Figure 3D–E ) , which is in agreement with previous work showing that dendrite outgrowth is dependent on the formation of excitatory synapses , the majority of which are contributed from granule cell parallel fibers in control animals ( Bradley and Berry , 1976; Pan et al . , 2009; Park et al . , 2019; Takeo et al . , 2021 ) . From these data , we conclude that Purkinje cells in the P14 En1Cre/+;Atoh1fl/- mice have less morphological complexity . This shows that in addition to causing gross morphological impairments of the cerebellum , loss of Atoh1 and a reduced number of excitatory cerebellar neurons also negatively affects the cellular level morphology of neurons . Next , we addressed whether the large reduction of excitatory cerebellar neurons changes the composition of cerebellar cortical afferents . First , we looked how the density of Vglut1+ synapses on Purkinje cells changed during the dynamic period . From P7 to P14 , a large population of excitatory ( Vglut1+ ) granule cells integrate into the cerebellar circuit and make direct contacts to Purkinje cells via their parallel fibers in the molecular layer ( Figure 4A ) . While some mossy fibers are also Vglut1+ , they do not directly contact mature Purkinje cells in control mice but instead target granule cells in the granule cell layer . Therefore , we only quantified the density of Vglut1+ synapses in the molecular and Purkinje cell layer . We saw an increase in the density of Vglut1+ synapses in the Purkinje cell layer and molecular layer from P7 to P14 in control mice ( Figure 4B , top row ) . Because we impaired the neurogenesis of Vglut1-expressing granule cells , we expected the density of Vglut1+ synapses to be lower in En1Cre/+;Atoh1fl/- mice and not increase between P7 and P14 . Indeed , we saw that the signal of Vglut1+ synapses in En1Cre/+;Atoh1fl/- mice did not change between P7 and P14 ( Figure 4B , bottom row ) and quantification showed that this signal was significantly lower at both time points in En1Cre/+;Atoh1fl/- mice compared to age-matched control mice ( Figure 4C ) . The remaining Vglut1+ synapses in the En1Cre/+;Atoh1fl/- mice likely originate from mossy fibers ( see Figure 2 ) , and the small number of granule cells and unipolar brush cells that escaped our genetic manipulation of rRL-derived cells ( Figure 1—figure supplement 2 , Figure 1—figure supplement 3 , and Figure 1—figure supplement 4 ) . However , the approximately nine-fold reduction in Vglut1+ signal in the cerebella of En1Cre/+;Atoh1fl/- mice further confirms that our conditional knockout strategy impairs the integration of granule cells into the cerebellar circuit . Taken together , we show that without excitatory cerebellar neurons , the density of Vglut1+ excitatory synapses in En1Cre/+;Atoh1fl/- mice is much lower when assessed on individual Purkinje cells ( Figure 3D–E ) as well as throughout the morphogenetically abnormal mutant cerebellum at large ( Figure 4C ) . Purkinje cells have a distinct firing profile characterized by spontaneously generated simple spikes and climbing fiber-induced complex spikes . In vitro recordings in rats ( Crepel , 1972; McKay and Turner , 2005 ) and mice ( Beekhof et al . , 2021 ) show that Purkinje cell firing properties change significantly during early postnatal development ( P7-P14 ) , but it is unclear how firing patterns evolve in vivo during this dynamic period of rewiring . Therefore , we set out to answer two questions: first , how do the in vivo Purkinje cell firing patterns change during the dynamic period ? Second , is Purkinje cell firing affected in En1Cre/+;Atoh1fl/- mice that have a reduced number of Vglut1+ excitatory inputs ? We hypothesized that the reduced number of Vglut1+ excitatory inputs affects the modulation of simple spike firing rate . Based on the time points we tested , and in line with the highest Vglut1+ density in P14 control mice , Purkinje cells recorded from these mice had the highest peak firing frequency , whereas Purkinje cells in P7 control mice had an overall lower firing rate ( Figure 5A–B ) . Interestingly , Purkinje cells in P14 controls fired in burst-like patterns that have unique powerfully represented features in their firing frequency and frequency mode ( or preferred frequency ) ( Figure 5B and D ) , whereas Purkinje cells in P7 control mice fired more continuously , yet at a lower frequency and with higher variability in their inter-spike intervals ( Figure 5A and E ) . We quantified the firing features with six parameters: frequency ( total spikes/recording time; Figure 5F ) , frequency mode ( most frequently observed frequency; Figure 5G ) , CV ( a measure for global regularity; Figure 5J ) , pause percentage ( defined as a discrete portion of the trace where no spikes were observed; Figure 5K ) , rhythmicity index ( the oscillatory properties in the auto-correlation of spike times; Figure 5L–N ) , and CV2 ( inter-spike interval irregularity; Figure 5O; Holt et al . , 1996 ) . Specifically , a faster firing pattern is represented as a higher frequency and frequency mode ( Figure 5D–G ) . A more burst-like , or globally irregular , firing pattern ( Figure 5H–I ) can be observed as a larger difference between the observed frequency and frequency mode ( Figure 5D–G ) , a higher CV ( Figure 5J ) , and a larger pause proportion ( Figure 5K ) . Additionally , the higher inter-spike interval ( local ) irregularity is reflected as a lower rhythmicity index ( Figure 5L–N ) and higher CV2 ( Figure 5O ) . We calculated these parameters for Purkinje cells recorded from control mice P7-8 , P9-10 , P11-12 , and P13-14 . We found that there is a statistically significant increase in frequency and frequency mode from P7-10 to P11-14 ( Figure 5F–G ) . We also found a statistically significant increase in CV and pause proportion from P7-12 to P13-14 ( Figure 5J–K ) . In addition , there was an increase in the rhythmicity index from P7-10 to P11-14 ( Figure 5N ) , but no change in the CV2 ( Figure 5O ) . Together , these results suggest that there are changes in the firing patterns that occur between P7-10 and P13-14 , and that these changes in activity involve a higher firing frequency , with more burst-like firing patterns , and a higher rhythmicity within the burst . The changes in firing patterns are further visualized in the firing frequency distributions for each single cell that we recorded for this study ( Figure 5—figure supplement 1 ) . Next , we set out to quantify the firing patterns in En1Cre/+;Atoh1fl/- mutants ( Figure 5C ) . We decided to measure spontaneous Purkinje cell signals at P10 and P14 , because P9-10 control cells were never statistically different from P7-8 in any of the parameters included in our study and because P13-14 control cells were different from P7-10 control cells in five out of six parameters we included . We found that spontaneous simple spikes in Purkinje cells recorded in P10 and P14 En1Cre/+;Atoh1fl/- mutants had a lower frequency and frequency mode than those recorded in Purkinje cells in P11-14 control mice ( Figure 5D–G ) . We also found that the CV and pause proportion was significantly lower in Purkinje cells recorded from P10 and P14 En1Cre/+;Atoh1fl/- mutants compared to the P13-14 control Purkinje cells ( Figure 5H–K ) . In addition , we found that Purkinje cells recorded from P10 and P14 En1Cre/+;Atoh1fl/- mutants had a lower rhythmicity index than the P11-14 control Purkinje cells ( Figure 5L–N ) and a higher CV2 than all control Purkinje cells ( Figure 5O ) . Finally , we found that Purkinje cell firing patterns in P10 and P14 En1Cre/+;Atoh1fl/- mutants never differed from each other in any of the six parameters that we included in our analysis . This observation is further supported by the similarity in firing frequency distributions of the single Purkinje cell recordings in P10 and in P14 in En1Cre/+;Atoh1fl/- mutants ( Figure 5—figure supplement 1 ) . Together , these results suggest that the changes in Purkinje firing patterns that occur between P7-10 and P13-14 in control animals does not occur in En1Cre/+;Atoh1fl/- mutants , and that Purkinje cells in P10 and in P14 in En1Cre/+;Atoh1fl/- mutants have a firing pattern that is highly similar to Purkinje cells in the P7-10 control mice . Based on the six parameters quantified in Figure 5 , P14 Purkinje cells in En1Cre/+;Atoh1fl/- mice appear more dissimilar from their age-matched controls than younger Purkinje cells in control mice . This is further supported when looking at representative traces from Purkinje cells recorded in P7 to P14 control mice and P10 and P14 En1Cre/+;Atoh1fl/- mice ( Figure 6A ) . However , no single parameter alone can fully describe the complex temporal dynamics of the transformation that occurs in Purkinje cell firing patterns during the second postnatal week . Therefore , we used several mathematical approaches to compare the firing patterns of the 149 single neuron recordings , based on all six firing properties described in Figure 5 . First , we performed an unbiased cluster analysis that groups together neurons that have the most similar firing properties ( Figure 6B ) . When we retrospectively label the age and genotype of each cell after unbiased clustering , we see that older control cells cluster together ( P11-14 in difference shades of green , clustered at the top ) and younger control cells cluster together ( P7-10 in different shades of purple , cluster at the bottom ) . Interestingly , Purkinje cells from both P10 and P14 En1Cre/+;Atoh1fl/- mice cluster with the young control cells ( orange , at the bottom ) ( Figure 6B ) . Second , we performed t-distributed stochastic neighbor embedding ( t-SNE ) analysis on the aforementioned firing parameters ( Figure 6C ) . After retrospectively color-coding the data points , we present a pseudo-timeline of the developmental transformation of firing patterns , with Purkinje cells from both P10 and P14 En1Cre/+;Atoh1fl/- mice clustered with the young ( P7-10 ) control cells in the bottom right and older ( P11-14 ) control cells on clustered together on the top left ( Figure 6C ) . Both the unbiased cluster and the tSNE analysis show some cells in an intermediate state ( between young and old ) ( Figure 6B–C , middle cells ) . This suggests the existence of a transformation that occurs between younger and older cells , but the transformation does not occur at the same time for all Purkinje cells , as is expected based on previous studies showing that the heterogeneity of Purkinje cell function is an inherent property of Purkinje cells that respects and spans the timelines of anatomical and functional development of Purkinje cells ( Beekhof et al . , 2021; McKay and Turner , 2005; van Welie et al . , 2011 ) . To further test whether there are general changes that occur at different time points or a transformation that occurs at a discrete age , we averaged the firing properties of all cells recorded in each group , allowing us to test for the relative contribution of specific firing properties in each age group , and performed an unbiased , hierarchical cluster analysis on the first two principal components of the averaged firing properties ( Figure 6D ) . We find that although the young control cells ( P7-10 ) are most similar to themselves and most dissimilar from older control cells ( P11-14 ) , there is a level of variability that likely reflects systematic differences for when the transformation occurs across different Purkinje cells . In agreement with our aforementioned findings , we found that the P10 and P14 Purkinje cells recorded in En1Cre/+;Atoh1fl/- mice cluster with the young control cells ( P7-10 ) and do not undergo the functional transformation in firing patterns that Purkinje cells in control mice do . Taken together , in a period of dynamic circuit development , Purkinje cells undergo a major developmental transformation in firing patterns between P7-10 and P11-14 . Furthermore , in agreement with the absence of change in gross cerebellar morphology or dynamics in Vglut1+ synapses , this developmental transformation does not occur between P10 and P14 in En1Cre/+;Atoh1fl/- mice . Next , we tested whether the altered neural activity reflects a delay in maturation of the firing patterns . We therefore recorded Purkinje cells in P18 control and En1Cre/+;Atoh1fl/- mice , which is a week after the changes in the electrophysiological signature occur in control mice and the latest timepoint at which we are confident that the general health of the En1Cre/+;Atoh1fl/- mice does not confound the quality of our recordings . We find that at P18 , Purkinje cells recorded in En1Cre/+;Atoh1fl/- mice still exhibit a lower firing frequency ( Figure 6—figure supplement 1A ) and preferred firing frequency ( Figure 6—figure supplement 1A ) , ( Figure 6—figure supplement 1B ) , no difference in pause proportion ( Figure 6—figure supplement 1C ) , lower rhythmicity index ( Figure 6—figure supplement 1D ) , lower global firing rate variability ( Figure 6—figure supplement 1E ) , and higher local firing rate variability ( Figure 6—figure supplement 1F ) compared to controls . We then reran the unbiased clustering analysis and tSNE analysis on all 179 single-cell recordings , and the hierarchical cluster analysis on the all group means to investigate whether P18 Purkinje cells in control and En1Cre/+;Atoh1fl/- mice more closely relate to Purkinje cells at the start ( P7-10 ) or end ( P11-14 ) of the dynamic developmental period . We found that Purkinje cells recorded in P18 control mice cluster with Purkinje cells recorded in P11-14 control mice , but Purkinje cells recorded in P18 En1Cre/+;Atoh1fl/- mice cluster with the younger Purkinje cells that were recorded at P7-10 in control mice , which were also similar to Purkinje cells recorded in P10 and P14 En1Cre/+;Atoh1fl/- mice ( Figure 6—figure supplement 1G–I ) . These analyses further underscore that developing Purkinje cells rely on the neurogenesis of their surrounding excitatory neurons to make a dynamic functional transformation in their firing properties and that Purkinje cells developing in an environment largely devoid of granule cells are maintained in a functionally immature state , rather than acquiring a new , pathophysiological firing pattern . Our results argue that excitatory neurons are essential for shaping the spontaneous firing properties of Purkinje cells . However , a subset of Purkinje cell action potentials is dynamically induced by excitatory inputs from climbing fibers originating in the inferior olive . Interestingly , during early postnatal development , multiple climbing fibers contact one Purkinje cell , but in a parallel-fiber/granule cell-dependent process , the additional climbing fibers are pruned away during the second and third weeks of life ( Crepel and Delhaye-Bouchaud , 1979; Hashimoto et al . , 2009; Kano and Hashimoto , 2012 ) . Immature Vglut2+/climbing fiber synapses are located around the Purkinje cell body in controls , but at around P9 only the ‘winner’ among these synapses translocates to the Purkinje cell dendrites in the molecular layer ( Kano et al . , 2018 ) . In addition to this change in Vglut2+-expressing climbing fibers , mossy fibers temporarily innervate Purkinje cells bodies ( Kalinovsky et al . , 2011; Sillitoe , 2016 ) and granule cells may transiently express Vglut2 ( Miyazaki et al . , 2003 ) . We therefore expected a reduction in Vglut2+-positive synapses around the Purkinje cell ( Figure 7A ) . Indeed , we observed a decrease in the density of Vglut2+ synapses in P14 control cerebella compared to P7 , likely due to a combined reduction in Vglut2+ expression in granule cells , displacement of mossy fibers to granule cells , and pruning of climbing fibers ( Figure 7B–C ) . At both timepoints , however , we observed a high density of Vglut2+ synapses in the cerebella of En1Cre/+;Atoh1fl/- mice ( Figure 7B–C ) . Given the large reduction in the total number of granule cells ( Figure 1—figure supplement 3 ) , it is likely that this high density of Vglut2+ synapses stems from mossy fibers and climbing fibers directly innervating Purkinje cell bodies . We tested this hypothesis by counting the number of Purkinje cells that had Vglut2+ puncta on their cell bodies . As expected , the number of soma-associated Vglut2+ contacts on Purkinje cells decreased between P7 and P14 in control mice , whereas all Purkinje cell somata in the En1Cre/+;Atoh1fl/- mice remained enriched with Vglut2+ synapses ( Figure 7D ) . These data show that the En1;Atoh1 lineage , excitatory cerebellar neurons are necessary for shaping the regional localization and cellular targeting of immature mossy fibers and climbing fibers that project to the Purkinje cells . Based on our findings that En1;Atoh1 lineage excitatory cerebellar neurons are necessary for the architectural maturation of mossy fibers and climbing fibers , we next wanted to investigate how climbing fiber-induced complex spikes changed between P7 and P14 control mice and if they were altered in En1Cre/+;Atoh1fl/- mice ( Figure 8A–C ) . Interestingly , the predominant shape of the complex spike was different between P7 and P14 . Purkinje cells in P14 control mice fired classical complex spikes , which are defined by a large sodium spike followed by a train of three to five smaller spikelets ( Davie et al . , 2008; Zagha et al . , 2008; Figure 8B ) . In contrast , Purkinje cells in P7 control mice fire ‘doublets’ , which are characterized by an initial simple spike-like action potential , followed by a smaller action potential that occurs within 20 ms ( a similar profile of doublets was previously reported in neonatal rats [Puro and Woodward , 1977; Sokoloff et al . , 2015 , Figure 8A] ) . Similar to Purkinje cells in P7 control mice , the Purkinje cells in P10 and P14 En1Cre/+;Atoh1fl/- mice also predominantly fired doublets instead of true complex spikes , suggesting that the extrinsically-induced action potentials in Purkinje cells are maintained in an immature state just like the firing patterns of spontaneous simple spikes ( Figures 5 and 6 ) . Furthermore , we found that some Purkinje cells fired both complex spikes and doublets , which may indicate two separate climbing fibers promoting distinct types of post-synaptic spikes ( example Figure 8D ) . We observed heterogeneity in complex spike type in about half of the Purkinje cells recorded in the youngest P7-10 control mice ( P7-8: 10/21 cells; P9-10: 19/30 cells ) and in about one third of Purkinje cells in the older P11-14 control mice ( P11-12: 9/35 cells; P13-14: 12/32 cells ) . However , the majority of Purkinje cells recorded from P10 En1Cre/+;Atoh1fl/- mice ( P10: 12/15 cells ) and all Purkinje cells recorded from P14 En1Cre/+;Atoh1fl/- mice ( P14: 15/15 cells ) showed complex spike heterogeneity . These data may collectively indicate a heterogeneity in climbing fiber innervation that is reduced within the second postnatal week and dependent on the integration of excitatory neurons within the cerebellar circuit . Another interesting observation we made was the occurrence of ‘complex spike trains’ , which we defined as three or more complex spikes that followed each other in close succession ( Figure 8E ) . These complex spike trains were observed in Purkinje cells recorded in control mice of all ages ( P7-8: 13/22 cells; P9-10: 8/30 cells; P11-12: 8/35 cells; P13-14: 9/32 cells ) as well as En1Cre/+;Atoh1fl/- mice ( P10: 5/15; P14: 15/15 cells ) and may be a necessary property that helps establish , stabilize , and/or maintain inferior olive to Purkinje cell climbing fiber synapses . When we quantified the number of different types of complex spikes , we found that classical complex spikes occurred less often in Purkinje cells of young control animals and En1Cre/+;Atoh1fl/- mutants . Specifically , we found that Purkinje cells recorded in P11-14 control mice had a statistically significant higher frequency of complex spikes than Purkinje cells recorded from P7 to P10 control mice and P10 to P14 En1Cre/+;Atoh1fl/- mutants ( Figure 8F ) . Conversely , the frequency of doublets was significantly higher in Purkinje cells recorded from P10 to P14 En1Cre/+;Atoh1fl/- mutants than in Purkinje cells in control animals of all ages ( Figure 8G ) . In addition , the number of all combined complex spikes was higher in Purkinje cells recorded from in P14 En1Cre/+;Atoh1fl/- mutants than that observed in Purkinje cells recorded in P7-8 , P9-10 , and P13-14 control animals ( Figure 8H ) . Together , these results indicate that the Purkinje cells in the mutant mice do not acquire a similar maturation in overall complex spike shape that the control Purkinje cells do and they have a high number of immature-like doublets , which may indicate an altered pruning process of immature climbing fibers . Finally , we wanted to investigate how the observed changes in cerebellar function are translated into behavioral abnormalities . Purkinje cells send projections to the cerebellar nuclei , which consists of both excitatory and inhibitory neurons , as well as some direct projections to the vestibular nuclei ( Sillitoe and Joyner , 2007 ) . These circuits are important for motor coordination and balance as well as social behaviors including ultrasonic vocalization ( USV ) in neonatal pups ( Fujita et al . , 2008; Lalonde and Strazielle , 2015 ) . Interestingly , the contribution of the cerebellum to these behaviors is initiated before circuit rewiring is completed ( Lalonde and Strazielle , 2015 ) . Therefore , we were curious to know whether the immature circuit of En1Cre/+;Atoh1fl/- mice was sufficient for performing a substantial repertoire of cerebellar-dependent behaviors . Observations of control and En1Cre/+;Atoh1fl/- mutant mice showed overt phenotypic differences in motor control ( Video 1 and Figure 9A ) similar to what was previously observed in other agranular mice ( Sidman et al . , 1962; Sidman et al . , 1965; Woodward et al . , 1974 ) . At P14 , control mice explore an open arena with smooth intentional motions , whereas En1Cre/+;Atoh1fl/- mice often fall on their backs . The frequent falling over onto their backs prevented us from performing classical assays of motor function such as rotarod or foot printing . Instead , we assayed the righting reflex , open field exploration , and USVs . We also tested for clinically relevant features such as tremor and dystonia-like postures , which often arise with cerebellar dysfunction . We found that En1Cre/+;Atoh1fl/- mice perform poorly compared to control littermates during the righting reflex . They were significantly slower in returning onto four paws when compared to Atoh1fl/+ control mice at P8 and P10 , and they were slower than heterozygous En1Cre/+;Atoh1fl/+mice only at P10 ( Figure 9B ) . Because all mice attempted to turn right-side-up immediately after being placed on their backs , it is likely that this delay in righting is the result of impaired motor coordination rather than an abnormal sense of gravity . Next , we tested whether En1Cre/+;Atoh1fl/- mice showed abnormal USVs when briefly separated from their mothers ( Figure 9C ) . We found that call time in En1Cre/+;Atoh1fl/- mice was shorter than those observed in control littermates and that En1Cre/+;Atoh1fl/- mice called less frequently than Atoh1fl/+ mice ( Figure 9C–E ) . We next quantified how En1Cre/+;Atoh1fl/- mice moved in an open field ( Figure 9F ) . The distance traveled or movement time in a 15-min-period was not significantly impaired ( total distance ( cm ) : Atoh1fl/+: 58 . 1±11 . 7; En1Cre/+;Atoh1fl/+: 35 . 7±9 . 9; Atoh1fl/-: 44 . 4±6 . 5; En1Cre/+;Atoh1fl/-: 43 . 1±12 . 4; Kruskal-Wallis test p=0 . 32; movement time ( s ) : Atoh1fl/+: 55 . 9±5 . 9; En1Cre/+;Atoh1fl/+: 42 . 1±7 . 2; Atoh1fl/-: 50 . 1±7 . 4; En1Cre/+;Atoh1fl/-: 83 . 5±17 . 7; Kruskal-Wallis test p=0 . 14 ) . However , En1Cre/+;Atoh1fl/- mutant mice traveled slower than Atoh1fl/- control mice and the En1Cre/+;Atoh1fl/+ mice made more isolated movements during their trajectory compared to all their littermate controls ( Figure 9F–H ) . Finally , we observed a tremor in the mutants and measured the severity with our custom-made tremor monitor ( Figure 9I; Brown et al . , 2020 ) . We found that En1Cre/+;Atoh1fl/- mice had a higher power tremor in the 12–16 Hz frequency range ( Figure 9J ) . This range corresponds to physiological tremor and indicates the presence of a pathophysiological defect that emerges from a rise in baseline values . The mutant mice also had a higher peak tremor power compared to all control littermates ( Figure 9K ) . Together , we uncovered that the lack of rRL-derived excitatory cerebellar neurons in developing En1Cre/+;Atoh1fl/- mice leads to multiple abnormal cerebellar-dependent behaviors that are observed in the first 2 weeks of postnatal life . In this paper , we used En1Cre/+;Atoh1fl/- mice as a model of cerebellar agranularity to test how cell-to-cell interactions impact the formation of functional circuits . Using this model with circuit-wide loss of granule cell neurogenesis , we uncovered how these late-born cells influence the functional development of their downstream synaptic partners , the Purkinje cells . We find that granule cell elimination stalls the anatomical and functional maturation of postnatal Purkinje cells . In humans , granule cell neurogenesis is impaired in premature infants with cerebellar hemorrhages , which is expected as proliferating granule cell precursors are highly vulnerable to the hemorrhage , likely because of their high metabolic demand ( Dobbing , 1974; Gano and Barkovich , 2019; Hortensius et al . , 2018 ) . Moreover , even a localized lesion that occurs during early cerebellar genesis can have major repercussions; loss of just a few precursors has exponential effects on the number of granule cells that integrate into the cerebellar circuit ( Corrales et al . , 2004; Corrales et al . , 2006 ) . Our findings show that manipulating granule cell neurogenesis not only causes major anatomical defects , but it also obstructs the remodeling and wiring of circuits that establish cerebellar cortical function during a dynamic postnatal period . Our data add to our understanding of how temporally defined insults to the developing cerebellum may result in downstream circuit dysfunctions that are associated with specific neurological disorders observed in premature infants ( Dijkshoorn et al . , 2020 ) . In our model of agranularity , we show that deletion of Atoh1 from the En1 domain impairs the development of the majority of granule cells . Even though some granule cells escape our genetic manipulation , we are confident that we impair the neurogenesis of the majority of granule cells ( Figure 1—figure supplement 3 ) . Granule cells are the major driver for increasing cerebellar size , foliation , and lamination . Accordingly , the cerebella of En1Cre/+;Atoh1fl/- mice do not undergo proper expansion , and are devoid of lobules and a densely packed granule cell layer ( Figure 1D–J ) . In addition , the granule cell marker , GABARα6 , is not reliably expressed in En1Cre/+;Atoh1fl/- mice ( Figure 1—figure supplement 2A ) and we find an approximate thousand-fold reduction in total En1;Atoh1 lineage neurons ( Figure 1—figure supplement 3 ) . Furthermore , the density of Vglut1+ terminals is decreased nearly nine-fold , with the remaining Vglut1 signal mainly representing mossy fiber terminals since any granule cell and unipolar brush cell escapees would only contribute minimally to the existing circuit in the mutants ( Figure 4 ) . Regardless , even if some granule cells are present , they are not sufficient to drive the cellular circuit reorganizations that are known to depend on granule cells ( Figure 2 , Figure 7 ) . In addition to impaired granule cell neurogenesis , we observe a lower number of unipolar brush cells and excitatory nuclei cells in our En1Cre/+;Atoh1fl/- mice , which could , in theory , contribute to our findings of abnormal Purkinje cell function by causing a lack of feedforward and feedback signals , respectively ( van Dorp and De Zeeuw , 2015; Gao et al . , 2016 ) . However , both excitatory nuclei cells and unipolar brush cells signal through the granule cells , so loss of granule cells alone would in theory have a similar effect on Purkinje cell function as the combined loss of all three subtypes . In addition , a lower number of excitatory nuclei cells can reduce the number of Purkinje cells ( Willett et al . , 2019 ) and a lower number of Purkinje cells was also noted in models of chimeric Atoh1 knockouts ( Jensen et al . , 2004 ) ; therefore , the reduction in cerebellar size that we reported in Figure 1 could be , at least in part , enhanced by a lower number of Purkinje cells . Regardless , the En1Cre/+;Atoh1fl/- cerebellum is filled with Purkinje cells , the predominant neuronal type comprising this massively altered cerebellum , making it an ideal environment for identifying and measuring synapse densities and easily isolating single-units during the in vivo recordings . Nevertheless , there are several caveats to using an agranular model to investigate circuitry . For instance , loss of morphogenetic processes that determine cerebellar architecture including its size ( Dahmane and Ruiz i Altaba , 1999 ) , foliation ( Corrales et al . , 2006 ) , and layering ( Miyata et al . , 2010 ) complicate interpretations of how Purkinje cells directly respond to granule cells . A previous study showed that impaired granule cell migration can also impair Purkinje cell morphology and monolayer formation , which in turn could change Purkinje cell function ( Adams et al . , 2002 ) . However , here , we observed that the firing patterns of young Purkinje cells ( P7-P10 ) do not differ between control and experimental mice , even though the defects in monolayer formation and foliation have already occurred by this time . This suggests that our observations of abnormal maturation in Purkinje cell firing patterns cannot be solely attributed to gross abnormalities in the anatomical structure of the cerebellum during morphogenesis in the En1Cre/+;Atoh1fl/- mice . In addition , changes in cerebellar shape and size could influence the function of forebrain regions ( Kuemerle et al . , 2007 ) , and the vast connectivity of the cerebellum with forebrain regions such as the hippocampus and prefrontal cortex could contribute to the non-motor vocalization defects that we observed ( Liu et al . , 2020; McAfee et al . , 2019 ) . The regional specificity of cerebellar circuitry that mediates non-motor connectivity is thus likely obscured in En1Cre/+;Atoh1fl/- mice ( Badura et al . , 2018; Stoodley and Limperopoulos , 2016 ) . Despite these caveats , much can be learned from our genetically precise agranular mouse model . Specifically , our mouse model has several advantages over previously described agranular mice when it comes to investigating the contribution of granule cell neurogenesis to cerebellar development and function . Because our model takes advantage of a conditional genetic strategy that only targets the rhombic lip lineage , our manipulation does not affect cell intrinsic developmental Purkinje cell programs ( Herrup , 1983; Miyata et al . , 2010; Sheldon et al . , 1997 ) . And unlike previous models , our approach is independent of procedural variations ( Altman and Anderson , 1971; Sathyanesan et al . , 2018; Yoo et al . , 2014 ) , targets the entire Atoh1 lineage in the cerebellum ( Ben-Arie et al . , 1997; Jensen et al . , 2002; Jensen et al . , 2004 ) , and has allowed us to study postnatal development . As a result , mutant mice with the En1Cre/+;Atoh1fl/- genotype have highly penetrant , consistent , and reliable anatomical and functional phenotypes , all of which have provided us with key insights into how cerebellar lineages shape circuit development and behavior . The cerebellum controls motor and non-motor behaviors ( Hull , 2020; Wagner and Luo , 2020 ) . Regardless of the specific behavior , Purkinje cells are always at the center of the responsible circuit . Interestingly , during the first 2 weeks of life in mice , motor control becomes more precise ( Lalonde and Strazielle , 2015 ) , in concert with the refinement of the cerebellar circuitry ( White and Sillitoe , 2013 ) . During this period , Purkinje cell innervation switches from climbing fibers and mossy fibers to climbing fibers and parallel fibers ( Mason and Gregory , 1984 ) , climbing fibers are pruned and a ‘winner’ establishes a single Purkinje cell target ( Kano et al . , 2018 ) , and Purkinje cell zones are sharpened ( White et al . , 2014 ) . Accompanying these changes are emergent properties of the two Purkinje cell spike profiles , simple spikes and complex spikes . In addition to the increased complexity of intrinsic cellular properties that have been defined using slice electrophysiology approaches ( McKay and Turner , 2005 ) , we postulated using our in vivo recordings that intercellular interactions between granule cells and Purkinje cells during development may support the maturation of Purkinje cell firing properties , both by providing direct inputs and by supporting anatomical maturation , including the outgrowth of Purkinje cell dendrites . In control mice , we observed dynamic changes in normal Purkinje cell firing between P7 and P14 . We found an increase in firing rate that was not uniformly acquired but was present during bursts of rapid firing that were interspersed with frequent pauses without Purkinje cell action potentials . We previously reported burst-like Purkinje cell firing from P15 to P19 , although by P30 the pattern acquires the regularity that is characteristic of adults ( Arancillo et al . , 2015 ) . Thus , burst-like firing occurs at intermediate stages of normal Purkinje cell development . Interestingly , bursting Purkinje cell firing patterns are also observed in mouse models of ataxia , tremor , and dystonia ( Brown et al . , 2020; Fremont et al . , 2014; LeDoux and Lorden , 2002; Miterko et al . , 2019; Miterko et al . , 2021; White and Sillitoe , 2017; White et al . , 2016 ) . The dynamically adapting circuit in control mice and the range of disease severities in disease models with bursting Purkinje cells raise the possibility that Purkinje cell firing is differentially decoded by downstream neurons based on the age of the mice . The data also indicate that the intermediate stages of Purkinje cell development not only highlight a developmental phase characterized by erratic neuronal activity , but that this mode of firing represents a pathophysiological hallmark that could be a default network state in different diseases . The higher level of baseline tremor in En1Cre/+;Atoh1fl/- mice suggests the possibility that perhaps the process of stabilizing neuronal activity is affected in our mutants . In contrast , alterations in other behaviors such as the USVs indicates that behaviors required during the postnatal period , such as communication with the dams , are also dependent of stage-specific neuronal properties . Therefore , the presence of granule cells in the cerebellar cortex impacts behaviors used at different ages as a consequence of how they influence Purkinje cell firing . The behavioral abnormalities in En1Cre/+;Atoh1fl/- mice are likely caused by a combined effect of Purkinje cell dysfunction and a reduction of excitatory neurons in the cerebellar nuclei . Given the broad network effects of our manipulation , it is not surprising that En1Cre/+;Atoh1fl/- mice have such severe behavioral abnormalities . However , our findings are very interesting given what we know about cerebellar function . First , while En1Cre/+;Atoh1fl/- mice often rotate onto their backs and have a slower righting reflex ( Figure 9A–B , Video 1 ) , they immediately initiate righting in both cases , suggesting that these mice have normal perception of gravitational force and that the observed phenotypes are mainly due to motor impairments . Second , because some extra-cerebellar structures are involved in airway pressure and breathing ( Bautista and Dutschmann , 2014; Chamberlin , 2004; Dutschmann and Herbert , 2006 ) , specifically the parabrachial nucleus and Kölliker Fuse ( van der Heijden and Zoghbi , 2018 ) , the defects in USVs could ultimately be due to alterations in multiple brain regions with particular effects on the duration of vocalizations ( Figure 9C–D ) . Yet , the respiratory cell types we have manipulated are unlikely to drive the initiation and number of vocalizations ( Figure 9E ) , suggesting that these abnormalities are cerebellum-driven . Finally , loss of Purkinje cell signaling is sufficient to reduce baseline and harmaline-induced tremor ( Brown et al . , 2020 ) , whereas we observe that loss of excitatory cerebellar neurons causes an enhanced tremor phenotype ( Figure 9I–K ) . These data show that unlike Purkinje cells , excitatory cerebellar neurons are not necessary for propagating tremor but they are nevertheless required for modulating it to achieve adequate control of the muscles . The severity of motor impairments and electrophysiological changes in Purkinje cells in our model and previously reported agranular mice remains in contrast to the modest changes in Purkinje cell firing after silencing granule cell synapses ( Galliano et al . , 2013 ) . However , a noteworthy distinction should be made that in our mice , granule cells are removed altogether , whereas Galliano et al . , 2013 blocked granule cell neurotransmitter release – the difference in functional outcomes could therefore include unique cell non-autonomous impacts on development induced by each type of manipulation and not solely depend on the primary manipulated cell type , the granule cells . Furthermore , in multiple models , impairing parallel fiber synapses results in motor impairments that can be assessed with the rotor rod assay ( Aiba et al . , 1994; Park et al . , 2019 ) , which we could not do due to the severity of motor impairment in En1Cre/+;Atoh1fl/- mice . Taking these results together , from both the technical and conceptual standpoints , when one seeks to resolve developmental mechanisms , we must not only consider what is manipulated , but also how it is manipulated . As such , the timing of neurogenesis is a primary consideration . Purkinje cells are generated between E10 and E13 ( Hashimoto and Mikoshiba , 2003 ) and granule cell progenitors from ~E13 onwards ( Machold and Fishell , 2005; Rose et al . , 2009; Wang et al . , 2005 ) . Whereas Purkinje cells migrate into the core of the cerebellar anlage upon their birth , granule cell precursors first migrate over the surface of the developing cerebellum and proliferate extensively in the external granular layer to increase the precursor pool ( Wingate and Hatten , 1999 ) . Only after this phase do they migrate radially past Purkinje cells , the first potential opportunity for direct cell-to-cell interactions . Based on our data , we argue that the initial communication between Purkinje cells and granule cells sets the efficiency of Purkinje cell function and the establishment of Purkinje cell spike properties through structural as well as synaptic signals . Thus , insults to granule cell proliferation and an obstruction of granule cell neurogenesis may have different , and perhaps more severe effects on downstream Purkinje cell function , compared to lesions of mature granule cells . This may explain why relatively mild but early lesions and structural changes , as observed in preterm infants with and without cerebellar hemorrhage , have large effects on cerebellar function . All mice used in this study were housed in a Level 3 , AALAS-certified facility . All experiments and studies that involved mice were reviewed and approved by the Institutional Animal Care and Use Committee ( IACUC ) of Baylor College of Medicine ( BCM ) . The following transgenic mouse lines were used for the experiments carried out in this study: Atoh1FlpO ( van der Heijden and Zoghbi , 2018 ) ; En1Cre ( En1tm2 ( cre ) Wrst/J , JAX:007916 ) ; Ai65 ( Gt ( ROSA ) 26Sortm65 . 1 ( CAG-tdTomato ) Hze , JAX:021875 ) ; Atoh1Flox ( Atoh1tm3Hzo , MGI:4420944 ) . Conditional knockout mice were generated by crossing En1Cre/+;Atoh1+/FlpO double heterozygote mice ( which can be functionally defined as En1Cre/+;Atoh1+/- mice ) with homozygote Atoh1fl/fl mice . En1Cre/+;Atoh1fl/- mice were considered experimental conditional knockout mice and all littermates with a different genotype were considered control mice . Specifically , we used Atoh1fl/+ , Atoh1FlpO/fl , and En1Cre/+;Atoh1fl/+ mice as controls . Ear tissue or tail clips were collected before weaning and used for genotyping and identification of the different alleles . For all experiments , we bred mice using standard timed pregnancies , noon on the day a vaginal plug was detected was considered embryonic day E0 . 5 and P0 was defined as the day of birth . Pups of both sexes were used in all experiments . Brain tissue was collected as described in our previous publications ( Zhou et al . , 2020 ) . First , we anesthetized mice with Avertin . Once the mice did not respond to toe or tail pinch , we accessed the chest cavity and then penetrated the heart with a butterfly needle for perfusions . The mice were perfused with 1M phosphate-buffered saline ( PBS pH 7 . 4 ) to remove blood from the tissue and 4% paraformaldehyde ( PFA ) to fix the tissue . The tissue was concomitantly post-fixed overnight in 4% PFA at 4°C . Tissue was cryoprotected in a sucrose gradient ( 10% → 20% → 30% sucrose in PBS ) at 4°C , each step lasting until the tissue sank to the bottom of a 15 mL tube . Tissue was frozen in optimal cutting temperature ( OCT ) solution and stored at −80°C until it was cut . All tissue were cut into 40 µm free-floating tissue sections and stored in PBS at 4°C until it was used for immunohistochemistry . Free floating tissue sections were stained according to the following protocol . Free floating tissue sections were blocked in 10% normal goat or donkey serum and 0 . 1% Triton-X in PBS ( PBS-T ) for two hours . Next , tissue sections were incubated overnight in primary antibodies in blocking solution . Tissue was washed three times for 5 min in PBS-T . For fluorescent staining , the tissue was incubated for 2 hr in PBS-T with the preferred secondary antibodies conjugated to an Alexa fluorophore . Finally , tissue sections were washed three times in PBS-T and mounted on electrostatically coated slides with hard-set , DAPI-containing mounting medium . Alternatively , for DAB staining , we incubated the tissue for 2 hr in PBS-T with the preferred secondary antibodies that was conjugated to horseradish peroxidase ( HRP ) . After washing three times in PBS-T , the tissue was incubated with diaminobenzidine ( DAB ) solution until the desired color intensity was reached . The DAB color reaction was stopped by washing tissue three times with PBS-T . The tissue was then mounted on electrostatically coated glass slides , dehydrated in an ethanol series ( 70% → 90% → 100% ) and then mounted using Xylene or histoclear . All steps of immunohistochemistry were performed at room temperature . All mounted slides were stored at 4°C until they were imaged . The following primary antibodies were used for the data described in this manuscript: guinea pig ( gp ) -α-Calbindin ( 1:1 , 000; SySy; #214004 ) ; rabbit ( rb ) -α-gamma-aminobutyric acid receptor α6 ( GABARα6; 1:500; Millipore Sigma; #AB5610 ) , rb-α-T-box brain protein 2 ( Tbr2; 1:500; Abcam; #AB23345 ) , mouse ( ms ) -α-Calretinin ( 1:500; Swant; #6B3 ) ; ms-α-Neurofilament Heavy ( NFH; 1:1 , 000; Biolegend; #801701 ) ; rb-α-Hyperpolarization Activated Cyclic Nucleotide Gated Potassium Channel 1 ( HCN1; 1:500; Alomone Lab; #APC-056 ) ; goat ( gt ) -α-RAR-related orphan receptor alpha ( RORα; 1:250; Santa Cruz; #F2510 ) ; rb-α-parvalbumin ( PV; 1:1 , 000; Swant; #PV25 ) ; rb-α-neurogranin ( 1:500; Millipore Sigma; #AB5620 ) ; ms-α-ZebrinII ( 1:500; kind gift from Dr . Richard Hawkes , University of Calgary , Calgary , Alberta , Canada ) ; rb-α- PLCβ4 ( 1:150; Santa Cruz Biotechnology; catalog #sc-20760 ) ; rb-α-Vglut1 ( 1:500; SySy; #135302 ) ; rb-α-Vglut2 ( 1:500; SySy; #135403 ) . The following secondary antibodies were used for immunohistochemistry: HRP-conjugated goat ( gt ) -α-mouse; gt-α-rabbit; and donkey ( dk ) -α-goat ( 1:200; DAKO ) . The following secondary antibodies were used for immunofluorescence: dk-α-mouse IgG Alexa Fluor 488 ( 1;1 , 500; Thermo Fisher Scientific; #A21202 ) and gt-α-gp IgG Alexa Fluor 488 ( 1;1 , 500; Thermo Fisher Scientific;#A11073 ) . Brain tissue sections were mounted on glass slides and then dried overnight . Slides were submerged in 100% histoclear and rehydrated in an ethanol series ( 100% → 90% → 70% ) . Then , the slides were submerged in cresyl violet solution for staining until sufficiently dark and then dehydrated in an ethanol series ( 70% → 90% → 100% ) . Finally , the slides were sealed with a coverslip using Cytoseal mounting media . All steps were performed at room temperature and the mounted slides were stored at 4°C until they were imaged . Anterograde neuroanatomical tracing of mossy fibers to the cerebellum was performed as described previously ( Lackey and Sillitoe , 2020; Sillitoe , 2016 ) . P12 pups were anesthetized with isoflurane on a surgery rig . Hair was removed and an incision was made in the skin over the lower thoracic/upper lumbar spinal cord , using the curvature of the spine as a guide . We used a Nanoject II to inject 0 . 2–1 µl of 2% WGA-Alexa Fluor 555 ( Thermo Fisher Scientific; #W32464 ) and 0 . 5% Fast Green ( Sigma-Aldrich; #F7252 , used for visualization ) diluted in 0 . 1 M phosphate-buffered saline ( PBS; Sigma-Aldric; #P4417; pH 7 . 4 ) . Tracer was injected 1 mm below the surface of the spinal cord , on the right side of the dorsal spinal vein . After tracer injection , we applied antibiotic ointment and closed the incision using VetBond ( 3M; #1469 SB ) and wound clips ( Fine Science Tools; #12032–07 ) . Pups were placed back with the mom after waking up from anesthesia . We placed soft food and hydrogel on the cage floor and monitored behavior closely for whether the mom accepted the pups back into the litter . Tissue was collected ( see section on Tissue processing above ) for tracer visualization 2 days after the surgery , at P14 . Golgi-Cox staining was performed according to previously described protocols ( Brown et al . , 2019 ) and the manufacturer’s instructions ( FD Neurotechnologies; #PK401 ) . Brains were dissected from the skulls and immediately emerged in the staining solution . After staining , tissue sections were cut at a thickness of 10 µm and directly mounted onto electrostatically coated glass slides . The tissue was then dehydrated in an ethanol series ( 70% → 90% → 100% ) , cleared with Xylene , and mounted with cytoseal . All slides were dried overnight before imaging and were kept at 4°C for storage . Photomicrographs of stained whole mount cerebella and DAB stained cerebellar tissue sections were acquired using Leica cameras DPC365FX and DMC2900 , respectively , attached to a Leica DM4000 B LED microscope . Photomicrographs of Golgi-Cox stained tissue sections and WGA-Alexa 555 tracing were captured using Zeiss cameras AxioCam MRc5 and AxiaCam Mrm , respectively , attached to a Zeiss Axio Imager . M2 microscope . Whole mount images were stitched together using Adobe Photoshop ( Adobe Systems 2020 ) Photomerge function . Color brightness and contrast were adjusted using ImageJ ( Schneider et al . , 2012 ) . Photomicrographs of images were cropped to the desired size using Adobe Illustrator ( Adobe Systems 2020 ) . We obtained high-magnification images from lobule IX in control animals or the ventrocaudal region of conditional knockout mice . We quantified the percentage of the area of interest occupied by the signal by transforming the signal into a binary signal with TdTomato pixel brightness cut off at 80/255 . We serially sectioned brains and kept every sixth ( coronal ) or eight ( sagittal ) section in an individual well of a 24-well plate . This allowed us to have an unbiased tissue collection system that spanned the entire cerebellum . We then stained all tissue sections collected in a single well for anti-Tbr2 ( unipolar brush cells ) or anti-NFH ( excitatory nuclei neurons and Purkinje cells ) plus anti-Calbindin ( Purkinje cells ) . Co-staining for NFH and Calbindin was necessary to distinguish between Purkinje cells ( Calbindin+ and NFH+ ) and excitatory nuclei cells ( Calbindin- and NFH+ ) in primarily the conditional knockout mice , as nuclei neurons and Purkinje cells did not have clear anatomical segregation in these mice . After staining our tissue sections as described above , we imaged whole tissue sections at a magnification that allowed us to identify and count single cells . Images were merged together in photoshop and all Tbr2+ cells and NFH+/Calbindin- cells were counted in all tissue sections from a single animal in ImageJ . The final cell-count was multiplied by six ( coronal ) or eight ( sagittal ) to obtain an estimate of the total number of unipolar brush cells or excitatory nuclei cells for each animal . A t-test was performed to investigate whether there was a statistically significant difference between cell number in control and En1Cre/+;Atoh1fl/- mice . Brightfield images of Golgi-Cox stains were imported into ImageJ and then converted into a binary image with brightness and contrast adjusted so that cellular processes were clearly visible . Then , Sholl analysis was performed using the build-in Sholl analysis module in ImageJ ( Ferreira et al . , 2014 ) and false positive intersections were manually subtracted from the counts . Next , we used a Zeiss Laser Scanner confocal Microscope model 710 to take high-magnification Z-stacked images ( 0 . 2 μm ) . We imported the images to imageJ/Fiji and used the Simple Neurite Tracer plug-in to track the length of a stained dendrite . We manually counted the number of synaptic boutons along the dendrite and reported this number as a density per dendrite length . We stained tissue sections obtained from P7 and P14 control and En1Cre/+;Atoh1fl/- mice with anti-Calbindin and anti-Vglut1 or Vglut2 antibodies as described above . Next , we imaged 3–5 tissue sections from three mice per age , per genotype . We imported tissue sections into ImageJ and defined our area of interest as the area of the image where Vglut1+ and Vglut2+ synapses could directly innervate Purkinje cells . This included the molecular and Purkinje cell layers in control tissue and the area occupied by Purkinje cell bodies in En1Cre/+;Atoh1fl/- tissue; both were identified as the area of the image positive for the Calbindin signal . We adjusted the brightness and contrast in the Vglut1 or Vglut2 channels so that individual synapses were visible and then transformed the image into a binary ( black and white ) image using the Threshold function in ImageJ . We quantified the percentage of area of interest covered by Vglut1 or Vglut2 synapses using the Analyze Particles tool in ImageJ . We also manually counted the number of Purkinje cells that received direct Vglut2+ inputs on their somas . Finally , we averaged the values of the area covered by Vglut1+ or Vglut2+ synapses across all tissue sections ( n=3–5 ) per animal ( N=3 ) and performed a two-way ANOVA ( genotypeage ) followed by a Tukey-Kramer post-hoc analysis to identify the statistical significance of observed differences in the density of Vglut1+ and Vglut2+ synapses in mice of different genotypes and ages . All in vivo , anesthetized experiments were performed as described in previous publications ( Arancillo et al . , 2015; White and Sillitoe , 2017 ) . Specifically , we anesthetized mice using a mixture of ketamine 80 ( mg/kg ) and dexmedetomidine ( 16 mg/kg ) . We held mice on a heated surgery pad . We removed hair from skull and made an incision in the skin over the anterior part of the skull . We stabilized the heads of our mice using ear bars and a mouth mount when animals were large enough ( most P11-P14 mice ) and otherwise fixed the mouse skull ( P7-P10 mice ) to a plastic mount that was attached to ear bars on our stereotaxic surgery rig to stabilize the head during recordings . Using a sharp needle or dental drill , we made a craniotomy in the interparietal bone plate , ~3 mm dorsal from lambda and ~3 mm lateral from the midline , with a diameter of ~3 mm . We kept our surgical coordinates consistent based on the distance from lambda across mice of all ages , as the skull undergoes significant growth during the ages at which we measured neural activity . After making a craniotomy , we recorded neural activity using tungsten electrodes ( Thomas Recording , Germany ) and then the digitized the signals into Spike2 ( CED , England ) . We recorded neural activity from cells that were 0–2 mm below the brain surface . All electrophysiological recording data were spike sorted in Spike2 . We sorted out three types of spikes: simple spikes , complex spikes , and doublets . Complex spikes were characterized by their large amplitude , and post-spike depolarization and smaller spikelets that follow . Doublets were characterized as action potentials that were followed by one or more smaller action potentials within 20 ms after the initial action potential . All other action potentials were characterized as simple spikes ( see examples in Figures 5 , 6 and 8 ) . We only included traces with clearly identifiable complex spikes or doublets and analyzed only cells from which we could obtain a sufficiently long and stable recording ( 186 ± 6 . 5 s; minimum = 75 s ) with an optimal signal to noise ratio . After spike sorting our traces in Spike 2 , we analyzed the frequency and regularity of firing patterns in MATLAB ( The MathWorks Inc , version R2018a ) . For this study , we defined ‘frequency’ as number of all spikes observed in the total analyzed recording time ( spikes/s ) . We determined the ‘frequency mode’ by populating the inverse of all observed interspike intervals ( ISI−1 , representing the frequency of a single spike ) into 2 . 5 Hz bins and defining the center of the bin with the largest proportion of spikes as the ‘frequency mode’ ( see Figure 5D and E for examples ) . Pause Percent was the proportion of the recording time during which the ISI was longer than five times the mean ISI for each independent cell , defined as the following: ( sum ( ISI>5mean ( ISI ) ) ) / ( total recording time ) . Our measures of global regularity or burstiness ( CV ) and regularity ( CV2 ) were based on the interspike intervals ( ISI ) between two adjacent spikes ( in seconds ) . CV = stdev ( ISI ) /mean ( ISI ) , and CV2 = mean ( 2\|ISIn-ISIn-1\|/ ( ISIn+ISIn-1 ) ) ( Holt et al . , 1996 ) . The rhythmicity index was based on oscillatory properties represented by the auto-correlogram of the ISI ( see Figure 5L and M for examples ) and based on previously described quantifications with minor adjustments ( Lang et al . , 1997; Sugihara et al . , 1995; White et al . , 2016 ) . The ‘rhythmicity index’ was calculated as the sum of the difference in peak and trough of the proportion of spikes at each delay from spike onset , divided by the baseline proportion of spikes . Baseline = ( total spike number ) 2/ ( recording time/bin width ) . We used a bin width of 5 ms . The first peak of each oscillation was determined as the highest bin between 10 ms and 1 . 5 times the mean ISI for a given cell , and the time of the peak was denoted as a1 . Each subsequent peak was determined as the highest bin between the delay-time of the previous trough and an+a1+10 ms , where an is the time of the previous peak . The first trough was determined as the lowest bin between the first peak ( a1 ) and an+a1 . Peaks and troughs were only accepted if their sum was higher than four times the standard deviation of auto-correlation between 0 . 96 and 1 s lag-time , or if the peak was higher than baseline + two times the standard deviation and the trough was lower than baseline – two times the standard deviation . To investigate and visualize how firing patterns changed over time , we performed additional computational analysis of the firing properties of single neurons ( n=179 neurons ) . We included the parameters summarized in Figure 5 in our analysis ( frequency , frequency mode , pause percent , rhythmicity index , CV , and CV2 ) . First , we performed an unbiased cluster analysis on the firing properties of single neurons using the MATLAB function ‘clusterogram , ’ which previously allowed us to visualize genotype-specific changes in Purkinje cell firing patterns ( van der Heijden et al . , 2021a ) . After analysis , we de-identified the cells and labelled them by age and genotype ( Figure 6B ) . Next , we performed a t-distributed stochastic neighbor embedding ( tSNE ) dimension reduction analysis to visualize the relative similarities of firing patterns between cells using the MATLAB function ‘tsne’ based on Euclidean distance . We plotted tSNE component 1 and 2 against each other ( tSNE ( 1 ) and tSNE ( 2 ) ) and color-coded the coordinates based on age and genotype of the cell recording to generate an electrophysiological pseudotimeline ( Figure 6C ) . Finally , we used the MATLAB function ‘linkage’ on the first two components of the principal component analysis on the average firing properties per age and genotype group to define how the firing patterns of each group clustered together . We visualized this unbiased linkage analysis using the MATLAB function ‘dendrogram’ ( Figure 6D ) . Righting reflex was measured on P6 , P8 , and P10 as follows . The mouse was placed on its back in a clean cage without bedding . One finger was used to stabilize the mouse on its back . The timer was set the moment the experimenter removed their finger , and time was recorded until the mouse righted itself up onto its four paws . All mice were tested twice at each age . A ‘failed’ trial was defined as when the mouse did not right itself within one minute ( sixty seconds ) . At P7 , we recorded pup vocalizations as described previously ( Yin et al . , 2018 ) . Pups were placed in an anechoic , sound-attenuating chamber ( Med Associates Inc ) . The pup was placed in a round plastic tub that was positioned near a CM16 microphone ( Avisoft Bioacoustics ) that was located in the center of the chamber . Sound was amplified and digitized using UltraSoundGate 416H at a 250 kHz sampling rate and bit depth of 16 . Avisoft RECORDER software was used to collect the recordings . Ultrasonic vocalizations were monitored for 2 min for each animal . We also performed an open-field assay at P13 , as previously described ( Alcott et al . , 2020 ) . Mice were habituated to a room with the light set to 200 lux and ambient white noise to 60 dB . We placed each mouse in the center of an open field ( 40x40x30 cm chamber ) . The chamber has photobeams that records movement . Each mouse was tested for 15 min and activity was recorded using Fusion software ( Accuscan Instruments ) . We used the standard settings for movement segmentation . We analyzed the data for total distance traveled , movement time , speed , and total movements during the 15-min test period . We measured tremor using our custom-built tremor monitor ( Brown et al . , 2020 ) . Each mouse was placed in the tremor chamber , which is a translucent box with an open top that is suspended in the air by eight elastic cords that are attached to four metal rods . An accelerometer is attached to the bottom of the box . Mice were allowed to habituate to the chamber for 120 s prior to initiating the tremor recordings . The mice are free to move around in the box . Power spectra of the tremor recordings were assessed using Fast Fourier transform ( FFT ) with a Hanning window in Spike2 software as previously described ( Brown et al . , 2020 ) . FFT frequency was targeted to ~1 Hz per bin . Statistical analyses were performed in MATLAB . For the Purkinje cell morphology data , we performed a linear mix-model analysis with genotype as the fixed variable and mouse number as the random variable and accepted p<0 . 05 as statistically significant . For Vglut1/2 quantification , we averaged measurements from three to five tissue sections per mouse and performed a two-way ANOVA ( genotypeage ) followed by a Tukey-Kramer post-hoc analysis to define statistical significance between independent groups . For electrophysiology data ( frequency ( complex spikes and simple spikes ) , frequency mode , pause percent , rhythmicity index , CV , CV2 , and doublet frequency ) , analysis was performed using an ANOVA with six total groups ( four control groups , two mutant groups ) , followed by a Tukey-Kramer post-hoc test . For behavioral data , we performed a Kruskal-Wallis test followed by a Tukey-Kramer post-hoc test to define significance between independent groups . For these tests , we accepted p<0 . 05 as statistically significant .	Preterm infants have a higher risk of developing movement difficulties and neurodevelopmental conditions like autism spectrum disorder . This is likely caused by injuries to a part of the brain called the cerebellum . The cerebellum is important for movement , language and social interactions . During the final weeks of pregnancy , the cerebellum grows larger and develops a complex pattern of folds . Tiny granule cells , which are particularly vulnerable to harm , drive this development . Exactly how damage to granule cells causes movement difficulties and other conditions is unclear . One potential explanation may be that granule cells are important for the development of Purkinje cells in the brain . The Purkinje cells send and receive messages and are very important for coordinating movement . To learn more , van der Heijden et al . studied Purkinje cells in mice during a period that corresponds with the third trimester of pregnancy in humans . During this time , the pattern of electrical signals sent by the Purkinje cells changed from slow and irregular to fast and rhythmic with long pauses between bursts . However , mice that had been genetically engineered to lack most of their granule cells showed a completely different pattern of Purkinje cell development . The pattern of electrical signals emitted by these Purkinje cells stayed slow and irregular . Mice that lacked granule cells also had movement difficulties , tremors , and abnormal vocalizations . The experiments confirm that granule cells are essential for normal brain development . Without enough granule cells , the Purkinje cells become stuck in an immature state . This discovery may help physicians identify preterm infants with motor disorders and other conditions earlier . It may also lead to changes in the care of preterm infants designed to protect their granule cells .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "neuroscience" ]	2021	Maturation of Purkinje cell firing properties relies on neurogenesis of excitatory neurons
Piezo2 ion channels are critical determinants of the sense of light touch in vertebrates . Yet , their regulation is only incompletely understood . We recently identified myotubularin related protein-2 ( Mtmr2 ) , a phosphoinositide ( PI ) phosphatase , in the native Piezo2 interactome of murine dorsal root ganglia ( DRG ) . Here , we demonstrate that Mtmr2 attenuates Piezo2-mediated rapidly adapting mechanically activated ( RA-MA ) currents . Interestingly , heterologous Piezo1 and other known MA current subtypes in DRG appeared largely unaffected by Mtmr2 . Experiments with catalytically inactive Mtmr2 , pharmacological blockers of PI ( 3 , 5 ) P2 synthesis , and osmotic stress suggest that Mtmr2-dependent Piezo2 inhibition involves depletion of PI ( 3 , 5 ) P2 . Further , we identified a PI ( 3 , 5 ) P2 binding region in Piezo2 , but not Piezo1 , that confers sensitivity to Mtmr2 as indicated by functional analysis of a domain-swapped Piezo2 mutant . Altogether , our results propose local PI ( 3 , 5 ) P2 modulation via Mtmr2 in the vicinity of Piezo2 as a novel mechanism to dynamically control Piezo2-dependent mechanotransduction in peripheral sensory neurons . Our sense of touch relies on mechanotransduction , that is the conversion of mechanical stimuli to electrical signals in primary afferent sensory neurons of the somatosensory system . Piezo2 ion channels have emerged as major somatosensory mechanotransducers and mediate rapidly adapting mechanically activated ( RA-MA ) currents in sensory neurons , such as those of dorsal root ganglia ( DRG ) ( Coste et al . , 2010 , 2012 ) . By now it has been established that Piezo2 is crucially involved in vertebrate light touch and proprioception ( Coste et al . , 2010 , 2012; Florez-Paz et al . , 2016; Ranade et al . , 2014; Woo et al . , 2015; Woo et al . , 2014 ) . Despite its importance , the regulation of native Piezo2 is only beginning to be elucidated . Mechanistically , several scenarios might be at play ( Wu et al . , 2017 ) : direct action of phospholipids , the modulation of local membrane properties and protein-protein interactions , just to name a few . Studies on other mechanosensitive ion channels such as the family of small conductance channels ( MscS ) ( Sukharev , 2002 ) as well as eukaryotic TRAAK , TREK1 ( Brohawn et al . , 2014a , 2014b ) and Piezo1 ( Cox et al . , 2016; Lewis and Grandl , 2015; Wu et al . , 2016 ) demonstrated their direct interplay with components of the lipid bilayer . In case of Piezo2 , RA-MA currents were shown to be inhibited through depletion of phosphatidylinositol 4 , 5-bisphosphate ( PI ( 4 , 5 ) P2 ) and phosphatidylinositol 4-monophosphate ( PI ( 4 ) P ) ( Borbiro et al . , 2015 ) , and also by depletion of cholesterol at the plasma membrane ( Qi et al . , 2015 ) . Besides , lipid- or cytoskeleton-induced changes in plasma membrane tension have been shown to impact somatosensory mechanotransduction ( Jia et al . , 2016; Morley et al . , 2016; Qi et al . , 2015 ) . In contrast to the vast knowledge of the molecular network governing mechanotransduction in the nematode C . elegans , to date only few protein-protein interactions relevant for Piezo2 physiology have been identified in vertebrates . These include stomatin-like protein STOML3 ( Poole et al . , 2014; Qi et al . , 2015; Wetzel et al . , 2007 ) , unidentified protein tethers to the extracellular matrix ( Hu et al . , 2010 ) and Pericentrin ( Narayanan et al . , 2016 ) . It is noteworthy that all of these exhibit pronounced effects on the magnitude or mechanical sensitivity of Piezo2 RA-MA currents . In order to advance the molecular understanding of Piezo2 regulation , we recently performed an interactomics screen , which revealed several additional binding partners of native Piezo2 in murine DRG ( Narayanan et al . , 2016 ) . A significantly enriched and prominent member of the Piezo2 interactome was myotubularin related protein 2 ( Mtmr2 ) ( Narayanan et al . , 2016 ) , a phosphoinositide phosphatase of the MTMR family ( Bolino et al . , 2002 ) . Interestingly , Mtmr2 was previously described to be highly expressed in DRG sensory neurons and Schwann cells ( Bolino et al . , 2002 ) . Functionally , Mtmr2 catalyzes the removal of a 3-phosphate group from its phosphoinositide ( PIPs ) substrates phosphatidylinositol 3-monophosphate PI ( 3 ) P and phosphatidylinositol 3 , 5-bisphosphate PI ( 3 , 5 ) P2 ( Begley et al . , 2006; Berger et al . , 2002 ) . Remarkably , PI ( 3 , 5 ) P2 is much less abundant than most other PIPs , for example PI ( 4 , 5 ) P2 ( Zolov et al . , 2012 ) , but can be rapidly and transiently regulated by a large enzymatic protein complex in response to cellular stimulation ( McCartney et al . , 2014a; Vaccari et al . , 2011; Zolov et al . , 2012 ) . Hence , PI ( 3 , 5 ) P2 is exquisitely suited to control rapid cellular signaling events ( Ho et al . , 2012; Ikonomov et al . , 2007; Li et al . , 2013a; McCartney et al . , 2014a; Zhang et al . , 2012 ) , and also the activity of receptors and ion channels ( Dong et al . , 2010; Ho et al . , 2012; Klaus et al . , 2009; McCartney et al . , 2014a; McCartney et al . , 2014b; Tsuruta et al . , 2009 ) . Altogether , this raises the question whether Mtmr2 and its PIP substrates are implicated in Piezo2-mediated mechanotransduction and generally , in somatosensory mechanosensation . Here , we show that Mtmr2 levels control Piezo2-mediated RA-MA currents . While elevated Mtmr2 expression attenuated Piezo2 RA-MA currents , siRNA-mediated knockdown of Mtmr2 resulted in Piezo2 RA-MA current potentiation . Interestingly , heterologously expressed Piezo1 and other known subtypes of MA currents in DRG were largely unaffected . Mechanistically , our experiments with catalytically inactive Mtmr2 , pharmacological inhibitors , and osmotic stress suggest that changes in the levels of PI ( 3 , 5 ) P2 regulate Piezo2 RA-MA currents . In line with these findings we uncovered a previously unknown polybasic motif in Piezo2 that can bind PI ( 3 , 5 ) P2 and confers Piezo2 sensitivity to Mtmr2 . Collectively , our study reveals a link between Mtmr2 activity and PI ( 3 , 5 ) P2 availability to locally control Piezo2 function . Our previous work revealed Mtmr2 as a highly enriched member of the native Piezo2 interactome in DRG ( significance of identification: p=0 . 00030 , unpaired t-test; enrichment factor: log2 9 . 96 ) ( Narayanan et al . , 2016 ) . We first validated the reported expression of Mtmr2 in peripheral sensory neurons of DRG ( Previtali et al . , 2003; Vaccari et al . , 2011 ) including those neurons expressing Piezo2 ( Figure 1—figure supplement 1a , b ) . For a more detailed subcellular analysis we used the proximity ligation assay ( PLA ) . In this way we could show the close vicinity of Piezo2 and Mtmr2 in both , somata and neurites of cultured DRG neurons , and upon co-transfection in HEK293 cells ( Figure 1a–d and Figure 1—figure supplement 1c , d ) . It is important to note here that the PLA technique is prone to high background upon heterologous expression as shown by our additional control experiments in HEK293 cells ( Figure 1—figure supplement 1d ) . In these , we co-overexpressed Piezo2 with TRPA1 and Vti1b , respectively . Both of these controls exhibited clear PLA signal ( likely attributable to massive overexpression upon transfection ) , though less than co-overexpression with Mtmr2 ( Figure 1—figure supplement 1d ) . Following , we wanted to assess whether Mtmr2 affects Piezo2 function . To this end we performed electrophysiological measurements of Piezo2-mediated RA-MA currents in HEK293 cells , which represent a well-defined heterologous system to study Piezo2 function ( Coste et al . , 2010; Poole et al . , 2014 ) . Interestingly , Mtmr2 co-expression led to a pronounced reduction of Piezo2-mediated RA-MA currents compared to mock conditions ( Figure 1e , f ) . Moreover , the displacement threshold of RA-MA currents was significantly increased upon Mtmr2 overexpression indicating the requirement of stronger mechanical stimulation to reach a threshold current amplitude ( Supplementary file 1; please see Materials and methods for details on the calculation of the displacement threshold ) ( Eijkelkamp et al . , 2013; Morley et al . , 2016; Narayanan et al . , 2016 ) . Importantly , the inactivation time constant of Piezo2 currents remained unchanged upon Mtmr2 co-expression ( Supplementary file 1 ) . These data indicate that Mtmr2 overexpression suppresses Piezo2 currents in HEK293 cells while maintaining their defining property , that is rapid adaptation . To exclude that Mtmr2 overexpression renders cells unhealthy and might therefore unspecifically suppress Piezo2 currents , we co-expressed Mtmr2 with Kv1 . 1 and measured voltage-gated currents , which were similar to mock transfected controls ( Figure 1—figure supplement 2a ) . In addition , we asked whether Mtmr2 may modulate MA currents generated by Piezo1 , the homologue of Piezo2 ( Coste et al . , 2010 ) . Remarkably , we did not observe any differences in Piezo1 currents upon overexpression of Mtmr2 ( Figure 1—figure supplement 2b and Supplementary file 1 ) . These results suggest a certain degree of specificity of the functional Piezo2-Mtmr2 interaction . Next , we aimed at assessing whether Mtmr2 can modulate native Piezo2 RA-MA currents in cultured DRG neurons , as well . DRG cultures allow for the targeted manipulation of protein levels by nucleofection and concomitant assessment of Piezo2-mediated RA-MA currents ( Coste et al . , 2010; Narayanan et al . , 2016 ) . It is also noteworthy , that the DRG culture system has been employed for the original discovery ( Coste et al . , 2010 ) and further characterization of Piezo2 RA-MA currents ( Dubin et al . , 2012; Eijkelkamp et al . , 2013; Jia et al . , 2016; Narayanan et al . , 2016; Poole et al . , 2014; Qi et al . , 2015 ) ; hence it serves as a model for sensory transduction processes ( Coste et al . , 2007; Coste et al . , 2010; Lechner and Lewin , 2009 ) . In order to test whether the effect on Piezo2 RA-MA currents observed in HEK293 cells can be recapitulated , we overexpressed Mtmr2 in DRG cultures and measured RA-MA currents . Similar to HEK293 cells , Mtmr2 overexpression suppressed native RA-MA currents in DRG cultures when compared to mock transfection ( Figure 1g , h ) . The displacement threshold was unaffected upon Mtmr2 overexpression ( Supplementary file 1 ) , which is in contrast to our results in HEK293 cells potentially reflecting differences in Piezo2/Mtmr2 stoichiometry or the contribution of unknown neuronal modulators . Also , the inactivation time constant of RA-MA currents was unchanged upon overexpression of Mtmr2 in DRG neurons ( Supplementary file 1 ) . We went on to investigate whether Mtmr2 downregulation could potentiate native RA-MA currents in DRG . Successful knockdown of Mtmr2 was achieved after 72 hr and evaluated by quantitative PCR ( Figure 2—figure supplement 1a ) . We assessed Piezo2 mRNA levels , Piezo2 membrane expression and the percentage of Piezo2-positive neurons and did not observe any changes upon Mtmr2 knockdown ( Figure 2—figure supplement 1b–d ) . However , when we measured RA-MA currents in Mtmr2 siRNA-nucleofected DRG cultures , we observed a significant augmentation in current amplitude whereas the displacement threshold and inactivation time constant were unchanged ( Figure 2a–b; Supplementary file 1 ) . These results are in agreement with our data on Mtmr2 overexpression ( Figure 1 ) and suggest that Mtmr2 levels can modulate Piezo2 function: decreased expression of Mtmr2 potentiated , while increased expression suppressed Piezo2 RA-MA currents . Besides Piezo2-mediated RA-MA currents , cultured DRG neurons display two other major types of MA currents: intermediately- ( IA ) and slowly-adapting- ( SA ) MA currents categorized based on their inactivation time constant ( please see Materials and methods for details ) ( Coste et al . , 2007; Coste et al . , 2010; Hu and Lewin , 2006; Lechner and Lewin , 2009 ) . These seemed to be largely unaltered by Mtmr2 knockdown ( Figure 2c , d; Supplementary file 1 ) indicating that mechanotransduction is not generally compromised . Remarkably though , upon Mtmr2 knockdown we detected a significant redistribution of the number of cells exhibiting MA subpopulations: a moderate increase in the proportion of RA-MA cells paralleled by a decrease of the IA-MA population ( Figure 2e ) . To date the interpretation of the latter is difficult and requires yet to be obtained insights into the molecular nature of IA-MA currents . Next we asked how , on a mechanistic level , Mtmr2 could regulate Piezo2 activity given that neither Piezo2 mRNA levels nor membrane expression seemed to be affected ( Figure 2—figure supplement 1b , c ) . Mtmr2 is a phosphatase that catalyzes the removal of the 3-phosphate group from PI ( 3 ) P as well as PI ( 3 , 5 ) P2 ( Berger et al . , 2002 ) . Under resting conditions PI ( 3 , 5 ) P2 is present at low levels ( Jin et al . , 2016; McCartney et al . , 2014a ) but is transiently and steeply generated upon a diverse range of cellular stressors . This is in contrast to intensely studied and highly abundant PI ( 4 , 5 ) P2 known to regulate ion channels ( Gamper and Shapiro , 2007 ) including both , Piezo1 and Piezo2 ( Borbiro et al . , 2015 ) . To explore a potential role of Mtmr2 phosphatase activity and the resulting change in PIP levels ( Laporte et al . , 1998; McCartney et al . , 2014b; Mironova et al . , 2016; Previtali et al . , 2007 ) for the regulation of Piezo2 currents , we generated a catalytically inactive Mtmr2 mutant ( Mtmr2C417S ) by substituting Cysteine 417 for Serine ( Berger et al . , 2002 ) . If the catalysis of PIPs was essential to the functional interaction of Piezo2-Mtmr2 , the catalytically inactive mutant should fail to suppress Piezo2 currents when overexpressed in HEK293 cells ( please see our data on wild type Mtmr2 in Figure 1e , f ) . This was indeed found to be the case . Mtmr2C417S co-expression did not suppress Piezo2-mediated currents in HEK293 cells as determined by comparison of stimulus-current curves ( Figure 3a ) with mock transfected cells . Instead , we observed a trend towards moderate potentiation of Piezo2 currents upon co-expression of Mtmr2C417S ( especially at low stimulus magnitudes; Figure 3a ) . This result strongly suggests that the catalytic activity of Mtmr2 and the consequential alteration of PIP levels ( Laporte et al . , 1998; McCartney et al . , 2014b; Mironova et al . , 2016; Previtali et al . , 2007 ) may underlie the regulation of Piezo2 by Mtmr2 . Importantly , enzymatically inactive Mtmr2C417S was abundantly expressed and remained capable of associating with Piezo2 in close proximity as suggested by PLA upon co-overexpression of Piezo2 and Mtmr2C417S in comparison to wild type Mtmr2 ( Figure 3b–c ) . Based on the results obtained from experiments with Mtmr2C417S we then proceeded to manipulate neuronal PIP levels . Figure 4a illustrates a schematic view of the PIP synthesis and turnover pathway Mtmr2 is involved in and also indicates pharmacological inhibitors known to intervene with this pathway ( Vaccari et al . , 2011 ) . According to this scheme and a wealth of previous work ( Laporte et al . , 1998; McCartney et al . , 2014b; Mironova et al . , 2016; Previtali et al . , 2007; Vaccari et al . , 2011 ) , knockdown of Mtmr2 would increase the levels of PI ( 3 ) P and even more PI ( 3 , 5 ) P2 ( Vaccari et al . , 2011 ) . We tried to experimentally mimic heightened levels of these PIPs by inclusion of exogenous PIPs in the intracellular recording solution ( Dong et al . , 2010 ) ; however , we did not see any change in RA-MA currents ( Figure 4—figure supplement 1a ) . Due to technical factors that may confound this data ( e . g . rapid breakdown of exogenous PIPs by intracellular phosphatases ) , we opted to perform additional experiments . We reversed the described accumulation of these two PIPs upon Mtmr2 knockdown ( Laporte et al . , 1998; McCartney et al . , 2014b; Mironova et al . , 2016; Previtali et al . , 2007; Vaccari et al . , 2011 ) by applying commonly-used inhibitors of PIP synthesis , that is Wortmannin , an inhibitor of the class III PI 3-kinase , ( Messenger et al . , 2015 ) , and Apilimod , an inhibitor of PIKfyve ( Cai et al . , 2013; Vaccari et al . , 2011 ) , respectively ( please see scheme in Figure 4a ) . If elevated levels of PI ( 3 ) P or PI ( 3 , 5 ) P2 were implicated in Mtmr2-knockdown-induced potentiation of Piezo2 , the presence of the corresponding inhibitor in the recording solution would be expected to counteract this potentiation . Wortmannin application only marginally altered the increase in RA-MA currents upon Mtmr2 knockdown in neurons ( Figure 4b ) . Apilimod , on the other hand , significantly diminished the magnitude of RA-MA currents in Mtmr2 siRNA-treated neurons ( Figure 4b ) . In line with a reversal of Mtmr2-induced potentiation , the displacement threshold of RA-MA currents was significantly increased upon Apilimod treatment ( Supplementary file 1 ) . Interestingly , Apilimod treatment in wild type DRG neurons did not affect RA-MA currents ( Figure 4—figure supplement 1b; Supplementary file 1 ) , as expected given the low cellular expression and tight regulation of PI ( 3 , 5 ) P2 under physiological conditions ( Jin et al . , 2016; McCartney et al . , 2014a ) . Taken together , our results point towards a role of PI ( 3 , 5 ) P2 for the functional interaction of Piezo2 and Mtmr2 . We then intended to assess the significance of PI ( 3 , 5 ) P2 availability for Piezo2 function in a more physiological setting . In peripheral sensory neurons changes in cellular osmolarity cause activation of diverse ion channels and receptors followed by initiation of various signaling pathways involved in volume regulation ( Lechner et al . , 2011; Liu et al . , 2007; Quallo et al . , 2015 ) . Interestingly , previous work has indicated that also Mtmr2 , PI ( 3 , 5 ) P2 ( Berger et al . , 2003; Dove et al . , 1997 ) and Piezo2 ( Jia et al . , 2016 ) can be modulated by osmotic stress . Upon hypoosmotic stress Mtmr2 trafficking was altered and PI ( 3 , 5 ) P2 levels were reported to be elevated in eukaryotic cells ( Berger et al . , 2003; Dove et al . , 1997 ) . In the case of Piezo2 , hypoosmotic stress was shown to potentiate Piezo2 RA-MA currents , which was independent of the prominent osmosensor TRPV4 ( Jia et al . , 2016 ) . Therefore , we employed osmotic stress as a physiological stimulus to investigate the link between Mtmr2 , PI ( 3 , 5 ) P2 and Piezo2 . First , we confirmed the previously described ( Jia et al . , 2016 ) potentiation of Piezo2 RA-MA currents by application of extracellular hypotonic stress to DRG cultures ( Figure 4—figure supplement 1c ) . We then postulated that an increase of PI ( 3 , 5 ) P2 levels by extracellular hypotonicity ( Berger et al . , 2003; Dove et al . , 1997 ) should counteract the RA-MA suppression upon Mtmr2-overexpression , which we described in Figure 1h above . Indeed , in Mtmr2 overexpressing DRG cultures we recorded significantly higher RA-MA currents under extracellular hypotonic conditions compared to isotonic conditions ( Figure 4c ) . Other RA-MA current parameters were unchanged ( Supplementary file 1 ) . We then tested whether the opposite was also true: Would an expected decrease of PI ( 3 , 5 ) P2 levels by intracellular hypotonicity ( Dove et al . , 1997 ) prevent RA-MA current sensitization upon Mtmr2 knockdown ? Conceptually , this experiment is analogous to Apilimod application in Figure 4b above , where we pharmacologically inhibited PI ( 3 , 5 ) P2 production in DRG cultures . As predicted , under intracellular hypotonic conditions RA-MA currents were significantly smaller in siRNA-treated cultures compared to isotonic conditions ( Figure 4d; Supplementary file 1 ) . In parallel , the displacement threshold was significantly increased , while other RA-MA current parameters remained unchanged ( Supplementary file 1 ) . As specific probes to assess the cellular distribution of PI ( 3 , 5 ) P2 are not available ( Hammond et al . , 2015; Li et al . , 2013b; McCartney et al . , 2014a; Nicot and Laporte , 2008; Previtali et al . , 2007 ) , we could not measure the actual levels of PI ( 3 , 5 ) P2 in neurons under different osmotic conditions . Even though , our pharmacological and osmotic experiments both suggest an interdependent contribution of Mtmr2 and PI ( 3 , 5 ) P2 to the modulation of Piezo2-mediated MA currents in DRG cultures . PIPs are essential membrane components and alterations of PIP availability could modify mechanical properties of cells and their membranes ( Janmey , 1995; Raucher et al . , 2000; Skwarekmaruszewska et al . , 2006 ) . In fact , several recent studies demonstrated the influence of membrane mechanics on mechanotransduction ( Brohawn et al . , 2014a; Brohawn et al . , 2014b; Cox et al . , 2016; Lewis and Grandl , 2015; Sukharev , 2002; Wu et al . , 2016 ) , particularly on Piezo2-mediated RA-MA currents ( Jia et al . , 2016; Qi et al . , 2015 ) . Therefore , we set out to test a possible impact of Mtmr2 knockdown upon mechanical properties of cultured DRG neurons by atomic force microscopy ( AFM ) ( Nawaz et al . , 2015; Qi et al . , 2015; Rehfeldt et al . , 2007 ) . However , our experiments did not show any significant differences between Mtmr2 siRNA-treated and control-treated DRG neurons . Neither the Young´s elastic modulus ( an indicator for cellular elasticity including membrane tension and cortex stiffness ) ( Morley et al . , 2016; Qi et al . , 2015 ) as determined from the indentation , nor the tether force ( an indicator for membrane tension and mechanical coupling to the cortex ) ( Qi et al . , 2015 ) obtained from the retraction portion of the force distance curves were altered ( Figure 4—figure supplement 2 ) . These results are in line with our aforementioned findings indicating that Mtmr2 knockdown does not fundamentally alter mechanotransduction in DRG cultures ( Figure 2c , d and Supplementary file 1 ) . Nevertheless , our AFM measurements cannot exclude possible small and local changes in membrane tension in the direct vicinity of Piezo2 . Based on the functional role of PI ( 3 , 5 ) P2 for Mtmr2-dependent Piezo2 regulation we investigated whether Piezo2 can bind PI ( 3 , 5 ) P2 . PIPs are known to bind to proteins through various domains such as the FYVE domain , WD40 domain , PHD domain or electrostatically via poly-basic regions with unstructured clusters of positively charged amino acid residues ( Lysine or Arginine ) ( Dong et al . , 2010; McCartney et al . , 2014a ) . Sequence analysis revealed a region in Piezo2 with considerable similarity to the proposed PI ( 3 , 5 ) P2 binding motif of the mucolipin TRP channel 1 ( TRPML1 ) ( Dong et al . , 2010 ) ( Figure 5a ) . We performed a peptide-lipid binding assay to test if this sequence in Piezo2 could bind to PI ( 3 , 5 ) P2 , which was indeed the case ( Figure 5b , d ) . In addition , we also observed that the Piezo2 peptide was able to bind PI ( 4 , 5 ) P2 and weakly to PI ( 3 , 4 ) P2 ( Figure 5b , d and Figure 5—figure supplement 1 ) . This is an intriguing result because Borbiro and colleagues showed that TRPV1 modulates Piezo2 currents through PI ( 4 , 5 ) P2 depletion , though the study did not report a PI ( 4 , 5 ) P2 binding region in the Piezo2 sequence ( Borbiro et al . , 2015 ) . In parallel we performed the binding assay with a mutated version of the Piezo2 peptide , in which positively-charged amino acid residues shown to be relevant for PI ( 3 , 5 ) P2 binding in TRPML1 were changed to neutral Glutamine ( Q; Piezo2 3Q mutant; QQILQYFWMS ) . This mutated peptide did not bind to any lipid ( Figure 5b , d ) . The here identified PI ( 3 , 5 ) P2-binding region in Piezo2 exhibits 50% sequence identity to Piezo1 with conservation of positively charged amino acid residues ( Figure 5a ) . Therefore , we also tested whether Piezo1 was able to bind PI ( 3 , 5 ) P2 , but did not find any evidence for this ( Figure 5c , d ) . These data suggest that not only positively charged amino acid residues , but also flanking amino acids in this Piezo2 region contribute to PIP2 binding in a yet to be explored manner . Further , these in vitro binding studies substantiate our functional data on the specific link between Piezo2 and the Mtmr2 substrate PI ( 3 , 5 ) P2 by identifying a PI ( 3 , 5 ) P2 binding domain in Piezo2 , but not in Piezo1 . It is important to note that the peptide region defined here may not be the only PI ( 3 , 5 ) P2 binding region in Piezo2 especially when considering the known diversity of PIP modules ( Dong et al . , 2010; McCartney et al . , 2014a ) . Due to the sheer size of Piezo2 a large-scale peptide-lipid binding screen was beyond the scope of this study . Next , we attempted to monitor the functional relevance of the here described PI ( 3 , 5 ) P2 binding domain for the Piezo2-Mtmr2 interaction . To this end we generated a Piezo2 P1 mutant by swapping the PI ( 3 , 5 ) P2 binding domain of Piezo2 with the corresponding domain of Piezo1 ( please see scheme in Figure 5a ) . Remarkably , MA currents of the Piezo2 P1 mutant were only slightly attenuated upon co-expression with Mtmr2 compared to mock-transfected controls ( Figure 5e ) . Also , the displacement threshold and inactivation time constant of MA currents remained unchanged ( Supplementary file 1 ) . This is in stark contrast to the pronounced Mtmr2-induced suppression of MA currents recorded from wildtype Piezo2 ( Figure 1f ) . Hence , our results suggest that the PI ( 3 , 5 ) P2 binding domain of Piezo2 identified here likely mediates the functional sensitivity of Piezo2 to Mtmr2-dependent changes in PI ( 3 , 5 ) P2 levels . Vertebrate somatosensory mechanotransduction entails a complex interplay of cellular components; however , their identity is far from being resolved . In our study , we demonstrate that Mtmr2 limits Piezo2 MA currents , and that this effect likely involves local catalysis of PI ( 3 , 5 ) P2 . Thus , our work elucidated a link between Mtmr2 and PI ( 3 , 5 ) P2 availability as a previously unappreciated mechanism how Piezo2-mediated mechanotransduction can be locally controlled in peripheral sensory neurons . Originally , we identified Mtmr2 as a significantly enriched member of the native Piezo2 interactome in DRG neurons ( Narayanan et al . , 2016 ) . Mtmr2 is an active phosphatase that catalyzes the removal of 3-phosphate from its substrates PI ( 3 ) P and PI ( 3 , 5 ) P2 . Hence , its subcellular localization is crucial as it dictates access to its substrates , which are embedded in membrane lipid bilayers . Interestingly , several reports suggest that membrane association of Mtmr2 is enhanced by hypotonic stress ( Berger et al . , 2003 ) and also by interactions with other members of the Mtmr family ( Kim et al . , 2003; Ng et al . , 2013 ) . For example , Mtmr13 and Mtmr2 reciprocally control their abundance at the membrane of cultured cell lines ( Ng et al . , 2013; Robinson and Dixon , 2005 ) and in uncharacterized endomembrane compartments in Schwann cells ( Ng et al . , 2013 ) . In addition , Mtmr2 enzymatic activity is augmented through interaction with Mtmr13 ( Berger et al . , 2006 ) . In this respect it is noteworthy that our interactomics screen ( Narayanan et al . , 2016 ) identified two additional Mtmr family members: Mtmr1 and Mtmr5 . Mtmr1 was shown to be similar to Mtmr2 in structure and substrate specificity ( Bong et al . , 2016 ) , but is much less studied than Mtmr2 . Mtmr5 is a catalytically inactive phosphatase which binds to Mtmr2 , increases its enzymatic activity and controls its subcellular localization ( Kim et al . , 2003 ) . The fact that Mtmr5 was previously reported as a binding partner of Mtmr2 further validates our published interactomics data ( Narayanan et al . , 2016 ) , and hints towards the intriguing possibility that a multiprotein complex of different Mtmr family members might affect Piezo2 function . Future experiments should focus on assessing the role of these two Mtmr family members for mechanotransduction in peripheral sensory neurons . Mechanistically , our in vitro data strongly suggest that enzymatic activity of Mtmr2 and consequently PI ( 3 , 5 ) P2 availability modulates Piezo2 RA-MA currents . Several lines of evidence support this conclusion . We found that , unlike wild type Mtmr2 , a catalytically inactive mutant of Mtmr2 ( C417S ) did not suppress Piezo2 currents in HEK293 cells . In sensory neurons expressing Mtmr2 siRNA , we could counteract RA-MA current potentiation by inhibiting PI ( 3 , 5 ) P2 synthesis in two ways: via Apilimod ( Laporte et al . , 1998; McCartney et al . , 2014b; Mironova et al . , 2016; Previtali et al . , 2007; Vaccari et al . , 2011 ) and via intracellular hypoosmolarity ( Dove et al . , 1997; McCartney et al . , 2014a ) , respectively . In analogy , increasing PI ( 3 , 5 ) P2 levels by extracellular hypoosmolarity ( Berger et al . , 2003; Dove et al . , 1997 ) attenuated the suppression of RA-MA currents upon Mtmr2 overexpression . Here , it is important to note that Mtmr2 activity usually serves a dual function: dephosphorylation of PI ( 3 , 5 ) P2 and of PI ( 3 ) P , albeit the latter with lower efficiency ( Berger et al . , 2002 ) . Yet , our results upon application of Wortmannin suggest only a marginal contribution of PI ( 3 ) P to altering Piezo2-mediated mechanotransduction . Wortmannin is commonly used to block class III PI 3-kinase ( please see scheme in Figure 4a ) , however , it also is reported to inhibit class I PI 3-kinase ( Messenger et al . , 2015 ) . Hence its action on class III PI 3-kinase might not have been efficient enough in our experiments . Biologically , many enzymes control PI ( 3 ) P synthesis ( Yan and Backer , 2007; Zolov et al . , 2012 ) and previous work on Mtmr2 has only shown minor modifications of PI ( 3 ) P levels in Mtmr2-deficient cells ( Cao et al . , 2008; Vaccari et al . , 2011 ) in line with its substrate preference ( Berger et al . , 2002 ) . Moreover , our lipid-peptide binding assays demonstrated that , in a cell-free system , a distinct region in Piezo2 , which is similar to the known PI ( 3 , 5 ) P2-binding motif of TRPML1 , can bind PI ( 3 , 5 ) P2 . Therefore , our data are highly indicative of a prominent involvement of the Mtmr2 substrate PI ( 3 , 5 ) P2 in controlling somatosensory mechanosensitivity . Whether or not the expected change of PI ( 5 ) P after PI ( 3 , 5 ) P2 catalysis ( please see scheme in Figure 4a ) plays a role could not be investigated due to the lack of tools and knowledge about PI ( 5 ) P-specific physiology ( McCartney et al . , 2014a; Zolov et al . , 2012 ) . Interestingly , the here identified PI ( 3 , 5 ) P2-binding region is conserved among Piezo2 in vertebrates and exhibits 50% of sequence identity to mouse Piezo1 with conservation of basic amino acid residues . Arthropods and nematodes , which only encode one Piezo protein , exhibit lower sequence similarity in this region , 40% ( D . melanogaster ) and 30% ( C . elegans ) , respectively . It would be of high interest to determine in future studies whether other members of the Piezo family also bind PI ( 3 , 5 ) P2 and are potentially regulated by Mtmr2 or its homologs . Yet , in mice the here described PI ( 3 , 5 ) P2-binding and inhibition via Mtmr2 seemed quite specific for Piezo2 as indicated by our peptide-lipid binding assays and functional experiments on the domain-swapped Piezo2 P1 mutant . In contrast to the majority of studies on the regulation of vertebrate mechanotransduction ( Borbiro et al . , 2015; Eijkelkamp et al . , 2013; Morley et al . , 2016; Poole et al . , 2014; Qi et al . , 2015 ) , we did not observe alterations in Piezo1-mediated MA currents in HEK293 cells , and the magnitude of IA- and SA-MA currents was largely unaffected in DRG cultures . Surprisingly though , we detected a mild increase in the proportion of RA-MA cells paralleled by a decrease of the IA-MA subpopulation . While there is some evidence that Piezo2 might also play a role for IA-MA currents ( Lou et al . , 2013 ) , to date the molecular identity of IA-MA currents has not been resolved ( Viatchenko-Karpinski and Gu , 2016 ) . For that reason the interpretation of this finding awaits further clarification of the molecular nature of IA-MA currents . In principle , the regulation of Piezo2 by lipids is not unexpected and our data significantly advance our knowledge about the link between mechanotransduction and components of the lipid bilayer ( Brohawn et al . , 2014a , 2014b; Cox et al . , 2016; Lewis and Grandl , 2015; Sukharev , 2002; Wu et al . , 2016 ) . In particular , PI ( 4 , 5 ) P2 and its precursor PI ( 4 ) P ( Borbiro et al . , 2015 ) as well as cholesterol ( Qi et al . , 2015 ) were recently found to be implicated in mechanotransduction mediated by both , Piezo1 and Piezo2 . In contrast to PI ( 4 , 5 ) P2 , our understanding of PI ( 3 , 5 ) P2 function , localization and regulation is limited to date ( McCartney et al . , 2014a ) . PI ( 3 , 5 ) P2 is much less abundant than most PIPs , for example 125-fold less than PI ( 4 , 5 ) P2 in mammalian cells ( Zolov et al . , 2012 ) , and tightly regulated by a large protein complex ( Vaccari et al . , 2011; Zolov et al . , 2012 ) ( please see scheme in Figure 4a ) . PI ( 3 , 5 ) P2 was believed to predominantly act in the endo-lysosome system of eukaryotes ( Di Paolo and De Camilli , 2006; Dong et al . , 2010; Ho et al . , 2012; Zolov et al . , 2012 ) . By now the picture is emerging that PI ( 3 , 5 ) P2 can serve a diverse range of cellular functions , such as autophagy , signaling in response to stress , control of membrane traffic to the plasma membrane as well as regulation of receptors and ion channels ( Dong et al . , 2010; Ho et al . , 2012; Ikonomov et al . , 2007; Klaus et al . , 2009; McCartney et al . , 2014b; Tsuruta et al . , 2009; Zhang et al . , 2012 ) . Moreover , depending on the cell type , PI ( 3 , 5 ) P2 synthesis and metabolism are dynamically regulated and subject to cellular stimulation and stressors , for example insulin-mediated PI ( 3 , 5 ) P2 changes in adipocytes ( Ikonomov et al . , 2007 ) , neuronal activity-dependent synthesis at hippocampal synapses ( McCartney et al . , 2014b; Zhang et al . , 2012 ) and the here exploited regulation of PI ( 3 , 5 ) P2 upon osmotic shock in mammalian cell lines ( Dove et al . , 1997; Nicot and Laporte , 2008; Previtali et al . , 2007 ) . These reports nourish the notion that changes of PI ( 3 , 5 ) P2 via Mtmr2 might contribute to regulating Piezo2 and , by extension , touch sensitivity . How could this be achieved mechanistically ? Instead of acting in a cell-wide manner , PI ( 3 , 5 ) P2 synthesis and turnover is confined to membrane microdomains . Despite their unknown composition and subcellular localization , these microdomains are believed to ensure dynamic and local control of PI ( 3 , 5 ) P2 levels ( Ho et al . , 2012; Jin et al . , 2016; McCartney et al . , 2014a ) . Concomitantly , the abundance of downstream effector proteins is likely to be altered , as well ( Ho et al . , 2012; Jin et al . , 2016; McCartney et al . , 2014a ) . Mtmr2 may physically bind to Piezo2 in order to ensure its enrichment in PI ( 3 , 5 ) P2 microdomains , so that local PI ( 3 , 5 ) P2 changes or yet to be identified effector proteins can efficiently modulate Piezo2 ( Figure 6 , working model ) . In these microdomains Mtmr2 would catalytically decrease PI ( 3 , 5 ) P2 levels , which in turn could inhibit Piezo2 function . Hence , a physical interaction between Mtmr2 and Piezo2 would selectively concentrate Piezo2 at the sites of PI ( 3 , 5 ) P2 depletion thereby allowing its inhibition in a defined membrane compartment . On the contrary , reduced Mtmr2 expression or activity would be expected to increase local PI ( 3 , 5 ) P2 levels ( Laporte et al . , 1998; McCartney et al . , 2014b; Mironova et al . , 2016; Previtali et al . , 2007; Vaccari et al . , 2011 ) and facilitate Piezo2 potentiation . It is conceivable that Mtmr2-controlled PI ( 3 , 5 ) P2 availability could in turn cause local changes in membrane tension ( Lewis and Grandl , 2015; Perozo et al . , 2002; Vásquez et al . , 2014; Wu et al . , 2017 ) . The latter has already been demonstrated to affect Piezo1 MA currents ( Cox et al . , 2016; Lewis and Grandl , 2015; Wu et al . , 2016 ) . Whether Piezo2 is also sensitive to local membrane tension and whether PI ( 3 , 5 ) P2 levels indeed influence membrane tension in sensory neurons remains to be investigated . Unfortunately , local changes in membrane tension in the vicinity of Piezo2 are likely too small to be captured by our AFM-based measurements . Nevertheless , the data presented here allow us to infer an attractive mechanism exquisitely suited for transient and compartmentalized control of Piezo2 function , that is physical vicinity to Mtmr2 , the activity status of Mtmr2 and consequently PI ( 3 , 5 ) P2 availability ( Figure 6 ) . We can further speculate that this local control of Piezo2 function may serve to tune the threshold and magnitude of neuronal activity to light touch . Ultimately , Mtmr2 and PI ( 3 , 5 ) P2 may represent a means by which touch sensitivity of an organism can be dynamically adjusted in response to diverse stimuli modulating Mtmr2 and PI ( 3 , 5 ) P2 levels ( Figure 6 ) . In this respect it is noteworthy that multiple Mtmr2 mutations – including those affecting its activity ( Berger et al . , 2002 ) – have been implicated in Charcot-Marie-Tooth type 4B1 ( CMT4B1 ) disease , a peripheral neuropathy characterized by abnormalities in myelination and nerve conduction ( Bolino et al . , 2004; Bolis et al . , 2005; Bonneick et al . , 2005 ) . An analysis of Piezo2 mechanotransduction , light touch as well as tactile hypersensitivity in Mtmr2 knockout ( Bolino et al . , 2004 ) and Mtmr2 mutant mice ( Bonneick et al . , 2005 ) would be warranted to thoroughly examine the potential role of Mtmr2 for ( patho ) physiological aspects of vertebrate mechanosensation . Several important questions remain to be investigated . While we identified a PI ( 3 , 5 ) P2 binding motif in Piezo2 in vitro , we can neither gauge the affinity of the fully assembled Piezo2 trimer for PI ( 3 , 5 ) P2 nor assess whether PI ( 3 , 5 ) P2 and Mtmr2 bind allosterically , competitively or independently of each other . Moreover , our peptide-lipid binding assay shows that the extremely abundant PI ( 4 , 5 ) P2 and to a lesser extent PI ( 3 , 4 ) P2 could bind the same motif as PI ( 3 , 5 ) P2 . This raises the question how PIP2 specificity can be achieved by Piezo2 when embedded in cellular membranes . Piezo2 might be differentially localized in membranes dependent on PI ( 3 , 5 ) P2 levels and Mtmr2 abundance . We did not observe any differences in the overall abundance of Piezo2 channels at the plasma membrane of sensory neurons . Yet , given that PI ( 3 , 5 ) P2 acts in membrane microdomains , it would be desirable to visualize whether Piezo2 is localized in subcellular membrane compartments , that is ( i ) in PI ( 3 , 5 ) P2-enriched versus PI ( 3 , 5 ) P2-depleted and/or ( ii ) Mtmr2-harboring versus Mtmr2-negative microdomains . It remains to be seen whether these membrane compartments are confined to the plasma membrane and/or to intracellular membranes such as endo-lysosomes , which are known to contain the majority of PI ( 3 , 5 ) P2 ( Di Paolo and De Camilli , 2006; Dong et al . , 2010; Zolov et al . , 2012 ) . Along these lines it is worth mentioning that AMPA receptor abundance at hippocampal synapses has been shown to be regulated by PI ( 3 , 5 ) P2-controlled cycling through early and late endosomes ( McCartney et al . , 2014b ) . While unknown so far , localization of Piezo2 to intracellular membranes would not be unexpected ( Coste et al . , 2010 ) since its family member Piezo1 has originally been described to reside in the endoplasmic reticulum ( McHugh et al . , 2010 ) . Thus , exploring endocytosis and intracellular trafficking of Piezo2 may offer novel insights into its regulation by the intracellular membrane pool of PI ( 3 , 5 ) P2 . To address some of these issues the development of high-affinity Mtmr2 and Piezo2 antibodies as well as PI ( 3 , 5 ) P2-specific probes ( Hammond et al . , 2015; Li et al . , 2013b; McCartney et al . , 2014a; Nicot and Laporte , 2008; Previtali et al . , 2007 ) , which faithfully represent their subcellular spatial and temporal dynamics , would be required . In contrast to PI ( 4 , 5 ) P2 , such probes have not yet been successfully designed for PI ( 3 , 5 ) P2 , and the value of the few existing probes is questionable due to spatial restrictions and limited specificity for PI ( 3 , 5 ) P2 ( Hammond et al . , 2015; Li et al . , 2013b; McCartney et al . , 2014a; Nicot and Laporte , 2008; Previtali et al . , 2007 ) . Future studies using these tools combined with high-resolution microscopy have the potential to address fundamental aspects of Piezo2 trafficking , localization , and function . Taken together , our data present Mtmr2 as a novel modulator of the mechanosensory apparatus , and we provide evidence for the functional convergence of Mtmr2 enzymatic activity and PI ( 3 , 5 ) P2 availability onto Piezo2-mediated mechanotransduction in peripheral sensory neurons . In essence , we propose the Mtmr2-Piezo2 interaction as a previously unappreciated mechanism to locally and dynamically regulate Piezo2 function and , consequently , the organism´s response to light tough . Therefore , our study significantly advances our understanding of the complex molecular machinery underlying somatosensory mechanosensitivity in vertebrates . Preparation and culture of mouse DRG neurons were performed as described previously ( Coste et al . , 2010; Narayanan et al . , 2016 ) . Throughout the study , DRG were isolated from 9 to 10 week old male C57BL/6J mice or , in case of experiments in Figure 1a , Figure 1—figure supplement 1a , c , Figure 2—figure supplement 1c , d from Piezo2GFP mice ( Woo et al . , 2014 ) . In brief , DRG neurons were promptly isolated and digested with collagenase ( Thermo Fisher Scientific , Germany ) and papain ( Worthington , Lakewood , USA ) . Neurons were plated on poly-D-lysine ( 1 mg/mL , Merck Millipore , Germany ) coated coverslips , which were additionally coated with laminin ( 20 µg/mL , Thermo Fisher Scientific ) . Growth medium ( Hams F12/DMEM 1:1 ratio with L-glutamine; Gibco , Germany ) was supplemented with 10% horse serum ( Thermo Fisher Scientific ) and 100 ng/ml NGF , 50 ng/ml GDNF , 50 ng/ml BDNF , 50 ng/ml NT-3 , and 50 ng/ml NT-4 ( all growth factors were procured from R&D Systems , Germany ) . Transfection of neurons was achieved by nucleofection of siRNA or plasmid into freshly isolated DRG neurons using the P3 Primary Cell 4D Nucleofector X Kit with the 4D-Nucleofector X Unit according to the manufacturer's instructions ( Lonza , Germany ) . 500 nM FlexiTube GeneSolution Mtmr2 siRNA ( Qiagen , #GS77116 , Germany ) or AllStar Negative control siRNA ( CTRL , Qiagen , #SI03650318 ) for knockdown in DRG neurons and 0 . 5 µg of pCMV6 Mtmr2-myc-DDK ( Origene ) or 0 . 5 µg of pCMVSport6 or pcDNA3 . 1 myc-His ( CTRL ) for overexpression in DRG neurons , was used . After nucleofection , neurons were allowed to recover in calcium free RPMI medium ( Thermo Fisher Scientific ) for 10 min at 37°C before plating in growth medium . Two hours after transfection half of the growth medium was exchanged with fresh medium and neurons were grown for 48–72 hr before being used for electrophysiology , immunostaining or qPCR . HEK293 cells were authenticated by ATCC ( Manassas , USA ) upon purchase . Thereafter , cell line identity was authenticated by regular morphological inspection . Symptoms for mycoplasma contamination were not observed and thus no test for mycoplasma contamination was performed . Cells were cultured in DMEM with Glutamax ( Thermo Fisher Scientific ) supplemented with 10% FBS ( Fetal bovine serum , Thermo Fisher Scientific ) and 5% Pen/Strep ( Thermo Fisher Scientific ) . Cells were grown up to 80–90% confluence before being used for transfection . Transfection was done using Fugene HD Transfection reagent ( Promega , Germany ) . Cells were plated on poly-D-lysine-coated coverslips and maintained in culture for 48 hr before being used for electrophysiology or proximity ligation assay ( PLA ) . pCMVSport6 Piezo2-GST-IRES GFP ( kind gift from Prof . Ardem Patapoutian , La Jolla , USA ) ; pCMV6-Entry Mtmr2-myc-DDK ( Origene , #MR215223 ) ; pCMV6-Entry Mtmr2C417S-myc-DDK ( mutation as described in ( Berger et al . , 2002 ) ; pCMV Sport6 Piezo1-753-myc-IRES GFP ( kind gift from Ardem Patapoutian ) ( Coste et al . , 2015 ) ; pGEM-Teasy Kv1 . 1-HA; pCMVSport6; pCDNA3 . 1-myc-His ( Invitrogen , #V80020 ) ; pCNDA3-GST ( kind gift from Prof . Ardem Patapoutian ) ; pCMVSport6 Piezo2 P1 mutant-GST-IRES GFP . Mtmr2C417S mutant and Piezo2 P1 mutant were prepared using Q5 Site-Directed Mutagenesis kit ( New England BioLabs ) . Mutagenesis was done according to manufactures instructions and all mutant plasmids were verified by sequencing . Primers used for the mutagenesis were as follows: Mtmr2 mutagenesis forward primer: GTGGTACACTCCAGTGATGGATG; Mtmr2 mutagenesis reverse primer: CACAGACGTCTTCCCAGA; Piezo2 mutagenesis forward primer: GTCTTCTGGTGGCTCGTGGTCATTTATACCATGTTGG; Piezo2 mutagenesis reverse primer: ACGCAGCAGCTTCCTCCACCACTCGTAGTGCAC . Total RNA was isolated from cultured DRG neurons ( transfected with Mtmr2 siRNA or CTRL , please see above ) , using NucleoSpin RNA XS ( Macherey-Nagel ) according to the manufacturer's instructions . First-strand cDNA synthesis was done using QuantiTect reverse transcription kit ( Qiagen ) . Mtmr2 and Piezo2 gene expression was assessed in both conditions by real-time qPCR using the SYBR green system ( Power SYBR Green PCR Master Mix; Thermo Fisher Scientific ) on a LightCycler 480 instrument ( Roche , Germany ) . The melting curve analysis of amplified products was used to confirm the specificity of qPCR assay . All samples were run in triplicate and negative control reactions were run without template . Threshold cycle ( Ct ) values , the cycle number in which SYBR green fluorescence rises above background , were normalized to two reference genes ( Actb and Gapdh ) and recorded as a measure of initial transcript amount . Relative quantification was performed using the ‘fit point’ as well as the ‘second derivative maximum’ method of the LightCycler 480 . Primer sequences 5′−3′ are the following: Mtmr2 ( fw: tgtaccccaccattgaagaaa; rev: taagagcccctgcaagaatg ) , Piezo2 ( fw: aggcagcacataggatggat; rev: gcagggtcgcttcagtgta ) , Actb ( fw: gatcaagatcattgctcctcctg; rev: cagctcagtaacagtccgcc ) , Gapdh ( fw: caatgaatacggctacagcaac; rev: ttactccttggaggccatgt ) . Of note , only data normalized to Actb are shown but data normalized to Gapdh gave similar results . Our qPCR results indicate successful siRNA-mediated knockdown of Mtmr2 across the whole coverslip , which also includes non-transfected neurons and glia cells . Therefore our data do not report on the transfection efficiency and extent of Mtmr2 knockdown in individual neurons . Whole-cell voltage clamp recordings were performed in transfected DRG cultures , wild type DRG cultures or transfected HEK293 cell cultures at room temperature as described in ( Narayanan et al . , 2016 ) . Briefly , ( protocol adapted from [Coste et al . , 2010] ) to elicit mechanically activated currents , the cell soma was mechanically stimulated using a blunt probe ( fire polished borosilicate glass capillary ) . The stimulation was delivered using a piezo-electrically driven micromanipulator ( Physik Instrumente GmbH and Co . KG , Germany ) . The probe was initially positioned ~4 µm from the cell body and had a velocity of 0 . 8 µm/ms during the ramp phase ( forward motion ) . The stimulus was applied for 150 ms with an inter-stimulus interval of 180 ms . Stimulus-current measurements were performed using mechanical stimulations from 0 to 6 µm in 1 µm increments at a holding potential of −70 mV in whole cell mode . All recordings other than specifically indicated were made in standard extracellular solution containing ( in mM ) 127 NaCl , 3 KCl , 1 MgCl2 , 2 . 5 CaCl2 , 10 Glucose and 10 HEPES; pH = 7 . 3; osmolarity = 285 mOsm ( Coste et al . , 2010 ) . To achieve hypotonicity the extracellular solution was adjusted to 160 mOsm . The intracellular solution for DRG neurons contained ( in mM ) 133 CsCl , 10 HEPES , 5 EGTA , 1 MgCl2 , 1 CaCl2 , 4 MgATP and 0 . 4 NaGTP; pH = 7 . 3; osmolarity = 280 mOsm ( Coste et al . , 2010 ) . Hypotonic intracellular solution contained ( in mM ) 65 CsCl , 10 HEPES , 5 EGTA , 1 MgCl2 , 1 CaCl2 , 4 MgATP and 0 . 4 NaGTP; pH = 7 . 3; osmolarity = 162 mOsm . Intracellular solution for HEK293 cells contained ( in mM ) 110 KCl , 10 NaCl , 1 MgCl2 , 1 EGTA and 10 HEPES; pH = 7 . 3 ( Poole et al . , 2014 ) . Based on protocols published elsewhere ( Jia et al . , 2016 ) recordings in extracellular hypotonic solutions were performed as follows: Cells were incubated with the hypotonic solution for 5 min at 37°C prior to recording , after which RA-MA currents were recorded in hypotonic extracellular solution between 0–20 min after initial application . For some experiments described in this study , DRG neurons were treated with chemical inhibitors of the phosphatidylinositol phosphate pathway . Cells were treated with 35 µM Wortmannin ( Sigma Aldrich; protocol adapted from [Mo et al . , 2009] ) in DMSO or 1 µM Apilimod ( Bertin Pharma , Germany ; protocol adapted from [Mironova et al . , 2016] ) in DMSO for 2 hr prior to recording . DMSO ( 0 . 08% for Apilimod and 0 . 35% for Wortmannin ) , was used as vehicle for each experiment . For PIP addition experiments , 1 µM PI ( 3 , 5 ) P2 ( Echelon ) or 1 µM PI ( 3 ) P ( Echelon ) were included in the isotonic intracellular solution ( protocol adapted from [Dong et al . , 2010] ) . Of note , the water-soluble diC8 form of the lipids was used as indicated elsewhere ( Dong et al . , 2010 ) . Data analysis and representation was done using Fitmaster ( HEKA Electonik GmbH , Germany ) and Igor Pro 6 . 37 ( WaveMetrics , USA ) . The current magnitude was calculated by measuring peak amplitudes ( at each stimulus point ) and subtracting leak current . The displacement threshold was defined as the minimum displacement required to elicit a visible RA-MA current , that is the displacement at which current values exceeded 100 pA in DRG ( Jia et al . , 2016; Narayanan et al . , 2016 ) and 50 pA in HEK293 cells ( Szczot et al . , 2017 ) . In HEK293 cells , this cutoff also served to prevent contamination of recordings by HEK293 endogenous Piezo1 ( Dubin et al . , 2017; Szczot et al . , 2017 ) . Of note , responses in HEK293 cells increased proportionally to stimulus strength and were absent in cells not transfected with Piezo1 or Piezo2 plasmids . For measurements of inactivation kinetics current traces reaching at least 75% of maximal current amplitude were fitted with a mono-exponential or bi-exponential equation and the fast time constant was used for analysis ( Coste et al . , 2010 ) . Data for the displacement threshold and inactivation time constant ( τ ) are provided in Supplementary file 1 . In DRG cultures neuronal soma size differs considerably within a coverslip . Therefore neurons have been traditionally categorized into small , medium and large diameter neurons ( Drew et al . , 2004; Hu and Lewin , 2006; Patapoutian et al . , 2009; Woo et al . , 2014 ) exhibiting well-reported differences in function and MA currents ( Drew et al . , 2004; Hu and Lewin , 2006; Patapoutian et al . , 2009; Poole et al . , 2014; Woo et al . , 2014 ) . To account for variability of our data due to soma size , we ( i ) recorded RA currents from visibly large diameter neurons ( cell capacitance >40 pF ) likely representing mechanoreceptors ( Drew et al . , 2004; Hu and Lewin , 2006; Poole et al . , 2014; Woo et al . , 2014 ) , ( ii ) analyzed SA currents from small diameter neurons ( cell capacitance <30 pF ) likely representing nociceptors ( Drew et al . , 2004; Hu and Lewin , 2006; Poole et al . , 2014; Woo et al . , 2014 ) and ( iii ) additionally normalized all obtained current data to the individual cell capacitance to obtain a value for current density ( pA/pF ) . The latter analysis ( data not shown ) yielded similar results as the presented analysis based on current amplitudes for all DRG datasets in this manuscript . IA currents were randomly recorded across all neuron sizes . Of note , current amplitudes and displacement thresholds cannot be compared among experimental datasets because of differential treatments ( e . g . nucleofection , inhibitors ) requiring measurements at different culture days in vitro ( DIV ) , and due to the inherent variability in DRG cultures dependent on mouse cohorts used throughout the course of this study . For this reason each dataset consists of experiments and respective controls measured in parallel ( i ) in the same mouse cohort and ( ii ) , where possible , on each experimental day . For each dataset ( experimental versus control conditions ) several coverslips from at least 4 ( range: 4–18 ) independent cell cultures or 2–12 independent platings of HEK293 cells were used . Immunostaining was carried out as described ( Narayanan et al . , 2016 ) . Briefly , Piezo2GFP mice ( Woo et al . , 2014 ) ( ages 8–9 weeks ) were euthanized with CO2 or perfused with 4% PFA ( Science Services ) . DRG neurons were isolated and cultured using the protocol described above or DRG were carefully dissected , collected in 4% PFA/1X PBS , and post-fixed for 30 min at 4°C . After overnight cryoprotection in 30% sucrose tissues were frozen in optimal cutting temperature medium , sectioned with a cryostat into 10 μm thick sections , mounted on SuperFrost Plus slides , and stored at −80°C . Frozen slides were thawed at room temperature for 30 min , washed thrice with 1X PBS , blocked for 30 min in 1X PBS containing 5% goat or donkey serum ( Dianova , Germany ) and 0 . 4% TritonX-100 ( Roth ) , and incubated with primary antibodies ( diluted in 0 . 1% TritonX-100% and 1% serum in 1X PBS ) , overnight at 4°C . The sections were then washed with 1X PBS and incubated for 2 hr at room temperature with secondary antibodies diluted 1:250 in 0 . 1% TritonX-100% and 1% serum in 1X PBS . Sections were then washed six times with 1X PBS and mounted in SlowFade Gold antifade reagent with DAPI ( Thermo Fisher Scientific ) . For cultured DRG neurons , coverslips containing cells were washed with 1X PBS and fixed in 4% PFA for 10 min . Thereafter cells were washed with 1X PBS and blocked with blocking solution containing 5% serum and 0 . 4% TritonX-100 . Cells were then incubated with primary antibodies ( diluted in 1% serum and 0 . 1% TritonX-100 ) , overnight at 4°C . Cells were then washed with 1X PBS and incubated with secondary antibodies ( diluted in 1% serum and 0 . 1% TritonX-100 ) for 2 hr at room temperature followed by additional washes . Coverslips were mounted onto SuperFrost Plus slides using SlowFade Gold reagent with DAPI ( Thermo Fisher Scientific ) . WGA staining to mark membranes of cells was done as follows; before fixation , the cells were treated with WGA-555 ( 1:200 ) for 15 min at 37°C . Cells were then washed three times with medium and the standard immunostaining protocol was performed . Imaging was done using a Zeiss Axio Observer Z1 inverted microscope or a Zeiss LSM 510 Meta Confocal microscope . All images were processed and analyzed by ImageJ , NIH ( Schindelin et al . , 2015 ) . For the analysis of Mtmr2 in cryo-frozen sections or primary cultures of DRGs from Piezo2GFP mouse , sections were stained with anti-GFP and anti-Mtmr2 antibodies . Positive cells were identified by setting the threshold as ‘mean intensity +3standard deviation’ of randomly selected ( ~10 ) negative cells . The numbers of positive and negative cells were counted using the ‘cell counter’ plugin of ImageJ . Only for presentation purposes brightness , contrast and levels of images were adjusted in Adobe Photoshop . In all cases image adjustments were applied equally across the entire image and equally to controls . For the analysis of membrane intensity of Piezo2-GFP upon Mtmr2 knockdown , the ‘analyze particle tool’ from ImageJ was used . The WGA staining was used to mark the region of interest around the cell membrane and for each ROI , the mean intensity ( arbitrary units , AU ) and area of positive signal was determined . The mean intensity of signal ( arbitrary units , AU ) was calculated by subtracting the threshold ( defined as ‘mean +3standard deviation’ of background ) from the total mean intensity . PLA was carried out as described ( Hanack et al . , 2015; Narayanan et al . , 2016 ) with minor modifications . HEK293 cells or DRG neurons were plated on MatTek dishes coated with poly-D-lysine ( and Laminin for DRG neuron cultures ) and transfected with appropriate plasmids ( details of plasmids are provided below ) . Cells were cultured for 48 hr before staining . Cells were washed with 1X PBS and fixed with 4% PFA for 10 min at room temperature . Thereafter cells were blocked in Duolink Blocking Solution ( Sigma Aldrich , Germany ) for 2 hr at room temperature . Cells were then incubated with primary antibodies ( diluted in Duolink Antibody Diluent ( Sigma Aldrich ) ) , overnight at 4°C . Cells were washed with wash buffer A ( 0 . 01 M Tris , 0 . 15 M NaCl and 0 . 05% Tween 20 , pH 7 . 4 ) and incubated with PLA probes ( PLUS and MINUS probes were diluted 1:10 in Duolink antibody diluent ( Sigma Aldrich ) ) for 1 hr at 37°C . Cells were washed again with wash buffer A and incubated with amplification mix ( amplification stock 1:5 and polymerase 1:80 in water ) for 100 min at 37°C . Cells were then washed with wash buffer B ( 0 . 2 M Tris , 0 . 1 M NaCl , pH7 . 5 ) and stored in 1X PBS before imaging . Secondary controls meant omitting all primary antibodies . Plasmids used: pCMVSport6 Piezo2-GST-IRES-GFP ( kind gift from Prof . Ardem Patapoutian ) ; pCMV6-Entry Mtmr2-myc-DDK ( Origene , #MR215223 ) ; pCMV6-Entry Mtmr2C417S-mycDDK , mutation as described by Berger and colleagues ( Berger et al . , 2002 ) , was custom-generated using the Q5 Site-Directed Mutagenesis kit ( New England BioLabs ) ; pCMVSport6; pCDNA3 . 1-myc-His ( Invitrogen , #V80020 ) ; pmaxGFPVector ( Lonza ) . The PLA was imaged using a Zeiss Axio Observer Z1 inverted microscope . The imaging settings were constant across all samples of the same experiments . Secondary controls were always imaged in parallel using the same settings . Image analysis was done using ImageJ . The background was determined as the ‘mean intensity +3standard deviation’ of randomly chosen negative cells per field of view and averaged for all images within one condition . The highest background value was then used as threshold for the analysis . GFP positive cells ( from Piezo2-GST-IRES-GFP expression or pmaxGFPVector ) were chosen for each field of view and the PLA signal was analyzed for these cells , using the ‘Analyze Particle’ tool of ImageJ . The number of PLA puncta and total area of positive PLA signal for each cell was measured . To account for variability in cell size , the total area of the cell was also measured and the PLA signal values were normalized to total cell area . Only for presentation purposes brightness , levels and contrast of images were adjusted in Adobe Photoshop . In all cases image adjustments were applied equally across the entire image and equally to controls . Experiments were performed on several coverslips of at least two independently transfected HEK293 and DRG cultures , respectively . DRG neurons were nucleofected with AllStar Negative control siRNA ( CTRL ) or Mtmr2 siRNA , as described above , and maintained in culture for 72 hr . Elasticity and tether force measurements were performed with an AFM ( MFP-3D extended head , Asylum Research , Germany ) mounted on an inverted microscope ( IX71 , Olympus , Germany ) using contact mode with a triangular cantilever comprising a pyramidal tip ( TR-400-PB , Olympus ) . During the measurements cells were maintained in growth media . The spring constant of the cantilever was determined using the built in thermal method ( 24–28 pN/nm ) . Indentation and retraction speed was kept constant at 5 µm/s , and force load of 200–1000 pN was used to measure the Young’s modulus of the cells . The effective Young’s modulus Eeff was fitted with a modified Hertz model using a self-written IGOR macro ( Rehfeldt et al . , 2007 ) . Tether forces were determined as the difference of pulling force before and after rupture of a tether from the AFM tip using a semi-automated step finding procedure ( Nawaz et al . , 2015 ) . TRPML1 is known to bind PI ( 3 , 5 ) P2 through a region in its N-terminus ( Dong et al . , 2010 ) . This region of TRPML1 ( NP_444407 . 1 ) was compared to mouse Piezo2 ( NP_001034574 . 4 ) and mouse Piezo1 ( NP_001032375 . 1 ) using NCBI protein Blast ( National Library of Medicine ( US ) , National Center for Biotechnology Information ) . The protocol was adapted from ( Berger et al . , 2002 ) with minor modifications . In brief , PIP strips ( Echelon Biosciences , Inc . , Salt Lake City , USA ) were washed once with PBS-T ( 1x PBS + 1% Tween ) and blocked with 3% fat free BSA ( bovine serum albumin; Sigma Aldrich ) in 1X PBS for 1 hr at room temperature . The membranes were then incubated with 0 . 5 µg/mL peptide solution ( peptide dissolved in 1% BSA ) for 2 hr at room temperature . Experiments and controls were processed in parallel . Membranes were washed with PBS-T three times for 7 min each and then incubated with primary antibody at room temperature for 2 hr . Membranes were then washed with PBS-T and probed with secondary antibodies coupled with Alexa680 for 1 hr at room temperature . Imaging was done on the Odyssey Infrared System ( LI-COR , Germany ) . Only for presentation purposes brightness , gradient levels and contrast of images were adjusted in Adobe Photoshop . In all cases image adjustments were applied equally across the entire image and equally to controls . The following peptides were used in this study ( all procured from GenScript , New Jersey , USA ) : Piezo2 ( 731-746 ) -FLAG tagged [EWWRKILKYFWMSVVIDYKDDDDKQNN]; Piezo2 3Q mutant ( 731-746 ) -FLAG tagged [EWWQQILQYFWMSVVIDYKDDDDKQNN]; Piezo1 ( 626-639 ) -FLAG tagged [TLWRKLLRVFWWLVDYKDDDDKqnn] . The following antibodies were used in this study: 1:100 rabbit anti-Mtmr2 ( Biotechne , Minneapolis , USA; #NBP1-33724 ) ; 1:500 chicken anti-GFP ( Thermo Fisher Scientific , #A10262 ) ; 1:250 ( Immunocytochemistry ) , 1:500 ( PLA ) rabbit anti-GST ( Santa Cruz , Santa Cruz , USA; #sc-459 ) ; 1:100 ( immunocytochemistry and immunoblotting ) , 1:750 , 1:500 ( PLA in HEK293 cells and DRG neurons respectively ) mouse anti-myc ( Santa Cruz , #sc-47694 ) ; 1:200 Rabbit anti-Piezo2 ( Novus Biologicals , Germany; #NBP1-78624 ) ; 1:500 mouse anti-FLAG ( Sigma Aldrich , #F1804 ) , Secondary antibodies conjugated to Alexa Fluor 488 , Alexa Fluor 546 , Alexa Fluor 647 , Alexa Fluor 680 ( Thermo Fisher Scientific ) , Duolink in situ PLA probes 1:10 anti-rabbit MINUS , 1:10 anti-mouse PLUS . Data was analyzed using GraphPad Prism 6 . 01 ( San Diego , USA ) . All data are represented as mean ± SEM ( standard error of mean ) unless indicated otherwise . All replicates were biological . All statistical tests are two-sided unless indicated otherwise . In all panels: ns > 0 . 05; p≤0 . 05; p≤0 . 01; p≤0 . 001; **p≤0 . 0001 ) . PLA data: Mann-Whitney test or Kruskal-Wallis test with Dunn’s Multiple Comparison test . Atomic force microscopy ( AFM ) data: Mann-Whitney test . qPCR: One sample t-test was used ( values were compared to a theoretical mean of 1 . 00 , that is mRNA expression in CTRL ) . Immunostaining: For membrane Piezo2 expression and number of Piezo2-positive neurons upon Mtmr2 knockdown , the Mann-Whitney test was used . Peptide-lipid binding assays: One-way ANOVA followed by Dunnett´s or Holm-Sidak’s multiple comparisons test was used as indicated . Electrophysiology: For the analysis of stimulus-current curves , 2-way ANOVA with Holm-Sidak’s multiple comparisons test was used . The P-value represents the results of 2-way ANOVA , testing the overall effect of the respective treatment . Results of the Holm-Sidak’s multiple comparisons test are represented by p-values and indicated in each legend . Outlier analysis was carried out using the Grubb’s test followed by testing whether the outlier value exceeded ‘mean +3 standard deviation’ . Outlier analysis was only performed on current values at maximal stimulation . Only if a value met both criteria ( Grubb’s outlier and >’mean +3*standard deviation’ ) the cell was excluded from further analysis . Datasets , where a single outlier was removed: Figure 1h , Figure 2b , Figure 4b , Figure 4d , Figure 4—figure supplement 1a , c . For the analysis of the displacement threshold and inactivation time constant the Mann-Whitney test was used , unless more than two groups were compared , for which one-way ANOVA or Kruskal-Wallis test followed by Dunn’s multiple comparisons test were used . For the analysis of mechanically activated ( MA ) current populations Chi-square test was used and data are represented as % of all analyzed cells .	We often take our sense of touch for granted . Yet , our every-day life greatly depends on the ability to perceive our environment to alert us of danger or to further social interactions , such as mother-child bonding . Our sense of touch relies on the conversion of mechanical stimuli to electrical signals ( this is known as mechanotransduction ) , which then travel to brain to be processed . This task is fulfilled by specific ion channels called Piezo2 , which are activated when cells are exposed to pressure and other mechanical forces . These channels can be found in sensory nerves and specialized structures in the skin , where they help to detect physical contact , roughness of surfaces and the position of our body parts . It is still not clear how Piezo2 channels are regulated but previous research by several laboratories suggests that they work in conjunction with other proteins . One of these proteins is the myotubularin related protein-2 , or Mtmr2 for short . Now , Narayanan et al . – including some of the researchers involved in the previous research – set out to advance our understanding of the molecular basis of touch and looked more closely at Mtmr2 . To test if Mtmr2 played a role in mechanotransduction , Narayanan et al . both increased and reduced the levels of this protein in sensory neurons of mice grown in the laboratory . When Mtmr2 levels were low , the activity of Piezo2 channels increased . However , when the protein levels were high , Piezo2 channels were inhibited . These results suggest that Mtmr2 can control the activity of Piezo2 . Further experiments , in which Mtmr2 was genetically modified or sensory neurons were treated with chemicals , revealed that Mtmr2 reduces a specific fatty acid in the membrane of nerve cells , which in turn attenuates the activity of Piezo2 . This study identified Mtmr2 and distinct fatty acids in the cell membrane as new components of the complex setup required for the sense of touch . A next step will be to test if these molecules also influence the activity of Piezo2 when the skin has become injured or upon inflammation .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2018	Myotubularin related protein-2 and its phospholipid substrate PIP2 control Piezo2-mediated mechanotransduction in peripheral sensory neurons
Actin has well established functions in cellular morphogenesis . However , it is not well understood how the various actin assemblies in a cell are kept in a dynamic equilibrium , in particular when cells have to respond to acute signals . Here , we characterize a rapid and transient actin reset in response to increased intracellular calcium levels . Within seconds of calcium influx , the formin INF2 stimulates filament polymerization at the endoplasmic reticulum ( ER ) , while cortical actin is disassembled . The reaction is then reversed within a few minutes . This Calcium-mediated actin reset ( CaAR ) occurs in a wide range of mammalian cell types and in response to many physiological cues . CaAR leads to transient immobilization of organelles , drives reorganization of actin during cell cortex repair , cell spreading and wound healing , and induces long-lasting changes in gene expression . Our findings suggest that CaAR acts as fundamental facilitator of cellular adaptations in response to acute signals and stress . Actin organization and dynamics are critical for most morphogenetic processes , including cell polarization , migration and division . The prominent role of the actin cytoskeleton is reflected in a host of regulators that mediate dynamic assembly of complex actin structures from filament bundles and networks . These structures in turn provide protrusive and contractile forces during physiological and pathological processes such as differentiation , wound healing and tumor metastasis ( Lecuit et al . , 2011; Pantaloni et al . , 2001; Pollard and Cooper , 2009 ) . While many studies in the past have focused on the molecular characterization of specific actin assemblies and their local function within cells , recent work has elegantly shown that a balance between different actin assemblies is established in cells by distinct actin nucleators . These nucleators constantly compete for a common pool of actin monomers ( Burke et al . , 2014 ) , and the balance between actin assemblies can be tightly controlled by regulatory factors such as profilin ( Rotty et al . , 2015; Suarez et al . , 2015 ) . In addition , cellular F- and G-actin levels have been linked to profound and long-lasting changes in cell physiology through regulation of transcriptional cofactors such as myocardin-related transcription factors ( MRTF ) , the Yes-associated protein ( YAP ) or transcriptional co-activator with PDZ-binding motif ( TAZ ) ( Halder et al . , 2012; Miralles et al . , 2003 ) . Finally , the rapidly expanding field of mechanobiology has highlighted global integration of actomyosin-mediated forces across the cortex of single cells ( Klingner et al . , 2014; Salbreux et al . , 2012 ) , as well as within cell groups and whole tissues ( Lecuit et al . , 2011; Trepat et al . , 2012 ) . In the context of widespread interdependence , competition and mechanical connectivity between actin assemblies , cells face formidable challenges when undergoing a global reorganization of their actin cytoskeleton . This is exemplified by the impact of moderate perturbations of the cortical actin cytoskeleton with low doses of latrunculin , which can result in a massive imbalance of actomyosin-mediated forces and lead to large-scale fluctuations in cell cortex organization ( Klingner et al . , 2014; Luo et al . , 2013 ) . How cells globally coordinate actin remodeling during rapid and profound morphological transitions has not been sufficiently addressed . Shear flow has been shown to induce rapid changes in cellular actin organization and mechanics ( Rahimzadeh et al . , 2011; Verma et al . , 2012 ) . In addition , a recent study reported calcium-mediated rapid and transient formation of actin filaments at the nuclear periphery of fibroblasts exposed to mechanical stress ( Shao et al . , 2015 ) . However , how these observations can be linked to cell cortex organization and overall cell physiology is not known . Here we show that calcium not only induces actin polymerization at the nuclear periphery , but causes a rapid reduction of cortical actin coinciding with the formation of new filaments at the entire endoplasmic reticulum ( ER ) . This global change in actin distribution is rapidly reversed within less than two minutes . Calcium-mediated Actin Reset ( CaAR ) occurs in a wide range of mammalian cell types and is initiated by a variety of physiological signals . Most importantly , we found that CaAR acts as a fundamental facilitator of acute cell responses . It is required for repair of cortical damage and coordinates the formation of actin-rich protrusions during cell spreading and cell migration . In addition , CaAR induces long-term changes in gene expression via MRTF and serum response factor ( SRF ) . Our results have important implications for a multitude of experiments where cells are exposed to acute stress as well as for calcium-regulated processes such as wound healing , inflammation , cell differentiation and cancer progression . To establish the cellular response to an acute mechanical stimulus we exposed Madin-Darby Canine Kidney ( MDCK ) epithelial cells to sudden shear flow . Prior to shear stress , MDCK cells stably expressing Lifeact-GFP exhibited a typical apical actin organization with clustered microvilli ( Klingner et al . , 2014 ) . Immediately upon exposure to fluid flow of 10–20 dyn/cm2 we observed the formation of a highly transient perinuclear actin ring , which only remained for a few minutes ( Figure 1A , Video 1 ) . A similar phenomenon was recently reported in fibroblasts ( Shao et al . , 2015 ) , but its consequences for cell organization were not further explored . To establish the reported relevance of calcium ( Shao et al . , 2015 ) for the actin reorganization in MDCK cells , we monitored Ca2+ levels with the fluorescent probe Fluo-4 . Upon induction of shear stress , actin reorganization was preceded by a strong intracellular Ca2+ pulse ( Figure 1A , B ) . To examine whether the elevation of Ca2+ levels is causally linked to actin reorganization , we treated MDCK cells with the calcium ionophore ionomycin . Indeed , exposure to as little as 0 . 3 µM ionomycin induced the formation of perinuclear actin rings and reduction of cortical actin within 60 s ( Figure 1C ) . Interestingly , in contrast to the previous report , we found a virtually simultaneous decrease of actin filaments at the apical cortex ( Figure 1C ) . As for the transient shear stress response , actin rapidly reverted to its cortical distribution within the following 60 s ( Figure 1C ) . To test the prevalence of calcium-mediated actin reorganization , we stably or transiently expressed Lifeact-GFP in a panel of mammalian cell lines , including epithelial , mesenchymal , endothelial and immune cells , and treated each with ionomycin . In all instances , rapid relocation of actin from the cell cortex to the nuclear periphery occurred within 60 s ( Figure 1—figure supplement 1A , Video 2 ) . We observed the strongest response for human MCF-7 breast cancer cells ( Figure 1D , Video 3 ) and therefore decided to focus on this cell line for further studies . Actin reorganization in MCF-7 cells was not influenced by Lifeact-GFP expression , as we observed an identical response in non-transfected cells stained with Alexa Fluor 647-phalloidin ( Figure 1—figure supplement 1B ) . Importantly , we found that , upon calcium influx , actin filaments formed not only at the nuclear periphery , but throughout the cell ( Figure 1—figure supplement 1C ) and all along the endoplasmic reticulum ( ER , Figure 1E ) . If cytosolic calcium is the key regulator of rapid actin reorganization , we reasoned that other signals that induced sufficient Ca2+ influx should be able to elicit the actin response . As expected , activation of calcium influx using physiological ligands for G-protein-coupled receptors , such as ATP or bradykinin efficiently induced actin rearrangement ( Figure 1F ) . In most cells release of calcium from ER stores activates store-operated calcium entry from the extracellular environment ( Hogan and Rao , 2015 ) . Hence , blocking uptake of Ca2+ into the ER with thapsigargin - in the presence of high extracellular Ca2+ - also induced actin rearrangement ( Figure 1F ) . Finally , perforation of the plasma membrane by localized mechanical disruption ( atomic force microscopy , AFM ) or laser-induced ablation efficiently induced cortex to ER actin reorganization ( Figure 1F ) . 10 . 7554/eLife . 19850 . 003Figure 1 . A calcium-mediated actin reset in mammalian cells . ( A , B ) MDCK cells labeled with Lifeact-mCherry ( Lifeact-RFP ) and Fluo4 were exposed to 10 dyn/cm2 shear flow . Regions used for intensity plots in ( B ) are indicated in ( A ) . ( C ) MDCK cells expressing Lifeact-GFP were stimulated with 1 µM ionomycin . ( D–F ) MCF-7 cells expressing Lifeact-GFP ( and ER-RFP in ( E ) ) were exposed to 1 µM ionomycin ( D , E ) , or to 50 µM ATP , 1 µM bradykinin or 1 µM thapsigargin or locally stressed with an AFM probe ( pointy tips ) or by laser ablation ( F ) ( asterisks at position of stimulus ) . Times in sec . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 00310 . 7554/eLife . 19850 . 004Figure 1—figure supplement 1 . A calcium-mediated actin reset in mammalian cells . ( A ) Treatment of indicated cell types expressing Lifeact-GFP with 1 µM ionomycin . ( B , C ) MCF-7 cells were treated with 1 µM ionomycin for 1 min , fixed and stained with phalloidin-Alexa647 . Images in ( C ) represent a series of z-planes acquired at 0 . 5 µm distance . Times in sec . Scale bars 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 00410 . 7554/eLife . 19850 . 005Video 1 . MDCK cells expressing Lifeact-mCherry and labeled with Fluo4 exposed to shear flow ( 10 dyn/cm2 ) . Corresponds to Figure 1A . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 00510 . 7554/eLife . 19850 . 006Video 2 . A panel of indicated cell types expressing Lifeact-GFP exposed to 1 µM ionomycin . Corresponds to Figure 1—figure supplement 1A . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 00610 . 7554/eLife . 19850 . 007Video 3 . MCF-7 cells expressing Lifeact-GFP exposed to 1 µM ionomycin . Corresponds to Figure 1D . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 007 In summary , we have identified a fundamental and conserved process of rapid , global and transient actin rearrangement , which occurs in a wide range of mammalian cells and can be induced by a variety of signals that raise cytosolic calcium levels . Due to the striking inversion of cellular actin organization that we observed ( cortex-ER-cortex , Figure 1C , D ) we decided to term this process ‘Calcium-mediated Actin Reset’ or CaAR . We proceeded to characterize the detailed kinetics and features of CaAR to gain insights into the underlying molecular mechanisms . In all cell types that we tested , CaAR was initiated within seconds after Ca2+ influx ( Figure 2—figure supplement 1A and not shown ) . Actin concentration at the ER reached its maximum within less than 1 min and reverted back to the original state within less than 5 min . Actin recruitment to the ER followed a typical three-phase kinetics characterized by a rapid increase , a plateau and a slower decrease ( Figure 2A ) . We considered serum-starved ( for 1 hr ) MCF-7 cells treated with 330 nM ionomycin at RT as a control condition . The actin increase at the ER was significantly faster for cells grown in serum or at 37°C . Also , the stimulation of MCF-7 and HeLa cells with 50 µM ATP led to faster increase than for ionomycin ( Figure 2B , Table 1 ) . In contrast , addition of serum or ATP had little effect on the rate of decrease or the amplitude of the reaction ( Figure 2B ) . Finally , the CaAR plateau was shortened at 37°C , for cells grown in serum and upon stimulation with ATP ( Figure 2B ) . Treatment of HeLa cells with ionomycin induced very strong and long-lasting CaAR response ( Figure 2B ) , indicating that these cells could not as efficiently remove Ca2+ from the cytosol . Most importantly , despite the specific differences described above , all kinetic parameters of CaAR remained within a three-fold range ( Figure 2B ) , highlighting the robust and stereotypic nature of the response . 10 . 7554/eLife . 19850 . 008Figure 2 . Quantitative analysis of CaAR . ( A , B ) Quantification of indicated parameters during CaAR . MCF-7 and HeLa cells expressing Lifeact-GFP were incubated at RT or 37°C and with serum or serum-free HBSS buffer . Cells were then treated with either 330 nM ionomycin or 50 µM ATP . Control corresponds to serum starved MCF-7 cells stimulated with 330 nM ionomycin at RT . GFP intensity was measured at the nuclear periphery and analyzed using customized Matlab scripts ( Supplementary material ) . All values are mean ± SD , n > 100 cells . See Table 1 for details . ( C ) Lifeact-GFP expressing MCF-7 cells were treated with the indicated drugs , ATP: 50 µM , ionomycin: 1 µM , thapsigargin: 1 µM in Ca2+-free medium . Intracellular calcium levels ( Fura2 ) and Lifeact-GFP intensities at the ER were monitored . Peak values: mean ± SD ( n > 30 ) . Numbers above bars indicate % cells exhibiting CaAR . ( D ) Cells treated with 1 µM ionomycin were followed over time by simultaneous fluorescence and atomic force microscopy . The relative Young’s modulus of whole cells was calculated from force-distance curves obtained with 10 µm beads ( mean ± SD , n = 24 ) . ( E ) Freezing of lysosomes labeled with Lysotracker Red in HeLa cells undergoing CaAR . Images correspond to the maximum projection of indicated periods in a time series . ( F ) Freezing of organelle motion during CaAR in MCF-7 cells . ER or mitochondria were fluorescently labeled with ss-RFP-KDEL or mitotracker Red , respectively . ( G ) Change in lysosome motility for control and blebbistatin-treated ( 50 µM ) HeLa cells . ( H , I ) Propagation of CaAR induced by laser ablation in the absence ( I ) and presence ( J ) of 50 µM ATP . Arrows: ablation sites , asterisks: cells reacting to stimulus . Times in sec after exposure to the stimulus . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 00810 . 7554/eLife . 19850 . 009Figure 2—figure supplement 1 . Quantitative analysis of CaAR . ( A ) MCF-7 cells expressing Lifeact-mCherry ( LA-RFP ) were labeled with Fluo4 and stimulated by laser ablation . Asterisks: first frame in which Fluo4 or CaAR response can be detected . ( B ) Incidence of CaAR ( in % ) induced in MCF-7 cells by the application of different loading forces with the AFM probe ( pointy tip ) . n = 10 cells per data point . ( C ) Efficiency of repeated cell stimulation with varying delays ( mean ± SD , n = 7 ) . ( D ) Repeated stimulation of cells with pointy AFM tip . ( E , F ) CaAR induction and propagation in MCF-7 cells following stimulation by AFM ( E ) or laser ablation ( F ) . Arrow: ablation site / AFM contact; asterisks: cells reacting to stimulus . Times in sec . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 00910 . 7554/eLife . 19850 . 010Table 1 . Quantitative analysis of CaAR . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 010Cell type condition MCF-7control MCF-7serum MCF-737°C MCF-7ATP HeLaIono HeLaATP 20 , 7414 , 098 , 2011 , 7815 , 595 , 34mean increasetime constant ( s ) 12 , 946 , 945 , 4211 , 1210 , 863 , 99stdev 980588102252371258n ******************ANOVA −41 , 47 −28 , 21 −17 , 15 −42 , 74 −119 , 30 −49 , 42 mean decreasetime constant ( s ) 26 , 3217 , 0611 , 7530 , 9040 , 0033 , 71stdev 1010586102279136193n ns ns ANOVA 3 , 292 , 512 , 993 , 526 , 972 , 91mean amplitudeintensity ( a . u . ) 1 , 500 , 711 , 361 , 963 , 151 , 34stdev 979588102252371255n ns ns * ANOVA 217 , 30137 , 0072 , 65162 , 70391 , 40115 , 20mean plateautime ( s ) 108 , 2061 , 4928 , 60110 , 50197 , 4062 , 77stdev 965586102238136193n ********************ANOVA Next , we investigated the role of calcium in more detail using the ratiometric dye Fura2 . We found that cytosolic Ca2+ levels in MCF-7 cells increased 2- to 4-fold upon exposure to ionomycin ( 3 . 59 ± 0 . 84 fold , n = 31 ) or ATP ( 2 . 82 ± 0 . 67 fold , n = 54 ) , which induced CaAR in virtually all cells ( Figure 2C ) . Induction of CaAR was completely prevented in Ca2+-free medium , but could be restored within 1 min by addition of Ca2+ ( not shown ) . In Ca2+-free medium , release of calcium from ER stores with thapsigargin could not be enhanced by store-operated calcium entry and led to a modest cytosolic Ca2+ increase of 1 . 26 ± 0 . 08 fold ( n = 31 ) . This was insufficient to induce CaAR ( Figure 2C ) . These results indicate that CaAR induction requires a threshold level of intracellular Ca2+ that can only be reached by influx from the extracellular environment . Observing such large-scale reorganization of actin , we wondered whether the mechanical properties of cells undergoing CaAR were altered . We therefore used atomic force microscopy ( AFM ) to probe cells with 10 µm beads attached to the cantilever . Despite the observed transient reduction of cortical actin , we found that the subcortical regions ( 800 nm below the plasma membrane ) of MCF-7 cells treated with ionomycin became markedly stiffer during CaAR , mirroring greater levels of actin recruitment at the ER ( Figure 2D ) . In addition , we observed that intracellular motility of organelles was transiently halted during CaAR . Random as well as directed motion of lysosomes ( Figure 2E ) , mitochondria and the ER network ( Figure 2F ) was abolished within a few seconds of CaAR onset and resumed after actin returned to the cell cortex ( Figure 2E , F ) . When simultaneously observing CaAR and lysosome motility we found the organelles trapped within a cytosol-filling actin mesh that corresponded to the transient ER-based actin filaments ( Video 4 ) . Lysosomes were also immobilized in cells treated with 50 µM blebbistatin arguing against a prominent role of actomyosin contractility for organelle freezing ( Figure 2G ) . 10 . 7554/eLife . 19850 . 011Video 4 . HeLa cell expressing Lifeact-GFP labelled with Lysotracker Red and stimulated by laser ablation . Corresponds to Figure 2E . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 011 When using pointy AFM tips to probe cells , we often were able to induce CaAR in individual cells ( Figure 1E , Figure 2—figure supplement 1B , C ) . Forces above 50 nN ( Figure 2—figure supplement 1B ) were able to robustly induce CaAR multiple times in a single cell ( Figure 2—figure supplement 1D , Video 5 ) with a refractory period of 30–60 s ( Figure 2—figure supplement 1C ) , consistent with the time scale of actin recruitment at the ER . When using localized CaAR stimulation by AFM we frequently observed that cells directly adjacent to the manipulated cells also reacted ( Figure 2—figure supplement 1E ) . To examine this behavior in more detail , we performed ablation experiments on MCF-7 monolayers . Strikingly , we found that nearly all cells within 60–150 µm of the ablation site exhibited CaAR ( Figure 2H ) . This was also observed for cells that were not in direct contact with the ablated cell ( Figure 2—figure supplement 1F , Video 6 ) . Previous reports have shown that Ca2+ signals can be propagated across tissue sections and cell layers via ATP ( Frame and de Feijter , 1997; Schwiebert , 2000 ) . Indeed , distant cells were no longer able to respond to cell ablation after the medium had been saturated with 100 µM ATP ( Figure 2I ) , indicating that ATP release and associated Ca2+ influx is responsible for the propagation of CaAR . 10 . 7554/eLife . 19850 . 012Video 5 . MCF-7 cell expressing Lifeact-GFP repeatedly stimulated by AFM ( asterisk ) . Corresponds to Figure 2—figure supplement 1D . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 01210 . 7554/eLife . 19850 . 013Video 6 . Propagation of CaAR in MCF-7 cells expressing Lifeact-GFP stimulated by laser ablation ( asterisk ) . Corresponds to Figure 2—figure supplement 1F . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 013 The rapid increase in actin localization at the ER indicated the involvement of a strong nucleator in CaAR . Accordingly , both G-actin sequestration by latrunculin A ( LatA ) and actin filament disruption by cytochalasin D ( CytoD ) completely blocked CaAR ( Figure 3A , Video 7 ) . The only actin nucleator that has been shown to be localized at the ER is inverted formin 2 ( INF2 ) ( Chhabra and Higgs , 2006 ) , and this nucleator has also been shown to mediate the formation of perinuclear actin filaments in 3T3 fibroblasts ( Shao et al . , 2015 ) . Indeed , in MCF-7 and HeLa cells , INF2 was localized to the ER , as shown either by immunofluorescence ( Figure 3B , Figure 3—figure supplement 1A ) or when expressing the full-length INF2-CAAX isoform fused to GFP ( Figure 3—figure supplement 1B ) . A constitutively active point mutant of INF2 ( A149D , ( Korobova et al . , 2013 ) induced the previously shown strong actin staining at the ER ( Ramabhadran et al . , 2013 ) , which was not further increased by the addition of ionomycin ( Figure 3C ) . To test a potential role of INF2 for CaAR we knocked down INF2 expression by RNA interference . While our attempts for INF2 knock down were unsuccessful in MCF-7 cells , we were able to efficiently abolish expression of INF2 in HeLa cells ( Figure 3D ) . Strikingly , CaAR was completely blocked in HeLa cells that were depleted for INF2 ( Figure 3E ) . This was especially apparent in areas where some cells still retained INF2 expression and therefore were still able to nucleate actin at the ER upon treatment with ionomycin ( Figure 3E , circled area ) . Notably , INF2 knock-down also blocked actin decrease at the cell cortex ( Figure 3E ) , increase in cell stiffness detected by AFM ( Figure 3—figure supplement 1C ) and freezing of organelle motility ( Figure 3—figure supplement 1D . To obtain stable cell lines without INF2 expression we knocked out INF2 using the CRSPR/Cas9 system . We obtained several independent HeLa INF2-KO clones that exhibited no detectable INF2 expression by Western blot ( Figure 3—figure supplement 1E ) . We then transiently expressed either the CAAX- or nonCAAX human isoform of INF2 in clone 18 ( Figure 3F ) . Upon induction of CaAR by laser ablation we observe no reaction in untransfected cells confirming that INF2 is essential for CaAR . Interestingly , both isoforms were able to rescue the knock out and to support efficient CaAR reactions ( Figure 3F ) , indicating that ER-localization was not essential for actin reorganization . 10 . 7554/eLife . 19850 . 014Figure 3 . CaAR is driven by INF2-mediated actin polymerization . ( A ) MCF-7 cells expressing Lifeact-mCherry were treated with ionomycin and the indicated drugs ( LatA: 400 nM latrunculin A , CytoD: 1 µM cytochalasin D ) . Plots show representative intensity profiles for Lifeact-GFP at the nuclear periphery and for intracellular Ca2+ ( Fluo4 ) . ( B ) MCF-7 cells were fixed at the indicated time points after addition of 1 µM ionomycin and stained with αINF2 antibody . ( C ) MCF-7 cells expressing Lifeact-mCherry were transfected with a constitutively active GFP-INF2 ( A149D ) construct and imaged before and after addition of 1 µM ionomycin . ( D , E ) siRNA-mediated knock-down of INF2 in HeLa cells expressing Lifeact-GFP . ( D ) Western analysis of INF2 after knock-down with two different siRNAs at indicated times . ( E ) 72 hr after siRNA transfection , HeLa cells were treated with 1 µM ionomycin and monitored by fluorescence microscopy . Cells were fixed , immunostained with anti-INF2 antibody and imaged again at the same positions . Dotted line surrounds residual INF2-positive cells . Plots show Lifeact-GFP intensity at the cortex ( green ) and the ER ( red ) . Values are mean ± SD , n = 30 . ( F ) Images of HeLa INF2 KO cells stably expressing Lifeact-mCherry ( derived from KO clone 18 ) and transfected with either GFP-INF2-CAAX or GFP-INF2-nonCAAX . Cells were stimulated by laser ablation outside the represented region . In each series one cell with INF2 expression ( 1 ) and a control cell without INF2 ( 2 ) are labeled . Corresponding intensity plots for ER ( red ) and cortical ( green ) regions are shown . Times in sec . Scale bars: 10 µm . ( G ) Co-precipitation analyses showing that GFP-INF2-CAAX expressed in HEK293 cells specifically interacts with immobilised calmodulin ( CaM ) . Comparison of pull down conditions with and without ( 1 mM EGTA ) 500 µM Ca2+ ( CaM activation at plateau ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 01410 . 7554/eLife . 19850 . 015Figure 3—figure supplement 1 . CaAR is driven by INF2-mediated actin polymerization . ( A ) HeLa cells expressing Lifeact-GFP were fixed at the indicated time points after addition of 1 µM ionomycin and stained with αINF2 antibody . ( B ) MCF-7 cells expressing Lifeact-mCherry were transfected with a GFP-INF2 expressing plasmid and induced with 1 µM ionomycin . Times in sec . Scale bars 10 µm . ( C ) Lifeact-GFP-expressing HeLa cells were transfected with scrambled siRNA or INF2 targeting siRNA1 , or left untreated ( control ) and analyzed by atomic force microscopy . The relative Young’s modulus of whole cells was calculated from force-distance curves obtained with 10 µm beads and plotted as mean ± SD ( n > 4 ) . ( D ) Normalized Lifeact-GFP intensities at the ER and the cortex and mobility of lysosomes labeled with Lysotracker Red of representative HeLa cells stimulated by laser ablation . Cells were observed 72 hr after transfection with scrambled ( scr ) or INF2 siRNA ( si1 ) . ( E ) Western blot with antiINF2 antibody showing loss of INF2 expression after CRSPR/Cas9-mediated knock out in different isolated clones . ( F ) CaAR in MCF-7 cells pretreated with 50 µM blebbistatin or 10 µM nocodazole for 20 min . Images were taken at the indicated times after addition of 1 µM ionomycin . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 01510 . 7554/eLife . 19850 . 016Video 7 . MCF-7 cells expressing Lifeact-GFP , either untreated or pretreated with 400 nM LatA or 1 µM CytoD were exposed to 1 µM ionomycin at t = 0 s . Corresponds to Figure 3A . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 016 Considering the rapid activation of INF2 by calcium and the lack of obvious calcium binding sites in the formin itself , we hypothesized that an abundant calcium sensor protein could be involved in the reaction . We therefore investigated binding of INF2 to the prototypical calcium regulator calmodulin . Co-precipitation analyses showed that GFP-INF2 expressed in HEK293 cells indeed specifically bound to immobilized calmodulin , and that this interaction was Ca2+-dependent ( Figure 3G ) . While this result provides a potential molecular basis for INF2 activation , additional work will be needed to clarify the specific role of calmodulin during CaAR . It is well documented that intracellular calcium influx also affects other cellular factors such as Myosin II , and INF2 has been linked to microtubule stabilization ( Andrés-Delgado et al . , 2012; Bartolini et al . , 2016 ) . These factors might therefore also play a role during CaAR . However , pretreatment of cells with the myosin II inhibitor blebbistatin or with the microtubule disrupting drug nocodazole did not inhibit or slow down CaAR ( Figure 3—figure supplement 1F ) . In summary , our results indicate that CaAR is largely dependent on a single factor – the calcium-regulated actin nucleator INF2 . Two remarkable characteristics of CaAR are the reciprocal behavior of actin at the ER and the cell cortex and the highly transient nature of the reorganization ( Figure 4A ) . Interestingly , INF2 is not only an actin nucleator and elongator , but also a very potent severing and actin-depolymerizing factor ( Chhabra and Higgs , 2006; Gurel et al . , 2014 ) . During CaAR , these activities should also be subject to regulation by calcium and therefore potentially facilitate the observed reciprocal and transient reaction . It is important to note that CaAR does not lead to the equal reduction of all cortical actin structures . Stable actin assemblies such as stress fibers were much less affected ( Figure 4—figure supplement 1A , Video 8 ) . Also , despite their opposite slopes , there was no apparent delay between ER and cell cortex reactions and they followed the calcium signal with the same offset ( Figure 4A ) . In addition , the duration of Ca2+ influx was directly correlated with the kinetics of actin reorganization at both locations ( Figure 4B , Figure 4—figure supplement 1B ) . These observations indicate that the actin filament assembly and disassembly at the ER and cell cortex are tightly coupled . A potential mechanism for such coupling would be simple competition between different actin nucleators . To test whether such a scenario could account for the observed CaAR kinetics , we developed a particle-based stochastic model of actin dynamics that explicitly considers actin nucleation , polymerization , depolymerization , capping and severing ( Figure 4C , Materials and Methods ) . We initially focused on a single actin population in equilibrium with the monomer pool . The model showed that actin turnover is mainly determined by the speed of actin depolymerization and modulated to a lesser extent by the rates of capping and severing ( Figure 4—figure supplement 1C ) . In addition , actin turnover was strongly dependent on the levels of actin monomers available ( Figure 4—figure supplement 1D ) . The experimentally measured t1/2 for recovery of actin-GFP at the cortex of MCF-7 cells was quite fast , at 14 . 48 ± 7 . 21 s ( n = 79 ) . This constrained the possible values for severing , depolymerization and ratios of F- to G-actin ( Figure 4—figure supplement 1D , E ) . We next included a second nucleation activity to represent INF2-mediated actin polymerization at the ER ( Figure 4C ) . By assuming that INF2 activity follows the intracellular Ca2+ curve ( See Materials and Methods ) , we were able to generate a transient actin peak at the ER with a corresponding decrease at the cortex ( Figure 4D ) . By increasing the INF2 nucleation rate or the amount of available G-actin in the system we were able to obtain kinetics that were reasonably close to the experimental observations ( dotted curves in Figure 4D ) . However , cortical actin never dropped below 50% of its original value , and we were unable to obtain complete reorganization of actin within less than 30 s , as seen with laser ablation ( Figure 4B ) . Interestingly , the model predicted that including Ca2+- or INF2-dependent actin depolymerization into the model strongly increased the amplitude and speed of actin reorganization at both cortex and ER ( solid curves in Figure 4D ) . To experimentally test this prediction we measured CaAR kinetics in HeLa KO cells expressing a WH2-mutant of INF2 ( 3L: L976A , L977A , L986A ) that was reported to have strongly reduced depolymerization activity ( Chhabra and Higgs , 2006 ) . Strikingly CaAR kinetics were strongly reduced in cells expressing the 3L mutant . The resulting changes were comparable to the situation without Ca2+-mediated depolymerization predicted by the model ( Figure 4D , E ) . The amplitude and the speed of reorganization were strongly reduced for both , ER and cell cortex ( Figure 4F , G ) . 10 . 7554/eLife . 19850 . 017Figure 4 . A stochastic model rationalizes CaAR features and kinetics . ( A , B ) Correlation analysis of actin and calcium dynamics in MCF-7 cells undergoing CaAR . ( A ) Temporal shift between half-maximal decay of calcium ( Ca ) , actin maximum at nuclear periphery ( ER ) and actin minimum at the cell cortex ( Co ) . ( B ) Correlation between Ca and ER ( red , Pearson correlation coefficient , r = 0 . 924 ) , Ca and Co ( green , r = 0 . 840 ) and between ER and Co ( black , r = 0 . 880 ) . Dotted lines: y = x . ( C ) Stochastic model for actin competition . Filaments at the cortex ( green ) and ER ( red ) are represented by arrows ( head: barbed end , circle: capped end ) . Total length of cortical ( LC ) and ER ( LER ) actin is indicated . Monomer pool is given by LG . Relevant parameters: Rates of severing ( rs per length l ) , capping ( rc ) , nucleation ( rn ) , velocities of elongation ( v+ ) and depolymerization ( v- ) . See Supplementary material for details of the model . ( D ) Evolution of F-actin concentration at ER ( red ) and cortex ( green ) generated by the model using a simple competition scenario ( dotted lines ) or including calcium-activated depolymerization of actin ( full lines ) . All simulations were run using the indicated Ca2+ curve ( black ) as input . ( E ) Average intensity curves of Lifeact-mCherry at ER ( red ) and cell cortex ( green ) of HeLa KO cells expressing wildtype or WH2 mutant GFP-INF2-CAAX ( 3L: L976A , L977A , L986A ) . Dotted red and green lines indicate SEM . ( F ) Effect of different depolymerization rates on maximal actin change at cortex ( green ) and ER ( red ) . Shown are predictions from simulations ( left ) and experimental data comparing wildtype and 3L-INF2 expressing HeLa cells ( data points , mean and SD ) . ( G ) As in ( F ) but showing the effects of varying depolymerization rate on the time until cortical minimum or ER-maximum is reached . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 01710 . 7554/eLife . 19850 . 018Figure 4—figure supplement 1 . A stochastic model rationalizes CaAR features and kinetics . ( A ) TIRF imaging of the basal surface of a Lifeact-GFP expressing MCF-7 cell after laser ablation . The dashed line indicates the position of the kymograph , the asterisk indicates the duration of CaAR . Times in sec , scale bars 10 µm , time arrow 60 s . ( B ) MCF-7 cells expressing Lifeact-GFP were labeled with Fluo4 and stimulated by laser ablation . Signal intensities for Fluo4 ( reflects intracellular calcium levels; dotted curves ) and Lifeact-mCherry in the perinuclear area ( solid curves ) and the cell cortex ( dashed curves ) were plotted for two cells with shorter ( red ) and longer ( black ) calcium peaks . Half maximal calcium levels and corresponding maxima/minima of actin are indicated by dashed vertical lines . ( C–E ) Impact of changes in the indicated parameters of the stochastic model describing actin reorganization during CaAR . See Materials and methods for details . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 01810 . 7554/eLife . 19850 . 019Video 8 . Cortical actin reorganization in an MCF-7 cell expressing Lifeact-GFP stimulated by laser ablation . Corresponds to Figure 4—figure supplement 1A . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 019 In summary , our simulations indicate that the observed kinetics of CaAR can be quantitatively explained by a competition scenario with constitutively active cortical actin nucleators and Ca2+-activated INF2 at the ER . Importantly , the depolymerization activity of INF2 provides additional optimization for CaAR kinetics . While the most apparent feature of CaAR is certainly the transient and reciprocal reorganization of actin , we wondered whether the global balance between G- and F-actin was also affected during or after the reaction . Indeed , fractionation of MCF-7 cell extracts revealed a modest increase in the global F/G-actin ratio during CaAR ( Figure 5—figure supplement 1A ) . In addition , the levels of actin at the cortex of MCF-7 cells were increased after completion of CaAR relative to those at the outset of CaAR ( Figure 5—figure supplement 1B ) . A very sensitive readout for an increased F/G actin ratio in cells is the release of transcriptional regulators , such as the serum response factor ( SRF ) co-factor MRTF-A , from sequestration in the cytosol ( Miralles et al . , 2003 ) . Indeed , we found that within a minute of Ca2+ influx and CaAR onset , MRTF-A translocated into the nucleus in >95% of MCF-7 cells , and remained there for up to 30 min ( Figure 5A , B ) . MRTF-A translocation also occurred in HeLa cells and was blocked upon INF2 knock-down ( Figure 5C ) . When we blocked CaAR by treatment with LatA , MRTF-A translocation was also inhibited ( Figure 5D ) . In contrast , treatment with CytoD induced CaAR-independent translocation of MRTF-A ( Figure 5E ) , consistent with previous reports ( Descot et al . , 2009 ) . We next tested whether CaAR-induced translocation of MRTF-A indeed caused SRF-mediated transcription . RNA levels of two known SRF targets , CTGF and NR4R3 , were strongly increased upon CaAR induction by ionomycin , reaching a maximum at 1 hr ( Figure 5—figure supplement 1C ) . We selected this time point for the analysis of CaAR-mediated changes in gene transcription . We treated MCF-7 cells with 1 . 5 µM ionomycin for 10 min after pre-incubation for 10 min with either LatA , CytoD or control buffer . After washout of the drugs and ionophore , we cultured the cells for another 50 min in 4-thio-uridine and collected newly synthesized mRNA ( Figure 5F ) . We found that a 10 min pulse of ionomycin was sufficient to decrease the expression of 478 genes and enhance the expression of 405 genes by more than 1 . 5 fold ( Figure 5—source data 2 ) . Treatment with LatA on its own had very little effect , but upregulation of 88 ionomycin-induced genes was strongly inhibited by LatA ( Figure 5—source datas 1 and 2 ) . Nearly half of these genes are known SRF targets and about one third were also induced by the SRF activator CytoD ( Figure 5G ) . In agreement with the reported functions of SRF , many of the identified transcriptional changes were associated with genes involved in cytoskeletal organization , calcium regulation , cell signaling and regulation of cell adhesion ( Figure 5G ) . More than 60% of the CaAR-regulated genes have been associated with cancer and more than one-third have been implicated in cellular stress responses , cell differentiation , cell migration or inflammation ( Figure 5G ) . SRF is known to respond to actin changes mediated through Rho GTPases ( Treisman et al . , 1998 ) . However , we found that CaAR-mediated induction of CTGF was not affected by the ROCK inhibitor Y27632 , the calcineurin inhibitor cyclosporin A ( CsA ) or by the CaM kinase inhibitors KN62 and KN93 ( Figure 5—figure supplement 1D ) . Importantly , all CaAR transcriptome experiments were performed in serum-starved cells and addition of serum did not induce CaAR in MCF-7 cells . In addition , treatment of cells with Y27632 , the Rho1 inhibitor C3 transferase or with CsA did not affect CaAR ( Figure 5—figure supplement 1E ) . Our results therefore indicate that CaAR mediates SRF activation through actin reorganization , but via a Ca2+-dependent pathway distinct from the previously described SRF activation involving Rho , ROCK and mDia ( Baarlink et al . , 2013; Copeland and Treisman , 2002 ) . 10 . 7554/eLife . 19850 . 020Figure 5 . CaAR induces transcription via MRTF-A and SRF . ( A ) Immunofluorescence detection of MRTF-A in MCF-7 cells at the indicated times after addition of 1 µM ionomycin . ( B ) Quantification of MCF-7 cells exhibiting CaAR ( red dotted line ) and weak ( light grey ) or strong ( dark grey ) nuclear MRTF-A accumulation ( mean + SD , n = 3 experiments with >100 cells per time point ) . ( C ) Quantification of HeLa cells exhibiting CaAR ( red dotted line ) and weak ( light grey ) or strong ( dark grey ) nuclear MRTF-A staining ( mean + SD , n = 3 experiments with >100 cells per time point ) . Graphs show values for cells treated with scrambled siRNA and two INF2-siRNAs . ( D , E ) Analysis of MRTF-A localization upon stimulation of MCF-7 cells with 1 µM ionomycin ( added at t = 0 ) . Cells were pretreated for 10 min with either 400 nM latrunculin A ( D ) or 1 µM cytochalasin D ( E ) . ( F ) Protocol for transcriptome analysis . ( G ) Major functional categories of CaAR-regulated genes , assembled from the manual curation of public databases and literature . Association categories are given for the known SRF regulation ( light grey ) , cellular processes ( medium grey ) and biological processes ( dark grey ) . Time in min . Scale bars 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02010 . 7554/eLife . 19850 . 021Figure 5—source data 1 . CaAR induces transcription via MRTF-A and SRF . Excel table with 88 genes differentially regulated in MCF-7 cells either treated for 10 min with ionomycin ( 1 . 5 µM ) or with ionomycin ( 1 . 5 µM ) and LatA ( 400 nM ) . Filled cells indicate inclusion into indicated categories . Orange: regulation through SRF or CytoD-induced MRTF-A translocation . Blue: cellular processes . Green: physiological processes . Total number and % of genes in each category are given at the bottom of the table . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02110 . 7554/eLife . 19850 . 022Figure 5—source data 2 . Differentially regulated genes . Excel table ( five pages ) with results from Affymetrix chip analysis for gene expression after treatment with Ionomycin and actin drugs . See Materials and Methods for details . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02210 . 7554/eLife . 19850 . 023Figure 5—figure supplement 1 . CaAR induces transcription via MRTF-A and SRF . ( A ) Ratio of F- to G-actin in MCF-7 cells during CaAR as determined by fractionation and Western blotting ( mean ± SD , n = 5 ) . ( B ) Lifeact-GFP intensity at the cortex of MCF-7 cells after stimulation by laser ablation . ( C ) Levels of CTGF and NR4A3 transcripts in MCF-7 cells at the indicated times after addition of 1 µM ionomycin ( at t = 0 ) as determined by quantitative RT-PCR ( mean ± SD , n = 2 , normalized against untreated control cells at each time point ) . ( D ) MCF-7 cells were pretreated with the indicated drugs ( ROCK inhibitor Y27632 , CaM-kinase inhibitors KN62 and KN93 , calcineurin inhibitor cyclosporin A ( CsA ) , actin inhibitors latrunculin A ( LatA ) and cytochalasin D ( CytoD ) ; concentrations are indicated in µM ) for 10 min before addition of 1 . 5 µM ionomycin , and levels of CTGF transcripts were determined . Values are mean ± SD , n = 3 . ( E ) Lifeact-GFP expressing MCF-7 cells were pretreated with the indicated inhibitors ( CsA: 1 µM , Y27632: 10 µM , C3: 2 . 5 µg/ml ) for at least 2 hr before the addition of 1 µM ionomycin to induce CaAR . Upper right: MCF-7 cells expressing Lifeact-GFP stimulated by the addition of 40% FBS and then monitored by fluorescence microscopy . Times in sec . Scale bars 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 023 We next wanted to understand the physiological consequences of transient actin reorganization during CaAR . When we examined the cellular consequences of CaAR in laser ablation experiments with HeLa cells , we observed accumulation of actin at the site of membrane damage in more than 90% of the cases ( Figure 6A , Video 9 ) . This accumulation was reminiscent of previously observed wound repair processes ( Clark et al . , 2009; McNeil , 2002 ) and typically occurred after completion of CaAR ( Figure 6B ) . Upon knock-down of INF2 , HeLa cells did not undergo CaAR and were also completely unable to recruit actin to the site of membrane wounding ( Figure 6C ) . Importantly , membrane damage as measured by uptake of propidium iodide or FM4-64 was repaired within a few seconds of damage induction ( not shown ) . Actin accumulation was therefore not required for membrane sealing but for a later step in wound repair , possibly cortex reassembly . Such repair mechanisms should be especially relevant for terminally differentiated cells that can no longer be replaced in vivo , such as podocytes ( Pavenstädt et al . , 2003 ) . We therefore examined CaAR in in vitro differentiated human AB8 podocytes ( Saleem et al . , 2002 ) . Similar to our observations in HeLa cells , laser ablation at the periphery of podocytes induced CaAR and led to subsequent actin accumulation at the wounding site , which was then efficiently sealed without apparent loss of cellular integrity ( Figure 6D , Video 10 ) . Interestingly , actin accumulation at the wound was accompanied by simultaneous induction of lamellipodia in the immediate vicinity ( Figure 6D ) . In both cell types examined above , the strong accumulation of actin at cortical wounding sites occurred right after completion of CaAR . Similarly , when we extended the period of observation in MCF-7 cells that had undergone CaAR upon ATP exposure , we found that they initiated extended basal protrusions that correlated with the end CaAR . Protrusions emanating from cell-cell junctions collapsed after only 5–10 min , but those appearing at free cell edges persisted for up to 1 hr ( Figure 6E , Video 11 ) . To study this phenomenon in more detail we ablated a single MCF-7 cell in a monolayer . CaAR was efficiently induced in all surrounding cells and we again observed short-lived protrusions at cell-cell junctions and longer-lived protrusions at free cell edges ( Figure 6F , Video 12 ) . In addition , some cells formed large lamellipodia , which rapidly closed the gap left by the ablated cell ( Figure 6F ) . Considering the extended activation of cell spreading and cellular protrusion after CaAR we wondered whether this would have noticeable consequences for collective migration in a typical wound healing setting . We therefore observed MCF-7 monolayers migrating into a free area after removal of a PDMS spacer ( Figure 7A ) . After 12 hr , untreated cells had moved in to the gap with an average speed of 3 µm/h , while cells migrating in the presence of 50 µM ATP ( and therefore exhibiting CaAR at the onset of the experiment ) covered a much larger area with an average speed of 7 µm/h ( Figure 7B ) . A more detailed analysis revealed that ATP treatment led to an acceleration for the initial 4 hr of migration and that cells then reverted to the speed of control cells ( Figure 7C , Video 13 ) . Our findings clearly show that ATP-mediated calcium influx and CaAR are associated with prolonged activation of protrusion and collective migration of MCF-7 cells . As we were not yet able to remove INF2 from MCF-7 cells , we cannot exclude at this time that the observed effects on protrusion could be due to a CaAR-independent effect of ATP . 10 . 7554/eLife . 19850 . 024Figure 6 . CaAR mediates acute cellular reorganization . ( A ) A HeLa cell expressing Lifeact-GFP was damaged by laser ablation and Lifeact-GFP intensity was monitored in different regions . Asterisk: ablation site . Regions for intensity measurements in ( B ) are indicated in corresponding colors . ( B ) Plots of Lifeact-GFP signal intensity at ER , cortex and at the ablation site of the cell shown in ( A ) . ( C ) Plots of Lifeact-GFP signal intensity in indicated regions of control and INF2-siRNA-treated cells . Quantification of peak intensities at the wounding site is shown as mean ± SD ( n > 9 ) . ( D ) A podocyte expressing Lifeact GFP was damaged by laser ablation and subsequently monitored by fluorescence microscopy . Asterisk indicates site of laser ablation . Red dotted line indicates the path of kymograph , a white dotted line indicates the outline of the cell before ablation , and arrows indicate instances of wound-repair and lamellipodia formation after CaAR completion . Lifeact-GFP intensity curves for ER , cortex and wound site are shown in the graph . Time arrow: 50 s . ( E ) Increase in cortical actin at the cell periphery upon activation of CaAR in MCF-7 cells with 50 µM ATP . The dotted lines indicate the position of the cell boundary before stimulation . ( F ) Changes in cortical actin dynamics after ablation of an MCF-7 cell within a monolayer . Kymographs are shown along the dotted lines for indicated positions ( red numbers ) . Time arrow: 20 min . At later time-points actin congresses into ring structures around gaps ( red arrow ) . Times in sec ( A–D ) and min ( E , F ) . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02410 . 7554/eLife . 19850 . 025Figure 7 . Wound healing following induction of CaAR and timeline . ( A ) Representative examples of MCF-7 cells expressing Lifeact-mCherry seeded on either side of a PDMS spacer . After removal of the spacer cells were cultured for 24 hr and then treated with either control medium ( left panels ) or medium containing 50 µM ATP ( right panels ) . Shown are the cell positions at the time of the medium exchange ( offset resulting from stitching ) , the kymographs along the indicated lines ( red numbers ) and the overlay between t0 ( red ) and t5h ( green ) . Time arrow: 4 hr . Scale bars: 100 µm . ( B ) Rate of cell front movement for control vs . ATP-treated cells . Student’s t-test p<0 . 05 for n = 5 experiments . ( C ) Graphs showing speed of closure in control cells vs . ATP-induced cells . ( D ) Timeline of CaAR and associated processes as discussed in the text . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02510 . 7554/eLife . 19850 . 026Video 9 . HeLa cell expressing Lifeact-GFP stimulated by laser ablation . Corresponds to Figure 6A . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02610 . 7554/eLife . 19850 . 027Video 10 . AB8 podocyte expressing Lifeact-GFP stimulated by laser ablation ( asterisk ) . Corresponds to Figure 6D . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02710 . 7554/eLife . 19850 . 028Video 11 . Spreading of MCF-7 cells expressing Lifeact-GFP stimulated by ATP . Corresponds to Figure 6E . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02810 . 7554/eLife . 19850 . 029Video 12 . MCF-7 cells expressing Lifeact-GFP stimulated by laser ablation of a single cell in a monolayer . Corresponds to Figure 6F . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 02910 . 7554/eLife . 19850 . 030Video 13 . MCF-7 cells expressing Lifeact-GFP migrating in to a gap in the absence ( left ) or presence ( right ) of 50 µM ATP . Corresponds to Figure 7A . DOI: http://dx . doi . org/10 . 7554/eLife . 19850 . 030 In summary , our results on plasma membrane sealing and transcriptional regulation indicate that CaAR plays an important role during acute morphogenetic adaptations , with implications for such diverse processes as cell migration , cancer progression and cell stress response . Calcium has long been known to transmit acute signals in processes such as inflammation , wound healing or stress responses ( Høyer-Hansen and Jäättelä , 2007; Wood , 2012 ) . How calcium signaling is linked to cellular morphogenesis is less well understood . Interestingly , a recent study reported formation of a Ca2+-dependent perinuclear actin ring in 3T3 fibroblasts ( Shao et al . , 2015 ) . We have now found that increased intracellular calcium levels lead to transient and global reorganization of actin in a wide range of epithelial , mesenchymal , endothelial and hematopoietic cell lines . This reorganization can be triggered by various mechanical and biochemical signals and is characterized by simultaneous actin polymerization at the ER and disassembly of cortical actin filaments . The ubiquity , sensitivity and conserved features of the observed Calcium-mediated Actin Reset ( CaAR , Figure 7D ) indicate that this is a fundamental response of mammalian cells . In agreement with this idea , we found that CaAR can act as a morphogenetic integrator during acute cellular perturbations , such as cell cortex damage and wound healing . The highly synchronous kinetics of actin polymerization at the ER and actin disassembly at the cell cortex strongly suggests that competition for a common pool of actin monomers occurs between actin nucleators at the two locations . Such competition between different actin structures within the same cell has recently been demonstrated for fission yeast cells ( Burke et al . , 2014; Suarez et al . , 2015 ) . Our stochastic model for CaAR shows that the extent and speed of actin reorganization at the cortex and ER can only be achieved if Ca2+ simultaneously increases the rate of nucleation at the ER and the overall rate of actin depolymerization . Interestingly , the key regulator of CaAR , INF2 , is a unique molecule that can at the same time stimulate actin nucleation , elongation , severing and depolymerization ( Gurel et al . , 2015 ) . We could validate our theoretical prediction , using a mutant INF2 with reduced depolymerization activity . Expression of this mutant in INF2 KO cells supported CaAR at much slower kinetics and reduced turnover . Ca2+-mediated activation of INF2 could therefore account for all the experimentally observed kinetics . How then does calcium regulate INF2 ? We have found that INF2 binds to calmodulin , a highly abundant cellular regulator of calcium signaling . It will be interesting to test whether the calmodulin-INF2 interaction occurs directly , or via additional adaptor proteins such as IQGAP1 . This protein has recently been shown to bind INF2 ( Bartolini et al . , 2016 ) and is a known interactor of calmodulin ( Jang et al . , 2011 ) . It is well established that SRF/MRTF-dependent transcription is activated by Rho GTPase and mDia-mediated actin dynamics ( Hill et al . , 1995; Miralles et al . , 2003 ) . We have found that CaAR provides an alternative , INF2-mediated , mechanism for MRTF activation . Interestingly , although CaAR is independent of both Rho and serum , we find a strong overlap between CaAR-regulated and classical serum-induced genes . Many of the CaAR-induced genes are associated with cytoskeletal organization , cell adhesion or signaling . This suggests a possible integration of transcriptional changes with the observed morphological effects of CaAR on actin protrusions , cortex repair and cell migration . In summary , we have identified CaAR as a fundamental process linking Ca2+ signaling , cell mechanics , actin dynamics and SRF-mediated transcription . Considering its basic constituents ( calcium , actin ) , this process likely influences a multitude of signaling pathways and physiological processes . Thus , the observed consequences of CaAR should be considered when studying cellular perturbations that are linked to Ca2+ influx . The strong association of CaAR-induced genes with cancer , inflammation and stress point to exciting opportunities for future studies of CaAR in living organisms . Of particular note is a potential role of CaAR in kidney pathology , as INF2 is known to regulate actin dynamics in podocytes , and INF2 mutations are linked to hereditary kidney diseases ( Brown et al . , 2010; Sun et al . , 2013 ) . Cells were grown at 37°C with 10% CO2 in Dulbecco’s Modified Eagle medium ( DMEM-Glutamax-I; Gibco , Carlsbad , CA , USA ) supplemented with 10% fetal bovine serum ( FBS; Gibco ) . Routinely , 2 × 104 cells/ml were seeded on glass-bottomed dishes ( Ibidi , Martinsried , Germany ) , 8-well slides ( Ibidi ) or 4-well LabTek dishes ( Nunc , Rochester , NY , USA ) and incubated for 24 hr or 48 hr before imaging . Live cell imaging was performed with cells seeded on glass bottom dishes and incubated in Hanks’ buffered salt solution ( HBSS ) supplemented with 10 mM HEPES ( pH 7 . 4 ) . Apart from the experiment in Figure 2B labeled ‘serum’ , all cells were starved in imaging buffer 1 hr prior to experiments . The following cell lines were used in this study: MDCK II ( ECACC 0062107 ) ; MDCK II cells stably expressing Lifeact-mCherry ( Klingner et al . , 2014 ) ; MCF-7 ( ECACC 86012803 ) ; MCF-7 cells stably expressing Lifeact-GFP or Lifeact-mCherry ( this study ) ; HeLa ( ECACC 93021013 ) ; HeLa cells stably expressing Lifeact-GFP or Lifeact-mCherry ( this study ) ; HeLa INF2 KO cells , and HeLa INF2 KO cells stably expressing Lifeact-mCherry ( this study ) ; NIH 3T3 ( ECACC 93061524 ) ; CCL-39 ( ATCC-CCL-39 ) ; PANC-1 ( ECACC 87092802 ) ; U-2 OS ( ECACC 92022711 ) ; COS-7 ( ECACC 87021302 ) ; GM7373 ( DSMZ ACC109 ) ; GM7373 stably expressing Lifeact-mKate2 ( Kronlage et al . , 2015 ) ; AB8 podocytes ( Saleem et al . , 2002 ) , AB8 cells stably expressing Lifeact-GFP ( this study ) ; HoxB8-immortalized mouse monocytes and neutrophils ( [Wang et al . , 2006] and this study ) ; HEK293 cells ( ECACC 85120602 ) . All cell lines were checked for identity by visual inspection of morphologies and tested negative in Mycoplasma tests using the following primers: RWS2534: 5’-CGCCTGAGTAGTACGTTCGC-3’; RWS2535: 5’-CGCCTGAGTAGTACGTACGC-3’; RWS2536: 5’-TGCCTGAGTAGTACATTCGC-3’; RWS2537: 5’-CGCCTGGGTAGTACATTCGC-3’; RWS2538: 5’-CGCCTGAGTAGTAGTCTCGC-3’; RWS2539: 5’-TGCCTGGGTAGTACATTCGC-3’; RWS2540: 5’-GCGGTGTGTACAAGACCCGA-3’; RWS2541: 5’-GCGGTGTGTACAAAACCCGA-3’; RWS2542: 5’-GCGGTGTGTACAAACCCCGA-3’ . Lifeact-acGFP , Lifeact-mCherry and Lifeact-mKate2 were expressed from pEFIRES and pGKIRES plasmid backbones . Beta-Actin ( ACTB-GFP ) was expressed from pEGFP-C1 . Lifeact-acGFP was cloned via NotI/PacI restriction sites into the retroviral expression vector pQXCIP for stable transfection of AB8 cells . pssRFP-KDEL ( Altan-Bonnet et al . , 2006 ) , pGFP-INF2-CAAX , pGFP-INF2-nonCAAX and pGFP-INF2 ( A149D ) -CAAX were described previously ( Ramabhadran et al . , 2011 ) . For expression in HEK293 cells , INF2 was first subcloned into pGADT7 . 3 ( BspEI/XmaI-XhoI ) and then into pEGFP-C3 ( EcoRI-SalI ) . The 3L ( L976A , L977A , L986A ) mutants were generated by site directed mutagenesis ( Stratagene Quickchange , Agilent , Santa Clara , CA , USA ) using either pGFP-INF2-CAAX or pGFP-INF2-nonCAAX as a template . Primers used for mutagenesis: RWS3036: 5’-GTTCAGCACGATGAAGGCCTTTAGGGACCTTTTCC-3’ siRNA resistant forward; RWS3037: 5’-GGAAAAGGTCCCTAAAGGCCTTCATCGTGCTGAAC-3’ siRNA resistant reverse; RWS3163: 5’-gtgtgtgtcatcgatgccGCgGCggctgacatcaggaaggg-3’ L976A and L977A forward; RWS3164: 5’-cccttcctgatgtcagccGCcGCggcatcgatgacacacac-3’ L976A and L977A reverse; RWS3165: 5’-catcaggaagggcttccagGCgcggaagacagcccggg-3’ L986A forward; RWS3166: 5’-cccgggctgtcttccgcGCctggaagcccttcctgatg-3’ L986A reverse . All sub cloning and mutagenesis steps were verified by sequencing . Cell transfections were performed using Fugene6 or LipofectamineTM 2000 ( Invitrogen ) according to the manufacturer’s instructions . To obtain stably transfected lines , cells were selected on 600 µg/ml hygromycin ( InvivoGen , San Diego , CA , USA ) and/or 600 µg/ml puromycin ( InvivoGen ) for 7–10 days under constant selection pressure . Antibiotics were omitted during drug treatments and imaging . For stable retroviral transduction of AB8 cells , GP2-293 cells were transfected using the calcium phosphate method with 5 µg pQXCIP-Lifeact-acGFP and 5 µg pVSV-G ( Clontech , Mountain View , CA , USA ) in a 10 cm dish . After 6 hr , the medium was replaced and cells were grown for another 72 hr . Virus-containing supernatant was filtered through a 0 . 45 nm filter ( Millipore ) and polybrene ( 8 µg/ml ) was added . AB8 cells in a 6-well dish were transfected with 2 ml of the virus-containing medium and 2 ml of fresh AB8 medium . After 24 hr , the medium was replaced and cells were allowed to recover for 24 hr . Transduced cells were selected with puromycin ( 2 µg/ml ) ( Schulze et al . , 2014; Wennmann et al . , 2014 ) . Epifluorescence imaging was performed on a fully automated iMIC-based microscope from FEI/Till Photonics , using an Olympus 100 × 1 . 4 NA objective and DPSS lasers at 488 nm ( Cobolt Calypso , 75 mW ) and 561 nm ( Cobolt Jive , 150 mW ) as light sources . Lasers were selected through an AOTF and directed through a broadband fiber to the microscope . A galvanometer-driven two-axis scan head was used to adjust laser incidence angles . Images were collected using an Imago-QE Sensicam camera . Acquisition was controlled by LiveAcquisition software ( Till Photonics ) . Fluorescence Recovery After Photobleaching ( FRAP ) of actin-GFP was performed using a third galvanometer-controlled mirror ( Polytrope ) to switch between wide-field and FRAP modalities . Ablation experiments were carried out on an iMIC setup equipped with a pulsed 355 nm picosecond UV laser ( Sepia , PicoQuant ) as previously described ( Raabe et al . , 2009 ) . Confocal microscopy was performed on an iMIC42 setup equipped with a spinning disk unit ( Andromeda ) using Olympus 20x air ( NA 0 . 75 ) and 60x oil immersion ( NA 1 . 49 ) objectives . Images were taken using typical filter settings for excitation and emission of fluorescence probes/proteins and recorded on EMCCD cameras ( Andor iXon Ultra 897 ) . Cells were treated with 50 µM blebbistatin ( Sigma-Aldrich , St Louis , MO USA ) to inhibit myosin II ATPase activity , 400 nM latrunculin A ( Enzo Life Sciences ) to sequester actin monomers , 1 µM thapsigargin to inhibit Ca2+ uptake into the ER , 1 µM cytochalasin D ( Sigma ) to depolymerize actin , 10–50 µM cyclosporin A ( Sigma ) to inhibit calcineurin , 1–10 µM KN93 or KN62 ( Tocris ) to inhibit Ca2+/calmodulin-dependent protein kinase II , 10–20 µM Y27632 ( Sigma ) to inhibit ROCK , 2 . 5 µg/µl C3 transferase ( Cytoskeleton ) to inhibit Rho1 , 1 µM nocodazole ( Sigma ) to depolymerize microtubules , 10–20 µM SMIFH2 ( Sigma ) to inhibit formins and 50–100 µM CK666 or CK869 ( Sigma ) to inhibit the ARP2/3 complex . Silencer Select ( 21 nt ) siRNAs were purchased from Ambion/Lifetechnologies: siRNA1 ( s34736 ) : sense sequence 5’-CCAUGAAGGCUUUCCGGGAtt-3’ ) ; siRNA2 ( s230622 ) : sense sequence 5’-GCAUUGUCAUGAACGAGCUtt-3’ . As a negative control , a random , non-targeting siRNA sequence was used . HeLa cells stably expressing Lifeact-GFP were seeded on coverslips or in 8-well Ibidi slides and transfected with siRNAs ( 30 nM ) on the next day , using Oligofectamine ( Invitrogen , Carlsbad , CA , USA ) according to the manufacturer’s instructions . Cells were incubated for 72 hr and either imaged live for CaAR after ionomycin addition ( Figure 3E , Figure 3—figure supplement 1C ) or laser ablation ( Figure 3—figure supplement 1D , 6C ) or fixed at various time-points after ionomycin addition and immunostained for MRTF-A ( Figure 5C ) . To knock out INF2 in human cell lines we obtained a mix of three different CRSPR/Cas9 plasmids and the corresponding HDR plasmids from Santa Cruz ( sc-410096 ) . gRNA sequences: A: Sense: GAGGAGCTGCTGCGAGTCTC; B: Sense: GGTCGACATGAGCAGCCACC; C: Sense: CAGCGACAACGTGCCCTACG . HeLa wt cells were co-transfected with both plasmid mixes using Fugene6 and grown for 24 hr without selection . We visually confirmed the appearance of RFP expressing cells indicated successful disruption of INF2 . We next grew cells for two weeks under selection pressure with 0 . 5 µg/ml puromycin . The population of stable puromycin resistant cells was then transfected twice with a Cre expression vector ( Santa Cruz ) using Fugene6 to remove the RFP and puromycin cassette . Finally , we selected individual INF2 KO clones by limited dilution in 96-well plates . All clones were characterized for INF2 expression by Western blot and immunofluorescence and for absence of CaAR induction ( Rhodamine-phalloidin staining ) upon ionomycin stimulation . For shear-flow experiments , 5 × 103 cells were seeded in µ-slide0 . 2 Luer flow chambers ( Ibidi ) , incubated for 48 hr and then connected to the Ibidi pump system , perfused with DMEM and subjected to 20 dyn/cm2 oscillatory shear stress at 0 . 2 Hz . For elasticity measurements and mechanical perturbation experiments ( AFM ) , cells were seeded ( at 2–5 × 105 cells/ml ) on 35 mm Fluorodish glass-bottomed dishes ( WPI ) 48 hr prior to experiments . All experiments were performed using a Nanowizard III AFM ( JPK Instruments , Berlin , Germany ) integrated into a TCS SP8 confocal laser scanning microscope ( Leica , Wetzlar , Germany ) . Gold-coated MSCT cantilevers with spherical ( 10 µm ) polystyrene probes ( Novascan , Ames , IA; USA ) with a nominal spring constant of 0 . 01 pN/nm were used to quantify cell elasticity . Force-distance curves were acquired with a z-length of 2 . 5 µm at a tip velocity of 1 µm/s and a retracted delay of 1 s ( ~1 force distance curve per 6 s ) . At a loading force of 0 . 75 nN the cantilever reached a maximal indentation of 1 µm ( Carl and Schillers , 2008 ) . Elasticity was measured as Young’s modulus based on Sneddon's model of nanoindentation using Protein Unfolding and Nano-Indentation Software ( PUNIAS , http://punias . voila . net ) . The last 200 nm of the force distance curve was analyzed to quantify the elasticity of the cell cytoplasm . All elasticity measurements were performed at RT . For all cell perforation experiments , Multi75-G Cantilevers ( BudgetSensors , Sofia , Bulgaria ) with spring constants ranging from 2 . 5 to 3 . 5 N/m were used . Cells were stimulated by placing the cantilever tip in the center of the nuclear area and indenting them with a speed of 1 µm/s . To determine the force threshold for CaAR induction , maximum loading forces ranging from 5 nN to 50 nN were used . To study repeated CaAR induction , cells were indented ten times with a maximum loading force of 50 nN and variable pauses between single indentations . Fluorescence images were acquired with a 63x HC PL APO CS2 oil immersion objective ( NA = 1 . 4 ) and hybrid detection system for photon counting ( Leica HD™ ) . Z-stacks of cells ( distance 500 nm ) were taken every 30 s . All confocal fluorescence images were analyzed and processed using Fiji . Fluorescence intensity was measured in unprocessed images . For presentation , grouped z-projections of three slices were cropped and contrast-adjusted . Images were processed using Fiji and Matlab ( Mathworks Inc . , Natrick , MA ) . Custom made Matlab scripts are included as Figure2-source codes 1–3 . Images were contrast-adjusted and zoomed for purposes of presentation in figures only . For image cleanup and denoising we routinely used the background subtraction algorithm in Fiji ( radius 50 pixel ) . Kymographs and intensity plots were created using the respective features in Fiji . Cells were loaded with Fluo-4 ( 5 µM , Thermo Fisher ) or Fura-2 ( 3 µM , Thermo Fisher ) for 15–30 min at 37°C . Fluo4 was excited at 488 nm . Fura-2 intensity was determined by ratiometric measurement using excitation at 340 and 380 nm ( detection at 500 nm ) . Fura-2 fluorescence was acquired using an Axiovert 200 ( Zeiss , Wetzlar , Germany ) equipped with a VisiChrome high speed polychromator system ( Visitron System , Puchheim , Germany ) , a CoolSNAP fx camera ( Photometrix ) and Metafluor imaging software ( Visitron System ) . Ca2+-dependent fluorescence was acquired every 10 s in alternation with imaging of Lifeact-GFP fluorescence . Cells were grown on glass coverslips , fixed with 4% paraformaldehyde in PBS for 20 min , washed with PBS and permeabilized with 0 . 1% Triton X-100 for 10 min prior to incubation with primary Ab for 1 hr . After incubation with secondary antibodies and/or Rhodamine-phalloidin ( Invitrogen , R415 ) for 1 hr in PBS , cells were washed and subsequently mounted on slides in Mowiol/Dabco ( Roth , 0713 and 0718 ) . Primary antibodies: rabbit anti-INF2 ( Chhabra et al . , 2009 ) , goat anti-MRTF-A ( SantaCruz: C-19; sc-21558 , RRID:AB_2142498 ) . Secondary antibodies: Alexa-Fluor 568 goat anti-rabbit/mouse , Alexa-Fluor donkey anti-goat 568 and Alexa-Fluor donkey anti-rabbit 647 ( all Invitrogen ) . LysoTracker Red ( Thermo , L7528 ) was used to label lysosomes . MitoTracker Red CMXRos ( Thermo , M7512 ) was used to label mitochondria . For detection of INF2 and actin in Western blots , equal amounts of cell lysates were separated by SDS-PAGE , transferred to ImmobilonTM-P-membrane ( Serva , 42581 . 01 ) , incubated in primary Ab ( rabbit anti-Inf2 , Proteintec: 20466–1-AP , RRID:AB_10694821 ) and ( Chhabra et al . , 2009 ) , ( mouse monoclonal anti-β-actin , Abcam: ab125248 , RRID:AB_11140352 ) , in the presence of 5% skim milk in TBS-T overnight and labeled with HRP-coupled secondary Ab for one hour . The signal was detected using ECL chemiluminescence on an Intas Imager ( Intas , Göttingen ) . Cell culture , transfection , coprecipitation with calmodulin and detection of associated proteins were conducted as described previously ( Hou et al . , 2015 PLoS Biol . ) . Briefly , HEK293 cells were cultured at 37°C under 5% CO2 in DMEM medium ( Gibco Life technologies ) additionally supplied with 10% fetal bovine serum , 0 . 1 mM gentamicin and 1x penicillin ( 100 units/ml ) -streptomycin ( 100 μg/ml ) . GFP and GFP-INF2-CAAX were transfected into HEK293 cells using TurboFect ( Thermo Scientific ) . 48 hr after transfection , cells were lysed in EGTA-free lysis buffer ( 10 mM HEPES , 1% ( v/v ) Triton X-100 , 100 mM NaCl , 0 . 1 mM MgCl2 , pH 7 . 5 ) supplied with 1x EDTA- free protease inhibitor cocktail and 200 μM calpain inhibitor I ( Sigma ) for 30 min and the supernatants containing soluble proteins were obtained after centrifugation ( 20 . 000 g , 20 min at 4°C ) . Cell lysates were supplied with 1 mM EGTA and 500 µM Ca2+ , respectively , then incubated with pre-washed 25 µl calmodulin-sepharose 4B ( GE Healthcare ) for 3 hr . calmodulin bound proteins were washed three times then eluted and denatured in 15 µl 8 M Urea and 15 µl 4x SDS-sample buffer by boiling for 5 min at 100°C . Protein samples were separated by SDS-PAGE using 9 . 5% acrylamide separation gels followed by western blot detection using anti-GFP antibodies ( Abcam , Ab290 ) . Results are representative of at least three experiments . MCF-7 cells were serum starved for 1 hr in HBSS and then treated with 50 µM ATP for 0 . 5 , 1 , 2 or 5 min . Cells were scraped off in ice-cold actin stabilization buffer ( 50 mM PIPES – pH9 , 50 mM NaCl , 5 mM MgCl2 , 5 mM EGTA , 5% glycerol , 0 . 1% NP-40 , 0 . 1% Triton-X-100 , 0 . 1% , Tween 20 , 0 . 1% 2-mercapto-ethanol , 0 . 001% antifoam C , 0 . 004 mM TAME , 15 µM leupeptin , 10 µM pepstatin A , 10 mM benzamidine ) . Cleared supernatants were centrifuged at 100 , 000 g for 1 hr at 4°C . High speed supernatant containing G-actin was recovered , and the pellet containing F-actin was re-solubilized with ice cold actin stabilization buffer . The F/G-actin ratio was determined by scanning densitometry using ImageJ software . For analysis of CTGF and NR4A3 mRNA levels , total RNA was extracted from 1 × 106 MCF-7 cells using the RNeasy Mini Kit ( Qiagen ) following the manufacturer’s protocol . cDNA synthesis was performed from 2 µg total RNA using cDNA Reverse Transcription Kit ( Thermo Fisher ) . QPCR was performed in triplicates from 3–4 biological replicates . QPCR primers are available upon request . mRNA levels were normalized to those of B2M and GAPDH ( Figure 5—figure supplement 1C , D ) . Treatment with drugs was performed at the indicated concentrations . Drugs were added for 10 min before stimulation with ionomycin . MCF-7 cells were starved in HBSS buffer for 1 hr before stimulation with 1 . 5 µM ionomycin for 10 min . After ionomycin washout , cells were incubated for an additional 50 min in the presence of 4-thiouridine . Cells were then lysed and total RNA was extracted from 1 × 107 MCF-7 cells using the RNeasy Mini Kit ( Qiagen ) following the manufacturer’s protocol . LatA ( 400 nM ) or CytoD ( 2 µM ) were added 10 min before ionomycin and removed together with ionomycin . 4-thiouridine-labeled RNA was chemically biotinylated ( Biotin HPDP , Thermo Fisher ) and purified using streptavidin-coated magnetic beads ( µMACS , Miltenyi ) . Subsequent labeling of samples for array analysis was performed with the GeneChip 3’IVT labeling assay ( Affymetrix ) . Samples were hybridized to the GeneChip Human Gene 2 . 0 ST Array following the instructions from the supplier ( Affymetrix ) . Quality control and array processing was done using GCRMA ( Wu and Irizarry , 2004 ) for expression and LIMMA ( Smyth and Speed , 2003 ) for elementary array comparisons and computation of fold-changes and p-values . Mean values , standard deviation ( SD and number of measurements ( n ) are provided for quantified results . Values were always pooled from at least three independent experiments . All replicates are biological replicates . Error bars in graphs are explained in the respective legends . Statistical comparison between conditions was performed using the unpaired t-test with Welch correction for single comparisons or ANOVA with Bonferroni post-test for multiple comparisons . For transcriptome analysis , genes were considered differentially regulated if expression levels deviated >1 . 5 fold from the mean of control cells and p value ( ANOVA ) was <0 . 05 ( Figure 5—source data 2 ) . Our initial hypothesis was that , after stimulation , a calcium-regulated polymerization activity at the endoplasmic reticulum ( ER ) begins to compete with actin polymerization at the cortex for a limited amount of actin monomers . We therefore considered two populations of polymerized actin ( F-actin ) coupled to a common pool of actin monomers ( G-actin , Figure 4C ) . We assumed that the available G-actin , characterized by the equivalent length LG of actin into which it can potentially polymerize , is homogeneously distributed in the cytoplasm . F-actin populations at the cortex and ER were modeled as ensembles of individual filaments . The state of these populations is characterized by their respective total lengths of polymerized actin , LC and LER . Initially , we consider the cortical population in steady-state equilibrium with the G-actin pool . Nucleation at the ER is then activated and regulated by the calcium influx signal . The ER-mediated decrease in monomer availability in turn reduces polymerization at the cortex . Ongoing actin turnover will then drive cortical actin disassembly . When the calcium signal has dissipated , nucleation at the ER stops and ER actin disassembles . This restores the G-actin pool , allowing cortical actin to recover again .	Our skeleton plays a vital role in giving shape and structure to our body , it also allows us to make coordinated movements . Similarly , each cell contains a microscopic network of structures and supports called the cytoskeleton that helps cells to adopt specific shapes and is crucial for them to move around . Unlike our skeleton , which is relatively unchanging , the cytoskeleton of each cell constantly changes and adapts to the specific needs of the cell . One part of the cytoskeleton is a dense , flexible meshwork of fibers called the cortex that lies just beneath the surface of the cell . The cortex is constructed using a protein called actin , and many of these proteins join together to form each fiber . When cells need to adapt rapidly to an injury or other sudden changes in their environment they activate a so-called stress response . This response often begins with a rapid increase in the amount of calcium ions inside a cell , which can then trigger changes in actin organization . However , it is not clear how cells under stress are able to globally remodel their actin cytoskeleton without compromising stability and integrity of the cortex . Wales , Schuberth , Aufschnaiter et al . used a range of mammalian cells to investigate how actin responds to stress signals . All cells responded to the resulting influx of calcium ions by deconstructing large parts of the actin cortex and simultaneously forming actin filaments near the center of the cell . Wales , Schuberth , Aufschnaiter et al . termed this response calcium-mediated actin reset ( CaAR ) , as it lasted for only a few minutes before the actin cortex reformed . The experiments show that a protein called INF2 controls CaAR by rapidly removing actin from the cortex and forming new filaments near a cell compartment called the endoplasmic reticulum . CaAR allows cells to rapidly and drastically alter the cortex in response to stress . The experiments also show that this sudden shift in actin can change the activity of certain genes , leading to longer-term effects on the cell . The findings of Wales , Schuberth , Aufschnaiter et al . suggest that calcium ions globally regulate the actin cytoskeleton and hence cell shape and movement under stress . This could be relevant for many important processes and conditions such as wound healing , inflammation and cancer . A future challenge will be to understand the role of CaAR in these processes .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2016	Calcium-mediated actin reset (CaAR) mediates acute cell adaptations
In fluctuating environments , switching between different growth strategies , such as those affecting cell size and proliferation , can be advantageous to an organism . Trade-offs arise , however . Mechanisms that aberrantly increase cell size or proliferation—such as mutations or chemicals that interfere with growth regulatory pathways—can also shorten lifespan . Here we report a natural example of how the interplay between growth and lifespan can be epigenetically controlled . We find that a highly conserved RNA-modifying enzyme , the pseudouridine synthase Pus4/TruB , can act as a prion , endowing yeast with greater proliferation rates at the cost of a shortened lifespan . Cells harboring the prion grow larger and exhibit altered protein synthesis . This epigenetic state , [BIG+] ( better in growth ) , allows cells to heritably yet reversibly alter their translational program , leading to the differential synthesis of dozens of proteins , including many that regulate proliferation and aging . Our data reveal a new role for prion-based control of an RNA-modifying enzyme in driving heritable epigenetic states that transform cell growth and survival . Cell size and proliferation are fundamental determinants of development , survival , and disease ( Su and O’Farrell , 1998 ) . Although they can be independently controlled—for example , when cell division is disrupted , cell size continues to increase , whereas restricting cell growth does not completely inhibit division ( Johnston et al . , 1977 ) —coupling is common due to their dependence on the same biochemical building blocks ( Su and O’Farrell , 1998; Turner et al . , 2012 ) . Changing cell size and proliferation genetically or through chemical perturbation—whether via diet , nutrient sensing ( e . g . , TOR signaling ) , or growth factor signaling and stress responses—also impacts lifespan across eukaryotes ( Fontana et al . , 2010; Kenyon , 2010; Pitt and Kaeberlein , 2015 ) . The importance of tightly controlling each of these properties is underscored by the many mutations affecting cell size and proliferation that are pathogenic , leading to cancer , developmental abnormalities , and numerous diseases of age ( reviewed in Hanahan and Weinberg , 2011; Saxton and Sabatini , 2017 ) . Organisms commonly alter the relationship between cell size and proliferation during development ( Su and O’Farrell , 1998 ) . In oocytes of Drosophila melanogaster , for example , a massive expansion in cytoplasmic volume without cellular division precedes a later step of numerous cellular divisions without cytoplasmic growth . The rates of cell size increase and proliferation can be strongly coupled to nutrient sensing and changes in metabolism ( Efeyan et al . , 2015; Turner et al . , 2012 ) . Thus , organisms can use genetically encoded signaling pathways to commit to different strategies depending on their needs and environment ( Ivanov et al . , 2015; Jung et al . , 2018 ) . Epigenetic tuning of cell size and proliferation could in principle provide a stable yet reversible mechanism to alter the relationship between these properties according to differing needs in fluctuating environments . Histone modifications can enable such adaptation in a way that can be heritable over several mitotic divisions . However , apart from a few notable exceptions ( Catania et al . , 2020; Grewal and Klar , 1996; Nakayama et al . , 2000 ) , most studied examples are erased during meiosis ( Heard and Martienssen , 2014; Moazed , 2011 ) . Prions are distinct among epigenetic mechanisms in that they are faithfully transmitted through both mitotic and meiotic cell divisions ( Brown and Lindquist , 2009; Cox , 1965; Garcia and Jarosz , 2014; Wickner , 1994 ) . The unusual folding landscape of prion proteins , which allows the recruitment of proteins from the naïve to the prion fold , promotes a mode of inheritance that is both stable and reversible ( Chakrabortee et al . , 2016a; McKinley et al . , 1983 ) . For example , transient perturbations in molecular chaperone activity ( Brown and Lindquist , 2009; Chernoff et al . , 1995 ) , specific environmental stressors ( Garcia et al . , 2016; Singh et al . , 1979; Tuite et al . , 1981 ) , or regulated proteolysis ( Ali et al . , 2014; Kabani et al . , 2014 ) can induce or eliminate prion states . It is now appreciated that this form of information transfer is far more common than previously realized , occurring not only in mammals but also in wild fungi , plants , and even bacteria ( Chakrabortee et al . , 2016b; Halfmann et al . , 2012; Yuan and Hochschild , 2017 ) . Here we investigate the inheritance of a prion-based epigenetic state that alters yeast cell physiology , potentiating a trade-off between proliferation and lifespan . We first describe the [BIG+] prion—driven by the conserved pseudouridine synthase Pus4 ( known as TruB in mammals and bacteria ) —which increases proliferation but shortens lifespan . We then quantitatively model the adaptive value of this ‘live fast , die young’ growth strategy in fluctuating environments . [BIG+] cells are larger and exhibit increased protein synthesis while maintaining Pus4-dependent pseudouridylation activity . Finally , we find evidence for analogous Pus4-dependent epigenetic control of cell size in wild yeast . The epigenetic inheritance of an altered form of an RNA-modifying enzyme over long biological timescales , as occurs in [BIG+] , reveals a new mechanism by which regulation of these activities can be perpetuated across generations . We recently discovered more than 40 protein-based epigenetic elements in Saccharomyces cerevisiae that are both heritable and reversible upon transient perturbation of protein chaperone function ( Chakrabortee et al . , 2016a ) . Many of the proteins that underlie this behavior have the potential to regulate growth . One of these epigenetic states was induced by transient overexpression of the highly conserved pseudouridine synthase PUS4/TRUB , which catalyzes the formation of a ubiquitous pseudouridine on U55 in tRNAs in bacteria , yeast , and humans ( Becker et al . , 1997; Gutgsell et al . , 2000; Zucchini et al . , 2003 ) . Mutation of U55 leads to large fitness deficits , second only to mutations in the tRNA anticodon loop , highlighting the functional importance of this nucleotide ( Li et al . , 2016 ) . Mutation of U55 is also linked to deafness and diabetes in humans ( Wang et al . , 2016 ) . We originally discovered this epigenetic state because transient overexpression of PUS4 led to an enduring and heritable growth improvement in medium containing zinc sulfate ( Chakrabortee et al . , 2016a ) . This phenotype could possibly be explained by the strong genetic interaction between PUS4 and the zinc homeostasis regulator ZRC1—among all genes , ZRC1 is the second strongest genetic interactor with PUS4 ( Koh et al . , 2010 ) . Upon closer examination , we noticed that cells harboring the Pus4-induced element also achieved an ~60% faster maximal proliferation rate than naïve cells in the absence of zinc stress in standard rich medium ( YPD; Figure 1A , p=0 . 0095 , unpaired t-test ) . We therefore initially named this mitotically heritable element ‘Big+’ for better in growth . We next tested whether this growth advantage from the Big+ epigenetic element was more pronounced during direct competition and oscillating nutrient availability—as a single-celled organism such as yeast often faces in nature . To do so , we performed a competition experiment that encompassed periods of abundant nutrient availability followed by starvation . Using resistance to canavanine as a marker to distinguish naïve and Big+ strains , we mixed equal numbers of cells from each strain . We propagated the mixed culture for approximately 100 generations , diluting into fresh medium and measuring the fraction of the population that harbored the canavanine resistance marker every 10 generations ( Figure 1—figure supplement 1A ) . In these experiments , cells harboring Big+ invariably outcompeted the genetically identical naïve cells that lacked it—they live fast—with a selection coefficient of nearly 1% ( Figure 1B ) . Competitions using reciprocally marked strains produced equivalent results . As a frame of reference , the fitness advantage that we measured for Big+ is larger than those conferred by >30% of non-essential genes ( Breslow et al . , 2008 ) , and the vast majority of natural genetic variants that have been quantified ( Jakobson and Jarosz , 2019; Jakobson et al . , 2019; Sharon et al . , 2018; She and Jarosz , 2018 ) . Given the close link between proliferation and aging in many organisms ( Bitto et al . , 2015 ) , we investigated whether Big+ also influenced lifespan . Studies of aging in budding yeast have led to the discovery of numerous genes with conserved roles in aging of metazoans ( Kaeberlein , 2010 ) . These studies have measured two types of aging , both of which we tested here—chronological lifespan and replicative lifespan ( RLS; Longo et al . , 2012 ) . Chronological lifespan is a measure of post-mitotic viability , during which cells cease division under starvation conditions until nutrients become available again . These conditions occur commonly in the natural ecology of budding yeast ( Landry et al . , 2006 ) . To investigate , we aged cultures of naïve cells and genetically identical cells harboring Big+ ( Figure 1—figure supplement 1B ) . Over the course of 80 days , Big+ cells had progressively lower viability than matched isogenic-naïve controls ( Figure 1C ) . RLS measures the number of cell divisions that yeast can undergo before death ( Figure 1—figure supplement 1B ) . These experiments revealed a significant difference in replicative potential between Big+ and naïve cells—the median survival rate is reduced by seven generations ( Figure 1D , p<0 . 0001 , Gehan–Breslow–Wilcoxon test , and Figure 1—figure supplement 1C ) . This degree of shortening is comparable to that of classic lifespan-altering mutants: yeast lacking SIR3 or SIR4 have reductions of ~3–4 generations , and yeast lacking SIR2 by ~10–11 generations ( Kaeberlein et al . , 2005; Kaeberlein et al . , 1999 ) . Thus , Big+ cells harboring this Pus4-induced element exhibit both a significantly decreased chronological lifespan and RLS: they die young . We next investigated the adaptive value of the Big+ phenotype by quantitatively modeling its fitness consequences in fluctuating environments , where committing to a single strategy can impose limits on the long-term fitness of a population . For example , when nutrient-rich periods tend to greatly exceed starvation periods , the rapid growth of the Big+ phenotype might be favored , despite its die-young phenotype ( Figure 2A ) . However , this same decision could be maladaptive if growth conditions skewed towards frequent and more extended periods of nutrient scarcity—where cells that grow slower and die older would instead have an advantage ( Figure 2B ) . To quantify these trade-offs , we considered a population in which individuals heritably adopted one of these two strategies ( live-fast-die-young , LF-DY , like Big+ cells; or live-slow-die-old , LS-DO , like naïve cells ) , modeling their fates in alternating nutrient environments ( Figure 2C ) . These simulations suggested that LS-DO cells should outcompete LF-DY cells when the periods of starvation are much longer than periods of nutrient abundance ( Figure 2—figure supplement 1A ) . By contrast , LF-DY cells come to dominate the simulated culture when periods of nutrient abundance are much longer than periods of starvation ( Figure 2—figure supplement 1B ) . When periods of starvation and nutrient repletion are of equal duration , both populations are equally fit ( Figure 2—figure supplement 1C ) . When we varied the growth advantage and lifespan cost of the LF-DY sub-population and the periods of nutrient availability and starvation , we noted that each phase space contained regimes in which either strategy could be advantageous ( Figure 2C and Figure 2—figure supplement 1D ) . Regimes in which either the LF-DY or LS-DO strategies would be strongly adaptive ( and the other maladaptive ) frequently arose under physiologically relevant environmental parameters . Oscillations within these regimes ( i . e . , between feast and famine ) are common in nature ( Broach , 2012 ) , and withstanding them is essential for survival . Theory predicts that reversible epigenetic mechanisms , such as prions , could confer a selective advantage in fluctuating environments ( King and Masel , 2007 ) . Importantly , our model demonstrates that this advantage could derive not only from a trade-off between improved stress resistance and impaired growth in normal conditions but also from a growth advantage in times of plenty coupled with a disadvantage under stress . Notably , the growth advantages we observed in our competition experiment were quantitatively consistent with Monte Carlo simulations sampled from our experimental measurements of the death rates ( from chronological lifespan measurements , Figure 2D ) and growth rates ( from proliferation rate measurements , Figure 2E ) of individual cultures in nutrient-starved and replete conditions , respectively ( Figure 2—figure supplement 1E ) . That is , the adaptive advantages that we measured in competition were equivalent to those that we predicted for a hypothetical LF-DY population after dozens of generations ( Figure 2F ) . Thus , selection on these properties could be alone sufficient to favor maintenance of the Big+ state under the conditions we examined . Positive correlations between growth rate and cell size have long been noted ( Johnston et al . , 1977; Schaechter et al . , 1958; Su and O’Farrell , 1998; Turner et al . , 2012 ) . To determine if the faster-growing Big+ cells were also larger , we examined them microscopically , employing a widely used image masking pipeline ( Carpenter et al . , 2006 ) . This approach allowed us to measure the sizes of thousands of cells from multiple biological replicates and define size distributions for both naïve and Big+ populations . During exponential growth , populations of Big+ cells had a similar size distribution to naïve control cells . However , as the cultures became denser , naïve control cells remained the same size whereas isogenic haploid Big+ cells became larger ( Figure 3A and Figure 3—figure supplement 1A ) . To simplify these comparisons , we scored distributions by the fraction of very large cells that we observed ( one standard deviation or larger than the mean naïve size; Figure 3B , n = 4678 and 5501 cells shown for naïve and Big+ , respectively ) . This increase in mean area from 22 . 01 µm2 to 25 . 36 µm2 corresponds to a 23% larger volume ( approximating the yeast cell as a sphere , naïve cell mean radius = 2 . 65 µm , Big+ cell mean radius = 2 . 84 µm ) . In Big+ cultures 33 . 1% ± 2 . 8% of cells exceeded this threshold , whereas 16 . 5% ± 1 . 8% did in naïve cultures ( p=0 . 0011 by unpaired t-test; Figure 3C ) . We note that this difference is not as large as that between haploid and diploid yeast; the naïve diploid mean area is 36 . 88 µm2 . Moreover , in mRNA-sequencing—discussed in more detail below—we observed expression profiles in Big+ cells consistent with being haploid and mating type Mat a ( like their naïve counterparts ) ( Figure 3—figure supplement 1B and C ) . We therefore conclude that the Big+ state does not represent a simple autodiploidization . Big+ was initially induced by transient PUS4 overexpression in a screen to identify prion-like epigenetic elements ( Chakrabortee et al . , 2016a ) . We therefore tested whether the increased cell size associated with this state was transmitted through genetic crosses with the unusual patterns of inheritance that characterize prion-based phenotypes ( Brown and Lindquist , 2009; Cox , 1965; Wickner , 1994 ) . We began by mating the large haploid Big+ cells to naïve haploids of the opposite mating type , selecting diploid cells , and measuring their size . The resulting diploids were significantly larger than those derived from control crosses with two naïve parents ( Figure 4A and B ) , establishing that the trait is dominant . We next investigated the meiotic inheritance of the large cell size trait . Because prions are not driven by changes in nucleic acid sequence , they have unusual patterns of inheritance that defy Mendel’s laws . In addition to dominance in genetic crosses , prion-based traits can be passed to all progeny of meiosis , in contrast to DNA-based traits , which are inherited by half ( Figure 4—figure supplement 1A; Garcia and Jarosz , 2014; Li and Kowal , 2012; Liebman and Chernoff , 2012; Wickner , 2016 ) . We first mated control naïve cells to naïve cells of the opposite mating type , sporulated the resulting diploids , and dissected their meiotic progeny . We then grew clonal cultures of these haploid meiotic progeny and examined their size distributions , which were indistinguishable from their haploid-naïve parents ( Figure 4C ) . In contrast , all cultures derived from the meiotic progeny of Big+× naïve crosses were large ( Figure 4C and Figure 4—figure supplement 1B ) . This non-Mendelian pattern of inheritance differs strongly from the expected behavior for genetic mutants or chromatin-based epigenetic elements but is consistent with a prion-based mechanism of transmission . We therefore term this state ‘[BIG+]’ ( with capital letters indicating dominance and brackets denoting its non-Mendelian pattern of segregation ) . In contrast to those driven by genetic mutations , the inheritance of prion-based phenotypes is strongly dependent upon the protein homeostasis network ( Garcia and Jarosz , 2014; Harvey et al . , 2018; Liebman and Chernoff , 2012; Shorter and Lindquist , 2005 ) . As a consequence , transient inhibition of molecular chaperones can lead to the permanent elimination of prion-based traits . We therefore examined whether the large size of [BIG+] cells also depended on protein chaperone activity ( Figure 4—figure supplement 1C ) . Transient inhibition of the Hsp104 disaggregase , which regulates the inheritance of many amyloid prions ( Chernoff et al . , 1995; Eaglestone et al . , 2000; Halfmann et al . , 2012; Shorter and Lindquist , 2004 ) , did not affect cell size in either naïve or [BIG+] cells ( Figure 4D ) . Transient inhibition of the Hsp90 foldase , which regulates the transmission of a different subset of prions ( Chakrabortee et al . , 2016a ) , also had no impact on [BIG+] transmission ( Figure 4E ) . By contrast , transient inhibition of Hsp70 via expression of a dominant-negative SSA1K69M allele ( Chakrabortee et al . , 2016a; Jarosz et al . , 2014; Lagaudriere-Gesbert et al . , 2002 ) caused [BIG+] cells to lose their large size phenotype permanently ( Figure 4F ) . Thus like other prions ( Brown and Lindquist , 2009; Chakrabortee et al . , 2016a; Chakravarty et al . , 2020 ) , and unlike genetic mutations , propagation of [BIG+] is dependent on the activity of this ubiquitous molecular chaperone . We note that Hsp70 expression drops dramatically as yeast reach saturation and begin to starve ( Werner-Washburne et al . , 1989; Werner-Washburne and Craig , 1989 ) and also decreases as cells age ( Janssens et al . , 2015 ) . These are two scenarios in which [BIG+] is disadvantageous . Therefore , environmental conditions that favor growth , during which Hsp70 is abundant , also favor prion propagation . By contrast , conditions known to reduce Hsp70 expression and thereby increase prion elimination are also those in which prion loss would be favored . Prion states often impact the localization of the proteins that encode them . To visualize endogenous levels of Pus4 expression in naïve and [BIG+] cells , we employed a strain in which an N-terminal GFP tag was appended at the endogenous PUS4 locus , and therefore subject to endogenous regulation ( i . e . , not overexpressed; Weill et al . , 2018; Yofe et al . , 2016 ) . We did not observe large fluorescent foci typical of amyloid prions ( Alberti et al . , 2009 ) , as might be expected given that Pus4 is not enriched in glutamine and asparagine residues typically abundant in amyloid prion proteins ( Chakrabortee et al . , 2016a ) . We did , however , observe altered localization of Pus4 . In naïve cells , Pus4 consistently localized to the nucleolus , as has been previously reported ( Huh et al . , 2003 ) . In [BIG+] cells , the Pus4 protein was also present in the nucleolus , but we also observed substantial fluorescence in a fragmented network throughout the cytoplasm ( Figure 4G ) . Although a high-resolution structure awaits determination , our data are consistent with an altered state of Pus4 in [BIG+] cells . Although [BIG+] was induced by a transient increase in Pus4 abundance and was stable after eliminating the inducing plasmid , we wanted to exclude the possibility that this prior overexpression event might have established a positive feedback loop leading to an enduring increase in Pus4 levels . We therefore constructed naïve and [BIG+] strains with a seamless N-terminal 3X-FLAG tag endogenously encoded at the PUS4 locus . Using immunoblots to detect the FLAG epitope in naïve and [BIG+] cells , we observed equivalent Pus4 levels , indicating that the phenotypes we observed in [BIG+] cells were not simply due to increased expression of this RNA-modifying enzyme ( Figure 4—figure supplement 1D ) . Prion proteins can be inherited through the cytoplasm and do not require the exchange of genetic material for propagation . To test this , we performed a cytoduction , in which we mated [BIG+] cells to naïve recipient cells of the opposite mating type carrying the kar1-∆15 mutation . Upon mating , this mutation prevents nuclear fusion , permitting the mixing of cytoplasm , but not nuclei , between donor and recipient cells ( Figure 4—figure supplement 1E , Materials and methods; Vallen et al . , 1992 ) . The transfer of [BIG+] cytoplasm into naïve recipient cells resulted in the transfer of the [BIG+] cell size phenotype ( Figure 4H ) . However , this was only the case for wild-type recipients . Naïve recipients lacking PUS4 did not acquire the [BIG+] cell size phenotype ( Figure 4I ) . It remained formally possible that a multi-protein prion state could be maintained by other cellular factors , even if prion-based phenotypes depended on Pus4 . Therefore , we tested whether transient loss of PUS4 was sufficient to eliminate [BIG+] permanently . We first deleted PUS4 from [BIG+] and naïve cells . Upon PUS4 deletion , the sizes of the mutants derived from naïve and [BIG+] parents were equivalent ( Figure 4—figure supplement 1F ) . Together with our cytoduction data , this suggested that continuous production of Pus4 is required to maintain this trait . Furthermore , PUS4 deletion did not increase the size of naïve cells , suggesting that [BIG+] does not inactivate Pus4 , in contrast to many well-characterized prions that phenocopy loss-of-function alleles of their underlying proteins ( Byers and Jarosz , 2014 ) . Finally , we restored the PUS4 gene to its native locus in these same cells by homology-directed integration . Even after the re-introduction of PUS4 , the size distributions of both populations remained equivalent ( Figure 4—figure supplement 1F ) . Thus , both the expression and the propagation of the [BIG+] phenotype require the continual presence of a PUS4 gene product . These various lines of evidence—transmission to all meiotic progeny , dependence on molecular chaperones , altered protein localization , and requirement for continuous expression of the protein that initiated the epigenetic trait—lead us to propose that [BIG+] is a protein-based element of inheritance , a prion , formed by the Pus4 pseudouridine synthase . Because loss of PUS4 did not phenocopy [BIG+] but did block propagation of the prion , we wondered whether the catalytic activity of Pus4 was maintained in [BIG+] cells . To measure pseudouridylation in naïve and [BIG+] cells , we employed a qPCR-based method that capitalizes on the enhanced susceptibility of pseudouridine to reacting with CMC ( 1-cyclohexyl- ( 2-morpholinoethyl ) carbodiimide metho-p-toluene sulfonate ) ; Figure 5A; Lei and Yi , 2017 . When pseudouridines are ‘labeled’ by CMC , and these RNAs are used as templates for replication by reverse transcriptase , the enzyme generates nucleotide deletions and other mutations at CMC-labeled sites that can be detected by differences in the melting temperature of the derived nucleic acid duplexes ( as compared to non-pseudouridylated or unlabeled controls ) . Using this approach , we examined the pseudouridylation of an archetypical Pus4 substrate , tRNAAGC ( Ala ) . When a pseudouridine is present and CMC is added , the melting curve shifts relative to an unlabeled control ( no CMC ) . We observed a similar leftward shift in melting curves for both naïve and [BIG+] cells ( Figure 5B–D ) . In contrast , control cells lacking the Pus4 protein , and therefore lacking pseudouridylation at U55 , did not produce this shift ( Figure 5B–D ) . Because tRNAs bear pseudouridines at other positions—catalyzed by different pseudouridine synthase enzymes—which could potentially interfere with our analysis of Pus4 activity , we also directly examined the U55 position . We cloned the PCR products into plasmids and then used Sanger sequencing to identify positions with mutations indicating a CMC adduct . Within the T-arm stem-loop of tRNAAGC ( Figure 5—figure supplement 1A ) , we saw mutations originating at U55 in CMC-labeled samples from both naïve and [BIG+] cells , but not in unlabeled samples or pus4Δ samples ( Figure 5—figure supplement 1B ) . These data establish that Pus4-dependent modification of tRNAs at U55 is maintained in [BIG+] cells . In addition to its ubiquitous pseudouridylation activity on all tRNAs , Pus4 also modifies some mRNAs ( Carlile et al . , 2014; Lovejoy et al . , 2014; Schwartz et al . , 2014 ) , which may impact mRNA stability ( Zhao et al . , 2017 ) . We performed mRNA sequencing to discern whether the phenotypes of [BIG+] cells might be due to relative changes in mRNA levels . We grew naïve and [BIG+] cultures in YPD medium until late exponential phase , extracted total RNA , and enriched these samples for polyadenylated RNAs . Comparing four replicates each of naïve and [BIG+] , the expression levels of only 47 genes changed significantly ( 28 decreased in expression and 19 increased as estimated using Kallisto and Sleuth; Bray et al . , 2016; Pimentel et al . , 2017; Benjamini–Hochberg-corrected q-value < 0 . 05 from the likelihood ratio test and \|log2 fold-change\| > 1 ) ( Supplementary file 2 ) . Spike-in controls demonstrated quantitative scaling in our estimates of gene expression across a broad range of transcript abundance ( Figure 5—figure supplement 1C ) . The upregulated genes in [BIG+] exhibited only a modest enrichment for water transport ( two aquaporin genes ) , while the downregulated genes were enriched for cell wall and plasma membrane components ( Supplementary file 2; Szklarczyk et al . , 2019 ) , perhaps because larger [BIG+] cells have less surface area per unit volume . We conclude that the major effects of [BIG+] during exponential growth ( e . g . , on growth rate and RLS ) likely do not occur via changes to steady-state mRNA levels . To investigate whether the relative abundance of tRNAs is perturbed in [BIG+] cells , we ran total RNA from naïve and [BIG+] cells ( grown to late-exponential phase ) on a nucleic acid fragment analyzer and quantified tRNA levels relative to a similarly abundant RNA that is not a target of Pus4 , 5 . 8 S rRNA ( 158 nt ) ( Figure 5—figure supplement 1D ) . We observed no significant differences . While we cannot exclude other effects on tRNA function or the relative abundance of particular tRNAs , our data show that bulk tRNA abundance differences are likely also not responsible for the phenotypes of actively growing cells containing [BIG+] . Because the increased cell size and proliferation and reduced lifespan of [BIG+] cells occur in the absence of major alterations to relative abundances of mRNA or tRNA , we wondered whether a change in protein synthesis might be responsible . To test this , we employed two inhibitors: ( 1 ) cycloheximide , an inhibitor of translational elongation ( Baliga et al . , 1969; McKeehan and Hardesty , 1969 ) , and ( 2 ) rapamycin , a natural product macrolide that inhibits the TOR kinase , blocking a conserved signaling cascade that promotes protein synthesis ( Beretta et al . , 1996; Chung et al . , 1992; Kuo et al . , 1992; Price et al . , 1992; Urban et al . , 2007 ) . [BIG+] cells grew nearly twofold better than naïve cells in a sub-inhibitory concentration of cycloheximide that was sufficient to impair proliferation ( 0 . 01 µg/mL , in which naïve cells needed 23 . 6 hours to reach their maximum proliferaton rate vs . 21 . 3 hours in YPD; for comparison , 100 µg/mL is typically used for experiments that rapidly and completely arrest translation; Figure 6A ) . [BIG+] cells also proliferated faster in a concentration of rapamycin ( 10 µM ) that inhibited growth ( Figure 6B ) , suggesting that the pathway may be more active in cells harboring the prion . We found additional evidence in support of this hypothesis by examining the localization of Par32 ( phosphorylated after rapamycin 32 ) ; the higher ratio of membrane-localized Par32 that we observed in [BIG+] cells is consistent with a more active TOR pathway ( Figure 6—figure supplement 1A; Varlakhanova et al . , 2018 ) . These latter observations also provide a potential explanation for the decreased longevity of [BIG+] cells: loss of TOR pathway function is associated with extended lifespan in yeast ( Dikicioglu et al . , 2018; Fabrizio et al . , 2001; Powers et al . , 2006 ) and many other organisms including nematodes , fruit flies , and mice ( Bjedov et al . , 2010; Harrison et al . , 2009; Robida-Stubbs et al . , 2012 ) . We next investigated whether these inhibitors impacted the size of [BIG+] cells . When grown in cycloheximide or rapamycin to saturation , [BIG+] cells were no longer large compared to naïve controls ( Figure 6C and D ) . Thus , unperturbed translation is necessary for the increased size of [BIG+] cells . These data could be explained by the inhibitors masking expression of the large cell trait or by reversion of the [BIG+] prion . To distinguish between these possibilities , we sub-cultured cells that had been treated with cycloheximide or rapamycin and allowed them to recover in rich medium for several dozen generations . We then examined their size distributions . Populations derived from [BIG+] ancestors were once again significantly larger than naïve controls that we subjected to the same propagation regime ( Figure 6C and D ) . We conclude that [BIG+] depends on protein synthesis to augment cell size but that the prion is stable to transient perturbations in translation . Translation is rate-limiting for growth in nutrient-rich conditions ( Kafri et al . , 2016 ) . We observed enhanced growth of [BIG+] cells in rich YPD medium ( Figure 1A and B ) . However , in synthetic defined medium with identical carbon source abundance ( 2% glucose ) but fewer amino acids , nucleosides , and other nutrients for optimum growth ( SD-CSM ) , [BIG+] cells did not grow faster than naïve controls ( Figure 6—figure supplement 1B ) . These data , combined with the resistance of [PRION+] cells to cycloheximide and rapamycin , suggested that protein synthesis might be enhanced in [BIG+] . Because a major component of translational regulation is the efficiency with which the ribosome translates each mRNA , we examined the impact of [BIG+] on mRNAs encoded with different codons using luciferase reporter assays . We transformed [BIG+] cells and isogenic naïve control cells with dual-luciferase plasmids encoding both Renilla and firefly luciferase genes . We tested two versions of the firefly luciferase gene: the first contained the standard suite of firefly mRNA codons; the second produced an identical protein product , but via codons that are rarer in S . cerevisiae , reducing steady-state protein levels by approximately fivefold ( Chu et al . , 2014 ) . Each plasmid contained an identical Renilla luciferase gene with its natural set of codons that served as an internal control ( normalizing for copy number and general transcriptional activity ) . The firefly reporter with normal codons did not produce more luciferase activity in [BIG+] cells than in naïve cells ( Figure 6E ) . However , the firefly reporter with rarer codons produced normalized luciferase levels approximately 50% higher in [BIG+] cells than in isogenic naïve control cells ( Figure 6E ) . These data suggest that [BIG+] cells may enhance the translation of some mRNAs , especially those containing a greater frequency of rare codons . We observed these effects in meiotic spores from naïve × [BIG+] crosses ( Figure 6E ) as well as in the original [BIG+] isolates ( Figure 6—figure supplement 1C ) , showing that the altered translation phenotype , like cell size , is inherited by all progeny of meiosis . We next investigated whether [BIG+] had global effects on the proteome by isolating total protein from cells harboring the prion and naïve controls . We reproducibly obtained more total protein per cell from the [BIG+] cultures ( Figure 6—figure supplement 1D ) . We next loaded an equal mass of protein lysate from naïve and [BIG+] cells onto a denaturing polyacrylamide gel and separated them by electrophoresis . Coomassie staining of these gels showed no pronounced differences in banding patterns ( Figure 6—figure supplement 1E ) , suggesting that [BIG+] does not exert substantial changes on the relative levels of the most highly expressed proteins , which are known to be efficiently translated ( Gingold and Pilpel , 2011; Plotkin and Kudla , 2011 ) . We next performed polysome gradient analysis to assess ribosome density on mRNAs , a measure of global translation activity . We observed no change in monosomes or disomes in [BIG+] samples compared to naïve controls . However , polysomes—which are responsible for most protein synthesis ( Noll , 2008; Warner and Knopf , 2002 ) —were increased in [BIG+] cells relative to naïve controls , with the polysome-to-monosome ratio increasing from ~2 . 9 to 3 . 2 ( Figure 6F ) . To further investigate translation in [BIG+] cells , we performed a pulse-labeling experiment to measure the rate of newly synthesized protein . We grew naïve and [BIG+] cells in medium containing [35S]-methionine and harvested protein lysates every 15 min for 1 hr , and then measured the extent of radiolabel incorporation by scintillation counting . These experiments established that the rate of newly synthesized total protein was higher in [BIG+] cells ( 10 , 179 cpm/min ) compared to naïve cells ( 8812 cpm/min ) ( Figure 6G; p<0 . 0011 , unpaired t-test ) . In summary , we found that [BIG+] cells have a higher translation rate , which can increase total protein output , including the levels of some proteins translated from mRNAs enriched with codons generally associated with lower translation efficiency . Conditions that enhance protein synthesis also tend to reduce the fraction of time that cells spend in the G1 stage of the cell cycle ( Jorgensen and Tyers , 2004 ) . This is because commitment to S phase entry—budding of a daughter yeast cell—depends on sufficient production of proteins needed to replicate the genome and essential cellular structures . Accelerating the production of these factors can thus shorten this period . We measured the fraction of naïve and [BIG+] cells in the G1 phase of the cell cycle by counting the fraction of unbudded cells ( i . e . , cells in G1 ) . For naïve cells , 36 . 2% were in G1 , whereas only 27 . 2% of [BIG+] cells were in G1 , a ~25% reduction ( Figure 6—figure supplement 1F , p=0 . 0017 , unpaired t-test ) . These data suggest that the cell cycle checkpoint for progression to S phase remains intact , and that a shortened G1 stage in [BIG+] cells is consistent with their increased protein synthesis . [BIG+] cells are not larger during exponential phase growth ( Figure 3—figure supplement 1A ) , suggesting that their cell size checkpoint remains intact . By contrast , cells in stationary phase , which do not have the nutrient content needed to progress to S phase , may continue to accumulate mass at a faster rate , contributing to their larger size . Because relative mRNA expression levels were only subtly altered in [BIG+] cells during exponential growth phase , but multiple measures of translation were altered , we considered the possibility that the phenotypes of the prion might be due to changes in protein levels of particular open reading frames . We tested this idea using two methods: ( 1 ) 2-D gel electrophoresis on pulse-labeled , newly synthesized proteins , and ( 2 ) a screen for differential expression of over 5000 GFP-tagged proteins . We pulse-labeled [35S]-methionine into newly synthesized proteins for 1 hr and then separated total protein using 2-D gel electrophoresis . Although labeling for less time than a cell division limited the number of newly synthesized proteins we could measure , we chose this time to minimize the contributions from factors other than protein synthesis , such as degradation . We observed [35S]-methionine counts of 9 . 7 cpm/µg and 11 . 5 cpm/µg protein from naïve and [BIG+] cells , respectively , indicating that cells harboring the prion synthesized proteins faster than naïve cells ( Figure 6G ) . We also analyzed the [35S]-methionine signal from 13 proteins that we could separate by 2-D gel electrophoresis . In [BIG+] cells , seven were increased , one was decreased , and five remained similar ( Figure 6H and I ) . These data suggest that [BIG+] cells have an altered translation program , with variations in even the most rapidly synthesized proteins . To identify specific proteins with altered levels in prion-containing cells , we capitalized on the dominance of [BIG+] in genetic crosses ( Figure 4B ) , mating haploid cells harboring the prion to a genome-wide collection of superfolder-GFP gene fusions ( ‘SWAT’ library; ~ 5500 ORFs; fused seamlessly to retain endogenous promoter/regulatory elements; Weill et al . , 2018 ) . Equivalent matings between naïve strains and this genome-wide collection served as controls . To control for the potentially larger size of the [BIG+] cells , we assessed protein levels in these diploid strains in terms of the relative GFP levels ( normalized by OD600 and Z-scored ) within the naïve and [BIG+] GFP-fusion collections separately . Mating and fluorescence measurements were performed in biological duplicate: the SWAT library was mated to two independent [BIG+] isolates alongside naïve controls . The reported OD600-normalized GFP levels are the mean of two technical duplicates of each biological duplicate . Protein levels measured in this way were generally well correlated between naïve and [BIG+] strains ( Pearson’s r = 0 . 76; p<10–16; Figure 6J ) , in concordance with our results from electrophoresis of total cellular protein . However , many proteins were up- or downregulated in [BIG+] cells . Of the 4233 fusions whose abundance could be robustly quantified across the four replicates ( Z-score < 1 for both naïve and [BIG+] ) , ~130 were differentially expressed in [BIG+] cells . Consistent with a bias toward enhanced translation , 81 were upregulated and 46 were downregulated ( Figure 6J and Supplementary file 3 ) . These proteins did not show any strong enrichment in physicochemical properties , nor were they enriched in Pus4-dependent pseudouridylation sites ( see Appendix 1 for further discussion ) . We also did not observe a widespread increase in the expression of ribosomal proteins . We did , however , observe that proteins that were increased in [BIG+] cells had a modest decrease in their codon adaptation index ( CAI ) at the 5' end of their mRNAs relative to all genes ( Figure 6K ) , indicating that cells harboring the prion might more efficiently translate these messages . The dip in CAI that is typical in the 5' end of all genes , known as ‘translational ramping , ’ is thought to reflect the bias toward translational control near the beginning of ORFs ( Frumkin et al . , 2018; Tuller et al . , 2010 ) . ( Lower CAI can both reduce the speed of elongation by requiring rarer tRNAs and lead to differences in mRNA structure that might also affect initiation or elongation efficiency . ) These data are consistent with our luciferase reporter data in which rarer codons throughout the message output more protein in [BIG+] than in naïve cells ( Figure 6E and Figure 6—figure supplement 1C ) , as well as the resistance of prion cells to the elongation inhibitor cycloheximide ( Figure 6A ) . Many hits were logically connected to the enhanced translation , increased size , and shortened lifespan of [BIG+] cells . The 81 upregulated proteins included multiple ORFs whose deletions are associated with decreased cell size ( such as PHO5 , MRPS28 , KAP122 , and SWE1; Harvey and Kellogg , 2003; Jorgensen et al . , 2002 ) , increased chronological lifespan ( DIG2 and UBX6; Garay et al . , 2014 ) , and extended RLS ( ENO1 , MUB1 , PHO87 , PGM2 , and YRO2; McCormick et al . , 2015 ) . Conversely , the 46 downregulated proteins included ORFs whose deletions are associated with increased cell size ( RNR4 and RPL15B; Jorgensen et al . , 2002; Perlstein et al . , 2005 ) and decreased chronological lifespan ( including BUD23 , SMI1 , MHP1 , and CLG1; Garay et al . , 2014; Marek and Korona , 2013 ) . Several proteins directly involved in translation control and ribosome biogenesis were also increased in [BIG+] cells , including FHL1 , MRPS28 , MRPS16 , RPL24B , RPS7A , and UTP10 ( Figure 6—figure supplement 1G ) . The [BIG+]-regulated proteins also included 29 ORFs associated with differential sensitivity to rapamycin ( including SAS4 , SNF1 , and HCA4; Butcher et al . , 2006; Dudley et al . , 2005; Kapitzky et al . , 2010 ) and 14 ORFs that are known genetic or physical interactors with TOR1 ( including GCD14 and PAR32; Krogan et al . , 2006; Varlakhanova et al . , 2018 ) . The functional breadth of these effects on the proteome and their logical connection to factors involved in the control of proliferation , cell size , and lifespan suggest that the phenotypes of [BIG+] are likely derived from an altered translational regulation program that favors increased cell size and proliferation at the expense of lifespan . Our finding that pseudouridylation on tRNA was maintained at similar levels in [BIG+] cells ( Figure 5B–D ) led us to wonder whether mRNA substrates of Pus4 were similarly modified . The best-documented mRNA target of Pus4 is the translation elongation factor TEF1/TEF2—whose position U239 is robustly pseudouridylated in a Pus4-dependent manner ( Carlile et al . , 2014; Lovejoy et al . , 2014; Schwartz et al . , 2014 ) . These paralogous genes encode identical copies of the eEF-1alpha translation elongation factor , which binds to aminoacylated tRNAs—specifically the T arm stem-loop that is pseudouridylated by Pus4 ( Dreher et al . , 1999 ) —and delivers them to the A-site of the ribosome during translation elongation ( Schirmaier and Philippsen , 1984 ) . We used three different methods to measure pseudouridylation of U239 in TEF1/TEF2 from naïve , [BIG+] , and pus4Δ cells . Sanger sequencing verified that the modified position was identical to that previously annotated in the literature ( Figure 7—figure supplement 1A ) . Next , we used a quantitative RT-PCR-based method , known as CMC-RT and ligation-assisted PCR analysis of Ψ modification ( CLAP ) ( Figure 7A; Zhang et al . , 2019 ) . We observed TEF1/TEF2 U239 pseudouridylation in both naïve and [BIG+] cells , in contrast to very low/background levels in pus4Δ cells ( Figure 7B ) . Finally , using the aforementioned qPCR-based method for detecting pseudouridylation ( Figure 5A ) , we found additional evidence that TEF1/TEF2 was pseudouridylated in our wild-type cells ( both naïve and [BIG+] ) in a Pus4-dependent manner ( Figure 7C ) . Most studies of yeast prions have characterized them as decreasing or eliminating activity ( Garcia and Jarosz , 2014 ) . However , we recently discovered one notable exception in which the [SMAUG+] prion can increase the activity of the protein that encodes it , Vts1 ( Chakravarty et al . , 2020 ) . We thus examined if there was an altered level of pseudouridylation of TEF1/TEF2 mRNA . If altered levels affected protein activity , this could be one possible mechanism linked to the altered translation program we found in [BIG+] cells . By quantifying differences in the melt curve shift after CMC labeling—an analysis made simpler for TEF1/TEF2 than for tRNAs by the absence of other pseudouridylated positions flanking U239—we observed an increase in the signal of pseudouridylation in [BIG+] cells relative to naïve ( Figure 7D and E and Figure 7—figure supplement 1B ) . Together these data demonstrate that the catalytic function of Pus4 is retained in [BIG+] cells and may be enhanced for specific substrates , contrasting with the classical view of prions as being loss-of-function protein conformations . They also provide a novel example of how RNA modification can be epigenetically controlled . Finally , we tested whether protein-based epigenetic control of cell size , protein synthesis , or localization is present in wild yeast populations . Protein chaperones are essential regulators of prion propagation in wild strains , just as they are in laboratory strains ( Halfmann et al . , 2012; Jarosz et al . , 2014 ) . To block the passage of prions in wild yeasts , we transiently expressed a dominant-negative variant ( SSA1K69M ) of Hsp70—the chaperone that is essential for [BIG+] propagation ( Figure 4F ) —in 20 wild S . cerevisiae strains isolated from various environments around the world ( Cubillos et al . , 2009; Itakura et al . , 2020 ) . We measured the size of cells before and after this chaperone curing and found that 20% ( four isolates out of twenty ) —'273614N’ ( clinical , Newcastle , UK ) ; ‘L-1528’ ( fermentation , Cauquenes , Chile ) ; ‘DBVPG1373’ ( soil , Netherlands ) ; ‘BC187’ ( barrel fermentation , Napa Valley , California ) —became smaller after curing ( Figure 8A ) . Given the wide array of genetic , epigenetic , and ecological diversity harbored in this collection , we also tested for three additional features associated with [BIG+] in laboratory strains: ( 1 ) increased protein synthesis , ( 2 ) differences in Pus4 localization between cured and uncured variants , and ( 3 ) a Pus4-dependent change in cell size . First , to test whether the four isolates that became smaller upon curing had altered protein synthesis , we transformed them with the same luciferase reporters we tested in laboratory [BIG+] cells . We normalized the firefly reporters with variable codons to the Renilla control for cured strains and then normalized these to the corresponding uncured strain values . Two of the four isolates showed significant changes to translation ( Figure 8B ) . BC187 yielded the largest change upon curing , an ~40% reduction in the translation of a firefly luciferase containing a ‘normal’ suite of codons , similar in magnitude to what we had observed with [BIG+] cells generated in the laboratory . Thus , since the loss of a chaperone-dependent epigenetic element reduced translational output from the reporter , the original , wild isolate ( uncured ) has higher protein synthesis , as we saw for [BIG+] in the BY4741 laboratory strains . Second , we examined the localization of Pus4 in BC187 and its Hsp70 cured derivative . We transformed uncured and cured strains with a plasmid expressing GFP-tagged Pus4 protein and imaged cells using epifluorescence microscopy . In uncured cells , we observed Pus4 in a fragmented network throughout the cytoplasm ( Figure 8C ) , similar to what we observed in laboratory [BIG+] cells ( Figure 4G ) . In contrast , in cured BC187 cells , the signal was more diffuse . Finally , we investigated the relationship between Pus4 and cell size in BC187 . The original [BIG+] trait identified in laboratory yeast strains ( this work and Chakrabortee et al . , 2016a ) was induced after transient overexpression of Pus4 protein . Therefore , we tested whether transient overexpression of Pus4 protein could re-establish the large-cell phenotype in BC187 derivatives that had been subject to curing by transient Hsp70 inhibition . Indeed , cured BC187 cells treated in this way became large ( Figure 8D ) . By contrast , transient overexpression of Pus4 did not increase the size of the already-large BC187 natural isolate ( Figure 8—figure supplement 1 ) . Thus , re-initiation of this epigenetic trait could be accomplished with transient Pus4 overexpression alone , thereby bearing strong similarities to [BIG+] in common laboratory strains and establishing Pus4 as a driver of epigenetic cell size control in this model eukaryote . Together , these observations suggest that epigenetic and potentially Pus4 prion-mediated control of mechanisms like [BIG+] that we have described in the laboratory is also likely present in nature . Epigenetic inheritance is most commonly thought to be driven by enzymes that modify chromatin and DNA . Here we show that an enzyme that catalyzes the epigenetic modification of RNA can itself be controlled by an extrachromosomal epigenetic process: a self-templating protein conformation that persists over long biological timescales . We found that prion cells not only maintain pseudouridylation activity but that activity can be increased , even without any detectable increase in PUS4 expression . Future studies should address whether Pus4-dependent modification of TEF1/TEF2 mRNA plays a role in the phenotypes of [BIG+] as this mRNA , which is the major non-tRNA substrate of Pus4 , encodes a central elongation factor that binds to and escorts tRNAs to the ribosome . The modest changes we observed in the relative abundance of a few dozen mRNAs in actively growing cells do not appear to drive these growth or aging phenotypes based on their known functions . Instead , we found that changes to the translational control of numerous genes are logically connected to these phenotypes . These include genes that affect cell size , chronological lifespan and replicative lifespan , translation control and ribosome biogenesis , and differential sensitivity to rapamycin ( Supplementary file 3 ) . Translation is a rate-limiting step for growth in many organisms ( Polymenis and Schmidt , 1997; Sonenberg , 1993 ) and is often activated in human cancers ( Sonenberg and Hinnebusch , 2009 ) . Other prions in yeast also affect translation , including [PSI+] and [MOD+] ( Baudin-Baillieu et al . , 2014; Suzuki et al . , 2012 ) . However , in contrast to [BIG+] , they lead to losses of their underlying protein activities , impairing translation and , in turn , growth in many conditions ( Baudin-Baillieu et al . , 2014; Cox , 1965; Suzuki et al . , 2012 ) . In contrast , in [BIG+] cells , translation appears to be amplified—it is faster , leading to higher expression levels of many proteins . We are unaware of an example in the literature of a mutation that enhances translation under nutrient-replete conditions . More investigation is needed to determine the precise mechanism by which translation is altered . Increased polysome density can be associated with increased translation ( more protein produced per mRNA ) or decreased translation ( due to defects in elongation or termination leading to accumulation of ribosomes ) . However , multiple lines of evidence lead us to favor the former interpretation . First , [BIG+] cells are resistant to the translation elongation inhibitor cycloheximide . Second , their rates of bulk translation as measured by [35S]-methionine incorporation are also higher than naïve cells . General defects in translation elongation or termination , by contrast , would be expected to reduce this activity . Our luciferase reporter data shows an improved translation of mRNAs containing rare codons , which slow translation elongation . Finally , when measuring the proteome-wide effects on translation using our GFP screen , not only did we observe logical connections between the gene functions of altered proteins and the observed [BIG+] growth phenotypes , but it was the proteins whose expression increased that contained a feature consistent with enhanced elongation of their mRNAs: their ORF 5′ ends had a lower CAI . Collectively , these data suggest an increase in translation elongation efficiency of many mRNAs as a parsimonious model . The various possible mechanisms need not be mutually exclusive—some mRNAs in [BIG+] cells may indeed be translated faster to produce more protein without any defects and perhaps exhibit elevated rates of elongation and termination . In contrast , other mRNAs in the same cells may be subject to deficiencies at any of these steps . Examining protein synthesis at higher resolution , using methods like ribosome profiling , and in response to perturbations to ribosome-associated quality control pathways ( Brandman and Hegde , 2016 ) , may reveal the landscape of changes to protein synthesis in [BIG+] cells in greater detail . Translation is also coupled to cell size , proliferation , and lifespan ( Ecker and Schaechter , 1963; Kaeberlein and Kennedy , 2007; Lloyd , 2013; Steffen and Dillin , 2016; Tanenbaum et al . , 2015 ) . Cell size , which is determined in large part by the activity of the TOR pathway ( Fingar et al . , 2002; Zhang et al . , 2000 ) , has been inversely correlated with lifespan ( Anzi et al . , 2018; He et al . , 2014; Yang et al . , 2011 ) , and older cells are larger ( Egilmez et al . , 1990 ) . Moreover , molecules that extend lifespan , such as rapamycin , also influence cell size and proliferation by restricting the cell’s translation capacity ( Beretta et al . , 1996; Terada et al . , 1994 ) . Here we describe an epigenetic paradigm that links all of these fundamental cellular properties: translation , cell size , proliferation , and lifespan . In future studies , we plan to explore to what extent the effect on lifespan may serve as material for selection to favor or disfavor [BIG+] . At present , we favor a hypothesis in which the aging defect of [BIG+] cells is due at least in part to pleiotropic consequences of their increased proliferation and translation capacity . Although these features have already been linked to aging in genetic studies , we note that such theories of antagonistic pleiotropy in aging are not without controversy ( Hughes and Reynolds , 2005 ) . The replicative lifespan of wild budding yeast strains varies widely ( Kaya et al . , 2015 ) and has been associated with changes in oxidative phosphorylation , respiration , and differences in metabolite biosynthesis . Genetic screening has also offered some insight into the genetic basis of this variability ( McCormick et al . , 2015 ) . Lastly , genetic mapping efforts have identified polymorphisms underlying natural variation in chronological lifespan variation ( Kwan et al . , 2011 ) . The genetic architecture of natural lifespan , both chronological and replicative , remains obscure , however . Our data further demonstrate that it can be subject to strong epigenetic control . We note that prior studies have characterized genetic links between cell size and lifespan in yeast—mutants that make cells larger tend to age faster , and older cells tend to be larger than younger cells ( Neurohr et al . , 2019; Yang et al . , 2011; Zadrag-Tecza et al . , 2009 ) . [BIG+] provides an epigenetic mechanism to alter these relationships heritably . Exerting epigenetic , rather than genetic , control over basic cell growth behaviors could be valuable in the face of fluctuations between nutrient-replete and nutrient-poor conditions . Theory predicts that such mechanisms can have substantial adaptive value when the frequency of environmental fluctuations is rare relative to the frequency of phenotypic change ( King and Masel , 2007 ) . In agreement with these inferences , our modeling quantitatively described the long-run selective advantage of this ‘live fast , die young’ prion state that we measured , illustrating the importance of considering not only steady-state phenotypes but also the ecological context in which prion states are expressed . Of particular relevance to this point is our data demonstrating that transient perturbation of Hsp70 activity can eliminate [BIG+] . Conditions in which the prion confers a growth benefit match those that promote its propagation . Conversely , conditions in which the prion is detrimental , shortening lifespan , are also known to reduce chaperone expression ( Janssens et al . , 2015; Werner-Washburne et al . , 1989; Werner-Washburne and Craig , 1989 ) and could thereby cure the prion . The strong dependence of [BIG+] on chaperones therefore means that it is a natural epigenetic sensor of its environment . Our data from wild yeast isolates demonstrates that cell size and protein synthesis can also be under epigenetic control in natural isolates and is influenced by Pus4 . Exploring whether a [BIG+]-like epigenetic mechanism that promotes proliferation in nutrient-replete conditions is conserved in metazoans is a major goal for the future . Indeed , cancer cells also experience frequent oscillations in their environments: as tumors grow and metastasize , cells are exposed to shifting gradients of oxygen and glucose that influence their proliferation rate ( Martinez-Outschoorn et al . , 2017; Schito and Semenza , 2016 ) . Regulation of cell physiology in these situations is best understood at the level of transcriptional changes due to the relative ease of profiling them . However , multiple lines of evidence suggest that translational hyperactivation can also fuel pathological proliferation ( Robichaud et al . , 2019 ) . Our discovery of a protein-based mechanism that changes cell growth via engagement of an altered translational program argues for greater investigation into the epigenetic control of post-transcriptional processes in both normal biology and disease . To model the fitness of an epigenetic element for which growth in nutrient-replete conditions is improved and survival in starvation conditions is worsened , we define the following growth equations in nutrient-replete conditions:x0=x00eμ0tx1=x10eμ1t where x0 represents the population of naïve cells and x1 represents the population of [BIG+] cells . μ0 and μ1 are the growth rates of naïve and [BIG+] cells , respectively . We neglect the lag and stationary phases of growth , as the ratio between populations does not change during this time . Likewise , in starvation , x0=x00e-δ0tx1=x10e-δ1t where δ0 and δ1 are the death rates of naïve and [BIG+] cells , respectively . We can furthermore define the times of nutrient repletion and starvation as τ1 and τ2 , respectively . Thus , for each cycle of nutrient/starvation , we can define the following recursion relations:x0 , i+1=x0 , ieμ0τ1x1 , i+1=x1 , ieμ1τ1x0 , i+2=x0 , i+1e-δ0τ2x1 , i+2=x1 , i+1e-δ1τ2 And we can write an analytical expression for the ratio of populations after one repletion/starvation cycle:x0t+τ1+τ2=x0teμ0τ1-δ0τ2x1t+τ1+τ2=x1teμ1τ1-δ1τ2x1x0t+τ1+τ2=x1tx0teμ1τ1-δ1τ2-μ0τ1-δ0τ2 For the purposes of the model , we consider only the ratio between populations in the two epigenetic states ( neglecting the total carrying capacity of the environment ) . The above defines a recursion relation from which we can predict the ratio of naïve and [BIG+] cells after N cycles of nutrient repletion and starvation . The free parameters defining the growth advantage and starvation disadvantage attributable to the epigenetic element in the model above ( μ0 and μ1; δ0 and δ1 ) were determined by Monte Carlo sampling of independent measurements of the growth and death rates ( Figure 2D and E ) . To generate the ensemble of competitive fitness predictions shown in Figure 2C , we randomly sampled growth and death rates for naïve and [BIG+] cells according to their experimentally determined distributions . The sampling was conducted 1000 times , and the prediction shown in Figure 2F is the median and 95% confidence interval of the resulting ensemble of model predictions . All MATLAB code and data required to generate the model predictions are available at http://github . com/cjakobson/liveFastDieYoung ( copy archieved at swh:1:rev:d86455c2e53f50644862690e22bfdcd080abdf63 , Jakobson , 2021 ) . Code to generate the figures is available upon reasonable request to jarosz@stanford . edu . Bacteria strains ( for plasmid propagation ) were cultured on LB agar or liquid ( Research Products International ( RPI ) , Mount Prospect , IL ) . Yeast strains were cultured on either YPD agar or liquid ( RPI ) or SD-Ura ( Sunrise Scientific , Knoxville , TN ) unless otherwise indicated . Strains were stored as glycerol stocks ( 25% glycerol [Amersco , Solon , OH] in appropriate media ) at –80°C and revived on YPD or amino acid dropout media before testing . Yeast were grown in YPD at 30°C on a TC-7 roller drum wheel ( New Brunswick ) unless indicated otherwise . Yeast transformations were performed with a standard lithium–acetate protocol ( Gietz et al . , 1992 ) . The pus4∆ strain was sourced from the BY4741 MATa haploid knockout library ( GE Dharmacon , Lafayette , CO ) . Most diploids were constructed by crossing indicated BY4741 haploids to the BY4742 parental strain ( ATCC , Manassas , VA ) by mixing a bead of cells of each strain ( from a single colony ) together on a YPD plate and growing overnight at 30°C . A small globule of this cell mixture was then restreaked to single colonies on SD-Lys-Met agar plates to select for diploids . Diploids constructed for the experiments presented in Figure 4G and Figure 6—figure supplement 1A were made by crossing [BIG+] or naïve haploids to the PCR-verified seamless GFP-Pus4 or seamless GFP-Par32 haploid strains , respectively . Due to overlapping auxotrophic makers in each parent strain , diploids could not be selected on dropout plates from cell mixtures . Instead , pools of the mated cells were grown in several successive competitions to allow diploids to outcompete haploid parents . Diploids were then isolated from single colonies , and verified using either microscopy to measure cell size ( GFP-Pus4 ) , which is significantly larger than either haploid parent , or PCR amplification of the GFP tag/ORF junction ( GFP-Par32 ) , which yielded two distinct bands matching the expected sizes of the wild-type allele and the tagged allele , in contrast to the haploid parental strains that yielded only one or the other band sizes . These isolated diploids were then imaged as described below . Sporulations were performed inoculating single diploid colonies in Pre-SPO liquid media ( YPD with 6% glucose ) for 2 days at room temperature on a roller drum wheel . Cells were then pelleted and washed twice in SPO media ( 1% potassium acetate [Sigma , St . Louis , MO] , 320 mg CSM-Met powder [Sunrise Scientific] , 20 mg methionine [Sigma] per liter ) , and then diluted 10-fold into 3 mL cultures of SPO . These cultures were incubated on a rotary wheel for 1 week at room temperature , before dissecting tetrads on a Singer Instruments MSM400 ( Somerset , England ) . The 3XFLAG-Pus4 strains were constructed using PCR combined with the ‘Delitto Perfetto’ method ( Storici and Resnick , 2006 ) . Canavanine mutants were constructed by pelleting 500 µL of a saturated YPD liquid culture , resuspending it in 100 µL YPD and plating this on SD-Arg agar plates containing 60 µg/mL canavanine ( Sigma ) , and growing for 2 days at 30°C . Single canavanine-resistant colonies were picked and re-tested for resistance before further testing . Cytoductions were performed as described in Chakrabortee et al . , 2016a . A BY4742 strain with a defective KAR1 allele ( kar1-15 ) was created as an initial recipient for cytoplasmic transfer . This allele prevents nuclear fusion during mating while permitting cytoplasmic transfer . The strain carries auxotrophic markers distinct from those in the putative [BIG+] or naïve donor strains , and was also converted to petite with growth on ethidium bromide ( strains were grown in YPD with 25 µg/mL ethidium bromide for approximately two-dozen generations before testing for growth on YP-glycerol ) . This allowed cytoplasmic transfer to be scored through the restoration of mitochondrial respiration , while selecting for auxotrophic markers unique to the recipient strain . The recipient and donor strains were mixed together on YPD-agar and grown overnight , followed by selection of heterokaryons and resulting haploid cytoductants on dropout media ( selecting for the BY4742 recipient strain markers ) containing glycerol as a carbon source . One more round of selection was used while replica-plating onto a dual-selection agar plate ( SD-Lys-Met ) to confirm that the colonies were not diploids . One additional round of propagation on a non-selective plate was performed before doing ‘‘reverse cytoductions , ’ which were performed in the same way except selecting for BY4741 auxotrophy in recipient naïve strain . In the reverse cytoductions , the donors were naïve or putative-[BIG+] BY4742 kar1-15 cytoductants from the first round , and the recipients were wild-type or pus4∆ naïve BY4741 petite cells . Transient PUS4 deletion experiment strains were made by the Delitto Perfetto method ( Storici and Resnick , 2006 ) . After deletion of PUS4 , strains were propagated for ~75 generations before phenotyping . The PUS4 gene was re-introduced by homologous recombination . Transient overexpression of Pus4 in haploid BC187 yeast was performed using PDJ541 ( CEN URA3 GAL::PUS4 ) ( Supplementary file 4 ) . Three independent transformants of BC187 ( previously cured by transient Hsp70 inhibition ) bearing this plasmid were grown in SGalactose-Ura for 48 hr at 30°C . Overexpression was stopped by outgrowth in non-selective YPD medium for 48 hr . Strains lacking the overexpression plasmid were isolated by plating to single colonies on media containing 5-FOA . Uracil auxotrophy was confirmed before subsequent measurements of cell size . Biological replicates of each yeast strain were pre-grown in rich media ( YPD ) . We then diluted these saturated cultures 1:20 in sterile water and then inoculated 3 µL into 96-well humidified plates ( Nunc Edge Plates [Thermo Scientific , Waltham , MA] ) with 150 µL of YPD or SD-CSM per well . Cycloheximide ( Sigma ) was added to growth media at either 0 . 01 µg/mL or 0 . 05 µg/mL , as indicated . Rapamycin ( LC Laboratories , Woburn , MA ) was added to growth media at 10 µM . Cell growth was monitored with continuous measurements of OD600 ( approximately every 10 min ) at 30°C over 96 hr using BioTek Eon or Synergy H1 microplate readers ( Winooski , VT ) . Timepoints plotted in bar graphs correspond to the maximum proliferation rates calculated from growth data . For each strain , four single colonies were picked from freshly streaked YPD plate and grown in 5 mL of pre-sporulation media for 3 days on a roller drum wheel ( New Brunswick Scientific , Edison , NJ ) at 30°C . Cultures were then pelleted and washed once with SPO media , and resuspended in 5 mL of SPO media , and placed back on the roller drum wheel at 30°C . On days indicated , a dilution was made of each replicate to achieve dozens to hundreds of colonies on a YPD plate , which were then counted using a colony counter ( Synbiosos Acolyte , Frederick , MD ) . Dilution was ~100 , 000× at early stages of the experiment and was later empirically determined due to significant cell death . We note that aging the cells in SPO did not lead to significant acidification ( pH of old cultures was found to be >5 ) , as has been reported for cells aged in YPD , which contains high levels of glucose that upon metabolism leads to the secretion of organic acids ( Murakami et al . , 2011 ) . RLS was assessed using the standard method of isolating virgin cells on agar YPD ( 2% glucose ) plates , and then separating their daughter cells at each cell division by micromanipulation and counting the total number of daughters produced by each mother cell ( Steffen et al . , 2009; Wasko et al . , 2013 ) . Strains were streaked from glycerol stocks onto YPD plates and allowed to grow at 30°C until individual colonies could be selected for each strain . Colonies were lightly patched onto fresh YPD overnight and 20 cells were isolated by microdissection from each patch . These cells were incubated at 30°C for ~2 hr until they had formed daughter cells , at which time individual virgin daughter cells were selected and arrayed as previously described ( Steffen et al . , 2009 ) for lifespan analysis . From these , daughter cells were removed by microdissection and counted approximately every 2 hr during the day . Plates were maintained at 30°C during the day and placed at 4°C overnight . At least four independent replicates ( arising from different colonies ) of 20 individual mother cells each were measured for each strain . Three regimes of chaperone inhibition were tested: ( 1 ) transient exposure to a dominant-negative version of Hsp70 ( Ssa1 ) to inhibit its activity ( Chakrabortee et al . , 2016a; Jarosz et al . , 2014 ) ; ( 2 ) transient exposure to radicicol ( RPI ) to inhibit Hsp90 activity ( Chakrabortee et al . , 2016a ) ; and ( 3 ) transient exposure to guanidinium hydrochloride ( Sigma ) to inhibit Hsp104 activity ( Ferreira et al . , 2001 ) . Regime 1 was performed by transforming cells with a plasmid , PDJ169 , harboring a dominant-negative version of Hsp70 ( Ssa1 ) as described previously ( Chakrabortee et al . , 2016a; Jarosz et al . , 2014; Lagaudriere-Gesbert et al . , 2002 ) . Transformants were picked and restreaked by hand or replica-pinned using a Singer HDA robot a total of 12 times on SD-Ura to promote Ssa1DN expression . ( Anecdotally , we note that prion phenotypes are frequently cured with fewer than 12 restreaks , but for reasons of technical throughput , 12 were used in this experiment . ) Then plasmids were eliminated by plating on media containing 5-fluoroortic acid ( SD-Ura +0 . 1% 5-FOA + 50 µg/mL uracil ) and plasmid loss was verified by replating on SD-Ura . Colonies were then tested for the elimination of prion phenotypes . Tested strains were compared to control strains that were restreaked in parallel on SD-CSM plates . Regime 2 was performed by replica-pinning cells six subsequent times on YPD agar plates containing 5 µM radicicol , using a Singer HDA robot . After replatings , cells were plated back onto YPD two subsequent times to facilitate recovery before being tested for the elimination of prion phenotypes . Tested strains were compared to control strains that were replica-pinned in parallel on YPD plates . Regime 3 was performed like regime 2 but with SD-CSM plates containing 0 . 5 g/L guanidinium hydrochloride . Tested strains were compared to control strains that were replica-pinned in parallel on SD-CSM plates . Wild yeast strains were cured by transforming a uracil-selectable 2micron plasmid ( PDJ1222 ) encoding the aforementioned dominant-negative version of Hsp70 ( Ssa1 ) , under control of a constitutive promoter ( pGPD ) . Transformants were passaged on selective media five times to allow the growth of single colonies . Transformants were then passaged three times on non-selective media ( YPD ) to permit plasmid loss , which was confirmed by the lack of growth on selective media ( SD-Ura ) . Single colonies were used to inoculate 5 mL YPD cultures that were grown for 3 days on a roller drum wheel at 30°C . Cells were diluted 1000-fold and then mixed in equal volumes to form 50:50 mixtures of either of the following: naïve CanS and [BIG+] CanR; or naïve CanR and [BIG+] CanS . Before mixing cells , saturated cultures were measured to have near equal cell densities , and ‘time zero’ measurement was made by plating the initial cell mixture and counting the number of canavanine-resistant colony-forming units ( CFUs ) relative to total CFUs . These initial strain mixtures were then grown for 2 days on a roller drum wheel at 30°C , after which cells were diluted 50 , 000-fold or 25 , 000-fold and plated on YPD or canavanine plates , respectively . These plates were grown at 30°C for 2 days before counting colonies . The liquid cultures were diluted 1:1000 in 5 mL of fresh YPD , and this process was repeated nine more times . Swapping of canavanine resistance between naïve and [BIG+] was done to correct for the potential of canavanine resistance to influence cell growth . However , in our experiments the differences were negligible . The numbers of canavanine-resistant and total colonies were compared relative to a number of cells plated to determine the number of naïve or [BIG+] colonies arising at each timepoint . Most microscopy was performed using a Leica inverted epifluorescence microscope ( DMI6000B ) with a Hamamatsu Orca 4 . 0 camera . Cells were imaged after 3 days of growth in 5 mL YPD at 30°C . Saturated cultures were diluted 10-fold with 1X PBS and briefly sonicated to break up cell clumps . Differential interference contrast ( DIC ) images were taken at 20 ms exposure time using a 63×/1 . 40 oil objective . Cell area was calculated using CellProfiler ( 3 . 1 . 5 ) image analysis software ( http://www . cellprofiler . org; Carpenter et al . , 2006 ) . The large cell threshold was set at one standard deviation above the cell area mean of the naïve cells . GFP microscopy data presented in Figure 4G were obtained from cells were grown in YPD for 24 hr . Aliquots of 1 mL were spun down and resuspended in 200 µL 1X PBS , 2 µL of which were prepared into an agar pad . Agar pad made with 2% agar and PBS . Fluorescent images were taken at 400 ms exposure time and a Z-stack consisting of 25 steps for a total length of 4 . 83 µm in each channel using a 100×/1 . 40 oil objective . Data presented in Figure 8C were imaged similarly to Figure 4G , except that cells were grown in SC-Ura media for retention of the GFP-Pus4 expression plasmid . This plasmid ( Supplementary file 4 ) overexpresses Pus4 protein via a strong promoter driving the expression of GFP-Pus4 constitutively on a centromeric plasmid . Data presented in Figure 6—figure supplement 1A were grown and imaged similarly to above , except cells were imaged in 1X PBS using a GE DeltaVision Ultra microscope ( Boston , MA ) . Cell size experiments using protein synthesis inhibitors ( Figure 6A–D ) : conditions were the same as above , with the following differences . Single colonies were inoculated into YPD , YPD + cycloheximide ( 0 . 05 µg/mL ) , or YPD + rapamycin ( 10 µM ) and grown for 4 days before imaging . ( Very similar results were observed after 3 days of growth . ) The following day , cultures were diluted into liquid YPD and grown for 3 days , after which they were restreaked once onto YPD agar . Single colonies were then used to inoculate liquid YPD cultures , grown for 3 days before imaging to test for the reappearance of the large cell size phenotype . ImageJ version 2 . 0 . 0-rc-69/1 . 52 p , Build 269a0ad53f . For GFP-Pus4 in laboratory S . cerevisiae and wild S . cerevisiae isolate BC187 ( Figures 4G and 8C ) : ( 1 ) using full-sized images , created a maximum projection of 25-step Z-stack; ( 2 ) in ImageJ , set 50 × 50 pixel square in the area between cells and measure average intensity; ( 3 ) subtracted this intensity from the total image using the ‘Math’ function; ( 4 ) added 5 µm scale bar calculated from the original voxel size of 0 . 065 µm; and ( 5 ) enlarged Photos 6X using Topaz Gigapixel AI software version 5 . 3 . 2 ( Topaz Labs ) . For GFP-Par32 microscopy ( Figure 6—figure supplement 1A ) , analysis was performed using ImageJ ( imagej . nih . gov ) . All images were adjusted to a uniform contrast and illumination corrected using standard tools in ImageJ . The background was calculated using a morphological opening with a disc of radius 75 pixels , such that the structuring element was larger than cells in the foreground . In these background-corrected images , we quantified the ratio of membrane to cytoplasmic Par32-GFP within each cell by measuring the average intensity for a region at the plasma membrane and the average intensity for an equivalently sized region approximately 0 . 2 μm from the plasma membrane . For each cell , the ‘Fraction membrane Par32 signal’ was calculated by dividing the average Par32-GFP intensity in the membrane region by the average Par32 intensity in the cytoplasmic region . Strains were transformed with PDJ512 and PDJ513 . To maintain the plasmids , we grew these cells in a synthetic complete medium containing nutrient levels between those in SD and YPD formulations . Four independent transformants for each sample were grown for 1 day in 150 µL SC-Ura ( Sunrise Scientific ) per well in 96-well plates at 30°C . Saturated cultures were then diluted 15× into fresh media in a new 96-well plate and grown until cultures reach OD600 0 . 6 , as determined by a BioTek Eon plate reader . 20 µL of each culture was added using a multichannel pipette into a white flat-bottom 96-well microplate ( E&K Scientific , Santa Clara , CA ) already containing 20 µL of room temperature 1X Passive Lysis Buffer from the Dual-Luciferase Reporter Assay System ( Promega , Madison , WI ) . Cultures were then lysed by shaking at 300 rpm for 25 min at room temperature . Renilla and firefly luciferase activity was measured using 75 µL injection volumes and otherwise default settings on a Veritas luminometer ( Turner Biosystems ) . pTH726-CEN-RLuc/minCFLuc ( PDJ512 ) and pTH727-CEN-RLuc/staCFLuc ( PDJ513 ) were gifts from Tobias von der Haar ( University of Kent; Addgene plasmids #38210 and #38211; Chu et al . , 2014 ) . Final luciferase values were normalized to OD600 measurements of cultures to account for cell density . We note , however , that [BIG+] cells did not have a general growth advantage over naïve cells in SC-Ura , and when comparing optical density measurements to those counting cells using a hemacytometer , we observed no perturbation in the relationship between cell number and optical density for [BIG+] cells . For wild strains , we considered the possibility that curing could reverse multiple epigenetic elements affecting plasmid copy number , transcription , or other elements of gene expression apart from protein synthesis . Indeed , after normalizing Renilla or firefly luciferase values to cell density , some strains have several-fold differences after curing , although they were closely correlated irrespective of which firefly codon variant was compared . Therefore , as for data presented in Figure 6E , for Figure 8B we also normalized firefly luciferase values to Renilla luciferase , which is expressed from the same plasmid . This normalization procedure thus tests for differences in protein synthesis that are codon-frequency dependent , that is , a measure of translational efficiency . Single colonies from two biological replicates per sample were used to inoculate 5 mL YPD cultures that were grown on a roller drum wheel at 30°C overnight . Saturated cultures were added to 95 mL of YPD in 500 mL flasks and shaken at 225 rpm at 30°C until cultures reached OD600 1 . 0 . Five minutes prior to harvesting cells , we added cycloheximide ( Sigma ) to a final concentration of 100 µg/mL to arrest translation , by adding 1 mL of a 10 mg/mL stock solution ( in ethanol ) per culture , then immediately swirling flask and putting back on a shaker for 5 min 225 rpm 30°C to permit the chemical to enter cells and arrest protein synthesis . Cultures were pelleted in 50 mL conical tubes for 3 min at 5000 rpm . After decanting the supernatant , pellets were quickly resuspended in ice-cold Polysome Lysis Buffer ( PLB; Jan et al . , 2014 ) ( 20 mM Tris pH 8 . 0 , 140 mM KCl , 1 . 5 mM MgCl2 , 100 µg/mL cycloheximide , 1% Triton X-100 , RNase-free reagents ) , 250 µL total PLB per sample . Resuspended pellets were then flash-frozen in liquid nitrogen . Pellets were weighed to ensure their weights were near equal and then thawed on ice . 250 µL of additional ice-cold PLB was added per sample , making slightly over 0 . 5 mL per sample . Samples were then flash-frozen in tiny pellets ( ‘yeast dippin’ dots’ ) by pipetting directly into a small dewar filled with liquid nitrogen and a wire mesh basket nested inside . Tiny pellets were then stored at –80°C until lysis . Samples were lysed using a Retsch Cryomill ( Haan , Germany ) with 25 mL canisters and the following program: pre-cool , then 12 cycles of 15 Hz × 3 min . Smears of lysate were stained with Trypan blue and imaged under a microscope to verify efficient lysis . ( We suspect with larger sample volume:canister volume ratios , fewer cycles would be necessary . ) Lysates were loaded onto 10–50% sucrose gradients pre-poured on a BioComp Gradient Master 108 ( Fredericton , ND , Canada ) . Lysates generally contained RNA concentrations around 12–18 µg/µL . 30 µL of lysate was carefully pipetted onto the top of the sucrose gradient , and samples were spun in a Beckman SW41 Ti Rotor for 2 . 5 hr at 4°C at 40 , 000 rpm . Gradients were analyzed on a Brandel fractionator ( Gaithersburg , MD ) . Technical replicates ( same lysate independently loaded onto separate gradients ) showed a very high degree of similarity , as did biological replicates . Adapted from Esposito and Kinzy , 2014 . Single colonies were inoculated into 5 mL complete media lacking methionine and grown overnight at 30°C . The saturated starter culture was then inoculated into 75 mL of complete media lacking methionine to an OD600 of 0 . 1 in a 250 mL flask . Flasks were grown at 30°C shaking at 225 rpm until an OD600 of 0 . 5–0 . 7 was reached at which point 37 . 5 µL of 100 mM ( cold ) methionine ( Sigma ) and 7 . 32 µL of 10 . 25 mCi/mL [35S]-methionine ( PerkinElmer ) was added for a final concentration of 50 µM methionine and 1 µCi/mL [35S]-methionine . At 15 min intervals from the time of methionine addition , OD600 was determined by cuvette and 1 mL aliquots were obtained for scintillation counting . To the 1 mL aliquot , 200 µL 50% TCA ( Sigma ) was added and incubated on ice for 10 min . The samples were then incubated at 70°C for 20 min and briefly spun down . In a 1225 Sampling Manifold ( MilliporeSigma XX2702550 ) , 25 mm glass microfiber filters ( GE Healthcare Life Sciences ) were arranged and pre-wetted with 5% TCA . Samples were then applied to the glass fiber filters under vacuum . Filters were then washed twice with 5 mL 5% TCA and once with 95% ethanol . Filters were then removed from the manifold , air-dried on Whatman paper , placed in scintillation vials with 5 mL ScintiSafe Econo 1 scintillation cocktail solution ( Fisher Scientific SX20-5 ) , and underwent six rounds of technical replicated counting on a Beckman LS 600SCS . Scintillation counts were then normalized to OD600 . Following 1 hr of pulse labeling with [35S]-methionine , 10 mL of cells were collected and centrifuged at 16k × g at 4°C for 10 min . The supernatant was removed and 300 µL of TCA buffer was added ( 10 mM Tris-HCl pH 8 . 0; 10% TCA; 25 mM NH4OAc; 1 mM Na2EDTA ) along with glass beads . Samples were then vortexed for 1 min bursts with 5 min rests for five times at 4°C . The slurry was transferred away from the glass beads to a fresh tube . 200 µL of fresh TCA buffer was added to beads to pick up residue . Samples were then split , one half destined for resuspension in SDS for protein quantification and the other in CHAPS for the 2-D gel . Samples were centrifuged at 16k × g at 4°C for 10 min . The supernatant was then removed from the resulting pellets and washed twice with 1 mL acetone . The samples were then resuspended in SDS Resuspension Solution ( 0 . 1 M Tris-HCl pH 11 . 0; 3% SDS ) or CHAPS Resuspension Buffer ( 8 M urea; 4% CHAPS; 2% IPG Buffer 3–10; 40 mM DTT ) . CHAPS samples were then adjusted to 1 µg/µL . First-dimension resolution was conducted with 250 µg with 13 cm 3–10 NL Immobiline DryStrip gels ( Cytiva ) . Rehydration protocol started with 12 hr rehydration followed by 500 V for 0 . 5 kVh , 1000 V for 1 . 0 kVh , 16 , 000 V for 14 . 5 kVh . First dimension strips were then trimmed to fit the second-dimension gel and equilibrated in Equilibration Buffer ( 6 M urea; 30% glycerol; 2% SDS , 50 mM Tris-HCl pH 8 . 8; 0 . 002% bromophenol blue; 65 mM DTT ) for 30 min . Strips were rinsed with MOPS running buffer and loaded onto 11 cm 4–12% Bis-Tris Criterion XT gels ( Bio-Rad ) . Strips were sealed in place with an agarose plug ( 0 . 5% agarose; 0 . 0002% bromophenol blue ) and ran at 200 V for 1 hr . Second-dimension gels were then stained in Coomassie and dried in a cellulose sandwich consisting of a plastic wrap on one side of the gel to facilitate the removal of one of the cellulose layers for autoradiography . X-ray film was then exposed for 9 days . Autoradiography spots were quantified in ImageJ by tracing out the spot and quantifying the integrated density of the signal . Densities of prion cells were then normalized to the naïve sample , as reported in Figure 6I . Naïve or [BIG+] cells were mated to the SWAT seamless-GFP library ( Weill et al . , 2018; Yofe et al . , 2016 ) on solid YPD agar plates in 384-spot format for 24 hr at room temperature . Diploids were selected on media lacking both lysine and methionine ( SD-Lys-Met ) and propagated for 48 hr at room temperature . Diploids were inoculated into 60 µL of liquid media lacking both lysine and methionine ( SD-Lys-Met ) in 384-well plates . All library manipulations were carried out using a Singer ROTOR HDA robotic pinning instrument . Cells were propagated in liquid medium for 24 hr at 30°C ( OD600 ~ 1 ) , at which time OD600 and green fluorescence were measured using a BioTek Synergy H1 plate reader . OD600 was adjusted based on known blank wells , and the GFP/OD600 measurements were normalized by Z-score ( [xi – µ]/σ ) within the naïve and [BIG+] populations independently . For SDS-PAGE , immunoblots , and protein yield measurements , cells were lysed using a Retsch Cryomill using the following program: six 3 min cycles at 15 Hz with 2 min cooling cycles in between . Cell lysates were loaded onto GenScript ExpressPlus SDS-PAGE 4–20% gels ( Piscataway , NJ ) and stained using Coomassie blue . For western blots , anti-FLAG M2 monoclonal antibody ( Sigma ) was used to detect 3XFLAG-Pus4 , and anti-PGK1 monoclonal antibody ( Invitrogen ) was used to detect the loading control . Four independent colonies of naïve and [BIG+] cells were grown in 5 mL liquid YPD cultures on a culture roller drum wheel overnight at 30°C . Saturated cultures were diluted 1:1000 in new 5 mL YPD cultures and grown until OD600 1 . 0 ( exponentially growing ) . Cultures were pelleted in a table-top centrifuge with the swing-bucket rotor at 4300 × g at 4°C for 5 min , washing once with ice-cold 1X PBS . RNA isolation , poly-A selection , the addition of ERCC spike-in controls , and sequencing were performed by the Beijing Genomics Institute . Analysis was performed using Kallisto ( Bray et al . , 2016 ) and Sleuth ( Pimentel et al . , 2017 ) . Transcripts Benjamini–Hochberg-corrected q-value <0 . 05 and \|log2 fold-change\| > 1 were considered differentially expressed . Raw data and other experimental information are available on the Gene Expression Omnibus ( Barrett et al . , 2013 ) , accession: https://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE176577 . Strains were grown up in 5 mL of YPD overnight at 30°C . Saturated cultures were then diluted 15× into fresh media and grown to an OD600 0 . 8 . Cells were then pelleted , and total RNA was extracted by phenol-chloroform extraction . RNA concentration for each sample was then analyzed by Nanodrop and diluted to equal concentrations in RNase-free water . The extracted RNA was then run on a nucleic acid fragment analyzer , and total tRNA abundance was quantified relative to the abundance of 5 . 8 S rRNA . Adapted from Zhang et al . , 2019 . Plots/graphs were made using PRISM 7/8/9 software ( GraphPad , San Diego , CA ) .	Cells make different proteins to perform different tasks . Each protein is a long chain of building blocks called amino acids that must fold into a particular shape before it can be useful . Some proteins can fold in more than one way , a normal form and a ‘prion’ form . Prions are unusual in that they can force normally folded proteins with the same amino acid sequence as them to refold into new prions . This means that a single prion can make many more that are inherited when cells divide . Some prions can cause disease , but others may be beneficial . Pus4 is a yeast protein that is typically involved in modifying ribonucleic acids , molecules that help translate genetic information into new proteins . Sometimes Pus4 can adopt a beneficial prion conformation called [BIG+] . When yeast cells have access to plenty of nutrients , [BIG+] helps them grow faster and larger , but this comes at the cost of a shorter lifespan . Garcia , Campbell et al . combined computational modeling and experiments in baker’s yeast ( Saccharomyces cerevisiae ) to investigate the role of [BIG+] . They found that the prion accelerated protein production , leading to both faster growth and a shorter lifespan in these cells , even without any changes in gene sequence . Garcia , Campbell et al . ’s findings explain the beneficial activity of prion proteins in baker’s yeast cells . The results also describe how cells balance a tradeoff between growth and lifespan without any changes in the genome . This helps to highlight that genetics do not always explain the behaviors of cells , and further methods are needed to better understand cell biology .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "genetics", "and", "genomics" ]	2021	A prion accelerates proliferation at the expense of lifespan
Divergence in gene regulation can play a major role in evolution . Here , we used a phylogenetic framework to measure mRNA profiles in 15 yeast species from the phylum Ascomycota and reconstruct the evolution of their modular regulatory programs along a time course of growth on glucose over 300 million years . We found that modules have diverged proportionally to phylogenetic distance , with prominent changes in gene regulation accompanying changes in lifestyle and ploidy , especially in carbon metabolism . Paralogs have significantly contributed to regulatory divergence , typically within a very short window from their duplication . Paralogs from a whole genome duplication ( WGD ) event have a uniquely substantial contribution that extends over a longer span . Similar patterns occur when considering the evolution of the heat shock regulatory program measured in eight of the species , suggesting that these are general evolutionary principles . Divergence in the regulation of gene expression has been repeatedly postulated to play a major role in evolution . Examples of regulatory differences between species were described in a wide range of species including bacteria ( McAdams et al . , 2004 ) , fungi ( Gasch et al . , 2004; Habib et al . , 2012 ) , flies ( Prud’homme et al . , 2007; Wittkopp et al . , 2008; Bradley et al . , 2010 ) , and mammals ( Khaitovich et al . , 2006; Odom et al . , 2007; Brawand et al . , 2011; Lindblad-Toh et al . , 2011; Perry et al . , 2012 ) . However , the mechanisms through which regulatory systems evolve are still only partially understood , and in most cases the adaptive importance of regulatory changes is unknown ( Lynch , 2007; Thompson and Regev , 2009; Wohlbach et al . , 2009; Baker et al . , 2012; Romero et al . , 2012 ) . In recent years , comparative genomics approaches have allowed us to begin to trace the evolution of gene regulation at different time scales ( Tuch et al . , 2008b; Weirauch and Hughes , 2010; Brawand et al . , 2011; Lindblad-Toh et al . , 2011; Romero et al . , 2012 ) , through two major approaches: ( 1 ) characterization of cis-regulatory elements in orthologous promoter sequences ( Gasch et al . , 2004; Tanay et al . , 2005; Bradley et al . , 2010; Lindblad-Toh et al . , 2011; Habib et al . , 2012 ) , and ( 2 ) comparative analysis of mRNA profiles and protein–DNA interactions measured across organisms ( Tirosh et al . , 2006 , 2011; Borneman et al . , 2007; Tuch et al . , 2008a; Schmidt et al . , 2010; Wapinski et al . , 2010; Brawand et al . , 2011; Romero et al . , 2012 ) . While studies relying on cis-regulatory sequences are more prevalent , functional studies of comparative gene regulation are beginning to shed light on how regulatory evolution is linked to functional changes . In particular , it has been suggested ( Lynch and Force , 2000; Gu et al . , 2004 , 2005; Teichmann and Babu , 2004; ; Conant and Wolfe , 2006; Tirosh and Barkai , 2007; Wapinski et al . , 2007b ) that gene duplication can promote regulatory divergence by either neo-functionalization or sub-functionalization of regulatory mechanisms of the two paralogs . Among eukaryotes , the Ascomycota fungi ( Figure 1A ) provide an excellent model to study the evolution of gene regulation ( Tsong et al . , 2003 , 2006; Ihmels et al . , 2005; Tanay et al . , 2005; Field et al . , 2008; Hogues et al . , 2008; Tirosh and Barkai , 2008; Tsankov et al . , 2010 , 2011; Baker et al . , 2012; Habib et al . , 2012 ) . They include the model organisms Saccharomyces cerevisiae , Schizosaccharomyces pombe and Candida albicans , as well as many non-model , genetically-tractable species with sequenced genomes . Species in the phylogeny diverged before and after a whole genome duplication event ( Wolfe and Shields , 1997; Kellis et al . , 2004 ) ( WGD , Figure 1A , star , ∼150 mya ) , allowing us to study the consequences of this evolutionary mechanism ( Wolfe and Shields , 1997; Kellis et al . , 2004; Wapinski et al . , 2007b ) . 10 . 7554/eLife . 00603 . 003Figure 1 . Ascomycota species in this study . ( A ) A phylogenetic tree of the 15 Ascomycota species in the study . Dark blue: respiro-fermentative; red: respiratory; green: obligate respiratory; light blue: intermediate between respiro-fermentative and respiratory . Star: a Whole Genome Duplication event ( WGD ) . ( B ) Growth rate ( log ( OD ) 600 , y axis ) of each species over time ( y axis ) during growth in the novel rich medium used in this study ( see ‘Materials and methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 00310 . 7554/eLife . 00603 . 004Figure 1—source data 1 . Evolutionary distance across the phylogeny of 15 species . Shown are the estimated branch lengths using PAML for our panel of 15 species . Each ’Sample’ represents the estimated branch length using a random subset of 1000 uniform orthogroups and the ‘Mean’ and ‘Stdev’ show the mean and standard deviations of these estimations . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 004 Comparative genomics of Ascomycota has already shed an important light on the evolution of gene expression . For example , studies in yeast showed that while co-expression of genes in modules can be conserved at substantial distances , the associated regulatory mechanisms often diverge , acquiring new regulators and losing ancestral ones , both for sequence-specific transcription factors ( Tsong et al . , 2003 , 2006; Tanay et al . , 2005; Hogues et al . , 2008; Lavoie et al . , 2010; Baker et al . , 2011 , 2012 ) and for chromatin organization ( Tirosh and Barkai , 2008; Tsankov et al . , 2010 , 2011 ) . In some cases , changes in gene expression and related mechanisms are clearly coupled to other adaptive changes in lifestyle ( Ihmels et al . , 2005; Field et al . , 2008; Tsankov et al . , 2010 ) , whereas in others they may be the result of neutral ‘regulatory drift’ ( Tsong et al . , 2003 , 2006; Lavoie et al . , 2010; Baker et al . , 2012 ) . Importantly , evolutionary changes in regulators , facilitated by protein modularity , in cooperative binding with other factors and shifts in protein–DNA interactions contribute toward different paths through a ‘hybrid’ regulatory state for the ancestral regulatory network to be resolved in to generate the diversity of regulatory network structures observed in modern species ( Baker et al . , 2011 , 2012; Tuch et al . , 2008a ) Despite these early successes , collecting experimental data across species has remained challenging , and hence most experimental studies rely on two to four species ( Tanay et al . , 2005; Tirosh et al . , 2006; Lelandais et al . , 2008; Wittkopp et al . , 2008 ) , with few exceptions ( Schmidt et al . , 2010; Tsankov et al . , 2010; Wapinski et al . , 2010; Brawand et al . , 2011; He et al . , 2011 ) . An important challenge is to collect experimental data in such a way that would minimize irrelevant differences , for example , due to growth conditions , and allow focusing on true evolutionary distinctions . Expanding the experimental scope to cover a broader phylogenetic range and density can help study the divergence of expression in individual genes and gene modules in order to answer questions on the extent of conservation of transcriptional programs , its relation to phylogenetic distance , the emergence of new regulatory patterns across modules of co-regulated genes , and the specific contribution of gene duplication and divergence—through both sporadic and whole genome duplication—to regulatory evolution . Here , we use comparative transcriptional studies across 15 Ascomycota species—spanning >300 million years of evolution ( Sipiczki , 2000 ) —to understand the evolution of modular gene regulation during batch growth on glucose and its depletion , a key physiological response . We optimized culture conditions across species , and collected ∼300 expression profiles at six physiologically comparable time points along each species’ growth: repletion ( lag phase ) , exponential growth ( ‘mid-log’ and ‘late log’ ) , the point of glucose depletion ( ‘diauxic shift’ ) and two later time points when the growth rate levels off ( ‘post shift’ and ‘plateau’ ) . To analyze the evolution of regulatory modules , we use a new algorithm , Arboretum ( Roy et al . , 2013 ) , to identify expression modules across species and to reconstruct their evolutionary history . We find that the degree of divergence of the transcriptional profiles correlates with phylogenetic distance , with the largest divergence in lag phase profiles . In all species , the transcriptional response involves five major transcriptional modules . While the module’s expression patterns are conserved across species , their gene membership diverges , proportionally to phylogenetic distance . Gene duplication events significantly contribute to regulatory divergence , in particular close to their phylogenetic point of duplication . This contribution is more pronounced and more prolonged for WGD paralogs . These patterns also characterize the evolution of the transcriptional response to heat shock , supporting their generality ( Roy et al . , 2013 ) . Our framework for comparative functional genomics is applicable to any complex phylogeny , and can help test these principles of regulatory evolution in other responses and species . We studied 15 yeast species whose genome is fully sequenced ( Figure 1A; ‘Materials and methods’; Table 1 and Figure 1—source data 1 ) , spanning >300 million years of evolution ( Sipiczki , 2000 ) . The species cover the different clades of the phylogeny well , with the exception of the filamentous Euascomycota , and have a range of phenotypes related to how they use glucose as a carbon source during aerobic growth . Species in the Kluyveromyces , Candida , and Yarrowia clades are respiratory and use oxidative phosphorylation ( Figure 1A , red and green ) . Conversely , a respiro-fermentative lifestyle—a preference to ferment glucose even in the presence of oxygen ( Piskur et al . , 2006 ) —has evolved independently at least twice in this phylogeny , once after the WGD ( Conant and Wolfe , 2007 ) and once in Schizosaccharomyces ( Rhind et al . , 2011 ) ( Figure 1A , dark blue ) . K . polysporus , the most basal post-WGD species ( Scannell et al . , 2007 ) has an intermediate phenotype between respiro-fermentative and respiratory ( Figure 1A , light blue ) . In contrast to the other respiratory species that can ferment , Y . lipolytica ( Figure 1A , green ) is an obligate respiratory species , but can uniquely use normal hydrocarbons and various fats as carbon sources ( Kurtzman , 2000 ) . 10 . 7554/eLife . 00603 . 005Table 1 . Number of genes and orthogroupsDOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 005SpeciesTotal genes in speciesTotal genes on arraysTotal orthogroups on arraysTotal genes with tree* , †Orthogroups available for analysis‡Genes available for analysis§Orthogroups analysis 1#Genes analysis 1#Orthogroups analysis 2¶Genes analysis 2¶S . cerevisiae6343625744245508440254642746274636763964S . paradoxus5512550443195256431252442577257734523720S . mikatae5697569342515094425150932513251333823618S . bayanus5489548342725191426951882555255534163679C . glabrata5338526941264909411948972534253433943614S . castellii5693568942775420425753622574257434613794K . polysporus53285324403945394027452525062506NANAK . waltii5198519443814849438148482560256034323497K . lactis5328532344354888442848792572257234553537S . kluyveri5321532043934879438648652496249633643444D . hansenii7938689340504635403446081903190325512634C . albicans6163610748585692485856922324232431103232Y . lipolytica6756667242604886425848742138213828552921S . japonicus5297514938634248386142461878187824872557S . pombe5068506042084751420847502001200124872746Shown are the total number of genes in each species ( defined as the sum of genes on arrays and with orthology , ‘Materials and methods’ ) . The number of genes , genes that have orthologs in another species , and the classes of genes that were measured on the species-specific arrays ( 1 ) total number of genes ( 2 ) total number of orthogroups ( 3 ) non-singleton ( those present S . cerevisae and in at least one other species ) . Also shown is the number of genes and orthgroups resulting after filtering based on a missing value cut of 50% ( see ‘Materials and methods’ ) . The number of genes and orthogroups per species used in the Arboretum analyses 1 and 2 ( without and with duplication ) . *Gene trees = orthogroups = orthology . †This class includes non-singletons that are represented on the microarray . ‡Orthogroups represented on the microarray and satisfy missing values cutoff ( 50% ) . §Genes represented on the microarray and satisfy missing values cutoff ( 50% ) . #Analysis 1: Figures 5–9 ( present in at least one species in addition to S . cerevisiae , and did not incur duplication ) . ¶Analysis 2: Figures 10–13 ( present in at least one species in addition to S . cerevisiae , and incurred at most one duplication ) . Total number of orthogroups: 7459 . Due to these lifestyle differences some of the species do not grow well in typical media formulations ( e . g . , YPD ) . We therefore first optimized our growth medium to minimize growth differences between species ( ‘Media tests’ under ‘Materials and methods’ ) . Our formulation boosts the growth of otherwise slow growers , without substantially impacting the growth of fast growers ( Figures 1B and 2A ) . 10 . 7554/eLife . 00603 . 006Figure 2 . Growth of species in published and novel growth media . ( A ) Performance of species in our optimized medium vs YPD medium , a common medium for S . cerevisiae . Shown are normalized saturation coefficients ( log2 ( OD600 ) during a 24-hr growth period , a measure of accumulated biomass ) of each species ( ‘Media tests’ under ‘Materials and methods’ ) in our panel ( rows ) in three media ( columns ) . ( B ) Choosing ‘physiologically comparable’ time points . Our experiments compare ‘physiologically analogous’ time points across all species ( see ‘Materials and methods’ ) . For example , shown is the growth curve ( x axis: time , minutes; y axis: growth rate , in log2 ( OD600 ) and glucose levels ( g/L , blue ) and ethanol levels ( g/L , orange ) for the relative slow growing species S . pombe ( left ) vs the growth curve for the faster growing C . glabrata ( right ) . Biological samples from each species were taken at the time points indicated by arrows . The Log phase time point ( shown in red ) used as the reference for microarray analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 00610 . 7554/eLife . 00603 . 007Figure 2—figure supplement 1 . Phenotypic characterization of each species . Shown are the growth curves ( log2 ( OD600 ) , purple ) , glucose levels ( g/L , blue ) and ethanol levels ( g/L , orange ) of two biological replicates for each species . Species name is noted on top of each panel . Y . lipolytica did not consume glucose despite a normal sigmoidal growth curve ( left ) , presumably due to a preference to consume lipids as a carbon source . When the duration of the experiment was extended ( right ) , this species consumed the glucose in the medium . Biological samples from each species were taken at the time points indicated by arrows at Lag , Log , Late log ( LL ) , diauxic shift ( DS ) , post-shift ( PS ) and plateau ( P ) . The Log phase time point ( shown in red ) used as the reference for microarray analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 007 Even in our new medium , there is still substantial variation in growth between species , likely indicating real physiological differences , inherent to each species ( Figure 2A; ‘Materials and methods’ ) . We therefore determined in real-time the growth rate , glucose , and ethanol levels for each species ( ‘Materials and methods’; Figure 2B , Figure 2—figure supplement 1 ) , and chose physiologically comparable ( but potentially physically different ) time points for each species for isolating RNA from lag , mid-log , late log , ‘diauxic shift’ ( the point at which glucose is depleted ) , post-shift , and plateau . We compared mRNA levels at each time point to those in a mid-log stage from the same time course in the same species ( Figure 3A; ‘Materials and methods’ ) , thus allowing us to compare differential ( relative ) expression levels in the response across species . We conducted all experiments in ≥2 biological replicates , which were highly reproducible ( ‘Materials and methods’; Figure 3—source data 1 ) . Furthermore , the values measured by arrays were highly consistent with those measured for a selected subset of samples by RNA-Seq , including for duplicated ( paralogous ) genes ( ‘Materials and methods’ ) . Given this high reproducibility , we present median values across replicates in subsequent analyses , for simplicity . 10 . 7554/eLife . 00603 . 008Figure 3 . Divergence in global expression profiles correlates with phylogenetic distance . ( A ) A comparative transcriptional compendium during growth on glucose . Shown are transcriptional profiles measured for each species ( tree , top ) , at six time points ( columns ) during growth on glucose: Lag , Late Log , Diauxic Shift , Post Shift and Plateau ( left to right ) . Genes ( rows ) are matched based on orthology and clustered ( ‘Materials and methods’ ) . Red: induced; blue: repressed; white: no change; grey: ortholog absent in species . ( B ) – ( F ) Correlation in expression decreases with phylogenetic distance . Shown are scatter plots relating—for each pair of species—their estimated phylogenetic distance ( y axis ) and the correlation between their matching global expression profile ( x axis ) at a matching physiological time point ( noted on top ) . The legend shows the clade to which the pair belongs ( if the same ) or ‘other’ ( if from different clades ) . Branch length was scaled by the maximum branch length to range from 0 to 1 . ( B ) Lag , ( C ) Late Log ( LL ) , ( D ) Diauxic Shift ( DS ) , ( E ) Post Shift ( PS ) , ( F ) Plateau ( PLAT ) . The line in each plot is the least squares fit . ( G ) Shown is the average Pearson’s correlation between pairs of species of the global expression profiles for each physiological time point . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 00810 . 7554/eLife . 00603 . 009Figure 3—source data 1 . Experimental design and correlation among biological replicates . Shown are the number of biological replicates per species per time point sampled , and the number of pairwise calculations used to calculate the correlation among biological replicates and the standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 00910 . 7554/eLife . 00603 . 010Figure 3—figure supplement 1 . Conservation of growth-rate regulated gene expression . ( A ) Expression of growth genes across species . Shown are the expression profiles across all species ( major columns ) and time points ( lag to plateau ) for gene orthologs ( rows ) whose expression was previously positively ( 257 ) and negatively ( 368 ) correlated with growth rate ( at 1 . 5 standard deviation ) in S . cerevisiae by Brauer et al . ( 2008 ) . Heatmap is laid out as in Figure 3 . ( B ) – ( F ) Correlations in expression profiles are maintained when growth genes are excluded . Shown are scatter plots relating—for each pair of species—their estimated phylogenetic distance ( Y axis ) and the correlation between their matching global expression profile with the growth-rate regulated genes removed ( X axis ) at a matching physiological time point ( noted on top ) . The legend shows the clade to which the pair belongs ( if the same ) or ‘other’ ( if from different clades ) . Branch length was scaled by the maximum branch length to range from 0 to 1 . ( B ) Lag; p≤1 . 14 × 10−4 , ( C ) Late Log ( LL ) ; p≤2 . 45 × 10−4 , ( D ) Diauxic Shift ( DS ) ; p≤1 . 5 × 10−20 , ( E ) Post Shift ( PS ) ; p≤8 . 22 × 10−26 , ( F ) Plateau ( PLAT ) ; p≤2 . 69 × 10−30 . The line in each plot is the least squares fit . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 010 To compare the extent of similarity in gene regulation , we calculated Pearson’s correlation coefficient between pairs of expression profiles ( using median values of biological replicates ) for each pair of species at physiologically-matched time points ( e . g . , post-shift in S . cerevisiae and C . albicans , ‘Materials and methods’ ) . We found that the degree of correlation is inversely related with the phylogenetic distance between the species ( r<−0 . 39 to −0 . 77 , p<3 × 10−5 , Figure 3B–F ) , such that the closer two species are in the phylogeny , the higher the correlation between their profiles . This inverse relation is apparent at each of the time points ( Figure 3B–F ) , and is consistent with what was recently reported in mammals ( Brawand et al . , 2011 ) . The extent of transcriptional divergence varies between stages . Profiles at diauxic shift , post shift and plateau are the most correlated between species ( Figure 3D–G ) , whereas the lag profiles , when cells ‘reset’ in response to nutrient repletion , are the most divergent ( Figure 3B , G ) , likely reflecting species-specific responses to nutrient signals shaped by niche adaptation . S . japonicus and S . castellii were the most divergent , such that in some cases , their Lag phase profiles were even anti-correlated to those of other species , suggesting distinct repletion programs . In principle , the lower conservation of the lag phase profiles in these and other species could have been merely due to the fact that this is the only phase sampled at the same absolute time ( 30 min ) . However , the repletion process is known to be fast ( Zaman et al . , 2008 ) , metabolite profiling in each species at this time point shows the least difference between species ( MS , DAT and AR , unpublished results ) , and additional profiling in some of the most divergent species at finer increments did not change this finding ( data not shown ) , ruling out this possibility . Nevertheless , some functional groups of genes do exhibit conserved induction or repression in the lag phase across most species ( e . g . , growth genes such as mRNA splicing genes are enriched among the highly induced genes [p<10−5 , FDR , ‘Materials and methods’] possibly to support splicing of transcripts encoding ribosomal proteins ) . To automatically trace the evolutionary trajectory of gene regulation , we used a novel probabilistic algorithm , Arboretum ( Roy et al . , 2013 ) ( ‘Materials and methods’ ) , to infer modules of co-expressed genes in each extant species and to reconstruct the ancestral modules from which they were derived . Arboretum uses a generative probabilistic model of the evolution of module membership of a given gene , starting from the last common ancestor ( LCA ) of a set of species and probabilistically propagating this membership down the branches of the phylogenetic tree to the leaf nodes . Arboretum uses the phylogeny to link the module IDs at each phylogenetic point to the modules of the LCA , and uses the structures of gene trees ( as defined from genome sequences; Wapinski et al . , 2007a ) during module identification . This allows Arboretum to simultaneously infer ancestral and extant module assignments , and to systematically handle complex orthology and paralogy relationships that arise from gene duplication and loss . Arboretum allows genes to change their module assignment during evolution , but at any ancestral or extant species , every gene ( if present in that species ) must be assigned to exactly one module for a particular response . We first applied Arboretum only to the transcriptional profiles of those genes that had no duplication events , but could have been lost or be lineage-specific ( Table 1; ‘Materials and methods’ ) , and found that in each species the data is best explained by five expression modules ( Figure 4 , Modules 1–5 , Figure 4—source data 1 ) , capturing 65–70% of the variation ( ‘Materials and methods’ ) : Module 1 ( strong repression , ‘growth’ genes , e . g . , for transcription and translation ) , Module 2 ( mild repression , cell division genes ) , Module 3 ( little or no change , cell morphogenesis genes ) , Module 4 ( mild induction , proteasomal and mitochondrial genes ) , and Module 5 ( strong induction , stress , respiration and carbohydrate metabolism genes , Supplementary file 1 ) . 10 . 7554/eLife . 00603 . 011Figure 4 . Arboretum reconstruction of expression module evolution ( Analysis 1 ) . ( A ) Five expression modules identified by Arboretum in the transcriptional response to glucose depletion . Each row corresponds to a species ( tree , left ) and each major column to a module ( 1–5 , labels top ) . Module labels are color coded by the regulation of the module’s genes following depletion , as noted on top , from bright blue ( Module 1 ) for strong repression to bright red ( Module 5 ) for strong induction . Each module’s height is proportional to the number of genes in the module . The five columns in each module are the expression levels at lag ( L ) , late log ( LL ) , diauxic shift ( DS ) , post-shift ( PS ) , and plateau ( P ) relative to mid-log phase . Red: induced; blue: repressed; white: no change . ( B ) – ( F ) Module assignments in all extant and ancestral species ( see Figure 5B for ancestral node assignment ) . Each matrix corresponds to the genes in one of the five modules in the LCA ( A14 ) ( B: Module 1; C: Module 2; D: Module 3; E: Module 4; F: Module 5 ) , and shows the module assignment of these genes in each of the extant and ancestral species from S . cerevisiae ( leftmost column ) to the LCA ( rightmost column ) . The biological functions listed at the top of each module are representative labels chosen based on Gene Ontology terms enriched in all species in that module ( Supplementary file 1 ) . The range of FDR p values and fraction of genes in each module are as follows: Module 1: Ribosome biogenesis , p<5 . 28 × 10−48 to 1 . 25 − 10−119 , fraction 37 . 3–61 . 6% . Module 2: cell division , p-value<3 . 51 × 10−02 to 4 . 52 × 10−02 , fraction 9–33 . 6% . Module 3: cell morphogenesis , p<4 . 64 × 10−02 to 4 . 95 × 10−02 , fraction 6 . 5–81% . Module 4: mitochondrial , p<3 . 20 × 10−02 to 4 . 90 × 10−02 , fraction 2 . 4–37 . 9%; proteasome , p<3 . 85 × 10−04 to 3 . 97 × 10−02 , fraction 1 . 6–15% . Module 5: respiration p<4 . 77 × 10−02 to 4 . 8 × 10−02 , fraction 32 . 6–58 . 9%; response to stress , p<4 . 75 × 10−02 to 4 . 86 × 10−02 , fraction 2 . 6–13 . 7% . Module assignment in each species is marked by a color code , as in the top of panel A ( bright blue: Module 1; light blue: Module 2; white: Module 3; pink: Module 4; red: Module 5 ) . Species are ordered by post-fix ordering ( left-child , right-child and parent ) of the species tree , as marked on the legend ( bottom ) . Black bars indicate points of phylogenetically coherent divergence in expression of orthologous genes , as discussed in the text . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 01110 . 7554/eLife . 00603 . 012Figure 4—source data 1 . Orthogroups included in the Aboretum run for analysis 1 . Shown are the genes used for Arboretum run excluding orthogroups with duplications ( Analysis 1 ) . Each column corresponds to an extant species . The string ‘Dummy’ separates the genes from different modules . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 012 Modules conserve their gene members relative to those of their immediately ancestral internal vertex in the tree ( ‘immediate ancestor’ ) , inversely to phylogenetic distance . To quantify this , we defined an Ancestral Module Conservation Index ( AMCI; Figure 5A; ‘Materials and methods’ ) based on the probability with which a gene in a species conserved its module assignment from its immediate ancestral module . Extant and ancestral species have relatively high AMCIs ( mean 0 . 82 ± 0 . 137 standard deviation [SD] ) , indicating conserved module assignment ( Figure 5B ) , and highlighting the consistency of our experimental design . AMCI values are inversely related to the length of the branch connecting a species and its immediate ancestor ( Figure 5C , Pearson’s correlation r = −0 . 68 , p<1 . 14 × 10−4 ) . 10 . 7554/eLife . 00603 . 013Figure 5 . Conservation of modular organization . ( A ) Module transition matrices . Shown are examples of transition matrices estimated by Arboretum for two species ( S . cerevisiae , top and S . pombe , bottom ) . Each matrix specifies , for each module in each child species ( columns ) , the probability with which a gene conserved its module assignment in that species’ immediate ancestor ( rows ) , or was reassigned to another module . Columns: modules of the child species , rows: modules of the ancestor species . Probabilities are color coded from black ( 1 ) to white ( 0 ) . Strong diagonal elements indicate high conservation with the immediate ancestor . The AMCI is calculated as the mean of the diagonal entries . ( B ) The Ancestral Module Conservation Index ( AMCI ) . Shown is the AMCI , ranging from 0: least conserved ( white circles ) to 1: most conserved ( black circles ) , for each extant and ancestral species . Tree is drawn to scale and species are color coded by carbon lifestyle as in Figure 1A . ( C ) AMCI decreases with increased phylogenetic distance . Shown is a scatter plot of the relationship , for each extant ( grey ) and ancestral ( black ) species , between its phylogenetic distance to its immediate ancestor ( branch length , y axis ) and its AMCI ( x axis ) . Branch length is scaled by the maximum value to range between 0 and 1 . The correlation between branch length and AMCI is −0 . 68 ( p≤1 . 13 −× 10−4 ) . The regression line is plotted . ( D ) and ( E ) Expansion and contraction of modules . Shown are the mean Module Contraction Index ( MCI , D ) and mean Module Expansion Index ( MEI , E ) for each Arboretum module ( x axis ) , based on the proportion of genes that respectively leave or join each module at each phylogenetic point . Blue and red indicate the modules from Arboretum runs with only no duplicates ( no paralogs ) and including duplicates ( with paralogs ) , respectively . Error bars were estimated from five Arboretum runs with different initializations . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 013 To assess the tendency of individual modules to diverge across the entire phylogeny , we traced the module membership of every gene through the phylogenetic tree from the LCA ( A14 ) to the extant species ( Figure 4B–F ) . Genes assigned in the LCA to Modules 1 ( Figure 4B , blue entries ) and 5 ( Figure 4F , red entries ) tend to persist in those modules ( Module Stability 94% ± 0 . 5% and 84% ± 1 . 2% , respectively , ‘Materials and methods’ ) . Conversely , Modules 2 , 3 , and 4 ( Figure 4C–E ) are less stable , with more frequent reassignment of their member genes to other modules . Module 1 and 5 are the least dynamic as reflected by the average proportion of genes that join ( 11% ± 1 . 6% and 17% ± 1 . 7% , respectively ) or leave each module ( 6% ± 0 . 4% and 16% ± 1% , respectively ) between each pair of ancestor and child species in the phylogeny ( Module Expansion Index and Module Contraction Index , respectively , ‘Materials and methods’; Figure 5D , E ) . Many coordinated phylogenetic transitions of functionally-related genes are consistent with differences in lifestyle between species . For example , cell cycle and DNA replication genes are reassigned in a coordinated fashion from Module 3 in the LCA to Modules 4 and 5 in Schizosaccharomyces ( Figure 6A ) , consistent with the unique nutrient control of mating in this clade ( Nadin-Davis and Nasim , 1990 ) . In another case , both ploidy- and meiosis-related genes ( Figure 6B ) are reassigned from the strongly induced Module 5 in the typically haploid pre-WGD species to the unchanged Module 3 or repressed Module 2 in the typically diploid post-WGD species . This is likely due to shifts in nutrient control of mating and meiosis post-WGD ( Barsoum et al . , 2011 ) . Arboretum infers that the LCA ( A14 ) followed the haploid , Module 5 , pattern ( Figure 6B , middle panel , A14 column ) . 10 . 7554/eLife . 00603 . 014Figure 6 . Conservation and rewiring of coherent functions across modules . Shown are expression ( left ) , Arboretum module assignments ( middle ) and a cartoon of the phylogenetic transition ( right ) for gene sets with coherent phylogenetic patterns . Each expression matrix is formatted as in Figure 3A , and each module assignment matrix as in Figure 4B–F . ( A ) Cell cycle genes , ( B ) mating and meiosis related genes , ( C ) mitochondrial genes , ( D ) oxidative phosphorylation genes , ( E ) amino acid and purine metabolism genes . Each module shows all the genes with a given phylogenetic pattern , and their labels ( e . g . , mitochondrial ) were manually generated based on enrichment of GO terms . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 01410 . 7554/eLife . 00603 . 015Figure 6—figure supplement 1 . Enrichment of Sfp1 binding sites . ( A ) in the promoters of genes with specific functions . Shown are the negative logarithm of the p value ( red intensity ) for a test of enrichment ( see ‘Materials and methods’ ) of the Sfp1 motif in the promoters of genes for mitochondrial , purine and amino acid metabolism and oxidative phosphorylation functions ( rows ) , across the 15 species ( columns ) . ( B ) Shown is the enrichment of the Sfp1 binding sites ( FDR < 0 . 05 ) in Arboretum Module 1 ( ‘growth module’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 015 Particularly substantial transitions in module assignment are associated with changes in carbon lifestyle during the transitions to respiro-fermentation post-WGD and in Schizosaccharomyces . These include genes with diverse mitochondrial functions , including mitochondrial ribosomal protein genes ( mRPs ) that are reassigned from ancestral Modules 1 or 2 to Modules 3 or 4 , independently at the WGD and in the Schizosaccharomces ancestor ( A13 ) ( Figure 6C ) . This is consistent with previous hypotheses that respiro-fermentation is the derived state ( Piskur et al . , 2006 ) . Similar changes have occurred in the expression of oxidative phosphorylation genes ( Figure 6D ) , albeit with a more gradual shift from repression in Y . lipolytica to late induction in Kluyveromyces to strong induction post-WGD ( Figure 6D , left panel ) . This is associated with concomitant changes in nucleosome organization and anti nucleosomal polydA:dT sequences in the promoters of oxidative phosphorylation and mRP genes ( Field et al . , 2008; Tsankov et al . , 2010 ) , which occurred convergently post-WGD and in Schizosaccharomyces ( Figure 7A , B ) . 10 . 7554/eLife . 00603 . 016Figure 7 . Changes in chromatin organization in mitochondrial , oxidative phosphorylation and amino acid metabolism genes . Shift in NFR occupancy in re-wired respiratory genes ( A and B ) . Shown are the logarithm of the p value of the KS-test ( y axis ) used to test if the genes in a given set ( mitochondrial genes , A , and oxidative phosphorylation genes , B ) have a significantly lower nucleosome occupancy at their 5′NFRs than that of all genome genes in each of 13 species ( x axis ) with nucleosome positioning data from Tsankov et al . ( 2010 ) and Xu et al . ( 2012 ) . ( C ) Evolutionary repositioning of binding sites for key amino acid TFs relative to NFRs . For each of 13 species ( columns , tree ) , shown are the enrichment ( yellow ) or depletion ( blue ) in NFRs of binding sites for several amino acid and purine metabolism TFs ( rows ) whose sites are depleted from NFRs in post-WGD species and enriched in pre-WGD species . The intensity of the color is proportional to the z-score estimated for each regulator from the fraction of all its binding sites that are in the NFR . Each row is centered by its mean value ( see ‘Materials and methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 016 In addition , the promoters of mRP and respiration genes have lost the binding site for the transcription factor Sfp1 ( previously identified as the ‘Rapid Growth Element’ , RGE; Ihmels et al . , 2005 ) in many respirofermentative species ( Figure 6—figure supplement 1A ) . The Sfp1 binding site is strongly conserved in ‘growth’ ( Module 1 ) genes across all species ( except S . japonicus ) and serves as an ancient regulatory tether of the growth module ( p<0 . 05 , Hyper-geometric test; Figure 6—figure supplement 1B ) . This site was lost from mRP gene promoters in post-WGD species ( Ihmels et al . , 2005; Tanay et al . , 2005 ) and S . japonicus ( Figure 6—figure supplement 1A ) . In S . pombe and K . polysporus some genes have lost the site whereas others retain it , where the number retained is an intermediate between that in post-WGD species and in respiratory species ( Jiang et al . , 2008 ) . The Sfp1 motif is also enriched in oxidative phosphorylation gene promoters in some of the respiratory species ( Hyper-geometric , p<0 . 05 , Figure 6—figure supplement 1A ) . Indeed , more surprisingly , purine metabolism and amino acid catabolism genes in pathways that feed nucleotide synthesis ( Figure 8A ) are reassigned from mildly repressed or unchanged Modules 2 and 3 pre-WGD to the strongly induced Module 5 post-WGD ( Figure 6E ) . These regulatory changes were not previously recognized , to the best of our knowledge , and are reminiscent of those observed in the Warburg effect ( Vander Heiden et al . , 2009 ) , a cancer metabolic state analogous to respiro-fermentation ( see ‘Discussion’ ) , including the induction of the glycine biosynthetic pathway that was recently shown to be correlated with proliferation rates across different cancers ( Jain et al . , 2012 ) . Unlike mRP and oxidative phosphorylation genes , this re-assignment is unique to the post-WGD species and does not occur in Schizosacchromyces ( Figure 6E ) . It is associated with the shifting of binding sites for key activators of amino acid and purine biosynthesis genes ( e . g . , Bas1 , Met32 , Gzf3 , Gln3 , Met28 , Met4 , and Gat1 ) from nucleosome-free to nucleosome-occluded positions ( Tsankov et al . , 2010 ) in rich media post-WGD ( Figure 7C ) . Conversely , the phylogenetic pattern of Sfp1 motif enrichment is less conclusive in these genes ( Figure 6—figure supplement 1A ) . 10 . 7554/eLife . 00603 . 017Figure 8 . Purine and amino acid metabolic pathways are linked to carbon metabolism . ( A ) Shown are the set of metabolic reactions in S . cerevisise associating purine biosynthesis and salvage and amino acid metabolism with carbon metabolism , and two key transcriptional regulators ( left ) . Mitochondrial genes link respiration to purine metabolism . Glycolysis is linked to purine salvage by the metabolic intermediate 3-P-glycerate . De novo purine metabolism is linked to the pentose shunt through ribulose-5-phosphate . The genes in red are induced post-shift in S . cerevisiae and other post-WGD species , but their orthologs are repressed in pre-WGD species . Both Schizosaccharomyces species have three copies of ZWF1 ( purple ) that are strongly induced . ( B ) Shown are the major carbon pathways involved in the fermentation or respiration of glucose and their interconnectivity . Both WGD and other duplicate genes in each pathway are indicated . The genes in red are induced post-shift in S . cerevisiae and most of the other post-WGD species while those in green are repressed similar to their pre-duplication orthologs . Differences in trans regulators may further contribute to the reassignment of their targets between modules . While many of the regulators of glucose repression in S . cerevisiae are present across the phylogeny ( Flores et al . , 2000 ) , the regulation of some has changed at the WGD and at the ancestor of the Schizosaccharomyces , consistent with the reassignment of their targets . For example , the glucose repressing MIG genes and the TUP1-CYC8 complex are strongly repressed following glucose depletion in most post-WGD species , whereas some respiration activators are strongly induced ( CAT8 and HAP2 , 4 , 5 and SIP2 post-WGD , HAP2 , MOT3 , and SIP2 in S . pombe , data not shown ) . We observed no such changes in the expression of known regulators of amino acid and purine metabolism ( data not shown ) . In some cases , duplication of key regulators followed by reassignment to a new module may have further contributed to new regulatory functions . For example , TPK1 and TPK3 are two WGD-derived paralogs encoding catalytic subunits of PKA , a major regulator of carbohydrate metabolism and stress responses ( Zaman et al . , 2008 ) . TPK1 in strongly induced in the sensu stricto species , as is the single TPK gene in the Schizosaccharomyces . TPK3 is repressed in those species , conserving the expression pattern of its ortholog in all the respiratory pre-duplication species ( data not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 017 Changes in expression following gene duplication have previously been proposed to contribute to evolutionary innovation ( Gu et al . , 2004; Ihmels et al . , 2005; Guan et al . , 2007; Tirosh and Barkai , 2007; Wapinski et al . , 2007b ) . In particular , several studies ( Teichmann and Babu , 2004; Gu et al . , 2004 , 2005; Conant and Wolfe , 2006 ) showed increased transcriptional divergence of paralogous genes in S . cerevisiae compared to non-duplicated genes , consistent with a model of accelerated transcriptional evolution following duplication ( Gu et al . , 2005 ) . However , since such studies typically compared the expression of paralogous genes in only one or two species ( Teichmann and Babu , 2004; Gu et al . , 2004 , 2005; Conant and Wolfe , 2006; Guan et al . , 2007; Tirosh and Barkai , 2007 ) , they could not reliably infer the ancestral state , and had to rely on different assumptions , for example that similarity between paralogs in a single species reflects their ancestral state ( Gu et al . , 2005; Teichmann and Babu , 2004 ) or that expression divergence should be interpreted within the context of sequence evolution ( Gu , 2004 ) . To systematically test the potential impact of gene duplication on the transcriptional response to glucose depletion , we applied Arboretum to an expanded set of 3676 orthogroups that incurred one duplication event ( ‘Materials and methods’; Table 1 ) . The modules identified after inclusion of paralogs exhibited the same major phenotypic patterns of expression ( Figure 9 , Figure 9—source data 1 , Supplementary file 2 ) , but had a lower AMCI ( paired t-test , p<0 . 03 , Figure 10A ) . 10 . 7554/eLife . 00603 . 018Figure 9 . Arboretum reconstruction of expression module evolution in the presence of paralogous genes ( Analysis 2 ) . ( A ) Five expression modules identified by Arboretum in the transcriptional response to glucose depletion , when paralogous genes are included in the run . Each row corresponds to a species ( tree , left ) and each major column to a module ( 1–5 , labels top ) . Modules labels are color coded by the regulation of the module’s genes following depletion , as noted on top , from bright blue ( Module 1 ) for strong repression to bright red ( Module 5 ) for strong induction . Each module’s height is proportional to the number of genes in that module . The five columns in each module are the expression levels at lag ( L ) , late log ( LL ) , diauxic shift ( DS ) , post-shift ( PS ) , and plateau ( P ) relative to mid-log phase . Red: induced; blue: repressed; white: no change . ( B ) – ( F ) Module assignments of all extant and ancestral species . Each matrix corresponds to the genes in one of the five modules in the LCA ( B: Module 1; C: Module 2; D: Module 3; E: Module 4; F: Module 5 ) , and shows the module assignment of these genes in each of the extant and ancestral species from S . cerevisiae ( leftmost column ) to the LCA ( rightmost column ) . The biological functions listed at the top of each module are general classifiers based on Gene ontology terms enriched in all species in that module ( Supplementary file 2 ) . The range of FDR p values and fraction of genes in each module are as follows: Module1: ribosome biogenesis , p<1 . 07 − 10−52 to 1 . 56 × 10−112 , fraction 32–53% . Module2: cell division , p<3 . 13 × 10−02 to 4 . 69 × 10−02 , fraction 10 . 2–32% . Module 3: cell morphogenesis , p<4 . 48 × 10−02 to 4 . 56 × 10−02 , fraction 22–78 . 7% . Module 4: mitochondrial , p<2 . 47 × 10−02 to 3 . 36 × 10−02 , fraction 2 . 3–36 . 2%; proteasome , p<2 . 7 × 10−03 to 5 . 48 × 10−03 , fraction 1 . 3–13 . 1% . Module 5: respiration , p<4 . 2 × 10−02 to 4 . 43 × 10−02 , fraction 34 . 9–55% . Module assignment in each species is marked by a color code , as in the top of panel a ( bright blue: Module 1 , light blue: Module 2 , white: Module 3 , pink: Module 4 , red: Module 5 ) . Species are ordered by post-fix ordering ( left-child , right-child and parent ) of the species tree , as marked on the legend ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 01810 . 7554/eLife . 00603 . 019Figure 9—source data 1 . Orthogroups included in the Aboretum run for analysis 2 . Shown are the genes used for Arboretum run including orthogroups in Analysis 1 as well as orthogroups with one duplication event ( Analysis 2 ) . The string ‘Dummy’ separates the genes from different modules . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 01910 . 7554/eLife . 00603 . 020Figure 10 . Regulatory evolution of paralogous genes . ( A ) Paralogous genes contribute to regulatory divergence . Shown in a scatter plot of the AMCI values for each extant ( blue ) and ancestral ( red ) species as estimated by Arboretum in a run without paralogs ( Analysis 1 ) ( y axis ) vs a run with paralogs ( x axis ) . Inclusion of paralogous genes lowers the AMCI , especially at the WGD and Schizosaccharomyces ancestors ( arrows ) . ( B ) Enrichment of paralogous genes among reassigned genes . Shown is for each species ( ancestral and extant ) the fold enrichment ( F ) of paralogs ( circle size ) among genes reassigned at that species . Only points at which there are significantly more paralogs that switch than expected by chance are shown ( Hyper-geometric p<0 . 05 ) . Circles are colored by the phylogenetic point of gene duplication ( cyan: A13 , black: A11 , purple: A10 , blue: A9 , white: WGD ancestor A5 ) . ( C ) Four possible regulatory fates of paralogous genes following duplication , relative to their immediate pre-duplication ancestor . Left: cartoon gene trees ( left ) and illustrative examples from our analysis ( right ) representing the module assignment ( circles ) of each paralog and their pre-duplication ortholog in each extant and ancestral species . Module assignment is color coded as in Figure 3 ( Bright blue , light blue , white , pink , red from Module 1 to 5 , respectively ) . Star: gene duplication . Lightning rod: gene loss . ( 1 ) Conserved: both paralogs ( UTP5 and UTP9 ) conserve the ancestral assignment ( Module 1 ) ; ( 2 ) Neo-functionalization: one paralog ( URA7 ) maintains the ancestral assignment ( Module 1 ) and the other ( URA8 ) is assigned to a different module ( Module 5 ) ; ( 3 ) Asymmetric divergence: both paralogs ( EUG1 , PDI1 ) are reassigned to distinct modules ( Module 3 , Module 4 ) than the ancestral one ( Module 5 ) . ( 4 ) Symmetric divergence: both paralogs ( SER3 , SER33 ) are reassigned to the same module ( Module 5 ) , distinct from the ancestral one ( Module 1 ) . ( C ) Cumulative distribution of module reassignment of genes before and after their duplication . Because after duplication there are two paralogs , each with its own re-assignment value , we compare the minimum ( red , p<1 × 10−4 ) , maximum ( green , p<1 × 10−66 ) , and average ( black , p<1 × 10−18 ) of the number of re-assignments after duplication , with the re-assignments before duplication ( blue ) . ( D ) Scatter plots showing for each gene its degree of module reassignment before duplication ( x axis ) vs the average degree of module reassignment of the two paralogs after duplication ( y axis ) . All module reassignments for a gene are normalized by the number the species in which the gene is present ( ‘Materials and methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 02010 . 7554/eLife . 00603 . 021Figure 10—source data 1 . Gene Ontology enrichment in sets of duplicate genes that diverged from the pre-duplication ancestor . Shown are the GO processes enriched in gene sets that switch their module assignment from their pre-duplication module . Each row corresponds to a GO process ( column 1 ) , and the subsequent columns indicate the phylogenetic point of duplication and numbers used to estimate the p value from the Hyper-geometric test . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 021 This increased divergence is specifically associated with paralogous genes . Genes that have a paralog were more likely to change their module assignment than those present in a single copy ( KS test , p<1 . 5 × 10−107 ) . Furthermore , controlling for any differences in function between duplicated and non-duplicated genes , paralogous genes are much more likely to be reassigned after duplication than their corresponding pre-duplication orthologs ( Figure 10D , E , KS test , p<1 × 10−18 ) . This increased divergence was most prominent at the WGD ancestor ( A5 ) and the Schizosaccharomyces ancestor ( A13 ) ( Figure 10A ) . Both ancestors also have a larger number of retained paralog pairs compared to other ancestral species and reside at the points of the ( independent ) evolution of respiro-fermentation . Consistently , the reassigned paralogs are enriched in carbon metabolism genes ( Figure 10—source data 1 ) . Most paralogs are reassigned within a short window following their duplication . In most cases , divergence in paralog assignment is already apparent at the immediate phylogenetic point of the duplication event ( at the phylogenetic resolution of this study ) . For example , ∼60% of the paralog pairs that arose at the ancestor of the Candida clade ( A10 ) and that will eventually change their assignment , do so at that ancestor . Similar trends are observed for paralogs that arose at most other ancestors ( e . g . , A11 , 54%; A9 , 48% ) . Overall , of all the paralog pairs that arose sporadically ( not in the WGD ) in the phylogeny , and where at least one member of the pair was subsequently reassigned to a different module , 53% do so ‘at’ the point at which they arose . Furthermore , such sporadically duplicated genes are significantly enriched ( Hyper-geometric test , p<0 . 05 , ) among all the reassigned genes ( duplicated or not ) at their point of duplication , but not at later phylogenetic points , when they behave comparably to non-duplicated genes ( Figure 10B ) . This suggests a short ‘window of opportunity’ in regulatory innovation , specifically facilitated by paralogous genes shortly after they are duplicated but not later , and provides strong experimental support to previous theoretical models ( Gu et al . , 2005 ) . Paralogs that arose in the WGD ( Figure 10B , A5 ) are a notable exception to some of these trends , and may affect divergence for a longer period . While a similar proportion ( 54% ) of those that diverge do so as early as they arise , they continue to contribute significantly to the divergence in module assignment at later points compared to non-duplicated genes ( Figure 10B , orange circles , Hyper-geometric test , p<0 . 05 ) . This distinction is not simply due to the finer phylogenetic resolution in the post-WGD clade in our study: the same trend is observed with eight of the 15 species with comparable resolution in the pre- and post-WGD clades ( Figure 11A , B ) . This extended window may be explained by the theoretical argument ( Lynch and Katju , 2004 ) that sporadically duplicated genes—but not WGD duplicated genes—may not be duplicated with the full set of ancestral regulatory elements and may move to novel genomic locations ( Lynch and Katju , 2004 ) , thus resulting in immediate divergence from the ancestral state . 10 . 7554/eLife . 00603 . 022Figure 11 . Similar evolutionary patterns in glucose depletion and heat shock . ( A ) Increased re-assignment of paralogous genes . Box-plots showing the fraction of module re-assignments for genes from orthogroups with duplication events ( Duplicate , left ) and without duplication events ( Singleton , right ) . Red plus: outliers that are ±2 . 7 SD from the mean . ( B ) Enriched re-assignment of paralogous genes at different phylogenetic points . Shown are the fold enrichment of paralogous genes among all the reassigned genes ( red , scale bar ) at different phylogenetic points ( rows ) for duplicates that arose at different ancestors ( columns ) for heat shock ( left ) and glucose depletion ( right ) . The number in each cell represents the number of paralogous genes that arose at a given phylogenetic point ( column ) and were reassigned at a phylogenetic point ( row ) . Numbers and fold enrichment are marked only at points with significantly more paralogs that are reassigned than expected by chance ( Hypergeometric p<0 . 05 ) . ( C ) – ( F ) correlation in expression decreases with phylogenetic distance . Shown are scatter plots relating—for each pair of species—their estimated phylogenetic distance ( y axis ) and the mean correlation between their matching global expression profiles ( x axis ) at matching time points ( labeled on top ) . Legend shows the clade to which the pair belongs ( if the same ) or ‘other’ ( if from different clades ) . Branch length was scaled by the maximum branch length to range from 0 to 1 . The line is the least squares fit . The Pearson correlation coefficient is shown on top ( C: p≤2 . 88 × 10−5; D: p≤2 . 86 × 10−5; E: p≤0 . 018; F: p≤0 . 19 ) . ( G ) Module divergence scales with phylogenetic distance . Shown is a scatter plot of the relationship , for each extant ( blue ) and ancestral ( red ) species , between its phylogenetic distance to its immediate ancestor ( branch length , y axis ) and its AMCI ( x axis ) . Branch length is scaled by the maximum value to range between 0 and 1 . The correlation between branch length and AMCI is shown at top ( p≤0 . 033 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 022 Paralogs may diverge either by generation of new functions ( neo-functionalization ) or by partitioning of several ancestral functions ( sub-functionalization ) between the paralogs through complementary degeneration ( the Duplicate-Degeneration-Complementation [DDC] model; Lynch and Force , 2000 ) . In the context of gene regulation , these events can in principle manifest either as changes in the expression of ‘targets’ ( e . g . , through cis-regulatory changes ) or by changes in regulator function ( e . g . , in a transcription factor’s responsiveness to signals or its binding site specificity ) . Previous studies of S . cerevisiae paralogs ( Teichmann and Babu , 2004; Conant and Wolfe , 2006; Tirosh and Barkai , 2007 ) showed indirect evidence for both sub- and neo-functionalization , but had to make strong assumptions on the ancestral state with little experimental data . To assess the extent of regulatory neo-functionalization of targets , we compared the module assignment of the two paralogs at the point of duplication to that of their immediate pre-duplication ancestor ( four possible fates , Figure 10C , 1–4 ) . We focused only on this earliest transition since the majority of paralogs ( 53% ) diverge first there , and since the pre-duplication module assignment provides a well-defined , identical reference point for both duplicates , whereas comparisons at later point do not afford a common reference . Our phylogenetic analysis allowed us to refine divergence events beyond the previously defined categories . Most paralogs ( 66% of all pairs ) have at least one member diverging from the ancestral module . This includes paralogs with distinct fates , either due to ‘classical’ neo-functionalization ( 27% of all pairs ) , where one paralog maintains the ancestral assignment and the other is assigned to a different module ( e . g . , the purine biosynthesis genes URA8 and URA7 , Figure 10C , case 2 ) , or due to asymmetric divergence ( 12% ) , where each paralog is reassigned to a distinct module , each different than the ancestral one ( e . g . , the disulphide isomerases EUG1 and PDI1 , Figure 10C , case 3 ) . In the remaining cases ( 27% ) there is symmetric divergence , where both paralogs are reassigned to the same module , distinct from the ancestral one ( e . g . , the ribonucleotide-diphosphate reductases SER3 and SER33 , Figure 10C , case 4 ) . Only in a minority of paralog pairs ( 34% ) the ancestral assignment of both copies is conserved ( e . g . , the ribosome biogenesis genes UTP5 and UTP9 , Figure 10C , case 1 ) . These classes could be easily confounded in previous studies relying on paralogs from a single species ( e . g . , ‘conserved’ and ‘symmetrically diverged’ would be indistinguishable ) . The pre-duplication orthologs of genes from the neo-functionalized class are most prominent in the ‘unchanged’ Module 3 and mildly induced Module 4 ( Hyper-geometric test , p<10−3 ) , suggesting that genes that were not regulated in glucose depletion prior to duplication , ‘gain’ such regulation de novo post-duplication , for at least one paralog . Neo-functionalization has also contributed to the regulatory divergence of key genes in carbon metabolism during the evolution of respiro-fermentation ( Figure 8B ) , consistent with previous studies ( Conant and Wolfe , 2007 ) . To assess the generality of our findings , we compared the key patterns discussed above to corresponding ones observed in the heat stress response in eight of the same species ( Roy et al . , 2013 ) . As in the response to glucose depletion , similarity in heat stress profiles is inversely proportional to phylogenetic distance ( Figure 11C–F ) , and the strongly repressed Module 1 , which consists of ‘growth’ genes , is the most conserved ( Roy et al . , 2013 ) . Furthermore , the evolutionary trajectories of paralogous genes—including the higher reassignment of paralogs , the distinct reassignment capacity of sporadic and large-scale duplications , and the preponderance of reassignment of only one of two paralogs—are highly similar between heat shock and glucose depletion ( Figure 11A , B ) . This suggests that these are general evolutionary principles , at least in this phylogeny . The additional condition allows us to further refine the classification of paralog fates ( Figure 12 ) . For example , some paralogs whose expression is ‘conserved’ in glucose depletion , may be reclassified as neo-functionalized when heat shock data is also considered , for example , due to a new regulation of one paralog in heat shock . Indeed , considering both responses , the conserved class was reduced from 34% to 28% , and the majority ( 70% ) of paralogs were in classes where either one ( 38% , Figure 12B , E ) or both ( 32% , Figure 12C , E ) were neo-functionalized in at least one response . Surprisingly , only 2% of all paralog pairs were in the sub-functionalized class ( Figure 12D , E ) . 10 . 7554/eLife . 00603 . 023Figure 12 . Regulatory evolution of paralogous genes in glucose depletion and heat shock . ( A ) – ( D ) several regulatory fates of paralogous genes following duplication , relative to their immediate pre-duplication ancestor in each of glucose depletion and heat shock . For each condition shown are cartoon gene trees ( left ) and illustrative examples from our analysis ( right ) representing the module assignment ( circles ) of each paralog and their pre-duplication ortholog in each extant and ancestral species . Module assignment is color coded as in Figure 3 ( Bright blue , light blue , white , pink and red from Module 1 to 5 , respectively ) . Star: gene duplication . Lightning rod: gene loss . ( A ) Conserved: both paralogs ( UTP5 and UTP9 ) conserve the ancestral assignment ( Module 1 ) in both responses; ( B ) Neo-functionalized , one paralog: one paralog ( URA7 ) maintains the ancestral assignment ( Module 1 ) and the other ( URA8 ) is assigned to a different module ( Module 5 ) in both responses; ( C ) Neo-functionalized , both paralogs: both paralogs ( POR1 , POR2 ) are reassigned to distinct modules than the ancestral one , but in different ways in each response . ( D ) Sub-functionalization: In glucose depletion , one paralog ( RFX1 ) maintains the ancestral assignment ( Module 2 ) and the other ( OG1201 ) is reassigned ( Module 3 ) . This pattern is reversed in heat shock . ( E ) Number of paralogs pairs in each of the classes . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 023 Here , we used comparative functional genomics of transcriptional profiles measured during growth of 15 species of Ascomycota on glucose to identify several principles in the evolution of gene expression in a complex phylogeny . Using yeasts allowed us to control growth conditions and closely monitor comparable physiological responses . The substantial overall conservation we observed in the response—exceeding that reported in other studies ( Tirosh et al . 2011 ) , indicates the robustness of our experimental strategy . However , despite this conservation , we found that the degree of correlation is inversely related with the phylogenetic distance between the species . Several lines of evidence suggest that these global differences in expression profiles are not simply dominated by the effect of growth rates or of ‘carbon metabolism’ lifestyle . First , the expression profiles in each species for a set of genes recently identified as regulated by growth rate in S . cerevisiae ( Brauer et al . , 2008 ) are similar across species in the phylogeny ( Figure 3—figure supplement 1A ) , consistent with our physiological sampling . Indeed , removal of these genes did not significantly change the correlation of global expression profiles ( p>0 . 1 , paired t-test ) at any of the points other than lag phase ( Figure 3—figure supplement 1B–F , paired t-test , p<2 . 02 × 10−8 ) . Notably , physiologically-matching profiles are significantly correlated even between distant species ( e . g . , Pearson’s r = 0 . 54 between S . cerevisiae and K . lactis at late log ) , supporting our experimental design . Second , the global profiles of post-WGD and Schizosaccharomyces species that share a respiro-fermentative lifestyle are nonetheless most distant from each other ( e . g . , Pearson’s r = 0 . 43 between S . paradoxus and S . pombe at late log ) . To systematically compare the responses of the different species to glucose depletion we developed Arboretum , a probabilistic algorithm , which incorporates the species and gene phylogenetic relationships to identify modules in each species and to reconstruct the evolutionary trajectory of each gene’s module assignment from the LCA ( Roy et al . , 2013 ) . During the learning process , Arboretum uses a soft probabilistic assignment of module membership allowing a gene to contribute to the mean expression profile of different modules . At the end of the learning , the gene is assigned to the single module with the highest probability . In the majority of cases , this ‘hard assignment’ is well supported in our dataset , such that the probability that a gene belongs to its assigned module is much higher than its probability to belong to the next-best module ( Figure 13A , B ) In some cases , however , a ‘soft’ probabilistic assignment may be preferable , to reflect the pleiotropic action of genes . A user can readily do this using the probabilities calculated by Arboretum . 10 . 7554/eLife . 00603 . 024Figure 13 . The per gene probability of Arboretum module assignments . ( A ) . Shown are the fraction of genes ( y axis ) that are assigned to the most likely module with probability of at least 0 . 5 , 0 . 7 or 0 . 9 in each species ( x axis ) . ( B ) . Shown are the fraction of genes ( y axis ) whose probabilities of the second most likely assignment is less 30% , 50% , or 70% of the most likely assignment , that is q/p<x% where q is the probability of the second most likely assignment and p is the probability of the most likely assignment . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 024 Another key parameter that could impact our results is the number of modules . We chose a relatively small number ( k = 5 ) as a conservative choice , which may under-estimate divergence relative to conservation , since genes with minor differences in expression between species would still likely belong to the same module . This choice increases our confidence in those divergence events that are highlighted , but may miss other divergence events . Several lines of evidence support the validity of our choice including the correspondence between the overall trends based on the direct expression levels ( e . g . , Figure 3B–F ) and those based on inferred modules ( e . g . , Figure 4C ) , the substantial proportion of variance ( 65–70% ) explained by 5 modules ( Figure 14 ) , and the lack of additional clear temporal patterns or distinct functional enrichments at higher k’s ( Figure 14—figure supplement 1 ) . Our companion manuscript ( Roy et al . , 2013 ) provides user guidelines for choosing k . Notably , Arboretum does not model dependencies between different time points , modeled each time point with a separate mean and variance . It is possible that by explicitly modeling temporal dependencies we can capture finer-grained dynamic differences . This is a direction of future research . 10 . 7554/eLife . 00603 . 025Figure 14 . Variance captured in Arboretum modules as a function of the number of modules . Shown are the mean and standard deviation of the coefficient of determination for each species , one per plot . Mean and standard deviation were calculated for different random initializations of Arboretum runs . Coefficient of determination ( y axis ) was measured for different values of the number of modules ( x axis ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 02510 . 7554/eLife . 00603 . 026Figure 14—figure supplement 1 . Mean expression of Aboretum modules as a function of different k values . Each plot is the mean expression profile of a module . Each row corresponds to different k's and each column corresponds to a species . DOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 026 Arboretum traces module ancestry and handles duplication events naturally using the tree structure as part of the probabilistic generative process of module assignments . In particular , the module assignments at the point of duplication are two independent draws of module membership , one each for copy , ensuring that the orthogroup has the same module assignment before the duplication and allowing it the freedom to change assignment after duplication . We used these features to characterize the role of gene duplication in the evolution of gene regulation , substantially expanding on previous analyses based on comparing paralog expression in on one or two species . We found that paralogs significantly contribute to divergence , mostly during a short window of opportunity after duplication , with a more prolonged contribution from WGD paralogs . This may be explained by the fact that segmental duplication rarely preserves the entire regulatory region resulting in immediate expression divergence ( Lynch and Katju , 2004 ) whereas duplicates from WGD events preserve the regulatory inputs and thus may take longer to diverge . Our study sheds important light on the regulatory and metabolic changes that accompanied the evolution of respiro-fermentation , which occurred twice independently in this phylogeny . In both cases , mitochondrial and respiratory genes were reassigned en masse , accompanied by convergent cis regulatory changes . However , changes in the regulation of amino acid , purine and sulfur metabolism genes occurred only following the WGD , possibly by repositioning of cis-regulatory elements relative to nucleosome free regions . The post-WGD event can be monitored in fine phylogenetic resolution , showing that many of the changes occurred gradually , with only some noticeable in K . polysporus , the first species to have diverged post-WGD . Our findings raise tantalizing analogies important for human physiology . Respiro-fermentation is functionally analogous to the ( somatically evolved ) Warburg Effect in cancer cells ( Vander Heiden et al . , 2009 ) , which relies—to some extent—on the ‘neo’-functionalization of metabolic enzymes . There are well-known similarities in glycolysis and respiration between the two states that enable both cancer and yeast cells to direct glucose to biosynthesis , supporting their rapid growth and proliferation . However , both the causes of this metabolic signature and its connection to biosynthesis are not fully understood . The observation that genes encoding enzymes in the nucleotide salvage and glycine biosynthesis pathways are strongly induced post-diauxic shift in respiro-fermentative post-WGD species is analogous to recently identified changes in these pathways in cancer cells that display the Warburg Effect ( Gaglio et al . , 2011; Jain et al . , 2012 ) , suggests that rewiring of nucleotide biosynthetic pathways contributes to this phenotype and points to fundamental metabolic principles that are independent of evolutionary descent . We used species specific Agilent microarrays , designed based on the ORF annotations of each species , as collected in the Fungal Orthogroup Repository ( Wapinski et al . , 2007b ) ( http://www . broadinstitute . org/regev/orthogroups/ ) . We conducted each experiment in at least biological triplicates ( Figure 3—source data 1 ) . However , since our sampling was done in ‘real time’ , concurrently with measuring growth rates and glucose levels , after the full time course was collected , we re-confirmed ( from the growth curve , glucose curve , etc . ) that the time sampled was consistent with its intended label ( e . g . , that diauxic shift indeed occurred at the time of glucose depletion ) . In a very few cases , we had to discard a planned replicate since its actual collection occurred before or after the intended physiological event ( see ‘Growth curve alignments’ below ) . The collected replicates were very highly reproducible ( median Pearson correlation between replicates is 0 . 97 , Figure 3—source data 1 ) . Furthermore , the microarray profiles are highly correlated to corresponding profiles measured using RNA-Seq data generated from RNA extracted from the same plateau and mid-log time point samples . This is true both when considering all genes ( 0 . 835 < r < 0 . 958 , mean r = 0 . 913 , 0 . 0362 SD ) and when considering only paralogs ( 0 . 849 < r < 0 . 965 , mean r = 0 . 924 , 0 . 0368 SD ) , indicating that our custom microarray designs accurately distinguish paralogs and estimate their expression levels . We used the following strains for each species: S . cerevisiae Bb32 ( 3 ) , S . paradoxus NCYC 2600 , S . mikatae IFO 1815 , S . bayanus CBS 7001 , S . bayanus uvarum CLIB 251 , C . glabrata CBS 138 , S . castellii CLIB 592 , K . lactis CLIB 209 , K . waltii NCYC 2644 , S . kluyveri NRRL 12651 , D . hansenii CLIB 195 , C . albicans SC 5314 , Y . lipolytica CLIB 89 , S . pombe SPY73h+ and S . japonicus YFS 275 . All species were grown in the following rich medium ( termed BMW ) : yeast extract ( 1 . 5% ) , peptone ( 1% ) , dextrose ( 2% ) , SC amino acid mix ( Sunrise Science ) 2 g/l , adenine 100 mg/l , tryptophan 100 mg/l , uracil 100 mg/l . The medium was chosen to minimize cross-species variation in growth . All strains were grown at 30°C except for S . castellii , which was grown at 25°C . D . hansenii is a halophilic yeast ( Kurtzman , 2000 ) and was grown in BMW media made in filtered sea water purchased from PETCO . We measured the growth of all our species in all published media formulations by determining the saturation coefficient ( Figure 2A ) . The saturation coefficient was determined by inoculating a 3 ml culture of each media type with 1 × 106 cells/ml of each species and measuring the OD600 using a Thermo Spectronic Genesys 20 spectrophotometer after 24 hr of growth at 30°C and 25°C . The BMW rich medium was chosen from saturation coefficient tests of over 50 formulations ( made by varying each of the components listed above ) as the formulation with the tightest distribution of saturation coefficients across all species ( Figure 2A ) . For each strain , cells were plated onto BMW plates from frozen glycerol stocks . After 2 days , cells were taken from plates and re-suspended into liquid BMW , and counted using a Cellometer Auto M10 . A 3 ml BMW culture was inoculated at 1 × 106 cells/ml and placed in a New Brunswick Scientific Edison model TC-7 roller drum on the highest speed until saturated ( 1–2 days ) . The saturated cultures were then used to inoculate 300 ml BMW batch cultures in 2-l Erlenmeyer flasks for the glucose depletion and repletion experiments described below . Flasks were transferred to New Brunswick Scientific Edison water bath model C76 shakers set to 200 rpm . High-resolution growth curve data were collected for each species . Cells were grown under the conditions described in ‘Basic experimental set up’ above . Saturated 3 ml cultures were counted , and an appropriate amount of cells was removed from each culture to inoculate 300 ml of BMW at 1 × 106 cells/ml . The OD600 was measured every 15–60 min using a Thermo Spectronic Genesys 20 spectrophotometer , and media samples were taken to measure glucose and ethanol levels on a YSI Biochemistry Analyzer Model 2700 . These data were used to determine when each species was at a particular phase of growth to choose comparable physiological time points . The Lag phase time point was taken 30 min after inoculation for all species . Log phase was defined as the mid-point of exponential growth . Diauxic Shift is the point at which glucose levels reached 0 . Two time points before the diauxic shift , Early Late Log , Late Log , and two time points after , Post Shift , Late Post Shift , were chosen for each species at a times proportional to the maximum growth rate in exponential phase in each species . Finally , the Plateau time point was defined as approximately 2 hr after the growth had plateaued ) . A pilot expression study was performed using the species-specific microarrays ( described below ) in which expression profiles were measured for each of the eight time points ( described above ) for one biological replicate for each species . These data were used to finalize the six time points used in the larger study ( Figure 2B , Figure 2—figure supplement 1 ) . Cultures were grown exactly as described above . The OD600 and glucose levels were measured throughout the day to ensure cultures were tracking with previously collected data . Samples were taken at: Lag , Log , Late log ( LL ) , Diauxic shift ( DS ) , Post shift ( PS ) , and plateau ( P ) . Samples were methanol quenched as described below and volumes removed for each sample were enough for gene expression analysis and for metabolite analysis ( not shown ) . The actual volumes removed are shown in Table 2 . 10 . 7554/eLife . 00603 . 027Table 2 . Sample volumes for RNA extraction and metabolite analysisDOI: http://dx . doi . org/10 . 7554/eLife . 00603 . 027RNA extractionMetabolite analysisPhaseTotal sample vol . ( ml ) No . tubesVol . /tube ( ml ) Vol . MeOH/tube ( ml ) Vol . /tube ( ml ) Vol . MeOH/tube ( ml ) Vol . Water/tube ( ml ) Lag504 ( 2 Met , 2 RNA ) 12 . 518 . 7512 . 5307 . 5Log605 ( 3 Met , 2 RNA ) 1522 . 5103010Early late log124 ( 3 Met , 1 RNA ) 69294Late log124 ( 3 Met , 1 RNA ) 69294Diauxic shift64 ( 3 Met , 1 RNA ) 34 . 5195Post shift64 ( 3 Met , 1 RNA ) 34 . 5195Late post shift64 ( 3 Met , 1 RNA ) 34 . 5195Plateau64 ( 3 Met , 1 RNA ) 34 . 5195Shown are the appropriate culture and methanol and water volumes used in the ‘cold’ methanol quenching procedure for cells prior to RNA and intracellular metabolite extraction . Samples were collected in 50-ml conicals filled with the appropriate amount of 100% methanol to produce a 60/40 mixture once the sample is added . The methanol-filled tubes were stored at −80°C until ready for use . During sample collection tubes were placed in a rack in a dry-ice ethanol bath kept at approximately −40°C . Once the sample was added to the methanol , the methanol and media were separated from the cells by centrifugation and poured off . The conicals containing a cell pellet were flash frozen in liquid nitrogen and then stored at −80°C until processed for permanent storage . For permanent storage , the cell pellets were then washed in 5 ml of nuclease-free water and spun for 5 min at 3700 rpm at 4°C . The supernatant was discarded and the pellet re-suspended in 2 ml of RNAlater ( Ambion , Life Technologies , Grand Island , NY ) and transferred to 2 ml Sarstadt tubes for storage . The samples were left at 4°C for 24 hr before being moved to a −80°C freezer . Total RNA was isolated using the RNeasy Midi or Mini Kits ( Qiagen , Valencia , CA ) according to the provided instructions for mechanical lysis . Samples were quality controlled with the RNA 6000 Nano ll kit of the Bioanalyzer 2100 ( Agilent , Palo Alto , CA ) . Total RNA samples were labeled with either Cy3 or Cy5 using a modification of the protocol developed by Joe DeRisi ( University of California at San Francisco ) and Rosetta Inpharmatics that can be obtained at http://www . microarrays . org ( Wapinski et al . , 2010 ) . For each time point , either two or three biological replicates were hybridized with the Log phase sample as the reference in all cases . We used two-color Agilent 55- or 60-mer oligo-arrays in the 4 × 44 K or 8 × 15 K format for the S . cerevisiae strain ( commercial array; four to five probes per target gene ) or the custom 8 × 15 K format for all other species ( two probes per target gene ) . After hybridization and washing per the manufacturer’s instructions , arrays were scanned using an Agilent scanner and analyzed with Agilent’s Feature Extraction software ( release 10 . 5 . 1 . ) . All the data has been deposited to the Gene Expression Omnibus and is available under accession GSE36253 . The median relative intensities across probes were used to estimate the expression values for each gene per replicate , and these median values across replicates were used to estimate the overall expression response per gene per time point , as previously described ( Wapinski et al . , 2010 ) . Since sampling time points were selected in real time during the experiment ( based on the growth curve data collected concurrently ) , after the data was collected we re-confirmed that the sample time points indeed matched their expected categorization . To this end , we used two methods to align the measured growth curves . In the first method , samples from different experiments in the same species were manually aligned by overlaying growth curves for each experiment . Samples were then categorized into time point classes ( LAG , LL , DS , PS , PLAT ) by their position on the growth curve and their expression profiles . In the second method , we applied two transformations to align growth curves . First , in each species sampling times for growth curves of biological replicates were shifted in order to align the exponential growth phase . As expected , the doubling time for each replicate was consistent . Next , a line was fitted to the exponential growth phase using all replicate data in order to get an average growth curve . This average growth curve for each species was then aligned to the S . cerevisiae growth curve , adjusting for the doubling time ( slope ) and speed ( shift along x axis ) during exponential growth . Finally , we plotted dextrose depletion and used it to manually align the DS time such that it matches S . cerevisiae . Sampling times were then extracted from the aligned growth curve . With few exceptions , the two approaches matched and were consistent with the original sampling choice . Expression data from different species were joined into a single matrix according to their orthology relationships from gene trees generated by the Synergy algorithm ( Wapinski et al . , 2007a , 2007b ) , as available from the Fungal Orthogroups Repository ( http://www . broadinstitute . org/regev/orthogroups/ ) . K . polysporus is not included in the repository and was added according to its orthology to S . cerevisiae from the Yeast Gene Order Browser ( YGOB; Byrne and Wolfe , 2005; http://wolfe . gen . tcd . ie/ygob/ ) . The resulting expression matrix was clustered using hierarchical clustering , weighted such that each clade is equally represented , each species within each clade is equal , and each time point within each species is equal . The correlation coefficient between a pair of species used in the clustering procedure was calculated by using the median expression values across five time points . We used a tree topology and branch length estimation generated as previously described ( Wapinski et al . , 2007a , 2007b ) . We used our previously published tree topology , which is based on protein sequences , except that the relative positions of two species ( C . glabrata and S . castellii ) have been flipped , as we previously described ( Wapinski et al . , 2007b ) . We estimated branch length by considering Uniform orthogroups , those where exactly one ortholog is present in every species in our panel . We randomly selected 1000 Uniform orthogroups and generated multiple sequence alignments of the genes each orthogroup using the MUSCLE program ( Edgar , 2004 ) . These alignments were concatenated and branch lengths were estimated using codeml of the PAML package ( Yang , 2007 ) . We repeated this procedure 10 times , selecting a random subset of 1000 uniform orthogroups each time , and used the average of the branch length values as branch length estimates . Average and standard deviations are reported in Figure 1—source data 1 . As a measure of expression divergence , we calculated Pearson’s correlation coefficient ( r ) between the expression profiles of each pair of species at physiologically matching time points: Lag , Late lag , Diauxic shift , Post-diauxic shift and Plateau , separately ( as in Figure 3B–G ) . For each pair of species , we considered only those genes that are present in both species . We used the sum of the branch lengths of each species in a pair to their nearest common ancestor as their sequence divergence , and normalized it to range from 0 to 1 by dividing by the maximum sequence distance . We next compared expression divergence and sequence divergence using linear regression . Arboretum is an algorithm that identifies modules of co-expressed genes across multiple species and infers their evolutionary histories ( Roy et al . , 2013 ) . Arboretum is based on a generative probabilistic model of module evolution via transition matrices ( e . g . , Figure 5A ) , and models expression generation at extant species via Gaussian mixture models , with a mean and variance for each time point of measurement in each species . Arboretum takes as input the genome-wide transcriptional profiles for all species measured , as well as a species tree , and all gene trees ( as defined from genome sequences ( Wapinski et al . , 2007a , 2007b ) . As output it provides module membership of genes in all extant and ancestral species , mean and variance of gene expression in each module in each extant species , and transition matrices at each species ( extant or ancestral , except the LCA ) that specify the probability of genes in each module to preserve or change their module membership in that species compared to their assignment in its immediate ancestor . During the module identification procedure , Arboretum uses the gene tree structure to trace the evolution of gene membership from ancestral modules to extant ones . If a gene changes it’s module membership compared to its immediate ancestor , the ancestral module is considered to have contracted , while simultaneously the new module of the gene is considered to have expanded . Explicitly incorporating the gene trees enables Arboretum to handle complex orthology and paralogy relationships that arise from gene duplications and losses . The model parameters are learned using an Expectation Maximization ( EM ) algorithm , which on convergence provides maximum likelihood estimates of the transition matrices , module mean and variance and module assignments in both extant and ancestral species . We applied Arboretum to two separate datasets ( Table 1 ) , the first ( Analysis 1 ) including orthogroups with no duplication events ( 2746 orthogroups ) and that are present in at least two species , one of which had to be S . cerevisiae , and the second ( Analysis 2 ) including orthogroups with at most one duplication event ( 3676 orthogroups ) and that are again present in at least one other species in addition to S . cerevisiae . Orthogroups with few duplications have the most reliable orthology relations and gene trees ( Wapinski et al . , 2007b ) . In both cases , the orthogroups could include losses , and could be lineage specific . Species and gene phylogenetic information were generated by Synergy ( Wapinski et al . , 2007a ) and obtained from the Fungal Orthogroup Repository ( http://www . broadinstitute . org/regev/orthogroups/ ) with the exception of K . polysporus , which was handled as described above . K . polysporus was excluded from the second set ( including duplicates ) , due to lack of reliable gene tree information for its orthogroups . Because Arboretum is based on the EM algorithm , final module inference and parameter estimation results may depend upon initial parameter settings . Furthermore , module assignments may get permuted during learning . For these reasons , we ran Arboretum with five different random initializations , where each run was seeded with an initial clustering with merged expression data across all species . We selected a particular random initialization for our final results based on manual inspection of the transition matrices , and did not have any arbitrary module permutations . Results for different random initializations were very comparable in terms of the module expression patterns and the transition matrices . In particular , the average F-score similarity of modules between any pair of random initializations ranged from 78% to 99% for the extant species , and from 88% to 96% for ancestral species for the non-duplication case . Similarly , for the case when we included duplications , the average F-score similarity of modules between any pair of random initializations ranged from 76% to 99% for the extant species , and ranged from 77% to 98% for ancestral species . A more detailed analysis of Arboretum’s stability and sensitivity is described in ( Roy et al . , 2013 ) . The transition matrices are also very similar across random initializations for both the no-duplication ( 0 . 73–0 . 99 ) and duplication cases ( 0 . 68–0 . 99 ) . The Gaussian parameters were initialized by projecting the clusters onto individual species . The transition matrices were initialized to have diagonal-heavy elements , by setting the diagonal elements to p , p>0 . 5 , and the off-diagonal elements to ( 1 − p ) / ( k − 1 ) where k is the number of modules . We experimented with different values of p , p∈ {0 . 5 , 0 . 6 , 0 . 7 , 0 . 8} , and used p=0 . 8 for the final results . Learned transition matrices are similar for p=0 . 6 and p=0 . 7 , and had a correlation of ( 0 . 7–1 . 0 ) , with the exception of Schizosaccharomyces and C . glabrata which exhibited a somewhat lower correlation ( 0 . 44–0 . 56 ) in the transition matrices initialized by the different p . In addition , we also compared the behavior of Arboretum using transition matrices initialized from the branch lengths . Specifically , the probability of a gene in species s to maintain it’s module membership was exp ( −m × ls ) where ls is the branch length from s to it’s immediate ancestor . We set m to 1 , 2 or 3 , but did not observe any significant differences in the learned transition matrices using these initializations vs initializations of p=0 . 8 . The transition matrices learned were not significantly different from those initialized uniformly with heavy diagonals ( and as we observed in the case of matrices initialized with p=0 . 5 , C . glabrata , and the Schizosaccharomyces species had the lowest correlation 0 . 44–0 . 56 with matrices initialized with p=0 . 8 ) . To select the number of modules in Arboretum we used a combination of penalized log likelihood and manual inspection . First we computed the penalized log likelihood using the minimum description length ( MDL ) as the penalty with different values for the number of modules , k = 5 , 7 , 9 , 11 . The MDL penalty increases as a function of the number of parameters in our model , which is equal to the number of means and variances we estimate for each extant species , and the size of the transition matrices in both extant and ancestral species ( see Roy et al . , 2013 for more details ) . Using this process the number of modules asymptotes at k = 9 . We next inspected the patterns associated with each module , which entailed plotting the mean expression profile of each module and selecting k to reflect the most distinct temporal patterns . Using this process we determined that k = 5 gave us the most meaningful and distinct set of profiles in terms of the GO processes enriched in different modules and the patterns associated with each module . Higher values of k did not improve the GO enrichment and was prone to module switching between closely related expression profiles . To measure the amount of signal captured by our modules , we estimated the percent variance explained , defined as 1−SerrStotal , where Serr is the sum of squared errors between the measured expression profile of a gene , and its predicted expression profile . The predicted expression profile is the mean of the module to which the gene is assigned . Stotal is defined as the total variance in the expression data . This is also called the coefficient of determination . The Ancestral Module Conservation Index ( AMCI ) is defined for every species ( except the LCA ) as the tendency of the genes in that species to conserve their module assignment compared to the assignment in its immediate ancestor . The AMCI for a species t is calculated as the average of the diagonal elements of t’s transition matrix . Because each element is a probability value , the AMCI is bounded between 0 and 1 . The closer it is to 1 , the more likely is the species to preserve the module assignments of its immediate ancestor , and the closer it is to 0 , the more likely it is to diverge from the module assignments of its immediate ancestor . We defined two metrics for measuring module divergence: the Module Contraction Index ( MCI ) and the Module Expansion Index . At each phylogenetic point , s , we estimate for each module m: ( 1 ) ‘innovations’: the number of genes for which the module assignment is m in s but not m in s’s immediate ancestor , and ( 2 ) ‘divergences’: the number of genes for which the module assignment is m in s’s ancestor but not in s . We define the global MCI of a module as the sum of divergences across all species ( extant or ancestral , except the LCA ) divided by a normalization term , Z , defined as follows: ∑s , t∈S , s≠tNst , where S is the set of all species other than the LCA , t is s’s immediate parent , and Nst is the number of genes for which we have a module assignment in both s and t . We define MEI as the sum of all innovations divided by Z . We define the stability of a module j by estimating the fraction of cluster assignments in a parent–child species pair that have the same assignment j in that pair of species . We use the FDR corrected hyper-geometric p-value to assess enrichment of GO processes in a given gene set or time point . We use the Benjamini–Hochberg method of FDR calculation . We use the Gene Ontology terms for S . cerevisiae downloaded from the Saccharomyces Genome Database ( SGD , http://www . yeastgenome . org/ ) . For all other species , we use orthology to transfer the S . cerevisiae annotations , as previously described ( Wapinski et al . , 2007b ) . The GO enrichments for each Arboretum module are listed in Supplementary files 1 and 2 . We used a database of species-specific motifs ( Habib et al . , 2012 ) to search for cis-regulatory elements in 600 bp upstream of the start codon of each gene in each species . Enrichment was assessed based on the p value from the Hypergeometric distribution . The species-specific motif library was created by starting from known position weight matrices in S . cerevisiae ( MacIsaac et al . , 2006; Zhu et al . , 2009 ) and refining them using an Expectation Maximization framework on individual species sequences . To find instances of motif matrices we used a procedure similar to Habib et al . ( 2012 ) . Briefly , we scan the upstream sequence of a gene summing over all instances of a motif to get a real-valued number per promoter that measures the affinity of a TF for a promoter . These numbers are ranked to select the top 40% genes with the highest affinities and these are considered as target genes for the motif . We assessed the contribution of gene duplication to module reassignment at several levels . First , we considered all orthogroups with duplications as a single set . To assess the tendency of a gene to be reassigned between modules before and after duplication , we compared the normalized number of reassignments of a gene pre-duplication to the minimum , maximum and average of the normalized number of reassignments of its two ‘descendant’ copies post-duplication . The normalization factor for the pre-duplication counts was the number of species before duplication , and the normalization factor for the post-duplication counts was the number of species including and after the duplication . In both cases we excluded a species from the normalization if it lost the gene or had a missing value . We then used the normalized reassignment counts to generate the cumulative distributions for number of reassignments pre-duplication , and minimum , maximum and average number post-duplication , and compared these using a Kolmogorov–Smirnov ( KS ) test . Second , we compared the cumulative distributions of normalized reassignment counts for genes that were never duplicated vs those that had a duplication event . For genes with duplications , we considered reassignments only at and after the point of duplication , and handled the paralogous copies as two data samples . We compared these cumulative distributions using a KS-test . Third , we analyzed the reassignment of genes at each phylogenetic point . We considered duplicates that arose at the Schizosaccharomyces ancestor ( A13 ) , the Hemiascomycota ancestor ( A11 ) , the Candida ancestor ( A10 ) , the Kluveromycetes and post-WGD species ancestor ( A9 ) , and the WGD ancestor ( A5 ) , because these are the phylogenetic points at which the vast majority of duplications have occurred in the considered orthogroups . For each of these phylogenetic points , s , we define three quantities: ( 1 ) X , the number of paralog pairs whose duplication occurred at s , ( 2 ) Y , the number of paralog pairs out of those in X of which at least one member is reassigned at any point in the phylogeny after duplication ( including s ) , and ( 3 ) W , the number of paralog pairs out of those in X of which at least one member is reassigned at s exactly . Note , each paralog pair is allowed to contribute at most once to these counts . We then estimate the fraction of paralogs that are reassigned first at their point of origin as W/Y , and the fraction of duplicates that are reassigned at all as Y/X . Fourth , we considered duplicates that arose at each point of origin s , and estimated their tendency to be reassigned at s and at all subsequent phylogenetic points , t . Specifically , we compared the number of duplicates that are reassigned at a given phylogenetic point to the total number of genes that are reassigned at that phylogenetic point . The statistical significance of the number of duplicates that are reassigned vs the number of all genes ( with or without duplication ) that are reassigned at that point is assessed using the Hyper-geometric distribution . As described in the main text , the module assignments of paralogs can have four possible fates right after duplication: conserved , neo-functionalized , symmetrically diverged and asymmetrically diverged . To assess if any of these classes were specifically associated with a particular expression module , we used a Hyper-geometric distribution that tests the probability that k or more genes from module i , have fate f , given that there are x genes in module i , t genes in all that are associated with fate f , and there are a total of T genes that can have any of the fates at the pre-duplication ancestral nodes . We performed this test at the pre-duplication ancestor and the post-duplication child node separately , as the total number of genes are different at these phylogenetic points . To analyze the nucleosome occupancy profiles of species-specific modules and gene sets that are reassigned in a coordinated way between modules , we used an approach similar to that previously described for functional gene sets ( Tsankov et al . , 2010 ) . Briefly , we compared the distribution of the 5′ Nucleosome Free Region ( NFR ) occupancy counts for a gene set with the distribution of 5′NFR occupancy of all the genes in the genome using the KS-test . We used the negative of the logarithm of the p value of the KS-test to quantify the difference in the two distributions . We used a collection of species-specific motifs ( Habib et al . , 2012 ) as well as motifs representing A or G rich k-mers ( e . g . , polyA and polyG ) ( Tsankov et al . , 2011 ) . Instances of these motifs were searched in 600-bp upstream of the ATG of each gene and each species . To assess if the instances of a particular motif m were repositioned with respect to the NFR in post-WGD species vs pre-WGD species , we used an approach as previously described ( Tsankov et al . , 2010 ) . For each motif m and species t , we computed the fraction of instances in NFR regions across all of t’s genes , fmt . We converted the fmt values to z-scores using the mean and standard deviation of all fmt estimated for t . To test if the instances of a motif m are re-positioned in the post-WGD species , we compared the set of z-scores for m in the post-WGD species to the set of z-scores for m in the pre-WGD species , using a t-test ( p<0 . 05 ) . To assess if the instances of a motif , especially polyA , were enriched or depleted in the NFR of a query set of genes , we used a Hyper-geometric test to compare the number of instances inside the NFR vs the number of instances that are in the NFR for all genes in the genome . Similarly , we calculated enrichment of instances outside the NFR by comparing the number of genes in our query set with instances that were outside the NFR to the total number of genes that had instances outside the NFR . Depletion of instances in NFR was ( 1 − p ) of enrichment of motif instances in the NFR region . Similarly depletion outside NFR was ( 1 − p ) of enrichment of instances outside NFR .	The incredible diversity of living creatures belies the fact that their genes are quite similar . In the 1970s Mary-Claire King and Allan Wilson proposed that a process called gene regulation—which determines when , where and how genes are expressed as proteins—is responsible for this diversity . Four decades later , the central role of gene regulation in evolution has been confirmed in a wide range of species including bacteria , fungi , flies and mammals , although the details remain poorly understood . In recent years it has been suggested that the duplication of genes—and sometimes the duplication of whole genomes—has had a crucial influence on the part played by gene regulation in the evolution of many different species . Ascomycota fungi are uniquely suited to the study of genetics and evolution because of their diversity—they include C . albicans , a fungus that is found in the human mouth and gut , and various species of yeast—and because many of their genomes have already been sequenced . Moreover , their genomes are relatively small , which simplifies the task of working out how it has changed over the course of evolution . It is also known that species in this branch of the tree of life diverged before and after an event in which a whole genome was duplicated . Ascomycota fungi use glucose as a source of carbon in different ways during aerobic growth . Most , including C . albicans , are respiratory and rely on oxidative phosphorylation processes to produce energy . However , a small number—including S . cerevisiae and S . pombe , two types of yeast that are widely used as model organisms—prefer to ferment glucose , even when oxygen is available . Species that favor the latter respiro-fermentative lifestyle have evolved independently at least twice: once after the whole genome duplication event that lead to S . cerevisiae , and once when S . pombe and the other fission yeasts evolved . Thompson et al . have measured mRNA profiles in 15 different species of yeast and reconstructed how the regulation of groups of genes ( modules ) have evolved over a period of more than 300 million years . They found that modules have diverged proportionally to evolutionary time , with prominent changes in gene regulation being associated with changes in lifestyle ( especially changes in carbon metabolism ) and a whole genome duplication event . Gene duplication events result in gene paralogs—identical genes at different places in the genome—and these have made significant contributions to the evolution of different forms of gene regulation , especially just after the duplication event . Moreover , the paralogs produced in whole genome duplication events have resulted in bigger changes over longer periods of time . Similar patterns were observed in the regulation of the genes involved in the response to heat shock in eight of the species , which suggests that these are general evolutionary principles . The changes in gene expression associated with the respiro-fermentative lifestyle may also have implications for our understanding of cancer: healthy cells rely on oxidative phosphorylation to produce energy whereas , similar to yeast cells , most cancerous cells rely on respiro-fermentation . Furthermore , yeast cells and cancer cells both support their rapid growth and proliferation by using glucose for biosynthesis to support cell division , although this process is not fully understood . Normal cells , on the other hand , use glucose primarily for energy and tend not to divide rapidly . Thompson et al . found that the genes encoding enzymes in two biosynthetic pathways—one that produces the nucleotides necessary for DNA replication , and one that synthesizes glycine—are induced in respiro-fermentative yeasts but repressed in respiratory yeast cells . The fact that similar changes are observed in the same two pathways when normal cells become cancer cells suggests that these pathways have an important role in the development of cancer . The framework developed by Thompson et al . could also be used to explore the evolution of gene regulation in other species and biological processes .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "evolutionary", "biology" ]	2013	Evolutionary principles of modular gene regulation in yeasts
Eukaryotes package DNA into nucleosomes that contain a core of histone proteins . During DNA replication , nucleosomes are disrupted and re-assembled with newly synthesized histones and DNA . Despite much progress , it is still unclear why higher eukaryotes contain multiple core histone genes , how chromatin assembly is controlled , and how these processes are coordinated with cell cycle progression . We used a histone null mutation of Drosophila melanogaster to show that histone supply levels , provided by a defined number of transgenic histone genes , regulate the length of S phase during the cell cycle . Lack of de novo histone supply not only extends S phase , but also causes a cell cycle arrest during G2 phase , and thus prevents cells from entering mitosis . Our results suggest a novel cell cycle surveillance mechanism that monitors nucleosome assembly without involving the DNA repair pathways and exerts its effect via suppression of CDC25 phosphatase String expression . Chromatin assembly during DNA replication is crucial for the repackaging of newly synthesized DNA and for maintaining or erasing histone modifications . During this process , pre-existing or so-called parental histones are recycled and assembled into nucleosomes together with de novo synthesized histones ( Alabert and Groth , 2012; Annunziato , 2012 ) . To compensate for the high demand of histone proteins during DNA replication , the canonical histones H1 , H2A , H2B , H3 , and H4 , which are encoded by multiple gene copies in higher eukaryotes , are highly and exclusively expressed in S phase of the cell cycle ( Marzluff et al . , 2008 ) . The assembly of chromatin is mediated by an interplay of components of the DNA replication machinery and histone chaperones , which mediate the deposition of histones into nucleosomes ( Alabert and Groth , 2012; Annunziato , 2012 ) . Apparently , the pace of DNA synthesis is tightly coupled to the assembly of newly synthesized DNA into chromatin . Multiple studies showed that the depletion of the histone chaperones Asf1 and CAF-1 results in a slow down of DNA synthesis during S phase ( Hoek and Stillman , 2003; Ye et al . , 2003; Nabatiyan and Krude , 2004; Groth et al . , 2007; Takami et al . , 2007 ) preceding the accumulation of DNA damage in mammalian cells ( Hoek and Stillman , 2003; Ye et al . , 2003 ) . Also , diminishing histone supply during S phase through knock down of SLBP , which is required for histone mRNA stability and translation , decreases the rate of DNA synthesis ( Zhao et al . , 2004 ) . A recent study that targeted SLBP together with FLASH , a factor that is required for histone mRNA transcription and processing ( Barcaroli et al . , 2006; Yang et al . , 2009 ) , revealed that replication fork progression depends on nucleosome assembly potentially through a mechanism based on a feedback from the histone chaperone CAF-1 to the replicative helicase and/or the unloading of PCNA from newly synthesized DNA upon nucleosome assembly ( Groth et al . , 2007; Mejlvang et al . , 2014 ) . The coupling of replication fork progression and nucleosome assembly might compensate for short-term fluctuations in histone availability ( Mejlvang et al . , 2014 ) . However , it is still unclear whether chromatin integrity is monitored after or during DNA replication . Genome integrity during S phase is governed by the ATR/Chk1 and ATM/Chk2 checkpoint mechanisms that sense replication stress and DNA damage , respectively ( Bartek and Lukas , 2007; Cimprich and Cortez , 2008 ) . Lack of CAF-1 or Asf1 function leads to accumulation of DNA damage and activation of the ATM/Chk2 pathway ( Hoek and Stillman , 2003; Ye et al . , 2003 ) . These findings led to the hypothesis that chromatin assembly is monitored indirectly through accumulation of DNA lesions in response to stalled replication forks . However , since these chaperones have multiple functions such as unwinding of DNA during replication , in DNA repair ( Gaillard et al . , 1996; Green and Almouzni , 2003; Schöpf et al . , 2012 ) as well as other nuclear processes ( Quivy et al . , 2004; Houlard et al . , 2006 ) . These multiple functions of these chaperones make it difficult to assess the direct effects of defective chromatin assembly . Taking advantage of a histone null mutation in a higher eukaryote that recently became available in Drosophila melanogaster ( Günesdogan et al . , 2010 ) , we directly addressed the requirement of canonical histone supply for DNA replication and cell cycle progression in a developing organism . By reintroducing a defined number of transgenic histone genes into the histone null mutant background , we show that the rate of DNA replication is coupled to the number of histone genes present in the genome and that histone supply is critical to coordinate S phase length with the developmental program . Surprisingly , cells that completely lacked de novo histone synthesis replicate DNA at a reduced rate , but complete S phase and arrest in cell cycle without accumulating DNA damage . This cell cycle arrest is mediated by suppressing the accumulation of transcripts encoding the CDC25 phosphatase String and provides evidence for a chromatin assembly surveillance mechanism that is independent of the known S phase checkpoints . The histone null mutation in D . melanogaster , called Df ( 2L ) HisC , lacks all genes encoding the canonical histones ( Günesdogan et al . , 2010 ) . Df ( 2L ) HisC homozygous mutant animals ( hereafter referred to as HisC mutants ) that are derived from heterozygous parents contain only maternal histone mRNA and proteins , which are sufficient to complete the first 14 cell division cycles of the embryo ( Günesdogan et al . , 2010 ) . HisC mutant embryos arrest before the onset of mitosis in cycle 15 ( M15 ) ( Günesdogan et al . , 2010 ) . This highly uniform phenotype is likely due to the degradation of maternal histone mRNAs during the first G2 phase of embryogenesis in cell cycle 14 ( Marzluff et al . , 2008; O'Farrell et al . , 1989 ) combined with the complete lack of zygotic histone gene expression during S phase of cell cycle 15 ( S15 ) ( Günesdogan et al . , 2010 ) . In order to verify that the lack of histone transcription also results in a diminished pool of histone proteins in S15 , we compared the protein levels of histone H2B and H3 of wild type embryos that are in S15 at 4–5 hr after egg laying ( AEL ) to sorted HisC mutant embryos that are still in S15 at 5 . 5–6 . 5 hr AEL ( see below and Günesdogan et al . , 2010 ) by quantitative Western blotting ( Figure 1A , B ) . The approximate twofold reduction in the histone levels of HisC mutant embryos is consistent with the fact that these embryos lack synthesis of new histones in S15 but still contain parental histones from chromatin that was assembled during cycle 14 . To test whether the reduced supply of histones in HisC mutant embryos leads to a decrease in nucleosome formation , we carried out Micrococcal Nuclease ( MNase ) digestion assays on chromatin from sorted HisC mutant and wild type sibling embryos ( Figure 1C , D , Figure 1—figure supplement 1 ) . The results show that chromatin from HisC mutant embryos is more accessible to MNase than control chromatin , leading to a more rapid generation of mononucleosomal DNA fragments and reflecting a decrease in nucleosome occupancy in chromatin of the HisC mutants . 10 . 7554/eLife . 02443 . 003Figure 1 . Nucleosome density is affected in HisC mutant cells . ( A ) Fluorescent Western blot analysis for histone H2B and H3 ( green ) using wild type embryos at 4–5 hr and sorted HisC mutant embryos at 5 . 5–6 . 5 hr AEL , respectively . α-Tubulin ( α-Tub; red ) was used as loading control . A dilution series of each extract was loaded ( 1 , 0 . 5 , 0 . 25 ) . ( B ) Quantification of Western blots as shown in ( A ) . Fluorescence measurements of the histone signal was normalised to the α-Tubulin signal and the wild type ( WT ) /HisC ratio is shown . The histone protein content is ∼twofold reduced in HisC mutant embryos as compared to wild type . Mean values from three independent experiments are shown . Error bars indicate standard error . ( C ) Gel electrophoresis of time-course ( 0′–5′ ) microccocal nuclease ( MNase ) digestions using sorted HisC mutant and wild type sibling embryos at 5 . 5–6 . 5 hr after egg laying , respectively . ( D ) Quantification of MNase digestion experiments as shown in ( C ) . HisC mutant chromatin is digested more rapidly into mononucleosomal DNA than control chromatin . Mean values from three independent experiments are shown . Error bars indicate standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 00310 . 7554/eLife . 02443 . 004Figure 1—figure supplement 1 . HisC mutant chromatin shows increased MNase sensitivity . Quantification of the time-course microccocal nuclease ( MNase ) digestion experiment shown in Figure 1C using the Tapestation analysis software ( Agilent Technologies ) . Shown are the individual electropherograms for each digestion ( 0′ , 1′ , 2′ , 3′ , 5′ ) , which display the size and concentration of DNA . Regions for quantification were defined as follows: 60–115 bp ( <115 bp ) , 115–270 bp ( mononucleosomes ) , 270–455 bp ( dinucleosomes ) , 455–650 bp ( trinucleosomes ) . Numbers above regions indicate the concentration as % of input . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 004 Arrested HisC mutant cells express high levels of mitotic Cyclin B suggesting a cell cycle arrest in G2 phase of cycle 15 ( G215 ) before mitosis ( Günesdogan et al . , 2010; Lehner and O'Farrell , 1990; Figure 2A ) . Consistent with this , we did not observe degradation of the mitotic Cyclin A or assembly of mitotic spindles in HisC mutant cells ( Figure 2—figure supplement 1 ) . To assess DNA replication in S15 , we used BrdU incorporation assays to label newly synthesized DNA ( Figure 2 ) . The results show that most cells of HisC mutant embryos entered S15 , but the spatial pattern of replicating cells was different from the highly stereotyped wild type pattern ( Figure 2B–D ) . In wild type embryos , about 5 hr AEL ( Figure 2C , E ) , the ventral epidermal cells incorporated BrdU in S15 , whereas the lateral cells entered G215 and stopped BrdU incorporation . Dorsal epidermal cells had already passed through M15 and incorporated BrdU in S16 . In contrast , the HisC mutant cells failed to reach G215 and the majority of the cells continued to incorporate BrdU at low levels in S15 ( Figure 2D , F ) . To test whether the extended S phase of HisC mutant cells is caused by a reduced rate of DNA synthesis , we shortened the BrdU labelling pulses from 15 to 5 min . In wild type cells , BrdU incorporation was detectable during S15 ( Figure 2I–K ) , whereas in HisC mutant cells BrdU incorporation was detected only in a few cells with low Cyclin B levels indicating that they were still in early S15 ( Figure 2L–N ) . Notably , in HisC mutants ventral cells showed a uniform pattern and dorsal cells a punctuate pattern of BrdU incorporation ( Figure 2G , H ) , which is characteristic for the replication of euchromatin and heterochromatin in early and late S phase , respectively ( Shermoen et al . , 2010 ) . Although we cannot exclude that the lack of histone synthesis in HisC mutants interfered with firing of individual origins , the results suggest that both early and late replication origins are activated in HisC mutants . In conclusion , the absence of de novo histone supply reduces the rate of DNA synthesis shortly after the initiation of DNA replication , resulting in an extended S phase in mutant cells . 10 . 7554/eLife . 02443 . 005Figure 2 . HisC mutant cells extend S phase and slow down DNA synthesis . ( A and B ) Schematic models of Cyclin B accumulation and BrdU incorporation during the embryonic cell cycle ( A ) and its spatial control ( B ) . AEL: after egg laying . ( C–N ) BrdU pulse labelling for 15 ( C–H ) or 5 min ( I–N ) and staining with antibodies against BrdU ( green in merge ) and Cyclin B ( red in merge ) . ( C and D ) BrdU was detected in the epidermis of HisC mutant embryos indicating DNA replication but the pattern of replicating cells was distinct from wild type . ( E and F ) Magnifications of an epidermal region as shown for the wild type embryo ( boxed area in C ) . ( E ) BrdU labelled wild type cells of the ventral epidermis in S15 ( below dashed lines ) and of the dorsal epidermis in S16 ( above dashed lines ) . Cells in G215 were BrdU negative with high levels of Cyclin B and located in the lateral epidermis ( between dashed lines ) . ( F ) In HisC mutant embryos , lateral and dorsal cells re-accumulated Cyclin B and were labelled for BrdU ( above dashed line ) , indicating that mutant cells still replicated DNA during S15 . ( G and H ) Punctate pattern of BrdU incorporation in cells that progressed into late S15 in the dorsal epidermis ( G ) and uniform incorporation pattern in cells of the ventral epidermis ( H ) . ( I–K ) In wild type embryos replicating cells in early ( ventral , below dashed line ) and late S15 ( dorsal , above dashed line ) were detected after a 5 min BrdU labelling pulse . Some patches of dorsal cells completed S15 and did not incorporate BrdU ( encircled ) . ( L–N ) In HisC mutant embryos BrdU was detected after a 5 min BrdU labelling pulse only in ventral cells that were in early S phase as indicated by low levels of Cyclin B ( arrowheads ) . Dorsal up ( E–N ) , scale bars: 100 µm ( C and D ) , 10 µm ( E–N ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 00510 . 7554/eLife . 02443 . 006Figure 2—figure supplement 1 . The cell cycle arrest of HisC mutant cells is before mitosis . ( A–H ) Immunofluorescent staining with antibodies against Cyclin A ( A and B ) and β-tubulin ( C–H ) . ( A ) Wild type embryos degraded Cyclin A in the dorsal epidermis during M15 ( below dashed line ) . ( B ) HisC mutant embryos failed to degrade Cyclin A . ( C–H ) Magnifications of epidermal cells from embryos . ( C–E ) Wild type embryos show mitotic spindles in ventral epidermal cells at M14 and dorsal epidermal cells in M15 as visualised by β-tubulin staining ( arrowheads , green in merge ) . ( F–H ) HisC mutant embryos show only mitotic spindles in the ventral epidermis during M14 ( arrowheads ) but not in the dorsal epidermis . Dorsal up in ( C–H ) , scale bars: 100 µm ( A and B ) , 10 µm ( C–H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 006 It was previously shown that depletion of histones in cultured cells leads to a slow down of replication fork movement and it was proposed that this mechanism might avoid chromatin assembly defects due to short-term fluctuations in histone availability ( Mejlvang et al . , 2014 ) . Thus , we asked whether S phase could be faithfully completed under diminished but constant histone supply and whether there is a direct dose dependent relation between histone synthesis and the length of S phase . We reintroduced defined numbers of histone gene units ( His-GUs ) into the genome of HisC mutant embryos . Embryos carrying either two ( 2xHis-GU ) or six ( 6xHis-GU ) units completed cell cycle 15 as shown by the degradation of Cyclin B in M15 ( Figure 3A , B ) . These embryos were not rescued and they died later towards the end of embryogenesis ( Figure 3—figure supplements 1 and 2 ) . Compared to wild type , the onset of M15 was delayed in embryos with 2xHis-GUs and 6xHis-GUs by 2 hr and 1 hr , respectively , which was due to an extended S phase 15 ( Figure 3A–E , Figure 3—figure supplements 1 and 2 ) . In contrast , embryos containing 12xHis-GUs were fully rescued and entered M15 at the same time as the wild type cells , that is , 4 . 5–5 hr AEL as shown previously ( Günesdogan et al . , 2010 ) . These results establish that the length of S phase directly correlates with the transgene-derived de novo histone supply . In addition , our data indicate that the mechanisms adjusting replication fork movement to the available histone supply allow the completion of S phase under conditions of permanently diminished new histone supply in vivo ( Figure 3F ) . 10 . 7554/eLife . 02443 . 007Figure 3 . Histone availability determines the rate of S phase progression but is not required for completion of DNA replication . ( A–D ) Immunofluorescent staining with antibodies against Cyclin B . ( A ) Cyclin B degradation at 6 . 5–7 hr AEL around the tracheal pits in 2xHis-GU embryos during M15 ( arrowheads ) shows that these embryos are lagging more than one cell cycle behind as compared to wild type ( see also Figure 3—figure supplement 1 ) . ( B ) Cyclin B degradation in dorsal cells of 6xHis-GU embryos at 5 . 5–6 hr AEL during M15 ( arrowheads ) , showing that 6xHis-GU reduced the delay in cell cycle progression compared to 2xHis-GUs ( see also Figure 3—figure supplement 2 ) . ( C ) Cyclin B degradation in dorsal cells of wild type embryos at 4 . 5–5 hr AEL during M15 ( arrowheads ) . ( D ) Cyclin B degradation at 5 . 5–6 hr AEL around the tracheal pits in wild type embryos during M16 ( arrowheads ) . ( E ) BrdU pulse labelling for 15 min of 2xHis-GU and 6xHis-GU embryos at indicated developmental stages and staining with antibodies against BrdU and Cyclin B . Shown is the quantification of embryos that completed S15 ( ‘S15 exit’ , based on lack of BrdU labelling ) and progressed into M15 ( ‘M15 entry’ , based on Cyclin B degradation ) , showing the interdependence of the number of histone genes and cell cycle progression . n: number of embryos . ( F ) Schematic model showing nucleosome assembly at the replication fork and its dependence on histone supply . ( G ) DNA quantification of single nuclei stained with DAPI . Wild type nuclei in G215 and early S16 defined 4N and 2N DNA content , respectively . HisC mutant nuclei show mean intensity value of 4N nuclei , suggesting that mutant cells completed genome duplication . n: number of nuclei , p: probability from Student's t test , n . s . : not significant . Scale bars: 100 µm ( A–D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 00710 . 7554/eLife . 02443 . 008Figure 3—figure supplement 1 . Cell cycle progression of HisC mutant embryos with two His-GUs ( 2xHis-GUs ) . Shown are representative time-matched ( as indicated on the left ) embryos that were incubated with BrdU for 15′ immediately before fixation and staining with antibodies against Cyclin B and BrdU , respectively . BrdU incorporation visualises DNA replication and Cyclin B degradation is a marker for mitosis ( Figure 1A ) . Columns 1–2 show the Cyclin B and BrdU staining in whole mount embryos . Columns 3–5 are magnifications of an epidermal region with single channel detections of Cyclin B ( CycB ) , BrdU and the merge of both channels ( BrdU in green , Cyclin B in red ) . Magnifications are oriented with dorsal side up . We previously described experiments showing the stereotyped cell cycle progression in wild type embryos ( Günesdogan et al . , 2010 ) . AEL: after egg laying , scale bars: 100 µm in columns 1–2 , 10 µm in columns 3–5 . 3 . 5–4 hr AEL: ( A ) dorsal epidermal cells degraded Cyclin B , completed M14 and entered S15 as shown by the incorporation of BrdU ( arrowheads ) . ( B ) Ventral cells underwent M14 and entered S15 ( arrowheads ) . Accordingly , all cells re-accumulated Cyclin B and incorporated BrdU . 4 . 5–5 hr AEL: ( C ) all epidermal cells undergo S15 and thus show Cyclin B staining and incorporation of BrdU . At this stage , dorsal epidermal cells of wild type embryos already entered M15/S16 ( Figure 2C ) . 5 . 5–6 hr AEL: ( D–F ) dorsal epidermal cells stopped incorporating BrdU and completed S15 ( arrowheads ) . 6 . 5–7 hr AEL: ( G and H ) dorsal epidermal cells degraded Cyclin B and entered M15 ( arrowheads ) . Some cells already went through M15 and showed BrdU incorporation , indicating that they had entered S16 ( arrowheads , G ) . The division pattern in wild type embryos is highly stereotyped and controlled by developmental genes . In wild type embryos , cells around the tracheal pits , which are structures in the posterior of each parasegment , enter M16 first ( Figure 2D ) . In 2xHis-GU embryos , M15 took place at a developmental stage , when wild type embryos progressed through or completed M16 in the epidermis . Interestingly , the M15 division pattern in 2xHis-GU embryos ( H ) resembled the M16 division pattern in wild type embryos ( Figure 2D ) , suggesting that the developmental program proceeds normally in 2xHis-GU embryos . 7 . 5–8 hr AEL: ( I ) after completion of M15 , cells entered S16 and showed BrdU incorporation ( arrowheads ) . ( J and K ) During and after germ band retraction , most epidermal cells did not incorporate BrdU , indicating that they aborted S16 . Wild type embryos at this developmental stage stop to proliferate after M16 by entering a G1/0 phase ( Knoblich et al . , 1994 ) . 2xHis-GU embryos apparently also enter a G1/0 phase , enforced by the ongoing developmental program . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 00810 . 7554/eLife . 02443 . 009Figure 3—figure supplement 2 . Cell cycle progression of HisC mutant embryos with six His-GUs ( 6xHis-GUs ) . Shown are representative time-matched ( as indicated on the left ) embryos that were incubated with BrdU for 15′ immediately before fixation and staining with antibodies against Cyclin B and BrdU , respectively . BrdU incorporation visualises DNA replication and Cyclin B degradation is a marker for mitosis ( Figure 1A ) . Columns 1–2 show the Cyclin B and BrdU staining pattern in whole mount embryos . Columns 3–5 are magnifications of an epidermal region with single channel detections of Cyclin B ( CycB ) , BrdU and the merge of both channels ( BrdU in green , Cyclin B in red ) . Magnifications are oriented with dorsal side up . We previously described experiments showing the stereotyped cell cycle progression in wild type embryos ( Günesdogan et al . , 2010 ) . AEL: after egg laying , scale bars: 100 µm in columns 1–2 , 10 µm in columns 3–5 . 3 . 5–4 hr AEL: ( A ) dorsal epidermal cells degraded Cyclin B , completed M14 , and entered S15 as shown by the incorporation of BrdU ( arrowheads ) . ( B ) Ventral cells underwent M14 and entered S15 ( arrowheads ) . Accordingly , all cells re-accumulated Cyclin B and incorporated BrdU . 4 . 5–5 hr AEL: ( C and D ) dorsal epidermal cells stopped incorporating BrdU and completed S15 ( arrowheads , C ) . A few of these cells degraded Cyclin B and entered M15 ( arrowheads , D ) . Thus , the delay during DNA replication in 6xHis-GU embryos is reduced as compared to 2xHis-GU embryos , which completed S15 at 5 . 5–6 hr AEL ( Figure 2—figure supplement 1D–F ) . 5 . 5–6 hr AEL: ( E and F ) dorsal epidermal cells degraded Cyclin B , completed M15 and some cells already entered S16 as shown by BrdU incorporation ( arrowheads ) . 6 . 5–7 hr AEL: ( G–I ) dorsal epidermal cells incorporated BrdU in S16 ( arrowheads , G ) . Ventral epidermal cells degraded Cyclin B and entered M15 ( arrowheads , H ) , followed by incorporation of BrdU in S16 ( arrowheads , I ) . 7 . 5–8 hr AEL: ( J–L ) some dorsal epidermal cells degraded Cyclin B and went through M16 ( arrowheads , J ) . During and after germ band retraction ( K and L ) , most cells did not reaccumulate Cyclin B or incorporated BrdU , suggesting that they stopped proliferating , which is similar to wild type at this developmental stage ( Günesdogan et al . , 2010 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 009 Several studies showed that mammalian tissue culture cells depleted for CAF-1 accumulate in S phase due to a decreased rate of DNA replication ( Hoek and Stillman , 2003; Ye et al . , 2003; Nabatiyan and Krude , 2004; Takami et al . , 2007 ) followed by accumulation of DNA damage and activation of the conventional DNA damage checkpoints ( Hoek and Stillman , 2003; Ye et al . , 2003 ) . However , it remained unclear whether this S phase arrest represents a direct consequence of a failure in chromatin assembly as yeast cells , for example , can complete one round of replication after the depletion of histone H4 ( Kim et al . , 1988 ) . In order to address whether DNA replication can be completed in the absence of de novo histone synthesis , we quantified the amount of nuclear DNA in DAPI-stained HisC mutant cells and compared it with wild type control cells during early S16 ( 2N ) and G215 ( 4N ) , respectively ( Figure 3G ) . The DNA content of HisC mutant cells at 6 . 5–7 . 5 hr AEL corresponded to the 4N value of wild type nuclei in G2 . This finding indicates that DNA replication in HisC mutant cells has essentially been completed . To further explore whether HisC mutant cells accumulate DNA damage and activate the DNA damage checkpoints , we stained for the phosphorylated histone variant H2Av ( γH2Av ) , which is like its vertebrate ortholog γH2AX a marker of DNA damage ( Madigan et al . , 2002 ) . We found that γH2Av can be induced by ionizing irradiation that causes double strand breaks ( DSBs ) both in wild type and HisC mutant embryos , showing that the ATM/Chk2 checkpoint mechanism is still functional in HisC mutant cells ( Figure 4A , B ) . We did not observe a difference with respect to γH2Av staining between non-irradiated HisC mutant and wild type cells in early S15 ( Figure 4C , D ) . However , we noted a slight increase in γH2Av staining betwen early and late S15 in HisC mutant embryos ( Figure 4—figure supplement 1 ) . To further investigate this increase , we performed Western blot analysis for γH2Av . Drosophila S2R+ cells showed a dramatic increase in γH2Av upon treatment with hyrdoxyurea ( HU ) ( Figure 4E , Figure 4—figure supplement 1 and see below ) , which was not detectable in HisC mutant embryos undergoing late S15 at 5 . 5–6 . 5 hr AEL when compared to wild type sibling embryos ( Figure 4E ) . 10 . 7554/eLife . 02443 . 010Figure 4 . The cell cycle arrest of HisC mutant cells does not depend on the ATM/Chk2 DNA damage checkpoint . ( A and B ) Wild type and HisC mutant cells responded to ionizing irradiation ( IR ) by phosphorylation of the variant histone H2Av , detected by a phosphospecific antibody against H2Av ( γH2Av ) . ( C and D ) Without irradiation , HisC mutant cells did not show elevated γH2Av staining compared to wild type , indicating that mutant cells did not accumulate DNA damage . ( E ) Western blot for γH2Av using untreated ( w/o HU ) and HU-treated ( HU ) S2R+ cells as controls ( two dilutions , 1 and 0 . 5 ) as well as HisC mutant embryos and wild type embryos at 5 . 5–6 . 5 hr AEL . α-Tubulin ( α-Tub ) was used as a loading control . HU treatment results in a significant increase of γH2Av , which was not observed in HisC mutant embryos . ( F–H ) BrdU pulse labelling for 15 min and staining with antibodies against BrdU and Cyclin B . lok , HisC double mutant embryos showed a similar phenotype as HisC mutant embryos ( see Figure 1F ) . Dorsal up in ( F–H ) , scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 01010 . 7554/eLife . 02443 . 011Figure 4—figure supplement 1 . HisC mutant embryos show a moderate increase of γH2Av during S15 progression . HisC mutant ( A–E ) and wild type sibling ( F–J ) embryos stained with antibodies against phosphorylated H2Av ( γH2Av ) , against β-Galactosidase ( lacZ ) and with DAPI to detect DNA . ( A–C ) HisC mutant embryos were identified by the lack of lacZ staining . ( D and E ) Late S15 cells in the dorsal epidermis of HisC mutant embryos show a moderate increase of γH2Av as compared to ventral cells that are in early S15 or still in M14 ( blue arrowhead ) . ( F–H ) Wild type sibling embryos of comparable age were identified based on lacZ staining . ( I and J ) In these embryos dorsal epidermal cells were already in S/G216 . The increase in γH2Av staining was less pronounced than in HisC mutant embryos . ( K–P ) Untreated ( w/o HU; K–M ) and HU-treated ( N–P ) S2R+ cells were stained for γH2Av and DNA showing an increase of γH2Av signal in HU-treated cells . Dorsal up in ( A–J ) , anterior to the left in ( A–C , F–H ) , scale bars: 100 µm ( A–C , F–H ) , 10 µm ( D , E , I , J ) , 20 µm ( K and N ) and 5 µm ( L , M , O , P ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 01110 . 7554/eLife . 02443 . 012Figure 4—figure supplement 2 . HisC mutant embryos do not show accumulation of DNA damage during early embryogenesis . ( A–H ) TUNEL assays with whole mount embryos to detect free 3′OH groups of DNA , which indicate DNA damage . ( A–C ) Wild type embryos at 4 . 5–5 hr AEL contained a few apoptotic cells ( arrowheads ) . In most nuclei , free 3′OH groups were not detected . ( D ) Whole mount wild type embryos at 9 . 5–10 hr AEL showed a few apoptotic cells . ( E–G ) HisC mutant embryos at 6 . 5–7 hr AEL also contained a few apoptotic cells ( arrowheads ) . However , the level of TUNEL staining was not uniformly elevated in the mutant cells , indicating that there is no increase of DNA damage compared to wild type . ( H ) HisC mutant embryos die around 9 . 5–10 hr AEL and show an increased number of apoptotic cells . Scale bars: 10 µm ( A–C ) , ( E–G ) , 100 µm ( D and H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 012 To independently address the extend of DNA damage accumulation in HisC mutant embryos , we used TUNEL assays which support that HisC mutant cells do not accumulate significant levels of DNA damage until much later in development ( ≥10 hr AEL ) when these embryos die ( Figure 4—figure supplement 2 ) . Finally , we performed genetic tests using mutants of loki ( lok ) , the Drosophila DNA damage checkpoint kinase chk2 ( Xu et al . , 2001 ) . Homozygous lok mutants are viable and fertile . In contrast , lok , HisC double mutant embryos exhibited the HisC phenotype ( Figure 4F–H ) . Hence , the cell cycle arrest in HisC mutant embryos is not mediated by the chk2-dependent DNA damage checkpoint pathway . Mutations in the Drosophila ortholog of Chk1 ( GRP ) show developmental defects prior to cell cycle 15 , excluding genetic experiments as we performed for lok ( Fogarty et al . , 1994; Su et al . , 1999 ) . Thus , we tested ATR/Chk1 checkpoint activation by using a phosphospecific antibody that recognizes the ATR-dependent phosphorylation of S345 in human Chk1 in response to replicative stress , for example , UV irradiation or HU treatment ( Zhao and Piwnica-Worms , 2001 ) . This antibody is expected to cross-react with Drosophila GRP due to sequence similarity and detected a single band in Western blots ( Figure 5—figure supplement 1 ) as well as a clear signal in immunofluorescence ( Figure 5—figure supplement 2 ) upon HU treatment of Drosophila S2R+ tissue culture cells . To test whether the ATR/Chk1 checkpoint is functional in HisC mutant embryos , we irradiated embryos with UV light ( 254 nm , UVC ) , which induces replication stress and replication fork uncoupling ( Byun et al . , 2005; Cimprich and Cortez , 2008 ) . Wild type embryos and HisC mutant embryos accumulated phosphorylated GRP protein ( pGRP ) in response to UVC , showing that the checkpoint response is functional in the mutant embryos ( Figure 5A–D ) . Without UVC treatment HisC mutant embryos did not display elevated pGRP levels as compared to wild type ( Figure 5E–H ) , which was also verified by Western blotting of extracts from sorted HisC mutant and wild type sibling embryos ( Figure 5I ) . In addition , treatment of HisC mutant embryos with the ATR inhibitor VE-821 ( Prevo et al . , 2012 ) or the Chk1 inhibitor CHIR-124 ( Tse et al . , 2007 ) did not result in a release of the cell cycle arrest ( Figure 5—figure supplements 3 and 4 ) . In addition to DSBs , replicative stress can induce phosphorylation of H2Av ( Figure 4—figure supplement 1K–P ) , either directly by ATR dependent phosphorylation ( Ward and Chen , 2001; Joyce et al . , 2011 ) or through interconversion of single-stranded DNA generated at stalled replication forks into DSBs ( Cimprich and Cortez , 2008 ) . Consistent with the notion that the DNA damage checkpoints are functional in HisC mutant embryos we found accumulation of γH2Av in UVC-treated embryos to levels well above the background levels detected in untreated HisC mutant embryos in late S15 ( Figure 5—figure supplement 5 ) . Interestingly , UVC-treated embryos were able to enter M15 while displaying levels of γH2Av comparable or above to what we observed in untreated HisC mutant embryos ( Figure 5—figure supplement 6 ) . Taken together , these results strongly suggest that HisC mutant cells complete DNA replication in S phase without inducing significant DNA damage or replication stress and that the cell cycle arrest at the G2/M transition in HisC mutant cells is not mediated by the conventional S phase checkpoints . 10 . 7554/eLife . 02443 . 013Figure 5 . HisC mutant cells do not activate the ATR/Chk1 DNA damage checkpoint . ( A–D ) Wild type and HisC mutant cells responded to UV irradiation ( UVC ) by phosphorylation of the Drosophila Chk1 ortholog GRP , detected by a phosphospecific antibody for Chk1 ( pChk1 ) . Embryos were counterstained for DNA to visualize nuclei ( B and D ) . ( E–H ) Without irradiation , HisC mutant cells did not show elevated staining for pChk1 compared to wild type , indicating that mutant cells did not activate the ATR/Chk1 checkpoint . ( I ) Western blot detecting pGRP by an antibody to phosphorylated Chk1 ( pChk1 ) and α−Tubulin ( α-Tub ) . Extracts were prepared from SR2+ tissue culture cells that were either untreated ( w/o HU ) or treated with HU ( HU ) . Embryos were either sorted wild type controls or HisC mutants . Two dilutions ( 1 and 0 . 5 ) of each sample were loaded . Dorsal up in ( A–H ) , scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 01310 . 7554/eLife . 02443 . 014Figure 5—figure supplement 1 . Detection of pGRP by a phosphospecific antibody for Chk1 by Western blotting . Western blot detecting pGRP with an antibody against phosphorylated Chk1 ( pChk1 ) and α−Tubulin ( α-Tub ) . Extracts were prepared from SR2+ tissue culture cells that were either untreated ( w/o HU ) or treated with HU ( HU ) . Serial dilutions of the respective extracts were loaded ( 100% to 3% ) . The apparent molecular weight of marker proteins is indicated to the left ( kD ) . The anti-Chk1 phospho-S345 antibody detects background bands in extracts from HU treated and untreated S2R+ cells , but only one band is induced by HU ( arrowhead ) . The molecular weight of the respective band is in the range of the expected molecular weight of GRP ( 58 kD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 01410 . 7554/eLife . 02443 . 015Figure 5—figure supplement 2 . Detection of pGRP by a phosphospecific antibody for Chk1 by immunofluorescence . S2R+ cells were either cultured in the presence of HU ( HU ) or without HU ( w/o HU ) and stained with an a phosphospecific antibody for Chk1 ( pChk1 ) and with DAPI ( DNA ) . ( A ) HU treatment induces a clear signal that is detected in S2R+ cells . HU induces replicative stress by depletion of intracellular dNTP pools . Therefore , only cells that were in S phase during the 12 hr treatment were susceptible to HU . ( B and C ) Single cell magnifications show that the signal detected by the anti-Chk1 phospho-S345 antibody localizes to the nucleus and resembles the focal staining pattern expected for phosphorylated and active GRP . ( D ) Untreated cells show only background signal for pChk1 . ( E and F ) Single cell magnifications show that the residual background signal detected in untreated cells is mainly cytoplasmic . Scale bars: 20 µm ( A and D ) , 5 µm ( B , C , E , F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 01510 . 7554/eLife . 02443 . 016Figure 5—figure supplement 3 . Treatment of HisC mutant embryos with the ATR inhibitor VE-821 does not release the cell cycle arrest . Wild type sibling ( A–E ) and HisC mutant ( F–J ) embryos stained with DAPI ( DNA ) and antibodies against Cyclin B ( CycB ) and β-Galactosidase ( lacZ ) . ( A–C ) Wild type sibling embryos were identified by lacZ staining . ( D and E ) Cyclin B degradation indicates that treatment with the VE-821 inhibitor did not block entry into mitosis . ( F–H ) HisC mutant embryos of comparable age were identified based on the lack of lacZ staining . ( I and J ) VE-821 treatment did not result in Cyclin B degradation in HisC mutant cells . Dorsal up in ( A–J ) , anterior to the left in ( A–C , F–H ) , scale bars: 100 µm ( A–C , F–H ) and 10 µm ( D , E , I , J ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 01610 . 7554/eLife . 02443 . 017Figure 5—figure supplement 4 . Treatment of HisC mutant embryos with the Chk1 inhibitor CHIR-124 does not release the cell cycle arrest . Wild type sibling ( A–E ) and HisC mutant ( F–J ) embryos stained with DAPI ( DNA ) and antibodies against Cyclin B ( CycB ) and β-Galactosidase ( lacZ ) . ( A–C ) Wild type sibling embryos were identified by lacZ staining . ( D and E ) Cyclin B degradation indicates that treatment with the CHIR-124 inhibitor did not block entry into mitosis . ( F–H ) HisC mutant embryos of comparable age were identified based on the lack of lacZ staining . ( I and J ) CHIR-124 treatment did not result in Cyclin B degradation in HisC mutant cells . Dorsal up in ( A–J ) , anterior to the left in ( A–C , F–H ) , scale bars: 100 µm ( A–C , F–H ) and 10 µm ( D , E , I , J ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 01710 . 7554/eLife . 02443 . 018Figure 5—figure supplement 5 . UV irradiation induces phosphorylation of H2Av . Wild type sibling ( A , B , E , F ) and HisC mutant ( C , D , G , H ) embryos stained with DAPI ( DNA ) and antibodies against phosphorylated H2Av ( γH2Av ) . ( A–D ) Wild type and HisC mutant cells responded to UV irradiation ( UVC ) by γH2Av . The lack of γH2Av in ventral cells that were still in G214 or in M14 indicates that the UVC treatment primarily affected cells that were in S15 at the time of irradiation or entered S15 during the recovery period after treatment ( 45 min ) . Notably , dorsal epidermal cells were either blocked or delayed in entry to M15 based on the absence of mitotic DNA structures . ( E–H ) Without irradiation , HisC mutant cells showed slightly elevated levels of γH2Av staining compared to cells from wild type siblings . Dorsal up in ( A–H ) , scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 01810 . 7554/eLife . 02443 . 019Figure 5—figure supplement 6 . Mitotic entry with elevated γH2Av levels . Wild type sibling ( A–C , G–I ) and HisC mutant ( D–F ) embryos stained with DAPI ( DNA ) and antibodies against phosphorylated H2Av ( γH2Av ) . ( A–C ) Wild type cells that were in S16/G216 display background levels of γH2Av staining . ( D–F ) Compared to wild type cells , HisC mutant cells in late S15 show increased γH2Av staining . ( G–I ) UVC-irradiated wild type cells show increased γH2Av staining . These irradiated cells entered into M15 ( blue arrowheads point at prophase structures , red arrowheads point at metaphase structures ) , indicating that they repaired DNA damage and were no longer arrested by the DNA damage checkpoints . The elevated γH2Av staining in HisC mutant cells compared to wild type cells is therefore unlikely to reflect significant DNA damage . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 019 Progression from the G2 phase into mitosis critically depends on the dephosphorylation and activation of Cyclin/Cdk complexes , which is accomplished by a single gene in Drosophila embryos , encoding the CDC25 phosphatase String ( Edgar and O'Farrell , 1990 ) . In wild type , string mRNA accumulates during G2 phase and becomes rapidly degraded after cells exit mitosis ( Edgar et al . , 1994 ) . string transcription is highly dynamic , dictates the pattern of cell divisions during embryogenesis and is controlled by the activity of developmentally regulated transcription factors binding to cis-regulatory sequences spread over >30 kb of the string locus ( Edgar et al . , 1994 ) . In contrast to wild type embryos which accumulate string mRNA in G215 cells of the dorsal epidermis , HisC mutant embryos failed to accumulate string in the corresponding cells although they showed normal upregulation of string expression in ventral epidermal cells prior to M14 ( Figure 6A , B ) . Given the complex developmental regulation of string , we asked whether string transcription is disturbed in HisC mutant embryos due to misregulation of key patterning genes . However , we observed normal temporal and spatial expression patterns of developmental genes such as the segmentation gene engrailed and the homeotic genes Ultrabithorax and Abdominal-B in HisC mutant embryos ( Figure 6—figure supplement 1 ) . Hence , the developmental programme progresses normally in HisC mutant embryos up to the late embryonic stage when they eventually die ( Figure 4—figure supplement 2 ) . Similar to HisC mutant embryos , string mRNA was not detectable in dorsal epidermal cells of 2xHis-GU embryos during their extended S15 ( Figure 6C ) . However , when wild type embryos undergo M16 at 6 . 5–7 hr AEL ( Figure 6D , E ) , HisC mutant embryos still failed to express string ( Figure 6F , G ) but 2xHis-GU embryos upregulated string expression ( Figure 6H , I ) . This result suggests that the failure of HisC mutant embryos to upregulate string after they finished replication in S15 is not simply a consequence of their extended S phase but rather due to a surveillance mechanism that blocks the G2/M transition because chromatin assembly is not completed . It is interesting to note that the pattern of M15 in 2xHis-GU embryos closely resembled the pattern of M16 in wild type embryos ( Günesdogan et al . , 2010; Figure 6D , H ) . This observation provides further support that the developmental programme of these embryos can progress normally , as string mRNA expression readjusts once DNA replication is completed . Thus , sufficient histone supply , as provided by multiple copies of His-GUs , is critical to the coordination of the developmental and cell division programmes during wild type embryogenesis of Drosophila . The results also explain why higher eukaryotes , which undergo rapid mitotic cell divisions during embryonic development , contain multiple His-GUs in their genomes . 10 . 7554/eLife . 02443 . 020Figure 6 . The cell cycle arrest of HisC mutant cells depends on string . ( A–I ) string RNA in situ hybridisation ( green in merge ) and staining with an antibody against Cyclin B ( red in merge ) . ( A–C ) Magnifications of epidermal cells from embryos at 4 . 5–5 hr AEL . ( A ) Wild type embryos upregulated string mRNA and accumulated Cyclin B in ventral epidermal cells during G214 ( below dashed lines ) and in dorsal epidermal cells during G215 ( above dashed lines ) . In lateral cells with low levels of Cyclin B , string is not expressed during S15 ( between dashed lines ) . ( B ) HisC mutant embryos failed to upregulate string in the dorsal epidermis ( above dashed line ) . ( C ) 2xHis-GU embryos failed to upregulate string in the dorsal epidermis ( above dashed line ) . ( D–I ) Whole mount embryos at 6 . 5–7 hr AEL . ( D and E ) Dorsal cells of wild type embryos progressed into G216 , upregulated string mRNA and degraded Cyclin B during M16 ( arrowheads ) . ( F and G ) HisC mutant embryos did not degrade Cyclin B and failed to accumulate string mRNA . ( H and I ) 2xHis-GU embryos accumulated string mRNA in G215 and degraded Cyclin B in M15 ( arrowheads ) . ( J–P ) HisC mutant embryos expressing UAS-EYFP and UAS-string under the control of prd-GAL4 stained with antibodies against Cyclin B and EYFP . ( M–P ) Magnifications of epidermal cells from an embryo ( boxed in J ) . ( J–P ) Epidermal cells within the EYFP , string expression domains degraded Cyclin B and entered M15/S16 . Arrowheads in ( O ) show mitotic cells . ( Q ) Mitotic progression in wild type M15 and M15 of HisC mutant cells rescued by string expression visualized by staining for α-Tubulin ( Tub ) , DNA , and Cyclin B ( CycB ) . M: metaphase , Aa: anaphase-a , Ab: anaphase-b , T: telophase , C: cytokinesis , I: interphase . Dorsal up ( A–C , M–P ) , scale bars 10 µm ( A–C , M–P ) , 100 µm ( D–L ) , 5 µm ( Q ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 02010 . 7554/eLife . 02443 . 021Figure 6—figure supplement 1 . The expression pattern of developmental genes is normal in HisC mutant embryos . ( A–L ) Fluorescent in situ hybridisations of whole mount embryos using probes for the segment polarity class gene engrailed ( A–D ) , the homeotic genes Abdominal-B ( E–H ) and Ultrabithorax ( I–L ) . ( A–D ) At embryonic stage 9 , engrailed is expressed in 14 parasegmental stripes ( 1–14 ) in the trunk and in additional expression domains in the head ( A ) . The same pattern was detected in HisC mutant embryos ( B ) . In embryonic stage 10 , engrailed is detectable in 14 parasegmental stripes in the trunk of wild type ( C ) and HisC mutant embryos ( D ) . ( E–H ) Abdominal-B is expressed in the most posterior parasegments 10–14 at embryonic stage 10–11 ( E ) . The same expression pattern was observed in HisC mutant embryos ( F ) . During germ band retraction at stage 12 of embryogenesis , the expression patterns are still very similar between wild type ( G ) and mutant embryos ( H ) . ( I–L ) Ultrabithorax is expressed in parasegments 5–12 at embryonic stage 9 ( I ) . The same expression pattern was observed in HisC mutant embryos at stage 9 ( J ) . Also , at stage 10–11 of embryogenesis , wild type ( K ) and mutant embryos ( L ) show a highly similar expression pattern . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02443 . 021 If string activity is the only limiting factor that restricts cell cycle progression in HisC mutant cells , its ectopic expression should drive the G2 to M transition . To test this hypothesis , HisC mutant embryos were forced to express string in a striped pattern under the control of the prd-GAL4 driver using the GAL4/UAS system ( Brand and Perrimon , 1993 ) , which was visualised by coexpression of a UAS-EYFP transgene . The string expressing HisC mutant cells entered M15 , degraded Cyclin B and reaccumulated Cyclin B after mitosis , whereas cells lacking string expression remained arrested ( Figure 6J–P ) . Mitotic progression in wild type embryos follows a stereotyped pattern characterized by ( i ) Cyclin B degradation and sister chromatid separation at the metaphase to anaphase transition , ( ii ) spindle elongation at the transition from anaphase-a into anaphase-b , ( iii ) the onset of chromatin decondensation in telophase , and eventually by ( iv ) cytokinesis ( Figure 6Q ) . The string-induced mitosis in HisC mutant embryos was normal up to the metaphase to anaphase transition . During anaphase-b and telophase , however , we observed lagging chromosomes or chromatin bridges in almost all of the cells ( 95 . 8% , n = 24 ) . These bridges were eventually resolved during cytokinesis , and the cells entered into interphase of cell cycle 16 ( Figure 6Q ) . Together , these data indicate that string transcription is indeed the limiting downstream factor that restricts cell cycle progression in the absence of de novo histone synthesis . In summary , our study shows that DNA replication and histone availability are tightly coupled . Lack of de novo histone synthesis causes a string-dependent cell cycle arrest in G2 phase , suggesting a novel chromatin assembly checkpoint monitoring chromatin integrity . We used a recently generated null mutation for canonical histones to address the consequences of histone deprivation during metazoan development . In addition to canonical histones , eukaryotes express histone variants that can replace canonical histones in a specific genomic context ( Banaszynski et al . , 2010 ) . Our results show that these histone variants do not compensate for the lack of canonical histone synthesis with regard to chromatin assembly and cell cycle progression . This could be due to insufficient expression of variant histones from their endogenous promoters as it has been shown for the variant histone H3 . 3 , which can fully replace its canonical counterpart , histone H3 , but only if it is expressed from within a histone gene unit like the canonical histone ( Hödl and Basler , 2012 ) . Alternatively , it could reflect structural divergence of the histone variants as in the case of His2Av ( van Daal et al . , 1988 ) and dBigH1 ( Perez-Montero et al . , 2013 ) . It will be interesting to test whether individual histone mutations , like a mutation in H2B which does not have a variant histone in Drosophila ( Talbert et al . , 2012 ) , will cause a similar cell cycle arrest as the histone null mutation HisC . Our results provide evidence that canonical histone supply directly affects the rate of DNA synthesis ( Figure 2L–N ) . This observation is in line with studies that targeted either histone chaperones ( Hoek and Stillman , 2003; Ye et al . , 2003; Nabatiyan and Krude , 2004; Groth et al . , 2007; Takami et al . , 2007 ) or histone mRNA through SLBP or FLASH ( Zhao et al . , 2004; Barcaroli et al . , 2006; Mejlvang et al . , 2014 ) to interfere with chromatin assembly in tissue culture cells . However , previous work on SLBP in multicellular organisms revealed pleiotropic effects ( Sullivan et al . , 2001; Lanzotti et al . , 2002; Pettitt et al . , 2002 ) . Our data illustrate that an extension of the S phase duration caused by diminished histone supply allows a faithful completion of S phase and transition from G2 into M phase of the cell cycle . This S phase extension is likely to be caused by a direct effect of lowered histone availability on replication fork progression ( Groth et al . , 2007; Mejlvang et al . , 2014 ) and not by a lack of origin firing , although we cannot exclude this possibility completely . It was previously shown that postblastodermal development in Drosophila embryos proceeds largely uncoupled from progression through cell cycles 14–16 ( Edgar et al . , 1994; Meyer et al . , 2002 ) . Therefore , histone availability limits S phase duration and appears to be a critical link between cell division and development . In the absence of de novo histone synthesis , we find that cells arrest in G2 phase of the cell cycle without activating the known ATM/Chk2 and ATR/Chk1 checkpoints . This observation is in contrast to previous studies on CAF-1 , which found that cells arrest in S phase and accumulate DNA damage ( Hoek and Stillman , 2003; Ye et al . , 2003 ) . This discrepancy might in part be explained by the fact that histone chaperones also have a direct function in DNA repair ( Schöpf et al . , 2012 ) ; and thus , in the presence of an intact DNA repair/chromatin assembly machinery in HisC mutants , DNA is replicated without the accumulation of damage , even when histone supply is restricted to the parental load of histones . Alternatively , the accumulation of DNA damage in histone chaperone-depleted cells might be the consequence of a prolonged replication slow down , since it was shown that neither ATM/Chk2 nor ATR/Chk1 are activated as an immediate consequence of histone deprivation but only after prolonged incubation times ( >48 hr ) ( Mejlvang et al . , 2014 ) . Based on our DNA quantification experiments , we found that the bulk of DNA replication in HisC mutants is completed by about 2 hr after entry into S phase , which might differ from the timeframe required to develop significant DNA damage . Interestingly , we find that HisC mutant cells become TUNEL positive by about 6 hr after they enter S phase 15 , which might reflect secondary DNA damage and/or cell death . Nevertheless , we found a moderate increase of γH2Av staining during late S phase in HisC mutant embryos . Our data indicate , however , that cells that resolved UVC-induced DNA damage , and therefore entered mitosis can do so with levels of γH2Av comparable to those we observe in HisC mutants . Thus , it is plausible that the slight increase in γH2Av in HisC mutants could result from incomplete turnover of γH2Av rather than directly reflect DNA damage that could activate the S phase checkpoints . Turnover of γH2Av was shown to require the Tip60 chromatin-remodelling complex ( Kusch et al . , 2004 ) , which may be affected by the altered chromatin structure in HisC mutants . Alternatively , H2Av was recently shown to be phosphorylated independent of ATM/ATR by the chromosomal tandem kinase JIL-1 ( Jin et al . , 1999; Thomas et al . , 2014 ) , which may also be influenced by the changed chromatin topology in HisC mutants . Both , the ATM/Chk2 and ATR/Chk1 checkpoints are known to act on CDC25 phosphatases by phosphorylation and protein destabilization ( Bartek and Lukas , 2007 ) and it was shown in Drosophila that string transcripts accumulate normally in embryos that suffered from DNA damage ( Su et al . , 2000 ) . In contrast , we find that HisC mutant cells fail to accumulate string transcripts when arrested in G2 . This finding was surprising since it was shown that the temporal and spatial expression pattern of string is essentially unchanged in embryos that are arrested in G2 by mutations in string or in mitotic Cyclins ( Edgar et al . , 1994 ) . Thus , this difference is likely due to the failure of HisC mutant embryos to assemble chromatin , resulting in a diminished nucleosome density as shown by the presence of excess MNase hypersensitive DNA . Although we cannot rule out that the lower abundance of histone proteins itself directly contributes to the G2 arrest , this possibility seems unlikely since histone levels rapidly decrease in G2 cells where the chromatin assembly surveillance should act ( Marzluff et al . , 2008 ) . It remains unclear how the presence of unassembled chromatin is linked to the regulation of string , but the effect is specific , since string transcript accumulation is the only limiting factor to overcome the G2 arrest in HisC-mutant embryos . The subsequent mitosis in HisC mutants is completed and cells enter into the next cell cycle . Given that HisC mutant cells enter mitosis with presumably about half of the nucleosomes present in wild type chromatin , the mitotic defects like lagging anaphase chromosomes appear surprisingly mild . These defects could reflect problems in loading of structural components that are required for chromosome condensation and sister chromatid cohesion , like Cohesins and Condensins , which are proposed to require contact to chromatin rather than naked DNA ( Bernard et al . , 2001; Nonaka et al . , 2002; Tada et al . , 2011 ) . Taken together , our results suggest that incomplete chromatin assembly is monitored by a novel surveillance mechanism that can block cell cycle progression at the G2/M transition in Drosophila . Our findings now pave the way to address key questions regarding the orchestration of DNA synthesis and chromatin formation as well as the control of chromatin integrity during cell cycle progression . Construction of Df ( 2L ) HisC and histone transgenes was described previously ( Günesdogan et al . , 2010 ) . The lokP6 Df ( 2L ) HisC double mutant was constructed from lokP6 ( Abdu et al . , 2002 ) by meiotic recombination . In order to obtain maternal and zygotic loki mutant embryos , lokP6 Df ( 2L ) HisC/lokP6 animals were crossed inter se . For string expression , we constructed Df ( 2L ) HisC , P{UAS-2xEYFP}AH2 by meiotic recombination and generated embryos of the genotype Df ( 2L ) HisC , P{UAS-2xEYFP}AH2/Df ( 2L ) HisC , P{GAL4-prd . F}RG1/P{UAS-stg . N}4 . For sorting of mutant embryos , we constructed Df ( 2L ) HisC , P{GAL4-twi . 2xPE}2 and selected EYFP-expressing embryos of the genotype Df ( 2L ) HisC , P{UAS-2xEYFP}AH2/Df ( 2L ) HisC , P{GAL4-twi . 2xPE}2 . In all other fly strains , Df ( 2L ) HisC was heterozygous over the balancer chromosome CyO , P{ftz/lacB}E3 to identify wild type sibling embryos . 6xHis-GU flies were homozygous for the M{3xHisGU . wt}ZH-86Fb transgene . 2xHis-GU flies were homozygous for the M{1xHisGU . wt}ZH-86Fb transgene ( Günesdogan et al . , 2010 ) . M{1xHisGU . wt}ZH-86Fb was constructed analogous to M{3xHisGU . wt}ZH-86Fb except that pENTRL4R1-T1 and pENTRR2L3-T2 that both contained ∼500 bp random DNA sequences instead of the His-GU were used for transgene construction . Time matched embryonic collections were obtained by restricting egg deposition to 30′ and subsequent aging of embryos at 25°C . Fixation , antibody staining , BrdU incorporation , and RNA in situ hybridisation procedures were described previously ( Günesdogan et al . , 2010 ) . Different from our standard protocol , we reduced the fixation time to 5 min and used 37% PFA instead of 4% PFA to fix embryos after UVC treatment . For experiments that required visualization of the tubulin cytoskeleton , we fixed as described ( Karr and Alberts , 1986 ) . Tissue culture cells were grown in µ-slide eight-well ibidi dishes ( IBIDI , Martinsried , Germany ) and fixed and treated in these wells by the same procedures as embryos ( Günesdogan et al . , 2010 ) . Primary antibodies used here were: rabbit anti-Cyclin B ( Jacobs et al . , 1998 ) ( 1: 3000 ) , chicken anti-β-Galactosidase ( 1:1000; Abcam , Cambridge , UK ) , H2Av pS137 ( γH2Av; 1:500; Rockland , Gilbertsville , PA ) , rabbit anti-Chk1 phospho-S345 ( 1:300; Abcam ) , mouse anti-α Tubulin DM1A ( 1:500; Sigma-Aldrich , Taufkirchen , Germany ) , chicken anti-Galactosidase ( 1:1000; Abcam ) , sheep anti-Digoxigenin ( 1:2000; Roche , Mannheim , Germany ) , and a mouse anti-BrdU antibody ( 1:80; Becton Dickinson , Heidelberg , Germany ) . Secondary antibodies were: goat anti-mouse IgG coupled to Alexa Fluor488 ( 1:400; Life Technologies , Paisley , UK ) , goat anti-rabbit IgG Alexa Fluor488 ( 1:400; Life Technologies ) , goat anti-rabbit IgG Alexa Fluor633 ( 1:400; Life Technologies ) , goat anti-rabbit IgG Alexa Fluor647 ( 1:400; Life Technologies ) , goat anti-chicken IgY Alexa 568 ( 1:400; Life Technologies ) , goat anti-chicken IgY Cy3 ( 1:400; Jackson ImmunoResearch , West Grove , PA ) , donkey anti-sheep IgG Biotin-SP ( 1:500; Jackson ImmunoResearch ) . DNA was stained with DRAQ-5 ( Biostatus Limited , Shepshed , UK ) or Vectashield DAPI ( Vector Laboratories , Burlingame , CA ) . Antisense string RNA was obtained by in vitro transcription and detected as described previously ( Günesdogan et al . , 2010 ) using biotinylated secondary antibodies . For detection , embryos were incubated for 45′ with ABC reagents ( Vector Laboratories ) , followed by a 5′ incubation with TSA Flourescein reagents ( Perkin Elmer , Waltham , MA ) diluted 1:50 . 50–100 wild type or HisC mutant embryos were collected and lysed in 50 µl lysis buffer ( 50 mM Tris pH8 , 150 mM NaCl , 0 . 5% Triton X-100 , 1 mM MgCl2 , 0 . 1 mM EDTA and protease inhibitor [Roche] ) . Phosphatase inhibitor ( Roche ) was added for Western blots stained for pChk1 and γH2Av , respectively . 50 µl 4× Laemmli buffer was added and the samples were sonicated for 7 min ( 30 s interval , power ‘low’ ) using a Bioruptor ( Diagenode , Liège , Belgium ) and centrifuged for 10 min at 4°C at maximum speed . The supernatants were denatured at 95°C for 10 min before loading on BioRad 4–20% Mini-PROTEAN TGX gels . PVDF membranes ( Merck Millipore , Billerica , MA ) were used for blotting . After blotting , membranes were washed with TBS-T ( Tris Buffered Saline with 0 . 1% Tween-20 ) and blocked for 1 hr in TBS-T and 5% dry milk powder . Primary antibodies diluted in blocking buffer were added and incubated over night at 4°C . Membranes were washed 3× for 5′ in blocking buffer and secondary antibodies were added and incubated for 1 hr at room temperature . Membranes were washed 3× for 5′ in TBS-T . For fluorescent Western blots , images were acquired using an Odyssey infrared imaging system ( LI-COR Biosciences , Lincoln , NE , USA ) . Quantifications were performed using the Image Studio Software ( LI-COR Biosciences ) . For pChk1 and γH2Av Western blots , the ECL Western Blotting Kit ( Pierce ) was used to detect signals . Primary antibodies were: rabbit anti-Histone H3 ( 1:20000; Abcam ) , rabbit anti-Histone H2B ( 1:1000; Abcam ) , mouse anti-α Tubulin DM1A ( 1:2000; Sigma-Aldrich ) , rabbit anti-Chk1 phospho-S345 ( 1:1000; Abcam ) , and H2Av pS137 ( γH2Av; 1:500; Rockland ) . Secondary antibodies were: goat anti-mouse IgG IRDye 680 ( 1:5000; LI-COR Biosciences ) , goat anti-rabbit IgG IRDye 800 ( 1:5000; LI-COR Biosciences ) , goat anti-rabbit IgG horseradish peroxidase conjugated , and goat anti-mouse IgG horseradish peroxidase conjugated ( Sigma-Aldrich ) . 70–100 control or HisC mutant embryos were collected and lysed in 50 µl micrococcal nuclease ( MNase ) digestion buffer ( 15 mM Tris , pH 7 . 5; 15 mM NaCl; 60 mM KCl; 0 . 34 M sucrose; 0 . 5 mM spermidine; 0 . 15 mM spermine; 1 mM PMSF; 0 . 5 mM DTT; 0 . 1% β-mercaptoethanol , protease inhibitor [Roche] ) . 200 µl MNase digestion buffer and 1 mM of CaCl2 was added . The suspension was divided in equal aliquots . 500 gel units of MNase ( NEB , Hitchin , UK ) were added and the samples were incubated at 32°C in a heat block . The reaction was stopped by adding 5 µl of both 0 . 5 M EDTA and 0 . 5 M EGTA . 2 . 5 µl of 10% SDS and 1 µl Proteinase K ( 10 mg/ml ) was added and incubated at 50°C over night . The DNA was purified using Ampure XP beads ( Beckman Coulter , Brea , CA ) . The samples were analysed using an Agilent 2200 Tapestation system with High Sensitivity D1000 screen tapes ( Agilent Technologies , Wokingham , UK ) . Genomic DNA screen tapes ( Agilent Technologies ) were used to determine the input ( time point 0′ ) concentration . The Tapestation analysis software was used for quantification of following fractions: 60–115 bp ( <115 bp ) , 115–270 bp ( mononucleosomes ) , 270–455 bp ( dinucleosomes ) , 455–650 bp ( trinucleosomes ) . Embryos were collected on apple-juice agar plates and aged to 4–5 . 5 hr AEL ( w1118 , wild type ) or 6–7 hr AEL ( HisC mutant embryos ) . This procedure yielded wild type embryos with cells in G215 and HisC mutant embryos that were arrested in cell cycle progression . After irradiation at 60 Gray in a Torrex 150D ( Astrophysics Research Corp . , City of Industry , CA ) embryos from both collections were aged for 20 min on the apple-juice agar plates and mixed before fixation . UV irradiation ( 254 nm ) was done at 200 mJ/cm2 as described ( Zhou and Steller , 2003 ) , and embryos were aged 45 min before fixation . HisC mutant embryos were identified in the stained samples based on their cell division arrest phenotype . For inhibitor treatment , we used the same procedure as for BrdU incorporation ( Günesdogan et al . , 2010 ) , replacing the BrdU with 10 µM VE-821 ( Selleckchem , Houston , TX ) or 10 µM CHIR-124 ( Selleckchem ) and incubation of 45 min at RT before fixation . S2R+ tissue culture cell were cultured in Schneiders Medium ( Life Technologies ) , with 10% Fetal Calf Serum . Hydroxyurea ( Sigma-Aldrich ) was added to a final concentration 10mM and incubated for 12 hr . Extracts were prepared in SDS sample buffer after addition of phosphatase inhibitors ( PhosSTOP , Roche ) and protease inhibitors ( cOmplete EDTA-free , Roche ) at approximately 5 × 108 cells per ml . Embryos were fixed by heat/methanol treatment and stained for Cyclin B . DNA was stained with DAPI ( 1:1000; Life Technologies ) . Cyclin B staining was used to distinguish homozygous mutant Df ( 2L ) HisC embryos and control embryos and to define the cell cycle stage of each cell . Stacks of nuclei were acquired with a 63× objective and a 10× optical zoom with a Leica TCS-SP5 AOBS confocal laser-scanning microscope ( z-axis increment: 0 . 1 µm , 8 bit images , 512 × 512 pixel , 400 Hz scan speed ) . The gain and offset were adjusted once and then used for one complete experiment , avoiding saturation . Fluorescent measurements were carried out using ImageJ software . To define nuclear circumferences , we used the ‘isodata thresholding’ algorithm followed by manual inspection . This threshold we used in a customized macro ( Source code 1 ) that utilizes the ‘connected threshold grower’ plugin of the ImageJ 3D toolkit to determine nuclear staining intensities in all slices of the z-stack . The absolute nuclear fluorescence intensities were calculated by integration of individual nuclear fluorescence distributed over the image stack . For background detection five regions in between nuclei were analysed and their average was used for background subtraction . TUNEL assays were done as described ( Arama and Steller , 2006 ) with some deviations . DIG-labelled nucleotides were detected with a sheep anti-DIG antibody ( Roche ) and a donkey biotinylated anti-sheep secondary antibody ( Jackson ImmunoResearch ) . For signal amplification , embryos were incubated for 45′ with ABC reagents ( Vector Laboratories ) , followed by a 5′ incubation with TSA Flourescein reagents ( Perkin Elmer ) diluted 1:50 . Embryos were mounted in ProlongGold ( Life Technologies ) .	As a cell prepares to divide , it goes through four distinct stages . First , it grows in size ( G1 phase ) ; next it copies its entire DNA content ( S phase ) ; then it grows some more ( G2 phase ) ; and , last , it splits into two new cells ( M phase ) . During S phase , groups of histone proteins that normally stick together to tightly package the DNA are pulled apart in order to make the DNA accessible for copying . After the DNA has been duplicated , both copies of the DNA strand need to be repackaged . Therefore , after copying the DNA the cell rapidly reassembles the DNA–histone complexes ( called nucleosomes ) , using a combination of old and newly synthesized histones to do so . A cell can adjust how quickly it copies DNA according to the availability of these histone proteins , which is important because copying DNA without the resources to package it could expose the DNA to damage . Here , Günesdogan et al . investigate how a cell controls these processes using a mutant of the fruit fly Drosophila melanogaster that completely lacks the genes required to make histones . Cells that lack histones copy their DNA very slowly but adding copies of histone genes back into these flies speeds up the rate at which DNA is copied . Günesdogan et al . ask whether the slower speed of DNA replication in cells without new histones is connected to preventing DNA damage . However , these cells can still copy all their DNA , despite being unable to package it , so the higher risk of making mistakes is not enough to stop S phase . In fact , indications suggest that DNA damage detection methods continue to work as normal in cells without histones: these cells can get all the way to the end of G2 phase without any problems . To go one step further and start splitting in two , a cell needs to switch on another gene , called string in the fruit fly and CDC25 in vertebrates , which makes an enzyme required for the cell division process . Normal cells switch on string during G2 phase , but cells that lack histones do not—and therefore do not enter M phase . Günesdogan et al . show that turning on string by a genetic trick is sufficient to overcome this cell cycle arrest and drive the cells into M phase . String could therefore form part of a surveillance mechanism that blocks cell division if DNA–histone complexes are not assembled correctly .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "cell", "biology" ]	2014	Histone supply regulates S phase timing and cell cycle progression
Emerging evidence indicate that the mammalian checkpoint kinase ATM induces transcriptional silencing in cis to DNA double-strand breaks ( DSBs ) through a poorly understood mechanism . Here we show that in Saccharomyces cerevisiae a single DSB causes transcriptional inhibition of proximal genes independently of Tel1/ATM and Mec1/ATR . Since the DSB ends undergo nucleolytic degradation ( resection ) of their 5′-ending strands , we investigated the contribution of resection in this DSB-induced transcriptional inhibition . We discovered that resection-defective mutants fail to stop transcription around a DSB , and the extent of this failure correlates with the severity of the resection defect . Furthermore , Rad9 and generation of γH2A reduce this DSB-induced transcriptional inhibition by counteracting DSB resection . Therefore , the conversion of the DSB ends from double-stranded to single-stranded DNA , which is necessary to initiate DSB repair by homologous recombination , is responsible for loss of transcription around a DSB in S . cerevisiae . DNA double-strand breaks ( DSBs ) are particularly dangerous for cells , since their inefficient or inaccurate repair can result in deletions and chromosomal translocations that can lead to cancer and/or severe developmental abnormalities in humans . DSB formation leads to activation of a complex DNA damage response ( DDR ) , whose key players are highly conserved protein kinases , which include human ATM and ATR , as well as their Saccharomyces cerevisiae orthologs Tel1 and Mec1 ( Gobbini et al . , 2013 ) . Once activated by DSBs , ATM/Tel1 and ATR/Mec1 promote DSB repair , delay cell cycle progression or trigger the elimination of genetically unstable cells by inducing cell death . One of the main mechanisms to repair DSBs is homologous recombination ( HR ) , which requires resection of the broken ends in order to generate 3′-ended single-stranded DNA ( ssDNA ) tails that invade the homologous undamaged template . In S . cerevisiae , DSB resection is initiated by the evolutionarily conserved MRX ( Mre11-Rad50-Xrs2 ) complex that , together with Sae2 , catalyzes the initial processing of the DSB ends . The 5′-ending strands can then be further degraded by two other machineries depending on Exo1 and Sgs1-Dna2 , respectively ( Symington and Gautier , 2011 ) . In eukaryotes , the DDR proteins function in the context of a highly organized chromatin environment that needs to be overcome to gain access to damaged DNA . Histone modifications and ATP-dependent chromatin remodelling proteins help to overcome this barrier by altering chromatin structure at the site of damage ( Price and D'Andrea , 2013 ) . One of the most characterized histone modifications is the phosphorylation of histone H2AX by ATM/Tel1 and ATR/Mec1 , which spreads away from the DSB into large domains of surrounding chromatin ( Rogakou et al . , 1999; Downs et al . , 2000; Burma et al . , 2001; Ward and Chen , 2001 ) . Other chromatin alterations detected at DSBs are associated with open and decondensed chromatin , indicating that chromatin at DSBs undergoes a transition to a more open , less compact conformation ( Price and D'Andrea , 2013 ) . However , several proteins associated with repressive or transcriptionally inactive chromatin , including HP1 , PcG ( Polycomb group ) proteins , PRMD2 methyltransferase , KAP-1 , su ( var ) 3-9 methyltrasferase variant ( SUV3-9 ) and the macrohistone variant macroH2A1 , are recruited to DNA lesions ( Soria et al . , 2012 ) , suggesting that repressive chromatin can be generated around DSBs . Consistent with this hypothesis , generation of several DSBs distal to the promoter of a reporter gene in mammalian cells leads to ATM-dependent transcriptional repression of this reporter gene ( Shanbhag et al . , 2010 ) , possibly through phosphorylation of the transcriptional elongation factor ENL ( Ui et al . , 2015 ) . Similarly , RNA polymerase I-mediated transcription of rDNA is inhibited in an ATM-dependent manner in the vicinity of DSBs ( Kruhlak et al . , 2007 ) . Furthermore , the steady state RNA levels of genes proximal to a single DSB have been observed to decrease by microarray analysis also in S . cerevisiae cells ( Lee et al . , 2000 ) . However , a different study in mammalian cells , where individual DSBs were induced at discrete endogenous sequences , showed that only transcription of the DSB-containing gene was affected ( Pankotai et al . , 2012 ) . Furthermore , non-coding RNAs that control DDR activation are induced in the surrounding of DSBs in both vertebrates and Arabidopsis ( Francia et al . , 2012; Michalik et al . , 2012; Wei et al . , 2012 ) , indicating that transcription can occur around DSBs . Given these apparently contrasting results , other studies are required to understand how DSBs affect transcription in their surroundings . Furthermore , as DSBs are resected to generate 3′-ended ssDNA , and ATM , macroH2A1 , PRMD2 and HP1 proteins are required for this process ( Jazayeri et al . , 2006; Soria and Almouzni , 2013; Ayrapetov et al . , 2014; Khurana et al . , 2014 ) , whether DSB resection has a role in the DSB-induced transcriptional inhibition needs to be investigated . By using the budding yeast HO endonuclease to create single DSBs at different chromosomal loci , we show that DSB induction causes Mec1- and Tel1-independent transcriptional inhibition of genes surrounding the DSB site . Failure to resect the DSB ends prevents this transcriptional inhibition , which is instead enhanced by accelerating the resection process . Altogether , these data indicate that loss of transcription around a DSB in S . cerevisiae cells is due to the conversion of DSB ends from double-stranded DNA ( dsDNA ) to ssDNA . To investigate changes in the transcription of genes surrounding DSBs , we took advantage of a yeast haploid strain ( JKM139 ) where a single DSB can be generated at the MAT locus by galactose-induced expression of the HO endonuclease . As the homologous donor sequences HML and HMR are deleted in this strain , this DSB cannot be repaired by HR ( Lee et al . , 1998 ) . We previously used this strain for total RNA-seq analysis of protein-coding gene expression upon induction of the HO-induced DSB to show that mRNA levels of the vast majority of protein-coding genes underwent no significant change upon DSB generation ( Manfrini et al . , 2015 ) . Here we focused on the MAT locus that was not analyzed in the previous study . We mapped the RNA-seq data from two biological replicates of the JKM139 wild-type strain grown in raffinose ( time zero , T0 ) and shifted to galactose-containing medium for 60 ( T60 ) and 240 ( T240 ) min on a restricted reference ‘genome’ corresponding to the MAT locus ±10 kb ( Figure 1A ) . Quantitative analysis of tag densities for annotated genes in this region revealed a moderate signal decrease for the DSB-proximal genes 60 min after HO induction , and a strong signal reduction along the whole region after 240 min ( Figure 1B ) . Strikingly , after 240 min of HO induction , density for the proximal genes BUD5 , HMRA2 , HMRA1 and TAF2 showed a ≥10-fold decrease compared to time zero , while density was only reduced by about twofold for the distal genes SNT1 , IMG1 and BUD23 ( Figure 1C ) . 10 . 7554/eLife . 08942 . 003Figure 1 . Decrease of RNA levels around a DNA double-strand break ( DSB ) . ( A ) Strand-specific RNA-seq data from the two biological replicates of wild type strain ( JKM139 ) before ( time zero , T0 ) and 60 ( T60 ) or 240 ( T240 ) min after HO induction , were uniquely mapped to the MAT locus ±10 kb . For each time point , densities ( tag/nucleotide , log2 scale ) from the two replicates were pooled and visualized as a heatmap with the upper and lower panels corresponding to the + and − strands , respectively . Black arrows represent annotated genes . ( B ) Density for genes along the MAT locus ±10 kb at T0 , T60 and T240 after HO induction . Mean values ±s . d . were calculated from the two biological replicates analyzed . RRT12 showed too low signal to be significantly quantified and was not included . ( C ) Ratio of density for genes along the MAT locus ±10 kb at T0 , T60 and T240 after HO induction . For each time point , densities were normalized on the values obtained at T0 . Mean values ±s . d . were calculated as above . RNA-seq data used to construct the graphs of Figure 1B , C are available in Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 00310 . 7554/eLife . 08942 . 004Figure 1—source data 1 . RNA-seq data used to construct the graphs of Figure 1B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 004 Validation of the above data by using quantitative reverse transcriptase PCR ( qRT-PCR ) confirmed the reduction in transcript levels for genes located around the HO-induced DSB 240 min after HO induction ( Figure 2A ) . The level of these mRNAs decreased progressively as the distance of the corresponding genes from the DSB diminished and such decrease was independent of the strand that was transcribed ( Figure 2A ) . These decreases measured 240 min after HO induction were not influenced by mRNA stability , as all the analyzed mRNAs have half-lives shorter than 50 min ( Geisberg et al . , 2014 ) . 10 . 7554/eLife . 08942 . 005Figure 2 . DSB-induced transcriptional inhibition at the MAT locus . ( A ) YEPR exponentially growing cell cultures of the JKM139 strain , carrying the HO cut site at the MAT locus , were transferred to YEPRG at T0 to induce HO . RNA levels of genes located at different distances from the HO cut site were evaluated by quantitative reverse transcriptase PCR ( qRT-PCR ) at T0 and T240 after HO induction . Results are presented as ratios between T240 and T0 . RNA levels were quantified using ∆∆Ct method and quantities were normalized to ACT1 RNA levels . The mean values ±s . d . are represented ( n = 3 ) . ( B ) Exponentially growing YEPR cell cultures of the JKM139 derivative strains expressing either a fully functional Rpb2-HA fusion protein or untagged Rpb2 were transferred to YEPRG at T0 to induce HO . Binding of Rpb2-HA at different distance from the DSB at T0 and T240 after HO induction was evaluated by ChIP and qPCR . Primers used were the same as in ( A ) . Results are presented as ratios between Rpb2-HA and untagged Rpb2 , both of which normalized against the corresponding input , at T240 relative to T0 . The mean values ±s . d . are represented ( n = 3 ) . Rpb2-HA binding at the ACT1 gene was used as internal control . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 005 Because both qRT-PCR and RNA-seq measured the steady state level of total RNAs , which is a balance between synthesis and degradation , we investigated whether the reduction of RNA levels for genes around the DSB was due to transcriptional inhibition . To this end , the binding of the Rpb2 second largest subunit of RNA polymerase II was measured at T0 and 240 min after HO induction at different distances from the DSB . Indeed , Rpb2 association in the surroundings of the DSB decreased and this effect progressively diminished as a function of the distance from the DSB ( Figure 2B ) , indicating that transcription was inhibited specifically around the DSB . To investigate whether the DSB-induced transcriptional inhibition was specific for the MAT locus , we analyzed RNA levels and RNA polymerase II binding in tGI354 and YFP17 strains , which carried the recognition site for the HO endonuclease at the ARG5 , 6 locus on chromosome V or at the LEU2 locus on chromosome III , respectively . Since tGI354 strain can use the uncleavable MATa-inc sequence on chromosome III as a donor to repair the HO-induced DSB by Rad51-dependent HR , this strain carried the deletion of RAD51 . After generation of the HO-induced DSB at the ARG5 , 6 locus , we found that both the RNA levels of genes located in the surrounding of the DSB ( Figure 3A ) and Rbp2 occupancy ( Figure 3B ) progressively decreased as the distance of the corresponding genes from the DSB diminished . Similar results were obtained when the HO-induced DSB was generated at the LEU2 locus ( Figure 3C , D ) . We conclude that inhibition of local transcription is a general response to DSB formation . 10 . 7554/eLife . 08942 . 006Figure 3 . DSB-induced transcriptional inhibition at different chromosomal loci . ( A ) YEPR exponentially growing cell cultures of tGI354 strain , carrying the HO site at the ARG5 , 6 locus and the deletion of RAD51 gene , were transferred to YEPRG at T0 to induce HO . RNA levels of genes located at different distances from the HO cut site were evaluated by qRT-PCR at T0 and T240 after HO induction , as described in Figure 2A . Results are presented as ratios between T240 and T0 . The mean values ±s . d . are represented ( n = 3 ) . ( B ) Binding of Rpb2-HA from samples collected in ( A ) was evaluated by ChIP and qPCR as described in Figure 2B . Primers used were the same as in ( A ) . The mean values ±s . d . are represented ( n = 3 ) . ( C ) YEPR exponentially growing cell cultures of YFP17 strain , carrying the HO cut site at the LEU2 locus , were transferred to YEPRG at T0 to induce HO . RNA levels were analyzed as in Figure 2A but normalized to the ALG9 gene transcript . Only transcription of genes located on the right side of the HO-induced DSB was analyzed because no transcription units are present on the left side of the break . The mean values ±s . d . are represented ( n = 3 ) . ( D ) Binding of Rpb2-HA from samples collected in ( C ) was evaluated by ChIP and qPCR as described in Figure 2B . Primers used were the same as in ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 006 In all eukaryotes , DNA DSBs induce a DDR that depends on the checkpoint kinases ATM/Tel1 and ATR/Mec1 ( Gobbini et al . , 2013 ) . As transcriptional repression in proximity to DSBs requires ATM in mammalian cells ( Kruhlak et al . , 2007; http://journal . frontiersin . org/Journal/10 . 3389/fgene . 2013 . 00136/full , Shanbhag et al . , 2010 ) , we asked whether Tel1 and/or Mec1 were necessary for transcriptional silencing of genes flanking the HO-induced DSB . HO expression was induced by galactose addition in mec1∆ , tel1∆ and mec1∆ tel1∆ cells ( mec1∆ and mec1∆ tel1∆ cells were kept viable by SML1 deletion ) and mRNA levels were analyzed 240 min after HO induction . As the lack of Mec1 causes cells to progress through the cell cycle even after DSB formation , HO was induced in cells arrested in G2 with nocodazole and kept arrested in G2 during HO-cut induction . As shown in Figure 4 , the amount of RNA transcribed from genes surrounding the DSB was reduced to quite similar extents in HO-induced wild type , mec1∆ sml1∆ , tel1∆ and mec1∆ tel1∆ sml1∆ cells , with mec1∆ sml1∆ cells showing a more pronounced decrease . This finding indicates that neither Tel1 nor Mec1 are responsible for DSB-induced transcriptional inhibition . 10 . 7554/eLife . 08942 . 007Figure 4 . DSB-induced transcriptional inhibition does not require Mec1 and Tel1 . YEPR exponentially growing cell cultures of the JKM139 strain , carrying the HO cut site at the MAT locus , were arrested in G2 with nocodazole and transferred at T0 to YEPRG in the presence of nocodazole . RNA levels were analyzed by qRT-PCR as described in Figure 2A . The mean values ±s . d . are represented ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 007 A single DSB triggers phosphorylation of histone H2A to form γH2A , which spreads on each side of the DSB to about 50 kb distance ( Rogakou et al . , 1999; Shroff et al . , 2004 ) . Interestingly , γH2A diminishes strongly over highly transcribed regions and H2A phosphorylation is rapidly restored if transcription is subsequently inhibited ( Lee et al . , 2014 ) . We therefore investigated whether γH2A plays a role in DSB-induced transcriptional inhibition by expressing HO in hta1-S129A cells , which produced a H2A variant carrying a non-phosphorylatable alanine residue replacing Ser129 . As histone H2A is encoded by the two genes HTA1 and HTA2 , hta1-S129A cells also carried the HTA2 deletion . Strikingly , not only RNA levels of genes proximal to the DSB still decreased in hta1-S129A cells , but this decrease was more severe than in wild type cells ( Figure 5A ) , indicating that γH2A counteracts repression of transcription around the DSB . 10 . 7554/eLife . 08942 . 008Figure 5 . The lack of γH2A enhances transcriptional inhibition around the DSB by accelerating resection . ( A ) YEPR exponentially growing cell cultures of the JKM139 derivative strains , carrying the HO cut site at the MAT locus , were transferred to YEPRG at T0 . RNA levels of genes located in the surroundings of the HO cut site at the MAT locus were analyzed at T0 and T240 after HO induction by qRT-PCR as described in Figure 2A . The mean values ±s . d . are represented ( n = 3 ) . ( B ) Method to measure DSB resection . Gel blots of SspI-digested genomic DNA separated on alkaline agarose gel were hybridized with a single-stranded MAT probe ( ss probe ) that anneals to the unresected strand . 5′–3′ resection progressively eliminates SspI sites ( S ) , producing larger SspI fragments ( r1 through r7 ) detected by the probe . ( C ) DSB resection . Genomic DNA prepared from samples collected in ( A ) was analysed for single-stranded DNA ( ssDNA ) formation at the indicated times after HO induction as described in ( B ) . The image that was used for the cropped final Figure 5C is available in Figure 5—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 00810 . 7554/eLife . 08942 . 009Figure 5—source data 1 . Image that was used for the cropped final Figure 5C . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 009 The ends of DSBs are nucleolytically processed to generate 3′-ended ssDNA that initiate HR . DSB resection is negatively regulated by the checkpoint protein Rad9 , which inhibits mainly the Sgs1-Dna2 resection machinery ( Bonetti et al . , 2015; Ferrari et al . , 2015 ) . The lack of γH2A reduces the amount of Rad9 bound at the DSB ends and , as a consequence , hta1-S129A cells resect the DSB ends more efficiently than wild type cells ( Eapen et al . , 2012; Clerici et al . , 2014 ) . Thus , we investigated whether the enhanced DSB-induced transcriptional repression observed in hta1-S129A mutant cells was due to their accelerated nucleolytic processing of the DSB ends . Resection of the DSB 5′ strand can be measured by following the loss of restriction fragments by Southern blot analysis with a ssRNA probe annealing on one side of the break ( Figure 5B ) . Strikingly , SGS1 deletion , which reduced the accelerated resection in hta1-S129A cells to almost wild type levels ( Figure 5C ) , also reduced their enhanced DSB-induced transcriptional inhibition ( Figure 5A ) . Furthermore , rad9∆ cells , which undergo accelerated DSB resection ( Lazzaro et al . , 2008 ) , showed transcription inhibition around the DSB to the same extent as hta1-S129A cells ( Figure 5A ) . Altogether these data indicate that γH2A and Rad9 limit the inhibition of transcription around the DSB ends by counteracting the resection process . We then investigated if generation of ssDNA was responsible for transcription inhibition at DSBs . Resection has been reported to initiate asynchronously from the ends of DSBs ( Shroff et al . , 2004 ) and to move about 4 kb/hr ( Vaze et al . , 2002 ) . After 60 min of HO induction , wild type cells accumulated mostly r2 resection products , indicating that DSB resection in most cells had not proceeded beyond 3 . 5 kb from the DSB ( Figure 6A ) . At the same time point , we detected a strong decrease of Rpb2 occupancy at the TAF2 and BUD5 genes , which are located within 4 kb from either side of the DSB ( Figure 6B ) . Concomitantly with the progression of 5′–3′ resection ( Figure 6A ) , also the binding of Rpb2 to distal genes progressively diminished ( Figure 6B ) , suggesting that the decrease of RNA polymerase II occupancy around the DSB correlates with the time it takes for a DNA-end to become single-stranded . 10 . 7554/eLife . 08942 . 010Figure 6 . Resection mutants fail to reduce transcription around the DSB . ( A ) DSB resection . YEPR exponentially growing cell cultures of JKM139 derivative strains , carrying the HO cut site at the MAT locus , were transferred to YEPRG at T0 . Formation of ssDNA was determined as described in Figure 5B . ( B ) Binding of Rpb2-HA in wild type cells from samples collected in ( A ) was evaluated by ChIP and qPCR as described in Figure 2B . The mean values ±s . d . are represented ( n = 3 ) . ( C ) Binding of Rpb2-HA in exo1∆ sgs1∆ cells from samples collected in ( A ) was evaluated by ChIP and qPCR as described in Figure 2B . The mean values ±s . d . are represented ( n = 3 ) . ( D ) RNA levels from samples collected in ( A ) were analyzed at T0 and T240 after HO induction by qRT-PCR as described in Figure 2A . The mean values ±s . d . are represented ( n = 3 ) . Contiguous images that were used for the cropped final Figure 6A are available in Figure 6—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 01010 . 7554/eLife . 08942 . 011Figure 6—source data 1 . Contiguous images that were used for the cropped final Figure 6A . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 011 DSB resection in S . cerevisiae is initiated by MRX and Sae2 , which catalyze an endonucleolytic cleavage of the 5′ strands . The resulting partially resected 5′ strand can be further processed by the nucleases Exo1 and Dna2 , the latter working in concert with the 3′–5′ helicase Sgs1 ( Mimitou and Symington , 2008; Zhu et al . , 2008 ) . DSB resection is severely affected by either simultaneous inactivation of Exo1 and Sgs1 or the lack of MRX , with the latter not only providing the endonuclease activity to initiate resection , but also promoting the loading of Exo1 , Sgs1 and Dna2 at DSB ends ( Shim et al . , 2010 ) . To assess whether DSB resection has a role in the DSB-induced transcriptional silencing , we asked whether exo1∆ , mre11∆ or exo1∆ sgs1∆ cells , which were impaired in resection to different extents , were still capable to repress transcription of genes in the vicinity of the HO-induced DSB . Indeed , neither the amount of Rpb2 bound around the HO-induced DSB ( Figure 6C ) nor RNA levels were decreased in exo1∆ sgs1∆ cells ( Figure 6D ) , which had the most severe DSB resection defect compared to mre11∆ and exo1∆ single mutants ( Figure 6A ) . RNA levels only slightly decreased in HO-induced mre11∆ cells ( Figure 6D ) , which were severely defective in the accumulation of resection products ( Figure 6A ) , while only the RNA levels of genes proximal to the break site were diminished in exo1∆ mutant cells ( Figure 6D ) , which initiated DSB resection but failed to accumulate resection products longer than 3 . 5 kb ( Figure 6A ) . Thus , loss of transcription around a DSB in S . cerevisiae is due to the conversion of dsDNA to ssDNA by the resection machinery . Here , we further extend the observation that induction of a DSB at the MAT locus in S . cerevisiae cells leads to a decrease of the RNA steady state levels of proximal genes ( Lee et al . , 2000 ) , by demonstrating that this decrease is not locus specific and is due to dissociation of RNA polymerase II from its template DNA . In contrast to the reported requirement of mammalian ATM for transcription silencing around a DSB ( Kruhlak et al . , 2007; Shanbhag et al . , 2010; Ui et al . , 2015 ) , neither Tel1 nor Mec1 appear to be necessary for transcriptional inhibition at the DSB site in S . cerevisiae . Instead , we found that mutants defective in DSB resection fail to inhibit transcription around a DSB , and the extent of this failure correlates with the severity of the DSB resection defect . Furthermore , accelerating DSB resection by preventing γH2A generation or by eliminating Rad9 enhances this DSB-induced transcriptional inhibition . Altogether , these data indicate that loss of transcription around a DSB in S . cerevisiae is not due to an active regulatory mechanism , but to the conversion of the DSB ends from dsDNA to ssDNA . Nucleolytic processing of the template DNA strand should stop transcription of the corresponding gene . However , we found that mRNA levels and RNA polymerase II occupancy around a DSB decrease independently of the strand that is transcribed , indicating that loss of non-template DNA strands also impedes transcription . Therefore , the most likely reason of why the conversion from dsDNA to ssDNA stops transcription is that the transcription machinery binds only dsDNA . Consistent with this hypothesis , it has been shown that within a region containing a stretch of dsDNA preceding a single-strand 3′ DNA end , purified RNA polymerase II transcribes within the dsDNA portion ( Lilley and Houghton , 1979; Kadesch and Chamberlin , 1982 ) . Furthermore , structural analyses of transcription initiation reveals that RNA polymerase II and its associated General Transcription Factors bind double-stranded promoter DNA to form a closed preinitiation complex ( PIC ) , where the transcriptional machinery interacts with both template and non-template DNA strands ( Liu et al . , 2013; Sainsbury et al . , 2015 ) . Moreover , removal of either the template or non-template strands prevents binding of T7 RNA polymerase to the promoter , supporting the importance of protein-dsDNA interaction in the spatial organization of the transcriptional initiation complex ( Maslak and Martin , 1993 ) . Interaction of the transcriptional machinery with dsDNA also allows transcription initiation and elongation . In fact , the double-stranded promoter DNA follows a straight path in the PIC complex and this rigidity allows the subsequent transition from a closed to an open promoter complex , where a central DNA region is melted leading to a transcription bubble in which the DNA template strand enters the RNA polymerase II cleft ( Murakami et al . , 2013 ) . Furthermore , competition between the non-template DNA strand and the RNA transcript for base-pairing with the DNA template strand was shown to be important for maintaining structure and function during elongation ( Kireeva et al . , 2011 ) , reinforcing the inhibitory effects of ssDNA on the transcription process . In summary , while DSB-induced transcriptional repression in mammals is reportedly an active mechanism controlled by ATM , the stop of transcription around a DSB in S . cerevisiae cells is due to the conversion of dsDNA to ssDNA that is necessary to initiate DSB repair by HR ( Figure 7 ) . In any case , while Mec1 and Tel1 are not required to resect a DSB , ATM and the chromatin compaction promoting proteins macroH2A1 , PRMD2 and HP1 promote end resection by facilitating the loading of BRCA1 at the DSB ends ( Soria and Almouzni , 2013; Ayrapetov et al . , 2014; Khurana et al . , 2014 ) . Thus , whether the nucleolytic processing of the DSB ends contributes to repress transcription around a DSB also in mammals is an important question that remains to be addressed . In fact , although in human cells inactivation of the end resection factor CtIP does not restore trascription around the DSB ( Shanbhag et al . , 2010 ) , other resection machineries are known to be active in resecting DSBs in CtIP-depleted cells ( Tomimatsu et al . , 2012 ) and could contribute to inhibit transcription around the DSB . 10 . 7554/eLife . 08942 . 012Figure 7 . Model for loss of transcription around a DSB . Resection of the 5′ strands at both DSB ends leads to release of the transcription machinery ( dashed lines ) and to subsequent transcription arrest independently of whether the degraded DNA strand acts as template or non-template . Since the RNA polymerase binds double-stranded promoter DNA , generation of ssDNA at the DSB ends prevents reinitiation events ( bar-headed line ) . Blue arrows indicate direction of transcription . DOI: http://dx . doi . org/10 . 7554/eLife . 08942 . 012 The yeast strains used in this study are derivatives of JKM139 , YFP17 or tGI354 strains and are listed in Supplementary file 1 . Cells were grown in YEP medium ( 1% yeast extract , 2% peptone ) supplemented with 2% glucose ( YEPD ) , 2% raffinose ( YEPR ) or 2% raffinose and 3% galactose ( YEPRG ) . DSB end resection at the MAT locus in JKM139 derivative strains was analyzed on alkaline agarose gels as previously described ( Clerici et al . , 2008 ) , by using a single-stranded probe complementary to the unresected DSB strand . This probe was obtained by in vitro transcription using Promega ( Madison , WI ) Riboprobe System-T7 and plasmid pML514 as a template . Plasmid pML514 was constructed by inserting in the pGEM7Zf EcoRI site a 900-bp fragment containing part of the MATα locus ( coordinates 200870 to 201587 on chromosome III ) . Total RNA-seq libraries were previously described ( Manfrini et al . , 2015 ) and data were retrieved from the Gene Expression Omnibus ( accession number GSE63444; Manfrini et al . , 2014 ) . Reads were uniquely mapped as described ( Manfrini et al . , 2015 ) on a reference ‘genome’ corresponding to the MATa sequence ±10 kb , that was manually reconstructed by replacing MATα-specific elements with the corresponding MATa-specific elements in the MATα locus sequence ±10 kb ( chr . 3 , positions 188671 to 211177 ) retrieved from SGD ( http://www . yeastgenome . org/ ) . Data were normalized using the normalization factors previously used for the whole transcriptome analysis ( Manfrini et al . , 2015 ) , based on the total number of reads that mapped on all the ORFs of the yeast genome . Total RNA was extracted from cells using the Bio-Rad ( Hercules , CA ) Aurum total RNA mini kit . First strand cDNA synthesis was performed with the Bio-Rad iScript cDNA Synthesis Kit . qRT-PCR was performed on a MiniOpticon Real-time PCR system ( Bio-Rad ) and RNA levels were quantified using the ∆∆Ct method . Quantities were normalized to either ACT1 or ALG9 RNA levels . Since RNR1 is transcriptionally induced immediately after HO-induced DSB , its RNA level at T240 was normalized on the value obtained 30 min after HO induction , when DSB resection was not proceeded beyond 1 . 7 kb . Primer sequences are provided in Supplementary file 2 . ChIP analysis was performed as previously described ( Viscardi et al . , 2007 ) . Chromatin extracts from both RPB2-HA and RPB2 strains were immunoprecipitated with anti-HA antibodies ( 12CA5 ) . Input and immunoprecipitated DNA were purified and analyzed by qPCR . Amplicons were chosen well within the coding region of the genes and of the highly transcribed ACT1 gene on chromosome VI as a control . The ratio between values obtained for Rpb2-HA and those obtained with the untagged Rpb2 immunoprecipitated samples , both of which normalized against the corresponding input , was calculated for each time point after HO induction . The obtained values were divided by the ratio calculated from uninduced cells , arbitrarily set to 1 , for each amplicon . Primer sequences are provided in Supplementary file 2 .	DNA is constantly under assault from harmful chemicals; some of which are produced inside the cell , while others come from outside of the cell . Breaks that form across both strands in a DNA double helix are considered the most dangerous type of DNA damage , and can cause a cell to die or become cancerous if they are not repaired accurately . ‘Homologous recombination’ is one of the main mechanisms used by cells to repair DNA double-strand breaks . This mechanism requires enzymes to eat away at the end of one of the DNA strands on each side of the double-strand break . This process is called ‘resection’ and it exposes single strands of DNA . These single-stranded DNA ‘tails’ are then free to interact with an intact copy of the same DNA sequence from elsewhere in the cell's nucleus , which is used as a guide when repairing the damage . The proteins involved in homologous recombination have to work around other processes that go on inside the nucleus , such as the transcription of DNA in genes into RNA molecules . Previous research has reported that forming a double-strand break in the DNA reduces the levels of transcription for the genes that surround the break , but it was not clear how this occurred . In mammalian cells , inhibiting the transcription of genes around a double-strand DNA break depends on a signaling pathway that is activated whenever DNA damage is detected . Manfrini et al . now show that this is not the case for budding yeast ( Saccharomyces cerevisiae ) . Instead , the experiments indicate that it is the resection of the DNA around a double-strand break to form single-stranded tails that inhibits transcription in budding yeast . One of the next challenges will be to see if the resection process makes any contribution to changes in the transcription of genes that surround a double-strand break in mammals as well .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression" ]	2015	Resection is responsible for loss of transcription around a double-strand break in Saccharomyces cerevisiae
Understanding how the brain recovers from unconsciousness can inform neurobiological theories of consciousness and guide clinical investigation . To address this question , we conducted a multicenter study of 60 healthy humans , half of whom received general anesthesia for 3 hr and half of whom served as awake controls . We administered a battery of neurocognitive tests and recorded electroencephalography to assess cortical dynamics . We hypothesized that recovery of consciousness and cognition is an extended process , with differential recovery of cognitive functions that would commence with return of responsiveness and end with return of executive function , mediated by prefrontal cortex . We found that , just prior to the recovery of consciousness , frontal-parietal dynamics returned to baseline . Consistent with our hypothesis , cognitive reconstitution after anesthesia evolved over time . Contrary to our hypothesis , executive function returned first . Early engagement of prefrontal cortex in recovery of consciousness and cognition is consistent with global neuronal workspace theory . The recovery of neurocognitive function after brain network perturbations such as sleep , general anesthesia , or disorders of consciousness is of both scientific and clinical importance . Scientifically , characterizing recovery processes after such perturbations might provide insight into the more general mechanisms by which consciousness and cognition are reconstituted after major network disruptions . The ability to recover cognitive function quickly after sleep , for example , likely confers a natural selection advantage . Moreover , understanding which brain functions are most resilient to perturbation could inform evolutionary neurobiology ( Mashour and Alkire , 2013; Kelz and Mashour , 2019 ) . Clinically , understanding the specific recovery patterns after pathologic states of unconsciousness could inform prognosis or therapeutic strategies . However , it is challenging to characterize differential cognitive recovery after sleep because of the rapidity of the process , whereas it can be impossible in pathologic states because of the unpredictable recovery . General anesthesia , by contrast , represents a controlled and reproducible method by which to perturb consciousness and cognition that is also amenable to systematic observations of the recovery process . Studying recovery of cognition after general anesthesia in humans is also of particular importance because animal studies suggest that general anesthetics have the potential to immediately and persistently impair cognition in the post-anesthetic period ( Culley et al . , 2004; Valentim et al . , 2008; Carr et al . , 2011; Callaway et al . , 2012; Zurek et al . , 2012; Jevtovic-Todorovic et al . , 2013; Zurek et al . , 2014; Avidan and Evers , 2016; Jiang et al . , 2017 ) , creating a potential public health concern for the hundreds of millions of surgical patients undergoing general anesthesia each year ( Weiser et al . , 2015 ) . To improve scientific understanding of recovery of consciousness and cognition after anesthetic-induced unconsciousness , we studied 30 healthy volunteers at three centers who were administered deep general anesthesia using isoflurane for 3 hr , with cognitive testing conducted at pre-anesthetic baseline as well as every 30 min for 3 hr after return of consciousness ( Figure 1 ) . We hypothesized that post-anesthetic recovery would be an extended process rather than a single point , commencing with return of responsiveness and concluding with return of executive function . We hypothesized that executive function would be the last to recover because there is evidence that neurologic recovery from general anesthesia occurs in a caudal-to-rostral direction ( Långsjö et al . , 2012; Reshef et al . , 2019 ) , suggesting that anterior structures mediating higher cognition would have the most prolonged recovery . To assess differential return of cognitive functions after the major perturbation of deep anesthesia , we assessed a neurocognitive battery of tests ( including the Psychomotor Vigilance Test ( PVT ) , Motor Praxis ( MP ) , Digit Symbol Substitution Test ( DSST ) , fractal 2-Back ( NBCK ) , Visual Object Learning Test ( VOLT ) , and Abstract Matching ( AM ) ; Table 1 ) at baseline and at multiple time points after the 3-hr period of anesthetic exposure . Isoflurane anesthesia was chosen because of its heterogeneous molecular targets , which affect multiple neural systems , and because its slower offset compared to other anesthetics would allow us to observe differential recovery of function ( Hemmings et al . , 2019 ) . A halogenated ether was chosen instead of propofol because of the greater diversity of molecular targets , which would be predicted to have a more profound effect on neural dynamics through multiple neurotransmitter receptor and channel systems . Clinical observations as well as clinical research comparing recovery from isoflurane vs . propofol support this interpretation ( Pollard et al . , 1994; Geng et al . , 2017 ) . The 3-hr duration of anesthesia was chosen based on clinical data related to recovery of surgical patients , the pharmacokinetics of isoflurane , and practical considerations for volunteers participating in day-long experiments . To control for the learning effects of repeated cognitive testing ( Basner et al . , 2018 ) , we also recruited 30 healthy volunteers who , instead of receiving anesthesia , were engaged in wakeful behavior for 3 hr and then underwent equivalent cognitive testing at time points corresponding to the cohort that underwent general anesthesia ( Figure 1 ) . All participants received actigraphy watches to monitor sleep-wake activity before and after anesthesia or the control condition , and all participants had electroencephalographic recording throughout the experiment to assess cortical dynamics of relevance to consciousness and cognition , with techniques that have been used to assess information processing in specific brain regions ( permutation entropy ) as well as more complex spatiotemporal patterns across the cortex ( Lempel-Ziv complexity ) . With the control group serving as a reference , the aims of the study were: ( 1 ) to determine whether emergence and cognitive recovery occurred at a point or , as we hypothesized , through a process; ( 2 ) assess the sequence of cognitive recovery following emergence from a prolonged state of unconsciousness with serial neurobehavioral assessments to test the hypothesis that higher executive functions reconstitute only after more primary functions; and ( 3 ) to measure correlated changes in cortical dynamics that might account for the hypothesized differential recovery of cognitive function . This multicenter study was reviewed and approved by the Institutional Review Board specializing in human subjects research at the University of Michigan , Ann Arbor; University of Pennsylvania; and Washington University in St . Louis . Volunteers were recruited through the use of fliers and were compensated for their participation at levels approved by ethics committees . Participation eligibility required that all subjects provide written informed consent , which was obtained after careful discussion , in accordance with the Declaration of Helsinki . The experimental design and data acquisition are summarized in Figure 1 . This was a within-group study of anesthetized participants with a primary outcome of the pattern of cognitive recovery after general anesthesia; a non-anesthetized cohort was included to control for the learning effects of repeated cognitive testing and circadian factors . Volunteers were randomly assigned to receive general anesthesia with isoflurane or to engage in waking activity on the study day and serve as experimental controls . Baseline cognitive assessment was performed after an initial screening . For anesthesia sessions , participants were closely monitored by two attending anesthesiologists during the study day . Attending anesthesiologists elicited a standard clinical preoperative history and physical examination , independently verified that volunteers met inclusion and fasting criteria , and safely conducted the anesthetic . Each subject underwent intravenous catheter placement . An appropriately fitted EEG head cap ( Electrical Geodesics , Inc Eugene OR ) was affixed to the scalp . Electrical impedances on each channel were kept under 50kOhms/channel whenever possible . Standard anesthesia monitors ( electrocardiogram , non-invasive blood pressure cuff , a pulse oximeter ) were applied , and capnography was measured . Subjects completed a second baseline round of neurocognitive testing with ongoing EEG recordings . Upon completing the neurocognitive battery , subjects were pre-oxygenated by face mask prior to induction of general anesthesia with a stepwise increasing infusion rate of propofol: 100 mcg/kg/min x 5 min , 200 mcg/kg/min x 5 min , and then 300 mcg/kg/min x 5 min . During this time , an audio loop issued commands every 30 s asking subjects to squeeze their left or right hand ( in random order ) twice . Loss of consciousness ( LOC ) was defined as the first time that a subject failed to respond to two sets of consecutive commands . After 15 min of propofol administration , subjects began inhaling isoflurane at 1 . 3 age-adjusted MAC ( minimum alveolar concentration ) ( Nickalls and Mapleson , 2003 ) . Thereafter , a laryngeal mask was inserted orally , a nasopharyngeal temperature probe was placed , and the propofol infusion was discontinued . Anesthetized subjects continued to inhale 1 . 3 age-adjusted MAC isoflurane anesthesia for 3 hr . Burst suppression , a sign of deep anesthesia ( Hemmings et al . , 2019 ) , was found to be associated with this concentration of isoflurane in this cohort ( Shortal et al . , 2019 ) . Blood pressure was targeted to remain within 20% of baseline pre-induction values using a phenylephrine infusion or intermittent boluses of ephedrine , as necessary . Pressure support ventilation was initiated with pressures titrated to maintain tidal volumes in the 5–8 ml/kg range while end tidal carbon dioxide levels were targeted to 35–45 torr . Surface warming blankets were utilized to maintain body temperature in the normal range . Subjects received 4 mg intravenous ondansetron 30 min prior to discontinuation of isoflurane for antiemetic prophylaxis . Isoflurane was discontinued at the end of the 3-hr anesthetic period . Verbal command loops were reissued every 30 s immediately upon cessation of isoflurane . The laryngeal mask was removed when deemed medically safe by the attending anesthesiologists . Recovery of consciousness ( ROC ) was defined as the earliest instance in which subjects correctly responded to two consecutive sets of audio loop commands . At this point , defined as time = 0 min , subjects restarted neurocognitive testing with a brief pause between consecutive rounds . Neurocognitive testing was repeated at t = 30 , 60 , 90 , 120 , 150 , and 180 min following emergence . Each battery of neurocognitive testing lasted approximately 15–25 min and was preceded by 5 min of eyes closed , resting state EEG data acquisition . Brief restroom or nutrition breaks were permitted between testing rounds , as necessary . Subjects were discharged according to standard post-anesthesia care unit discharge criteria after completing their final battery of neurocognitive testing . A study site coordinator contacted each subject within 24 hr of the study day to document any adverse events . A second group of healthy individuals ( n = 30 ) was recruited to participate in the same study design . These individuals also fasted overnight , but did not have intravenous lines inserted . Rather than being anesthetized , these volunteers remained awake ( by reading or watching television on a personal electronic device ) and continued fasting for 3 . 5 hr in order to control both for potential learning effects of repeated testing and also for circadian variability in testing performance ( McLeod et al . , 1982; Gur et al . , 2001; Van Dongen et al . , 2003; Van Dongen and Dinges , 2005; Jasper et al . , 2009; Tucker et al . , 2010 ) . We chose not to sedate these participants as a control for the anesthetized state because doing so would have obscured the predicted learning effect accompanying repeated neurobehavioral testing and would thus confound the normal performance standard that was required . Volunteers randomized to the restful group were instructed to avoid napping and were regularly monitored by a dedicated research assistant . A total of 60 , healthy American Society of Anesthesiologists physical status classification I or II volunteers were enrolled . The choice of study subject numbers was informed by several factors , including previous studies ( Eger et al . , 1997; Eger et al . , 1998 ) , biological plausibility ( our best estimates regarding effect size and standard deviation ) , and safety considerations ( exposing the minimum number of humans to general anesthesia in order to answer the questions of interest ) . The main factor that was considered in estimating our required sample size was the time difference in return of cognitive functions within the subjects receiving general anesthesia . Sample size calculation was modeled with various assumptions regarding the difference in recovery times between the first and last cognitive domains to return , and the standard deviations of these parameters . A range between 30 min and 90 min was considered for differences in recovery times between cognitive domains ( possible effect sizes ) . A range between 20 min and 40 min was considered for standard deviations of these parameters . Assuming relatively conservative estimates ( difference in recovery times = 30 min and standard deviation = 40 min ) , 30 subjects would provide >80% power with a two-sided alpha <0 . 05 , using an unpaired t test . With relatively liberal assumptions ( difference in recovery times = 90 min and standard deviation = 20 min ) , 30 subjects would provide >99% power with a two-sided alpha <0 . 001 , using an unpaired t test . Each study site ( University of Michigan , University of Pennsylvania , Washington University in St . Louis ) recruited 20 volunteers who met the inclusion criteria . Prospective volunteers were screened using a phone questionnaire administered by a study coordinator . Eligible subjects that consented to participate in this study underwent a baseline familiarization round of neurocognitive testing and were given rest-activity monitoring devices ( actiwatch ) 1 week prior to the study day . To assess the neural correlates of the anesthetized state and recovery , participants enrolled at the University of Pennsylvania and University of Washington in St . Louis ( n = 40 ) were fitted with 32 EEG scalp electrodes . Subjects at the University of Michigan ( n = 20 ) were fitted with 64 or 128 EEG scalp electrodes . EEG recordings began prior to the baseline neurocognitive testing on the study day and were continued with minimal interruption until the completion of the final neurocognitive test . The raw EEG signals were exported into MATLAB ( version 2015a; MathWorks , Inc , Natick , MA ) , down-sampled to 250 Hz ( resample . m function in Matlab signal processing toolbox ) , and re-referenced to the linked-mastoid reference . Electrodes on the lowest parts of the face and head were removed , leaving 21 channels on the scalp ( common to EEG montage for all participants ) for the analysis . Data segments with obvious noise or non-physiological artifacts were identified and removed by visual inspection of the waveform and spectrogram of the EEG signals . Prior to the analysis , the EEG signals were bandpass filtered at 0 . 5–30 Hz via butterworth filter of order 4 ( butter . m and filtfilt . m in MATLAB signal processing toolbox ) to remove the possible baseline drift and muscle artifacts . Ten 2 min epochs were selected during the seven resting-state eyes-closed sessions , and also during the exposure to anesthesia: ( 1 ) LOC - the first 2 min after loss of responsiveness; ( 2 ) Maintenance – the last 2 min before the discontinuation of isoflurane; and ( 3 ) Pre-ROC – the 2 min immediately preceding recovery of responsiveness . Burst-suppression patterns were present in the Maintenance epoch for six participants; to prevent the confounding effect of the suppression pattern on the EEG measures , we instead extracted 2 min continuous , non-suppression epochs in the last 10 min before the discontinuation of isoflurane ( n = 3 participants ) , or 7 min immediately after the discontinuation of isoflurane ( n = 3 participants ) , which showed similar spectral properties when compared to the other participants . Detailed information on the EEG sample size is listed in Supplementary file 1A . For completeness , seven 2 min epochs were selected during the seven resting-state eyes-closed sessions in the non-anesthetized group . Preliminary analysis with high-density EEG suggested that phase-based connectivity measures and graph-theoretical variables were not robust enough to capture the neurophysiologic dynamics of general anesthesia and recovery with appropriate temporal resolution ( Blain-Moraes et al . , 2017 ) . We therefore chose two measures of cortical dynamics that have been suggested to differentiate levels of consciousness and , unlike spectral analysis , also serve as a surrogate for the repertoire of brain states , which we would expect to increase with recovery of consciousness and cognitive function . We used permutation entropy ( PE ) to measure the local dynamical changes of EEG in frontal and posterior channels . PE quantifies the regularity structure of a time series , based on a comparison of the order of neighboring signal values , which is conceptually simple , computationally efficient , and artifact resistant ( Bandt and Pompe , 2002 ) , and has been successfully applied to the separation of wakefulness from unconsciousness ( Jordan et al . , 2013; Li et al . , 2008; Olofsen et al . , 2008; Ranft et al . , 2016 ) . The calculation of PE requires two parameters: embedding dimension ( dE ) and time delay ( τ ) . In line with previous studies , we used dE=5 and τ=4 in order to provide a sufficient deployment of the trajectories within the state space of the EEG beta activity during wakefulness and anesthesia ( Jordan et al . , 2013; Ranft et al . , 2016 ) . Supplementary analysis was performed to test the sensitivity of PE using alternative strategies of parameter selection . In the implementation , each 2 min epoch was divided into non-overlapping 10 s windows , the PE was calculated for each window , and the PE values were averaged across all the windows for each studied epoch and channel . The topographic maps of group-level PE value for each studied epoch was constructed using the topoplot function in the EEGLAB toolbox ( Delorme and Makeig , 2004 ) . For statistical comparisons , the averaged PE values were calculated over the frontal ( Fp1 , Fp2 , Fpz , F3 , F4 , and Fz ) and posterior ( P3 , P4 , Pz , O1 , O2 , and Oz ) channels at each studied epoch for each participant . Lempel-Ziv Complexity ( LZC ) was computed as a surrogate of complexity to reflect the spatiotemporal repertoire across scalp potentials . LZC is a method of symbolic sequence analysis that measures the complexity of finite length sequences ( Lempel and Ziv , 1976 ) , which has been shown to be a valuable tool to investigate brain states related to consciousness and cognition ( Casali et al . , 2013; Abásolo et al . , 2015; Schartner et al . , 2015; Hudetz et al . , 2016; Schartner et al . , 2017 ) . The calculation of LZC requires a binarization of the multichannel EEG data . In this study , we used the implementation as described in Schartner et al . , 2015; Schartner et al . , 2017 , and calculated the instantaneous amplitude from the Hilbert transformed EEG signal for each channel , which was binarized using its mean value as the threshold for the current channel ( supplementary analysis was performed to test the effect of threshold selection ) . The data segment was then converted into a binary matrix , in which rows represent channels and columns represent time points . LZC was computed by searching the spatiotemporal matrix time point by time point and counting the number of different spatial patterns across different time points . Thus , the resultant measure captures the complexity or diversity in both temporal and spatial domains . For implementation , the average signal was first subtracted from all channels in order to remove the effect of common reference , and then the multichannel EEG epochs were divided into non-overlapping 4 s windows to compute the LZC , with the resultant LZC values being averaged across all the windows for each studied epoch . In line with previous studies , we normalized the original LZC by the mean of the LZC values from N = 50 surrogate data sets generated by randomly shuffling each row of the binary matrix , which is maximal for a binary sequence of fixed length ( Schartner et al . , 2015; Schartner et al . , 2017 ) ( supplementary analysis was performed to test alternative methods in the generation of surrogate data ) . Statistical analyses were conducted in consultation with the Center for Statistical Consultation and Research at the University of Michigan . All EEG-derived PE and LZC values were exported to IBM SPSS Statistics version 24 . 0 for Windows ( IBM Corp . Armonk , NY ) . Statistical comparisons were performed using linear mixed models ( LMM ) , to test ( 1 ) the difference between the ten studied epochs for both PE and LZC measures and ( 2 ) the difference between PE values derived from frontal and posterior channels . In contrast to traditional repeated-measures ANOVA analysis , LMM analysis offers more flexibility in dealing with missing values ( see Supplementary file 1B ) and accounting for the within-participant variability by including a random intercept associated with each participant . The non-anesthetized group was included primarily to control for learning effects in repeated cognitive testing and thus , for EEG analysis , the statistical analysis was focused on anesthetized group . For the model of LZC values , the fixed effect is the studied epoch . For the model of PE values , the fixed effects include the studied epoch , region , and the interaction between them . We fitted the models with random intercept specific for each participant and used the default variance components as the covariance structure . We modeled the studied epoch as repeated effects and assumed each studied epoch was associated with different residual variance by using the diagonal structure as the covariance structure of the residuals . We employed restricted maximum likelihood estimation . The models described above were chosen by taking into account the information criteria and likelihood ratio test results in the comparisons , with alternative models including additional random effects and repeated effects , as well as different covariance structures ( Supplementary file 1B ) . For all post hoc pairwise comparisons , the Bonferroni corrected p-value along with the estimate and 95% confidence interval ( CI ) of the difference were reported . A two-sided p<0 . 05 was considered statistically significant . To explore whether the EEG-based measures of cortical dynamics are predictive of the performance of the cognitive functions in the post-anesthetic period , we examined the associations between EEG measures during pre-anesthetic baseline ( EC1 ) , Maintenance and pre-ROC and the impairment of cognitive performance at emergence ( just after recovery of consciousness ) . Spearman’s rank correlation was used to assess the relationships between each of the EEG measures ( frontal PE , posterior PE and LZC ) and the impairment of performance , in terms of both accuracy and response time , for each of the six tasks . Neurocognitive tests were selected from the cognition test battery ( Basner et al . , 2015 ) to reflect a broad range of cognitive domains , ranging from basic abilities such as sensory-motor speed to complex executive functions such as abstraction . The order of the six tests was randomized but balanced across subjects . Individual subjects took the tests in the same order ( except during familiarization , which occurred at least 1 week prior to the study day ) . In each test session , subjects repeated the first test after completion of sixth test . Therefore , the temporal resolution for one test in five control subjects and five experimental subjects was doubled . Each testing session required 15–25 min to complete , with a new session starting every 30 min after emergence . There was thus a total of six testing sessions in the 3-hr time period after emergence . The following six tests , adopted from Basner et al . , 2015 , were chosen for this study . See Table 1 for a summary of the tests , associated cognitive function , and associated neuroanatomical substrates . The Motor Praxis Task ( MP ) measures sensorimotor speed and validates that volunteers have sufficient command of the computer interface . Participants were instructed to click on squares that appear randomly on the screen , with each successive square smaller and thus more difficult to track . The test depends upon function of visual and sensorimotor cortices ( Gur et al . , 2001; Gur et al . , 2010; Neves et al . , 2014 ) . The Psychomotor Vigilance Test ( PVT ) measures a volunteer’s reaction times ( RT ) to visual stimuli that are presented at random inter-stimulus intervals over 3 min ( Basner et al . , 2011 ) . Subjects monitor a box on the computer screen , and press the space bar once a millisecond counter appears and begins timing response latency . In the well-rested state , or whenever sustained attention performance is optimal , right frontoparietal cortical regions are active during this task . Conversely , with sleep deprivation and other suboptimal performance , studies demonstrate increased activation of default-mode networks during this task , which is considered to be a compensatory mechanism ( Drummond et al . , 2005 ) . The Digit-Symbol Substitution Task ( DSST ) adapts the Wechsler Adult Intelligence Scale ( WAIS-III ) for a computerized presentation . The DSST required participants to refer to a continuously displayed legend that matches each numeric digit to a specific symbol . Upon presentation of one of the nine symbols , subjects must select the corresponding number as rapidly as possible . The DSST primarily recruits the temporal cortex , prefrontal cortex , and motor cortex . Activation of frontoparietal cortices during DSST performance has been interpreted as reflecting both on-board processing in working memory and low-level visual search ( Usui et al . , 2009 ) . The Fractal 2-Back ( F2B ) is an extremely challenging nonverbal variant of the Letter 2-Back test . N-back tests probe working memory . The F2B consisted of the sequential presentation of a set of fractals , each potentially repeated multiple times . Participants were instructed to respond when the current stimulus matched the stimulus previously displayed two images prior . The F2B is a well-validated task that robustly activates dorsolateral prefrontal cortex , cingulate , and hippocampus ( Ragland et al . , 2002 ) . The Visual Object Learning Test ( VOLT ) measures the volunteer’s memory for complex figures ( Glahn et al . , 1997 ) . Participants memorize 10 sequentially displayed three-dimensional figures . Subsequently , they were instructed to select the familiar objects that they memorized from a larger set of 20 sequentially presented objects that included the 10 memorized and 10 similar but novel objects . Visual object learning tasks have been shown to depend upon frontal and bilateral anterior medial temporal cortices as well as the hippocampus ( Jackson and Schacter , 2004 ) . The Abstract Matching ( AM ) test ( Glahn et al . , 2000 ) is a validated test of executive function . Subjects are presented with two pairs of objects at the bottom left and right of the screen whose perceptual dimensions ( e . g . color and shape ) vary . Subjects were presented with a target object in the upper middle of the screen and had to classify the target using its perceptual dimensions to one of the two pairs , based on a set of implicit , abstract rules . The abstract matching paradigm evaluates abstraction and cognitive flexibility and depends upon the prefrontal cortex ( Berman et al . , 1995 ) . Apart from the Bayesian analyses , statistical analyses were implemented by the SAS software version 9 . 4 . In view of the learning effect with repeated cognitive testing , it is difficult to pinpoint when any cognitive domain returns to baseline . We could therefore not simply compare recovery times between individual cognitive tests , as we had planned . Instead , in order to test whether there was a difference in recovery times between cognitive domains , we opted for a Bayesian regression approach using all the data from the anesthetized subjects as well as the non-anesthetized controls . For this analysis , we used the brms package in R ( R Foundation for Statistical Computing , Vienna , Austria ) , which uses Stan for full Bayesian inference . We simulated M samples from the ( posterior ) conditional distribution of the model parameters given the data , and for each set of simulated model parameters , we calculated the corresponding recovery time , thus we obtained M samples of recovery times . Then , based on these simulated recovery times , we further calculated the corresponding differences in recovery times between pairs of cognitive tests ( or domains ) , and evaluated the posterior probability of one cognitive test recovering more than 30 min before another test [P ( diff >30 min\|data ) ] by checking the sample proportion in the posterior sample . The recovery time involves both performance accuracy and performance speed . Both processes are defined by the time for anesthetized subjects to have the test score back to their respective accuracy/speed baseline scores . Regarding priors , we chose normal priors with a large variance , which is a routine choice for non-informative flat priors . Multiple statistical comparisons were separately conducted on the standardized accuracy and speed indices , which were two metrics to evaluate each task performance . We used nonlinear mixed-effects models ( NLMM ) based on a damped exponential in time to fit the data of each task at all time-points , that is yt=ybaseline+α+β*eγt , where yt is the task performance response at time t , ybaseline is the pre-treatment baseline response , α is the random intercept , β is the random slope , and γ is the coefficient of the fixed-effect time t . Based on our model , the recovery time was calculated by T=log⁡-αβ/γ . The random intercept α and the random slope β are independent , and distributed from normal distributions . Some NLMMs were degenerated to models with a fixed intercept or a fixed slope according to goodness of fit . The appeal of NLMM analysis is its flexibility in modeling the nonlinear trend of cognitive data with repeated measures over time and the ability to adjust the pre-treatment baseline performance . In order to test against the alternative hypothesis that there was a significant difference at the end of the 3-hr period between the anesthetized group and the control group , we performed model-based multiple testing on the least squares means of predicted response difference at 3 hr . For accuracy and speed of each task , the point estimates of the difference as well as the Bonferroni corrected confidence intervals ( CI ) of the difference with an overall significance level 0 . 05 were reported . The p-values of testing difference between groups should be compared to the corrected level 0 . 05/6=0 . 0083 . For both accuracy and speed estimates of recovery time ( measured in hours ) in each testing domain , we based our estimates on 10 , 000 Markov Chain Monte Carlo samples . For these simulations , we present posterior probabilities and corresponding 90% credible intervals for differences between cognitive domains in recovery times . In order to assess and control for differences in baseline sleep-wake rhythms and to potentially evaluate the effect of isoflurane anesthesia on subsequent rest activity behavior , participants were trained and instructed to wear a wrist GT3X + device ( ActiGraph ) on their nondominant wrist beginning at the conclusion of their baseline visit , 1 week before prior to and 1 week following their assigned study day . Actigraphy data were downloaded to a computer and GT3X + devices recharged on the study day and again at completion 1 week following the study day . Raw activity counts for each subject were binned into 1 min epochs and analyzed for bouts of inactivity using the Cole-Kripke scoring algorithm ( Cole et al . , 1992 ) , included in the ActiLife 6 . 7 . 2 software . For each subject , minutes of inactivity each hour were calculated . ActiLife’s wear time validation was employed using default settings to confirm that subjects used the GT3X + monitor as instructed . Hours in which wear time validation revealed that the watch was not worn were excluded from the analysis . Actigraphy data were imported into Prism 5 . 0d ( GraphPad ) and analyzed with a two-way ANOVA with Time in hours relative to the midnight before testing and Treatment Group ( Isoflurane Exposed or Awake Controls ) as the two factors . Effects of Time , Treatment Group , and the interaction between these two factors were considered to be significant for p values < 0 . 05 . Due to asynchrony in times during which the GT3X + device was not worn across individuals , it was not possible to conduct a repeated measures two-way ANOVA . To obtain a graphical best fit of actigraphy data , we performed a standard Cosinar analysis . The study received ethics committee approval at all three sites independently; written informed consent was obtained after careful discussion with each participant . The average age of all study participants was 27 ( ±4 . 5 ) years , with 50% females . There were no adverse events during the course of the study or at 1-week follow up at the completion of the study . We administered six distinct cognitive tests at baseline and twice per hour for 3 hr after exposure to general anesthesia or a comparable period without anesthetic exposure in the control participants . The order of the MP , PVT , DSST , NBCK , VOLT , and AM tests was randomized between subjects but consistent within subjects . In anesthetized volunteers ( with correction for learning based on results from tests taken at corresponding times by the non-anesthetized controls ) , the accuracy and speed of all six cognitive tests were significantly impaired compared with the pre-anesthetic baseline assessments ( all twelve statistical tests yielded p<0 . 008 , with adjustments for multiple comparisons ) . Thus , the first question answered is that all tests were impaired at initial recovery . The next question to be answered was whether rates of recovery differed among cognitive domains and whether the time to recovery exceeded 30 min when comparing among cognitive tests . Based on likelihood ratio tests , there were statistically significant differences in the rates of recovery of the six cognitive domains , both with regard to their accuracy and speed ( p<0 . 05 ) . Results from Bayesian analyses yielded posterior probability estimates that differences in recovery times between the various cognitive domains for accuracy and speed exceeded 30 min . Hence , recovery of cognition is a process that appears to evolve over time rather than one that occurs simultaneously . For accuracy ( Figure 2 ) , the probability was high that ( i ) recovery of AM occurred more than 30 min before NBCK ( 95% ) , DSST ( 77% ) , VOLT ( 72% ) , and MP ( 65% ) ; ( ii ) recovery of PVT occurred more than 30 min before NBCK ( 99 . 9% ) , DSST ( 81% ) , MP ( 76% ) , and VOLT ( 72% ) ; ( iii ) recovery of VOLT occurred more than 30 min before NBCK ( 95% ) , DSST ( 76% ) , and MP ( 68% ) ; and that ( iv ) recovery of MP occurred more than 30 min before NBCK ( 57% ) . The results from the 10 , 000 Markov Chain Monte Carlo samples yield the following additional insights with respect to accuracy of recovery . The accuracy in PVT is strongly impaired as most samples cannot recover . For the other five tests , the posterior probability that: ( a ) recovery of MP occurred more than 30 min before VOLT , NBCK and DSST are 93% , 97% , and 99% , respectively; the 90% corresponding credible intervals of the difference in recovery time are ( −2 . 82 , –0 . 27 ) , ( −3 . 6 , –0 . 77 ) , and ( −0 . 15 , 0 . 78 ) ; ( b ) recovery of VOLT occurred more than 30 min before NBCK and DSST are 55% and 50% , respectively; the 90% corresponding credible intervals of the difference in recovery time are ( −2 . 38 , 1 . 32 ) and ( −1 . 69 , 0 . 74 ) ; ( c ) recovery of AM occurred more than 30 min before VOLT , NBCK , and DSST are 98% , 99% , and 100% , respectively; the 90% corresponding credible intervals of the difference in recovery time are ( −2 . 92 , –0 . 47 ) , ( −3 . 66 , –0 . 91 ) , and ( −2 . 47 , –1 . 15 ) ; ( d ) recovery of DSST occurred more than 30 min before NBCK is 42%; the 90% corresponding credible interval of the difference in recovery time is ( −1 . 55 , 1 . 09 ) . For speed ( Figure 3 ) , the probability was high that ( i ) recovery of AM occurred more than 30 min before PVT ( 68% ) , VOLT ( 64% ) , and DSST ( 52% ) ; ( ii ) recovery of NBCK occurred more than 30 min before PVT ( 74% ) , VOLT ( 66% ) , and DSST ( 63% ) ; ( iii ) recovery of MP occurred more than 30 min before PVT ( 81% ) ; and that ( iv ) recovery of DSST occurred more than 30 min before PVT ( 53% ) . We had further hypothesized that , if there were differential rates of recovery in cognitive domains , higher order executive function ( as tested by AM ) would be most impaired and consequently would be the last to recover . Contrary to this expectation , AM was one of the least impaired of the tests at ROC . For both accuracy and speed measures , performance on AM quickly approached its pre-anesthetic level . The results from the 10 , 000 Markov Chain Monte Carlo samples yield the following additional insights with respect to speed of recovery . The recovery of NBCK ( as measured by speed to task completion ) is strongly impaired . Most samples did not recover over the experimental time course , a fact mirrored in the simulations . For the other five tests , the posterior probability that: ( a ) recovery of MP speed occurred more than 30 min before AM , PVT , and DSST are 54% , 91% , and 83% , respectively; the 90% corresponding credible intervals of the difference in recovery time are ( −1 . 31 , 0 . 24 ) , ( −3 . 95 , –0 . 02 ) , and ( −1 . 64 , –0 . 24 ) ; ( b ) recovery of VOLT speed occurred more than 30 min before MP , AM , PVT , and DSST are 82% , 98% , 100% , and 100% , respectively; the 90% corresponding credible intervals of the difference in recovery time are ( −1 . 97 , –0 . 12 ) , ( −2 . 35 , –0 . 79 ) , ( −4 . 84 , –1 . 17 ) , and ( −2 . 66 , –1 . 33 ) ; ( c ) recovery of AM speed occurred more than 30 min before PVT and DSST are 79% and 40% , respectively; the 90% corresponding credible intervals of the difference in recovery time are ( −3 . 49 , 0 . 10 ) and ( −0 . 92 , 0 . 10 ) ; ( d ) recovery of DSST occurred more than 0 . 5 hr before PVT is 64%; the 90% corresponding credible interval of the difference in recovery time is ( −3 . 02 , 0 . 47 ) . As expected , for all tests , the maximal degree of impairment was upon ROC . Information on the trajectory of recovery for an individual cognitive test is depicted in Figures 2 and 3 , and was assessed using a non-linear mixed effects model . In the 3 hr follow-up period , accuracy for those in the anesthesia group increased gradually for some tests ( DSST , PVT ) and more rapidly for others ( AM , NBCK , MP , VOLT ) . At 3 hr after ROC in the anesthetized group , accuracies in five out of six tests were not statistically significantly different from the control group . There remained a significant ( p<0 . 001 ) , albeit small , difference in accuracy performance between the anesthetized and the control group on the VOLT at 3 hr . Overall , within 3 hr of return of consciousness , the anesthetized group returned to an accuracy level that was not substantially different from that of participants who were not anesthetized . The comparison with the non-anesthetized control group is highly informative , because assessing performance on cognitive tests over time is confounded by learning . Had the anesthetized group simply returned to , or even exceeded , their own baseline performance after 3 hr of testing , this would not have provided sufficient evidence to conclude that their cognition had truly returned to baseline . The inclusion of an awake control group strengthens the conclusion that the anesthetized group did not experience a decrement in their performance that might have been masked by learning , which is known to occur with repeated testing . To account for a trade-off in accuracy versus speed , we also evaluated speed of task performance ( Figure 3 ) . Speed was also most strongly impaired at ROC , with a drop of more than 5 SD for all tests but the NBCK . At 3 hr after ROC in the anesthetized group , speed in two ( MP and PVT ) out of six tests were not statistically significantly different from the control group . There remained significant ( p<0 . 001 ) , albeit small , differences between the anesthetized and the control group in speed performance on four ( VOLT , NBCK , AM , DSST ) tests at 3 hr . The results of the nonlinear mixed-effects models for cognitive performance in anesthetized and non-anesthetized cohorts are summarized in Table 2 . We assessed cortical dynamics before , during , and after anesthetic exposure using local measures of permutation entropy ( PE ) and global measures of Lempel-Ziv complexity ( LZC ) . Using a linear mixed model , the PE demonstrated significant differences associated with behavioral states ( F9 , 86 = 42 . 423 , p<0 . 001 ) , brain regions ( F1 , 257 = 4 . 275 , p=0 . 040 ) , and the interaction between them ( F9 , 85 = 2 . 750 , p=0 . 007 ) . As compared to the baseline condition of eyes-closed resting state , frontal PE decreased at propofol-induced loss of consciousness ( LOC ) , further decreased during maintenance of the anesthetized state with isoflurane anesthesia ( p<0 . 001 , –0 . 160 [-0 . 185 to −0 . 135] , maintenance vs . EC1 ) , and returned to or even exceeded the baseline level just before the recovery of consciousness ( ROC ) ( p=0 . 002 , 0 . 036 [0 . 018 to 0 . 055] , pre-ROC vs . EC1 ) ( Figure 4A and B ) . Posterior PE did not show significant changes at LOC but was decreased during the maintenance phase ( p<0 . 001 , –0 . 110 ( -0 . 135 to −0 . 085 ) , maintenance vs . EC1 ) , and then returned to baseline level just before ROC ( Figure 4A and B ) . The topographic maps of PE exhibited region-specific patterns , in which frontal channels demonstrated significantly higher PE values as compared to posterior channels at the eyes-closed resting state directly after emergence ( EC2 , p=0 . 002 , 0 . 032 [0 . 016 to 0 . 048] , frontal vs . posterior ) ( Figure 4A ) . The LZC demonstrated significant state-dependent differences ( F9 , 50 = 59 . 364 , p<0 . 001 ) . As compared to baseline consciousness , LZC declined at LOC and decreased further during the maintenance phase ( p<0 . 001 , –0 . 258 [-0 . 285 to −0 . 231] , maintenance vs . EC1 ) , returning to baseline level just before ROC ( Figure 4C ) . Thus , both the local ( PE ) and global ( LZC ) metrics of cortical dynamics recover just as the brain is recovering consciousness . Similar state-dependent changes were observed despite different strategies in the parameter selection for PE ( Figure 4—figure supplement 1 ) , the threshold of binarization , and the method of generating surrogate data for LZC ( Figure 4—figure supplement 2 ) . Additionally , as expected , the non-anesthetized control group showed no differences among the seven resting-state eyes-closed epochs ( Figure 4—figure supplement 3 ) . We explored further whether the EEG measures during EC1 , Maintenance or pre-ROC were associated with the impairment of cognitive performance at emergence , where the maximal degree of impairment occurred . We found no evidence that the EEG measures during EC1 is associated with the impairment in accuracy or response time for all tasks ( Figure 4—figure supplement 4 ) . Of note , the LZC during Maintenance showed some correlation with the impairment of response time for MP ( r = 0 . 522 , p=0 . 004 ) and DSST ( r = 0 . 446 , p=0 . 016 ) tasks ( Figure 4—figure supplement 5 ) . Moreover , both the frontal PE and LZC were found to be weakly correlated with the impairment of accuracy for DSST task ( r = 0 . 381 , p=0 . 046 and r = 0 . 479 , p=0 . 011 ) ( Figure 4—figure supplement 6 ) . These data must be interpreted with caution given the multiple statistical comparisons and weak correlations . Average rest-activity patterns for all study subjects are shown in Figure 5A . As expected , time of day significantly affected inactivity in both groups ( F129 , 6867 = 46 . 24 , p<0 . 0001 ) . Cosinar analysis demonstrated that peak inactivity occurred between 3 and 4 am for both groups . Importantly , two-way ANOVA analysis revealed that over the week prior to the study day , there were no significant differences between the participants who would subsequently be anesthetized and those who would not ( F1 , 6865 = 3 . 70 , p>0 . 05 ) . Moreover , there were no significant interactions between time and group ( F149 , 6865 = 1 . 18 , p>0 . 05 ) . Analysis of rest activity resumed upon completion of the experiment on the study day . Actigraphy revealed a small yet statistically significant effect of anesthetic exposure , accounting for only 0 . 19% of the total variance ( F1 , 3576 = 12 . 55 , p=0 . 0004 ) . This was attributable to an increase in inactivity in the isoflurane-exposed volunteers during the initial 2 hr of the early evening ( Figure 5B ) . Activity patterns were not distinguishable after that time period . As before the study day , time remained as a highly significant predictor of inactivity accounting for 42% of the variance ( F77 , 3576 = 35 . 78 , p<0 . 0001 ) . This is the most comprehensive and controlled study assessing cognitive recovery from the anesthetized state in healthy humans . Although there have been numerous studies of recovery from anesthetic-induced unconsciousness ( Långsjö et al . , 2012; Purdon et al . , 2013; Chennu et al . , 2016; Kim et al . , 2018; Banks et al . , 2020; Scheinin et al . , 2021 ) , our investigation was unique in that the length and depth of general anesthesia was consistent with surgical conditions but the experimental paradigm did not include any surgical intervention . This allowed us to study cognitive reconstitution after a major perturbation of arousal state while also informing the ongoing controversy regarding the effects of general anesthesia on human cognition . We have demonstrated that reconstitution of cognition after general anesthesia is a process that unfolds over time rather than a discrete , singular event . Cognitive recovery also follows a counterintuitive sequence . Executive function was found to recover early , whereas tests of processing speed , attention , and reaction time had a more prolonged recovery . Furthermore , EEG-based measures of cortical dynamics return to baseline just prior to the recovery of consciousness after general anesthesia , with entropy in frontal cortex statistically significantly higher than posterior cortex just after emergence . There were not , however , strong predictors of cognitive recovery based on EEG dynamics . Finally , sleep-wake activity patterns are essentially unperturbed after anesthetic exposure . These findings are summarized in Figure 6 . As described in the introduction , our hypothesis was motivated by preliminary evidence in the literature that neurocognitive recovery after general anesthesia proceeds in a caudal-to-rostral direction ( Långsjö et al . , 2012; Reshef et al . , 2019 ) , which would imply that executive functions mediated by anterior cortical structures would be the last to return . However , the finding that performance on the abstract matching test displayed more robust recovery than other cognitive functions indicates that there is early engagement of prefrontal cortex after recovery of consciousness . Furthermore , EEG analysis demonstrated that both anterior and posterior cortices were active in association with recovery of consciousness , although there was no statistically significant relationship between neurophysiologic complexity in prefrontal cortex and subsequent executive function performance . Collectively , our results are aligned with the theoretical framework of global neuronal workspace theory , which attempts to describe the neurocognitive mechanisms of conscious access ( Dehaene et al . , 1998; Dehaene and Changeux , 2011; Mashour et al . , 2020 ) . Access consciousness refers to the broad cognitive availability of sensory information , which often is measured based on report or task performance ( as in this study ) , and is distinct from phenomenal consciousness , which has been defined as experience itself and hypothesized to be independent of post-perceptual processing in anterior cortex ( Block , 2005; Raccah et al . , 2021 ) . According to workspace theory , a reverberant network that includes both prefrontal and posterior parietal cortex serves to sustain , amplify , and make available neural representations , with a ‘broadcasting’ effect that allows access by otherwise non-conscious cognitive processors ( Mashour et al . , 2020 ) . The active frontal-parietal networks at return of consciousness and early recovery of a cognitive test associated with dorsolateral prefrontal cortex are consistent with at least a partially reintegrated global neuronal workspace . However , a fully intact workspace requires exquisite temporal coordination . We hypothesize that the differential temporal recovery of cognitive functions is a manifestation of differential temporal dynamics in the neural structures and cognitive processors across the workspace , which appear to return to , or be asymptotically close to , baseline approximately 3 hr after recovery of consciousness . It is important to note that the first evidence of conscious access during recovery was hand squeezing to command , which is a simple , binary task unlike the more challenging cognitive battery with continuous performance measures such as accuracy and speed . Thus , hand squeezing to command would not be robust evidence for a full reintegration of the global neuronal workspace . Further work is needed to determine if global neuronal workspace theory is sufficient to describe the reconstitution of consciousness and cognition after a major perturbation such as a general anesthesia , or whether other cognitive theories need to be employed . This study is consistent with our prior analysis of source-localized alpha oscillations and graph-theoretical variables such as global efficiency and modularity , which were identified in a subset of participants who had high-density EEG and which also recovered within 3 hr after emergence from general anesthesia ( Blain-Moraes et al . , 2017 ) . From a clinical perspective , it is remarkable that within 3 hr of recovery after a prolonged and deep general anesthetic , participants were performing a variety of complex cognitive tasks with similar accuracy and speed in comparison to participants who had not been anesthetized . On a shorter timescale , both local and global dynamic measures of cortical activity returned to baseline levels just prior to the return of responsiveness after general anesthesia . On a longer timescale , there was no evidence of disrupted sleep in the days following anesthetic exposure compared to non-anesthetized controls . This latter finding is striking considering that actigraphy measures of sleep are sensitive to stress , low alcohol exposure , and even sedative effects of non-alcoholic beer ( Mezick et al . , 2009; Geoghegan et al . , 2012; Franco et al . , 2012 ) . The data suggest that , in healthy humans , higher cognition and arousal states recover uneventfully after deep general anesthesia . The recovery of cortical dynamics in anterior and posterior cortical areas just prior to the return of consciousness in our experimental paradigm should be considered in light of prior studies focused on emergence from general anesthesia . One study of healthy volunteers anesthetized with either propofol or dexmedetomidine and imaged using positron emission tomography found that those responsive to command exhibited activation of subcortical arousal centers with only limited frontal-parietal cortex involvement ( Långsjö et al . , 2012 ) a recent study with a more sophisticated pharmacologic dosing strategy has largely confirmed these findings ( Scheinin et al . , 2021 ) . These data are consistent with a study of functional magnetic resonance imaging that demonstrated a transient activation of brainstem loci upon emergence from propofol sedation ( Nir et al . , 2019 ) as well as a positron emission tomography study revealing subcortical and posterior cortical activation in association with the reversal of propofol sedation using the acetylcholinesterase inhibitor physostigmine ( Xie et al . , 2011 ) . However , it is important to note that all of these studies involved experimental conditions in which there was ongoing exposure to a sedative-hypnotic , coupled with pharmacological or behavioral stimulation . By contrast , our paradigm reflects spontaneous emergence with residual isoflurane levels predicted to be 1 to 4 orders of magnitude below those required for hypnosis , which likely accounts for evidence of the robust return of cortical dynamics . However , it is worth noting that—in addition to the subcortical sites identified in human and animal studies ( Kelz et al . , 2019 ) —the prefrontal cortex might play a critical role in the control of arousal states of relevance to general anesthesia . One animal study demonstrated that cholinergic stimulation of medial prefrontal cortex , but not two sites in posterior parietal cortex , was sufficient to reverse the anesthetized state despite continuous administration of clinically relevant concentrations of sevoflurane ( Pal et al . , 2018 ) . Thus , the prefrontal cortex might play a critical role in recovery of both consciousness ( medial prefrontal ) and cognition ( dorsolateral prefrontal ) . Strengths of this study include: ( 1 ) surgically relevant anesthetic concentrations and duration of anesthetic exposure , in the absence of the confound of surgery itself; ( 2 ) multicenter design with substantially more participants than are typically found in studies of anesthetic-induced unconsciousness; ( 3 ) a non-anesthetized population to control for learning effects of cognitive assessment as well as comparison of sleep-wake patterns; and ( 4 ) complement of neurophysiological , cognitive , and behavioral measures on different time scales . Limitations of this study that constrain interpretation include: ( 1 ) young healthy population that precludes extrapolation of findings to older , younger , or sicker surgical patients encountered in clinical care; ( 2 ) only one anesthetic regimen tested , albeit a clinically common one , with unclear relevance to other perturbations such as sleep or pathological disorders of consciousness; ( 3 ) inability to blind participants to anesthetized and non-anesthetized conditions , which could potentially influence results; ( 4 ) relatively sparse EEG channels that were found to be a reliable sources of data across all participants; ( 5 ) no assessment of sleep macroarchitecture ( i . e . rapid eye movement sleep vs . slow-wave sleep ) ; ( 6 ) no longer term cognitive or behavioral assessment beyond the first three days of post-anesthetic actigraphy; and ( 7 ) inability to assess source-localized brain regions , subcortical regions , or resting-state networks . In conclusion , this study establishes neurophysiologic , cognitive , and behavioral recovery patterns after a surgically relevant general anesthetic in human volunteers . The rapid recovery of cortical dynamics just prior to recovering consciousness , the restored accuracy of executive function and multiple cognitive functions within 3 hr of emergence , and the normal sleep-wake patterns in the days following the experiment provide compelling evidence that the healthy brain is resilient to the effects of even deep general anesthesia . The findings also suggest that the immediate and persistent cognitive dysfunction identified after general anesthesia in healthy animals ( Culley et al . , 2004; Valentim et al . , 2008; Carr et al . , 2011; Callaway et al . , 2012; Zurek et al . , 2012; Jevtovic-Todorovic et al . , 2013; Zurek et al . , 2014; Jiang et al . , 2017 ) does not necessarily translate to healthy humans and that postoperative neurocognitive disorders might relate to factors other than general anesthesia , such as surgery or patient comorbidity ( Wildes et al . , 2019; Krause et al . , 2019 ) . Finally , our data are consistent with the global neuronal workspace theory of conscious access .	Anesthesia is a state of reversable , controlled unconsciousness . It has enabled countless medical procedures . But it also serves as a tool for scientists to study how the brain regains consciousness after disruptions such as sleep , coma or medical procedures requiring general anesthesia . It is still unclear how exactly the brain regains consciousness , and less so , why some patients do not recover normally after general anesthesia or fail to recover from brain injury . To find out more , Mashour et al . studied the patterns of reemerging consciousness and cognitive function in 30 healthy adults who underwent general anesthesia for three hours . While the volunteers were under anesthesia , their brain activity was measured with an EEG; and their sleep-wake activity was measured before and after the experiment . Each participant took part in a series of cognitive tests designed to measure the reaction speed , memory and other functions before receiving anesthesia , right after the return of consciousness , and then every 30 minutes thereafter . Thirty healthy volunteers who did not have anesthesia also completed the scans and tests as a comparison group . The experiments showed that certain normal EEG patterns resumed just before a person wakes up from anesthesia . The return of thinking abilities was an extended , multistep process , but volunteers recovered their cognitive abilities to nearly the same level as the volunteers within three hours of being deeply anesthetized . Mashour et al . also unexpectedly found that abstract problem-solving resumes early in the process , while other functions such as reaction time and attention took longer to recover . This makes sense from an evolutionary perspective . Sleep leaves individuals vulnerable . Quick evaluation and decision-making skills would be key to respond to a threat upon waking . The experiments confirm that the front of the brain , which handles thinking and decision-making , was especially active around the time of recovery . This suggests that therapies targeting this part of the brain may help people who experience loss of consciousness after a brain injury or have difficulties waking up after anesthesia . Moreover , disorders of cognition , such as delirium , in the days following surgery may be caused by factors other than the lingering effects of anesthetic drugs on the brain .	[ "Abstract", "Introduction", "Materials", "and", "methods", "Results", "Discussion" ]	[ "medicine", "neuroscience" ]	2021	Recovery of consciousness and cognition after general anesthesia in humans
Unpaired ligands are secreted signals that act via a GP130-like receptor , domeless , to activate JAK/STAT signalling in Drosophila . Like many mammalian cytokines , unpaireds can be activated by infection and other stresses and can promote insulin resistance in target tissues . However , the importance of this effect in non-inflammatory physiology is unknown . Here , we identify a requirement for unpaired-JAK signalling as a metabolic regulator in healthy adult Drosophila muscle . Adult muscles show basal JAK-STAT signalling activity in the absence of any immune challenge . Plasmatocytes ( Drosophila macrophages ) are an important source of this tonic signal . Loss of the dome receptor on adult muscles significantly reduces lifespan and causes local and systemic metabolic pathology . These pathologies result from hyperactivation of AKT and consequent deregulation of metabolism . Thus , we identify a cytokine signal that must be received in muscle to control AKT activity and metabolic homeostasis . JAK/STAT activating signals are critical regulators of many biological processes in animals . Originally described mainly in immune contexts , it has increasingly become clear that JAK/STAT signalling is also central to metabolic regulation in many tissues ( Dodington et al . , 2018; Villarino et al . , 2017 ) . One common consequence of activation of JAK/STAT pathways in inflammatory contexts is insulin resistance in target tissues , including muscle ( Kim et al . , 2013; Mashili et al . , 2013 ) . However , it is difficult to describe a general metabolic interaction between JAK/STAT and insulin signalling in mammals , due to different effects at different developmental stages , differences between acute and chronic actions , and the large number of JAKs and STATs present in mammalian genomes ( Dodington et al . , 2018; Mavalli et al . , 2010; Nieto-Vazquez et al . , 2008; Vijayakumar et al . , 2013 ) . The fruit fly Drosophila melanogaster has a single , well-conserved JAK/STAT signalling pathway . The unpaired ( upd ) genes upd1-3 encode the three known ligands for this pathway; they signal by binding to a single common GP130-like receptor , encoded by domeless ( dome ) ( Agaisse et al . , 2003; Brown et al . , 2001; Chen et al . , 2002 ) . Upon ligand binding , the single JAK tyrosine kinase in Drosophila , encoded by hopscotch ( hop ) , is activated; Hop then activates the single known STAT , STAT92E , which functions as a homodimer ( Binari and Perrimon , 1994; Chen et al . , 2002; Hou et al . , 1996; Yan et al . , 1996 ) . This signalling pathway plays a wide variety of functions , including segmentation of the early embryo , regulation of hematopoiesis , maintenance and differentiation of stem cells in the gut , and immune modulation ( Amoyel and Bach , 2012; Myllymäki and Rämet , 2014 ) . Importantly , several recent studies indicate roles for upd cytokines in metabolic regulation; for example , upds are important nutrient-responsive signals in the adult fly ( Beshel et al . , 2017; Rajan and Perrimon , 2012; Woodcock et al . , 2015; Zhao and Karpac , 2017 ) . Here , we identify a physiological requirement for Dome signalling in adult muscle . We observe that adult muscles show significant JAK/STAT signalling activity in the absence of obvious immune challenge . Plasmatocytes are a source of this signal . Inactivation of dome on adult muscles significantly reduces lifespan and causes muscular pathology and physiological dysfunction; these result from remarkably strong AKT hyperactivation and consequent dysregulation of metabolism . We thus describe a new role for JAK/STAT signalling in adult Drosophila muscle with critical importance in healthy metabolic regulation . To find physiological functions of JAK/STAT signalling in the adult fly , we identified tissues with basal JAK/STAT pathway activity using a STAT-responsive GFP reporter ( 10xSTAT92E-GFP ) ( Bach et al . , 2007 ) . The strongest reporter activity we observed was in legs and thorax . We examined flies also carrying a muscle myosin heavy chain RFP reporter ( MHC-RFP ) and observed co-localization of GFP and RFP expression in the muscles of the legs , thorax and body wall ( Figure 1—figure supplement 1 ) . We observed strong , somewhat heterogeneous reporter expression in all the muscles of the thorax and the legs , with strong expression in various leg and jump muscles and apparently weaker expression throughout the body wall muscles and indirect flight muscles ( Figure 1A ) . dome encodes the only known Drosophila STAT-activating receptor . To investigate the physiological role of this signal , we expressed domeΔ , a dominant-negative version of Dome lacking the intracellular signalling domain , with a temperature-inducible muscle specific driver line ( w;tubulin-Gal80ts;24B-Gal4 ) ( Figure 1—figure supplement 1 ) ( Brown et al . , 2001 ) . Controls ( 24B-Gal80ts/+ ) and experimental flies ( 24B-Gal80ts > dome∆ ) were raised at 18°C until eclosion to permit Dome activity during development . Flies were then shifted to 29°C to inhibit Dome activity and their lifespan was monitored . Flies with Dome signalling inhibited in adult muscles were short-lived ( Figure 1B , Figure 1—figure supplement 1 ) . This effect was also observed , more weakly , in flies kept at 25°C ( Figure 1—figure supplement 1 ) . Upd-JAK-STAT signalling is important to maintain gut integrity , and defects in gut integrity often precede death in Drosophila; however , our flies did not exhibit loss of gut integrity ( Figure 1—figure supplement 1 ) ( Jiang et al . , 2009; Rera et al . , 2012 ) . To determine whether Dome inhibition caused meaningful physiological dysfunction , we assayed climbing activity in 24B-Gal80ts/+ control flies and 24B-Gal80ts > dome∆ flies . 24B-Gal80ts > dome∆ flies showed significantly impaired climbing compared to controls ( Figure 1C ) . Adult muscle-specific expression of domeΔ with a second Gal4 line ( w;tub-Gal80ts;Mef2-Gal4 ) gave a similar reduction in lifespan and decline in climbing activity , confirming that the defect resulted from a requirement for Dome activity in muscle ( Figure 1—figure supplement 1 ) . Impaired muscle function is sometimes accompanied by lipid accumulation ( Baik et al . , 2017 ) . Therefore , we stained thorax muscles with the neutral lipid dye LipidTox . In 14 day old flies , we detected numerous small neutral lipid inclusions in several muscles , including the large jump muscle ( TTM ) , of 24B-Gal80ts > dome∆ flies ( Figure 1D ) . Having observed lipid inclusions in adult muscles , we analysed the systemic metabolic state of 24B-Gal80ts > dome∆ flies . We observed significant reductions in total triglyceride , glycogen and free sugar ( glucose + trehalose ) in these animals ( Figure 1E , F ) . Reduced free sugar could result from increased cellular sugar uptake . Increased uptake of sugars by peripheral tissues should be reflected in increased metabolic stores or metabolic rate . Since metabolic stores were decreased in our flies , we tested metabolic rate by measuring respiration . CO2 production and O2 consumption were both significantly increased in 24B-Gal80ts > dome∆ flies , indicating an overall increase in metabolic rate ( Figure 1G ) . dome acts via hop to regulate AKT activity with little effect on other nutrient signalling pathways . The observed metabolic changes imply differences in activity of nutrient-regulated signalling pathways in 24B-Gal80ts > dome∆ flies . Several signalling pathways respond to nutrients , or their absence , to coordinate energy consumption and storage ( Britton et al . , 2002; Lizcano et al . , 2003; Ulgherait et al . , 2014 ) . Of these , insulin signalling via AKT is the primary driver of sugar uptake by peripheral tissues . We examined the activity of these signalling mechanisms in legs ( a tissue source strongly enriched in muscle ) from 24B-Gal80ts > dome∆ flies . We found an extremely strong increase in abundance of the 60 kDa form of total and activated ( S505-phosphorylated ) AKT ( Figure 1H , I ) . This change was also seen in legs from Mef2-Gal80ts > dome∆ flies , confirming that dome functions in muscles ( Figure 1—figure supplement 1 ) . We also saw this effect in flies carrying a different insertion of the dome∆ transgene , under the control of a third muscle-specific driver , MHC-Gal4 , though the effect was weaker; the weakness of this effect may be a consequence of the fact that the MHC-Gal4 driver is not expressed in visceral muscle ( Bland et al . , 2010 ) ( Figure 1—figure supplement 1 ) . These MHC-Gal4 >dome∆ ( II ) animals were also short-lived relative to controls ( Figure 1—figure supplement 1 ) . Elevated total AKT could result from increased transcript abundance or changes in protein production or stability . We distinguished between these possibilities by assaying Akt1 mRNA; Akt1 transcript levels were elevated in 24B-Gal80ts > dome∆ muscle , but only by about 75% , suggesting that the large effect on AKT protein abundance must be , at least in part , post-transcriptional ( Figure 1—figure supplement 1 ) . Similarly , AKT hyperactivation could be driven by insulin-like peptide overexpression; however , we assayed the expression of Ilp2-7 in whole flies and observed that none of these peptides were significantly overexpressed ( Figure 1—figure supplement 1 ) . Next , we analysed if the feeding behaviour is affected by the muscle-specific dome loss , but we could not find a difference in food uptake in 24B-Gal80ts > dome∆ flies compared to controls ( Figure 1—figure supplement 1 ) . Unlike AKT , the amino-acid-responsive TORC1/S6K and the starvation-responsive AMPK pathway showed no significant difference in activity in 24B-Gal80ts > dome∆ flies ( Figure 1K , L ) . However , flies with AMPK knocked down in muscle did exhibit mild AKT hyperactivation ( Figure 2—figure supplement 1 ) . To identify signalling mediators acting between Dome and AKT , we first tested activity of the MAPK-ERK pathway , which can act downstream of the JAK kinase Hop ( Luo et al . , 2002 ) . We found an insignificant reduction in ERK activity in 24B-Gal80ts > dome∆ flies ( Figure 1M ) . We then assayed survival and AKT activity in flies with hop ( JAK ) , Dsor1 ( MEK ) and rl ( ERK ) knocked down in adult muscle . rl and Dsor1 knockdown gave mild or no effect on survival and pAKT ( Figure 2—figure supplement 1 ) . In contrast , hop knockdown gave a mild phenocopy of dome∆ with regards to survival and pAKT ( Figure 2—figure supplement 1 ) . We further analysed the requirement for hop in muscle dome signalling by placing 24B-Gal80ts > dome∆ on a genetic background carrying the viable gain-of-function allele hopTum-l . Flies carrying hopTum-l alone exhibited no change in lifespan , AKT phosphorylation , or muscle lipid deposition ( Figure 2A–C ) . However , hopTum-l completely rescued lifespan and pAKT levels in 24B-Gal80ts > dome∆ flies ( Figure 2D , E ) , indicating that the physiological activity of muscle Dome is mediated via Hop and that signal is required , but not sufficient , to control muscle AKT activity . The phenotype of 24B-Gal80ts > dome∆ flies is similar to that previously described in flies with loss of function in Pten or foxo ( Demontis and Perrimon , 2010; Mensah et al . , 2015 ) , suggesting that AKT hyperactivation might cause the dome loss of function phenotype; however , to our knowledge , direct activation of muscle AKT had not previously been analysed . We generated flies with inducible expression of activated AKT ( myr-AKT ) in adult muscles ( w;tubulin-Gal80ts/+;24B-Gal4/UAS-myr-AKT [24B-Gal80ts > myr-AKT] ) ( Stocker et al . , 2002 ) . These animals phenocopied 24B-Gal80ts > dome∆ flies with regards to lifespan , climbing activity , metabolite levels , metabolic rate , and muscle lipid deposition ( Figure 3A–F ) . We concluded that AKT hyperactivation could cause the pathologies seen in 24B-Gal80ts > dome∆ flies . Therefore , we tested whether reducing AKT activity could rescue 24B-Gal80ts > dome∆ flies . We generated flies carrying muscle-specific inducible dominant negative dome ( UAS-dome∆ ) with dsRNA against Akt1 ( UAS-AKT-IR ) . These flies showed significantly longer lifespan than 24B-Gal80ts > dome∆ and 24B-Gal80ts > AKT IR flies , similar to all control genotypes analyzed ( Figure 3G ) . Dome and AKT antagonism synergised to control the mRNA level of dome itself , further suggesting strong mutual antagonism between these pathways ( Figure 3—figure supplement 1 ) . AKT hyperactivation should reduce FOXO transcriptional activity . To test whether this loss of FOXO activity caused some of the pathologies observed in 24B-Gal80ts > dome∆ flies , we increased foxo gene dosage by combining 24B-Gal80ts > dome∆ with a transgene carrying a FOXO-GFP fusion protein under the control of the endogenous foxo regulatory regions . These animals exhibited rescue of physiological defects and lifespan compared to 24B-Gal80ts > dome∆ flies ( Figure 3H–J ) . They also exhibited increased dome expression ( Figure 3—figure supplement 1 ) . The effects of these manipulations on published foxo target genes were mixed ( Figure 3—figure supplement 1 ) ; the strongest effect we observed was that Dome blockade increased upd2 expression ( Figure 3—figure supplement 1 ) , consistent with the observation that FOXO activity inhibits upd2 expression in muscle ( none of the other genes tested have been shown to be FOXO targets in muscle ) ( Zhao and Karpac , 2017 ) . This may explain some of the systemic effects of Dome blockade . The effect of the foxo transgene was stronger than expected from a 1 . 5-fold increase in foxo expression , so we further explored the relationship between FOXO protein expression and AKT phosphorylation . We found that 24B-Gal80ts > dome∆ markedly increased FOXO-GFP abundance , so that the increase in total FOXO was much greater than 1 . 5-fold ( Figure 3—figure supplement 1 ) . This drove an apparent feedback effect , restoring AKT in leg samples of foxoGFP;24B-Gal80ts > dome∆ flies to near-normal levels ( Figure 3—figure supplement 1 ) . We also analysed expression of the catabolic hormone Akh and its putative targets bmm , Hsl , plin1 and plin2 in 24B-Gal80ts > dome∆ animals ( Figure 3—figure supplement 1 ) . We observed no clear regulation of Akh itself or of Hsl , bmm , or plin2; plin1 was reduced in expression by expression of dome∆ . We conclude that the primary effect of muscle dome is on AKT-foxo signalling . Plasmatocytes—Drosophila macrophages—are a key source of upd3 in flies on high fat diet and in mycobacterial infection ( Péan et al . , 2017; Woodcock et al . , 2015 ) . Plasmatocytes also express upd1-3 in unchallenged flies ( Chakrabarti et al . , 2016 ) . We thus tested their role in activation of muscle Dome . We found plasmatocytes close to STAT-GFP-positive leg muscle ( Figure 4A , B ) . This , and the prior published data , suggested that plasmatocytes might produce relevant levels of dome-activating cytokines in steady state . We then overexpressed upd3 in plasmatocytes and observed a potent increase in muscle STAT-GFP activity ( Figure 4C ) , confirming that plasmatocyte-derived upd signals were able to activate muscle Dome . To determine the physiological relevance of plasmatocyte-derived signals , we assayed STAT-GFP activity in flies in which plasmatocytes had been depleted by expression of the pro-apoptotic gene reaper ( rpr ) using a temperature-inducible plasmatocyte-specific driver line ( w;tub-Gal80ts;crq-Gal4 ) . These animals exhibited a near-complete elimination of plasmatocytes within 24 hr of being shifted to 29°C ( Figure 4—figure supplement 1 ) . STAT-GFP fluorescence and GFP abundance were reduced in legs of plasmatocyte-depleted flies ( crq-Gal80ts > rpr ) compared to controls ( crq-Gal80ts/+ ) ( Figure 4D , E ) . Activity was not eliminated , indicating that plasmatocytes are not the only source of muscle STAT-activating signals , although these animals did exhibit a significant reduction in climbing activity ( Figure 4—figure supplement 1 ) . We then examined the lifespan of flies in which we had depleted plasmatocytes in combination with various upd mutations and knockdowns . Plasmatocyte depletion gave animals that were short-lived ( Figure 4F ) . ( This effect was different from that we previously reported , possibly due to changes in fly culture associated with an intervening laboratory move [Woodcock et al . , 2015] ) . Combining plasmatocyte depletion with null mutations in upd2 and upd3 did not significantly further reduce lifespan; upd2 upd3 mutants with plamatocytes intact exhibited near-normal lifespan ( Figure 4F ) . Similarly , plasmatocyte depletion drove muscle lipid accumulation , and upd2 upd3 mutation synergised with plasmatocyte depletion to further increase muscle lipid inclusions ( Figure 4G ) . Plasmatocyte depletion reduced free sugar levels as well as glycogen levels in the whole fly ( Figure 4—figure supplement 2 ) , but did not reduce the abundance of stored triglycerides ( Figure 4—figure supplement 2 ) . However , depleting plasmatocytes in upd2 upd3 mutants failed to recapitulate the effects of muscle Dome inhibition on whole-animal triglyceride , free sugar , and glycogen levels ( Figure 4—figure supplement 2 ) . This could be due to antagonistic effects of other plasmatocyte-derived signals . We attempted to pinpoint a specific Upd as the relevant physiological ligand by examining STAT-GFP activity , first testing mutants in upd2 and upd3 because upd1 mutation is lethal . However , these mutants , including the upd2 upd3 double-mutant , were apparently normal ( Figure 4—figure supplement 2 ) . We then tested plasmatocyte-specific knockdown of upd1 and upd3; these animals were also essentially normal ( Figure 4—figure supplement 2 ) , and plasmatocyte upd1 knockdown did not reduce lifespan ( Figure 4H ) . However , plasmatocyte-specific upd1 knockdown gave significant compensating increases in expression of upd2 and upd3 ( Figure 4I ) . In keeping with this , combining plasmatocyte-specific upd1 knockdown with mutations in upd2 and upd3 reduced lifespan ( Figure 4J ) and also reduced STAT-GFP activity in these flies ( Figure 4—figure supplement 2 ) . Our results indicate that plasmatocytes are an important physiological source of the Upd signal driving muscle Dome activity in healthy flies , and suggest that upd1 may be the primary relevant signal in healthy animals . However , plasmatocytes are not the only relevant source of signal , and Upd mutual regulation prevents us from pinpointing a single responsible signal . Here we show that upd-dome signalling in muscle acts via AKT to regulate physiological homeostasis in Drosophila . Loss of Dome activity in adult muscles shortens lifespan and promotes local and systemic metabolic disruption . Dome specifically regulates the level and activity of AKT; AKT hyper-activation mediates the observed pathology . Plasmatocytes are a primary source of the cytokine signal . In healthy adult flies , insulin-like peptides are the primary physiological AKT agonists . The effect we observe thus appears to be an example of a cytokine-Dome-JAK signal that impairs insulin function to permit healthy physiology . Our work fits into a recent body of literature demonstrating key physiological roles for JAK-STAT activating signals in Drosophila . Upd1 acts locally in the brain to regulate feeding and energy storage by altering the secretion of neuropeptide F ( NPF ) ( Beshel et al . , 2017 ) . Upd2 is released by the fat body in response to dietary triglyceride and sugar to regulate secretion of insulin-like peptides ( Rajan and Perrimon , 2012 ) . More recently , muscle-derived Upd2 , under control of FOXO , has been shown to regulate production of the glucagon-like signal Akh ( Zhao and Karpac , 2017 ) . Indeed , we observe that upd2 is upregulated in flies with Dome signalling blocked in muscle , possibly explaining some of the systemic metabolic effects we observe . Plasmatocyte-derived Upd3 in flies on a high fat diet can activate the JAK/STAT pathway in various organs including muscles and can promote insulin insensitivity ( Woodcock et al . , 2015 ) . Our observation that Upd signalling is required to control AKT accumulation and thus insulin pathway activity in healthy adult muscle may explain some of these prior observations and reveals a new role for plasmatocyte-derived cytokine signalling in healthy metabolic regulation . Several recent reports have examined roles of JAK/STAT signalling in Drosophila muscle . In larvae , muscle JAK/STAT signalling can have an effect opposite to the one we report , with pathway loss of function resulting in reduced AKT activity ( Yang and Hultmark , 2017 ) . It is unclear whether this difference represents a difference in function between developmental stages ( larva vs adult ) or a difference between acute and chronic consequences of pathway inactivation . Roles in specific muscle populations have also been described: for example , JAK/STAT signalling in adult visceral muscle regulates expression of Vein , an EGF-family ligand , to control intestinal stem cell proliferation ( Buchon et al . , 2010; Jiang et al . , 2011 ) ; the role of this system in other muscles may be analogous , controlling expression of various signals to regulate systemic physiology . Importantly , though we do not observe loss of gut integrity in our flies , it remains possible that the gut is an important mediator of some aspect of the physiological unpaired signal we document—either acting as an endocrine relay or via more subtle effects on gut physiology that could affect nutrient absorption . This would fit our observation that expression of dominant-negative Dome under the control of Mhc-Gal4 ( which does not express in visceral muscle ) gives a weaker effect on survival and AKT abundance than other muscle drivers and is particularly of interest given the documented role of plasmatocytes in regulation of gut homeostasis ( Ayyaz et al . , 2015 ) . The roles of mammalian JAK/STAT signalling in muscle physiology are more complex , but exhibit several parallels with the fly . In mice , early muscle-specific deletion of Growth Hormone Receptor ( GHR ) causes several symptoms including insulin resistance , while adult muscle-specific GHR deletion causes entirely different effects , including increased metabolic rate and insulin sensitivity on a high-fat diet ( Mavalli et al . , 2010; Vijayakumar et al . , 2013; Vijayakumar et al . , 2012 ) . GHR signals via STAT5; STAT5 deletion in adult skeletal muscle promotes muscle lipid accumulation on a high-fat diet ( Baik et al . , 2017 ) . Other STAT pathways can also play roles . For example , the JAK-STAT activating cytokine IL-6 , which signals primarily via STAT3 , increases skeletal muscle insulin sensitivity when given acutely but can drive insulin resistance when provided chronically ( Nieto-Vazquez et al . , 2008 ) . STAT3 itself can promote muscle insulin resistance ( Kim et al . , 2013; Mashili et al . , 2013 ) . The relationship between these effects and those we have shown here , and the mechanisms regulating Upd production by plasmatocytes during healthy physiology , remain to be determined . All fly stocks were maintained on food containing 10% w/v Brewer’s yeast , 8% fructose , 2% polenta and 0 . 8% agar supplemented with propionic acid and nipagin . Crosses for experiments were performed at 18°C ( for crosses with temperature inducible gene expression ) or 25°C . Flies were shifted to 29°C after eclosion where relevant . Male flies were used for all experiments . All flies were backcrossed onto our laboratory isogenic w1118 genetic background , with the exception of VDRC knockdown lines ( these lines are also on a uniform genetic background and could be compared with one another ) . All crosses were performed using driver females so that the male progeny used for experiments would have a uniform X chromosome . The following original fly stocks were used for crosses: Fly stocksDescription and originw1118; tubulin-Gal80ts/SM6a;24B-Gal4/TM6c , Sb1Temperature sensitive muscle specific driver line; 24B-Gal4 a gift of Nazif Alicw1118; tubulin-Gal80ts/SM6a;Mef2-Gal4/TM6c , Sb1Temperature sensitive muscle specific driver line; Mef2-Gal4 a gift of Michael Taylorw1118;;UAS-dome∆/TM6c , Sb1Line for expression of a dominant-negative dome , gift of James Castelli-Gair Hombríaw1118;UAS-dome∆/CyOLine for expression of a dominant-negative dome , gift of James Castelli-Gair Hombríaw1118;;UAS-myr-AKT/TM6c , Sb1Line for over-expression of a constitutive active ( myristoylated ) AKT , gift of Ernst Hafenw;UAS-AMPKα-IRVDRC KK106200w;UAS-AMPKβ-IRVDRC KK104489w;UAS-rl-IRVDRC KK109108w;UAS-Dsor1-IRVDRC KK102276w1118;foxoGFPExpresses GFP-tagged foxo fusion protein ( genomic rescue construct inserted at AttP40 ) . Bloomington Drosophila Stock Center ( BDSC ) 38644w;UAS-AKT-IRVDRC KK103703w1118;10xSTAT92E-GFPSTAT-GFP reporter line ( Bach et al . , 2007 ) . BDSC #26197w1118;MHC-Gal4 , MHC-RFP/SM6aMuscle specific driver line and muscle specific reporter line . Derived from BDSC #38464w upd2 Δ upd3Δ;;;Gift of Bruno Lemaitrew1118;;crq-Gal4/TM6c , Sb1Plasmatocyte specific driver line , gift of Nathalie Francw1118;tub-Gal80ts;TM2/TM6c , Sb1Line for ubiquitous expression of Gal80ts , BDSC #7108w1118;;UAS-rpr/TM6c , Sb1Line for over-expression of the pro-apoptotic protein rpr . Derived from BDSC #5824w1118;UAS-CD8-mCherryLine for overexpression of a CD8-mCherry fusion protein . Derived from BDSC #27391w1118;;srpHemo-3xmCherry/TM6c , Sb1Plasmatocyte reporter line ( Gyoergy et al . , 2018 ) w;UAS-hop-IRVDRC GD40037w;UAS-upd1-IR/SM6aVDRC GD3282w;UAS-upd3-IRVDRC GD6811w1118;;UAS-upd3/TM6c , Sb1Line for overexpression of upd3 , gift of Bruce Edgarw1118;UAS-2xeGFP/SM6aLine for expression of bicistronic GFP , BDSC #6874w1118 hopTum-L/FM7hGain-of function mutant of hop; derived by backcrossing from BDSC 8492 onto our control w1118 background Genotype abbreviations were used for the different experimental flies in this study , in the following table the complete genotypes are indicated: Genotype abbreviation of flies used in the manuscriptComplete genotype of flies used in the manuscript10XSTAT92E-GFP/MHC-RFPw1118;10xSTAT92E-GFP/MHC-Gal4 , MHC-RFP24B-Gal80ts/+w1118;tub-Gal80ts/+;24B-Gal4/+24B-Gal80ts > dome∆w1118;tub-Gal80ts/+;24B-Gal4/UAS-dome∆24B-Gal80ts > myr-AKTw1118;tub-Gal80ts/+;24B-Gal4/UAS-myr-AKT24B-Gal80ts > AMPKα-IRw1118;tub-Gal80ts/UAS-AMPKα-IR;24B-Gal4/+24B-Gal80ts > AMPKβ-IRw1118;tub-Gal80ts/UAS-AMPKβ-IR;24B-Gal4/+24B-Gal80ts > rl-IRw1118;tub-Gal80ts/UAS-rl-IR;24B-Gal4/+24B-Gal80ts > Dsor1-IRw1118;tub-Gal80ts/UAS-Dsor1-IR;24B-Gal4/+24B-Gal80 > hop-IRw1118;tub-Gal80ts/UAS-hop-IR;24B-Gal4/+hoptum-L;24B-Gal80 > dome∆w1118 hoptum-L;tub-Gal80ts/+;24B-Gal4/UAS-dome∆24B-Gal80ts > AKT-IRw1118;tub-Gal80ts/UAS-AKT-IR;24B-Gal4/+24B-Gal80ts > AKT-IR;dome∆w1118;tub-Gal80ts/UAS-AKT-IR;24B-Gal4/UAS-dome∆MHC-Gal4/+w1118;MHC-Gal4 , Mhc-RFP/+;MHC-Gal4 > dome∆ ( II ) w1118;MHC-Gal4 , MHC-RFP/UAS-dome∆;foxoGFP;24B-Gal80ts/+w1118;foxoGFP;tub-Gal80ts/+;24B-Gal4/+foxoGFP;24B-Gal80ts > dome∆w1118;foxoGFP;tub-Gal80ts/+;24B-Gal4/UAS-dome∆UAS-dome∆/+w1118;;UAS-dome∆/+UAS-AKT-IR/+w1118;UAS-AKT-IR/+;UAS-AKT-IR;dome∆/+w1118;UAS-AKT-IR/+; UAS-dome∆/+Mef2-Gal80ts/+w1118;tub-Gal80ts/+;Mef2-Gal4/+Mef2-Gal80ts > dome∆w1118;tub-Gal80ts/+;Mef2-Gal4/UAS-dome∆srpHemo-3xmCherryw1118;; srpHemo-3xmCherry/+crq-Gal4/+w1118;;crq-Gal4/+crq-Gal80ts > rprw1118;tub-Gal80ts/+;crq-Gal4/UAS-rpr or w1118;tub-Gal80ts/+;crq-Gal4 , UAS-CD8-mCherry , 10xSTAT92E-GFP/UAS-rprcrq-Gal80ts/+w1118;tub-Gal80ts/+;crq-Gal4/+ or w1118;tub-Gal80ts/+;crq-Gal4 , UAS-CD8-mCherry , 10xSTAT92E-GFP/+crq-Gal4/+w1118;;crq-Gal4 , UAS-CD8-mCherry , 10xSTAT92E-GFP/+crq-Gal4 > upd1-IRw1118;UAS-upd1-IR/+;crq-Gal4 , UAS-CD8- mCherry , 10xSTAT92E-GFP/+crq-Gal4 > upd3-IRw1118;UAS-upd3-IR/+;crq-Gal4 , UAS-CD8-mCherry , 10xSTAT92E-GFP/+crq-Gal4 > upd3w1118;;crq-Gal4 , UAS-CD8-mCherry , 10xSTAT92E-GFP/UAS-upd3upd2 Δ upd3Δ;crq-Gal80ts/+w upd2 Δ upd3Δ;tub-Gal80ts/+;crq-Gal4/+upd2 Δ upd3Δ;crq-Gal80ts > rprw upd2 Δ upd3Δ;tub-Gal80ts/+;crq-Gal4/UAS-rprupd2 Δ upd3Δ;upd1-IR/+w upd2 Δ upd3Δ;UAS-upd1-IR/+upd2 Δ upd3Δ;crq-Gal4/+w upd2 Δ upd3Δ;;crq-Gal4/+upd2 Δ upd3Δ;crq-Gal4 > upd1-IRw upd2 Δ upd3Δ;UAS-upd1-IR/+;crq-Gal4/+MHCYFP; srpHemo-3xmCherryw1118; MHCYFP/+;srpHemo-3xmCherry/+10xSTAT92E-GFP; srpHemo-3xmCherryw1118; 10xSTAT92E-GFP/+;srpHemo-3xmCherry/+ Male flies were collected after eclosion and groups of 20–40 age-matched flies per genotype were placed together in a vial with fly food . All survival experiments were performed at 29°C . Dead flies were counted daily . Vials were kept on their sides to minimize the possibility of death from flies becoming stuck to the food , and flies were moved to fresh food twice per week . Flies were transferred into new vials without CO2 anaesthesia . Male flies were collected after eclosion and housed for 14 days in age-matched groups of around 20 . The assay was performed in the morning , when flies were most active . Flies were transferred without CO2 into a fresh empty vial without any food and closed with the open end of another empty vial . Flies were placed under a direct light source and allowed to adapt to the environment for 20 min . Negative geotaxis reflex was induced by tapping the flies to the bottom of the tube and allowing them to climb up for 8 s . After 8 s the vial was photographed . This test was repeated 3 times per vial with 1 min breaks in between . The height each individual fly had climbed was measured in Image J , and the average between all three runs per vial was calculated . Male flies were aged for 7 days and changed on food supplemented with 0 . 1% bromophenol blue and 0 . 5% xylene cyanol . Control flies for each genotype were maintained on non-blue food for background subtraction . Flies were left on the blue or non-blue food for 4 hr at 29°C . After 4 hr flies were anaesthetized with CO2 and decapitated to avoid interference of the eye pigment with the measured absorbance . five flies were homogenized in 100 µl PBS per sample . The fly torsos were homogenized and centrifuged for 20 min at 12 . 000 rpm . The supernatant was collected and absorbance at 620 nm was analysed with a plate reader . For immunofluorescent staining of thorax muscles , we anaesthetized flies and removed the head , wings and abdomen from the thorax . Thorax samples were pre-fixed for 1 hr in 4% PFA rotating at room temperature . Thoraces were then halved sagitally with a razor blade and fixed for another 30 min rotating at room temperature . Samples were washed with PBS + 0 . 1% Triton X-100 , then blocked for 1 hr in 3% bovine serum albumin ( BSA ) in PBS + 0 . 1% Triton X-100 . For Lipid-Tox staining , samples were washed with PBS and stained for 2 hr at room temperature with HCS Lipid Tox Deep Red ( Thermo Fisher #H34477; 1:200 ) . For Phalloidin labelling , the samples were washed in PBS after fixation and stained for 2 hr at room temperature with Alexa Fluor 488 Phalloidin ( Thermo Fisher #A12379 , 1:20 ) . Afterwards the samples were washed once with PBS and mounted in Fluoromount-G . All mounted samples were sealed with clear nail polish and stored at 4°C until imaging . Imaging was performed in the Facility for Imaging by Light Microscopy ( FILM ) at Imperial College London and in the Institute of Neuropathology in Freiburg . A Leica SP5 and SP8 microscope ( Leica ) were used for imaging , using either the 10x/NA0 . 4 objective , or the 20x/NA0 . 5 objective . Images were acquired with a resolution of either 1024 × 1024 or 512 × 512 , at a scan speed of 400 Hz . Averages from 3 to 4 line scans were used , sequential scanning was employed where necessary and tile scanning was used in order to image whole flies . For imaging of whole live flies , the flies were anaesthetized with CO2 and glued to a coverslip . Flies were kept on ice until imaging . For measuring mean fluorescence intensity , a z-stack of the muscle was performed and the stack was projected in an average intensity projection . Next the area of the muscle tissue analyzed was defined and the mean fluorescent intensity within this area was measured . Images were processed and analysed using Image J . For RNA extraction three whole flies or three thoraces were used per sample . After anaesthetisation , the flies were smashed in 100 µl TRIzol ( Invitrogen ) , followed by a chloroform extraction and isopropanol precipitation . The RNA pellet was cleaned with 70% ethanol and finally solubilized in water . After DNase treatment , cDNA synthesis was carried out using the First Strand cDNA Synthesis Kit ( Thermo Scientific ) and priming with random hexamers ( Thermo Scientific ) . cDNA samples were further diluted and stored at −20° C until analysis . Quantitative Real-time PCR was performed with Sensimix SYBR Green no-ROX ( Bioline ) on a Corbett Rotor-Gene 6000 ( Corbett ) . The cycling conditions used throughout the study were as follows: Hold 95°C for 10 min , then 45 cycles of 95°C for 15 s , 59°C for 30 s , 72°C for 30 s , followed by a melting curve . All calculated gene expression values were measured in arbitrary units ( au ) according to diluted cDNA standards run in each run and for each gene measured . All gene expression values are further normalized to the value of the loading control gene , Rpl1 , prior to further analysis . The following primer sequences have been used in this study: Gene nameForwardReverseAkt15’-ctttgcgagtattaactggacaga-3’5’-ggatgtcacctgaggcttg-3’Ilp25’-atcccgtgattccaccacaag-3’5’-gcggttccgatatcgagtta-3’Ilp35’-caacgcaatgaccaagagaa-3’5’-tgagcatctgaaccgaact-3’Ilp45’-gagcctgattagactgggactg-3’5’-tggaccggctgcagtaac-3’Ilp55’-gccttgatggacatgctga-3’5’-agctatccaaatccgcca-3’Ilp65’-cccttggcgatgtatttcc-3’5’-cacaaatcggttacgttctgc-3’Ilp75’-cacaccgaggagggtctc-3’5’-caatatagctggcggacca-3’dome5’-cggactttcggtactccatc-3’5’-accttgatgaggccaggat-3’upd15’-gcacactgatttcgatacgg-3’5’- ctgccgtggtgctgtttt −3’upd25’-cggaacatcacgatgagcgaat-3’5’-tcggcaggaacttgtactcg-3’upd35’-actgggagaacacctgcaat-3’5’-gcccgtttggttctgtagat-3’Pepck15’-ggataaggtggacgtgaag-3’5’-acctcctgcgaccagaact-3’Thor5’-caggaaggttgtcatctcgga-3’5’-ggagtggtggagtagagggtt-3’InR5'-gcaccattataaccggaacc-3'5'-ttaattcatccatgacgtgagc-3'Akh5’- agccgtgctcttcatgct-3’5’-aaaggttccaggaccagctc-3’Hsl5’-cttggaaatacttgaggggttg-3’5’-agatttgatgcagttctttgagc-3’bmm5’-gtctcctctgcgatttgccat-3’5’-ctgaagggacccagggagta-3’plin15’-gcgttctatggtagccttcag-3’5’-gcgtccggatagaaagctg-3’plin25’-gcagaatggcaagagttctga-3’5’-actgtgtgtaggactggatcctc-3’Rpl15’-tccaccttgaagaagggcta-3’5’-ttgcggatctcctcagactt-3’ Smurf assays with blue-coloured fly food were performed to analyse gut integrity in different genotypes . Normal fly food , as described above , was supplemented with 0 . 1% Brilliant Blue FCF ( Sigma Aldrich ) . Experimental flies were placed on the blue-coloured fly food at 9AM and kept on the food for 2 hr at 29°C . After 2 hr , the distribution of the dye within the fly was analysed for each individual . Flies without any blue dye were excluded , flies with a blue gut or crop were identified as ‘non-smurf’ and flies which turned completely blue or showed distribution of blue dye outside the gut were classified as ‘smurf’ . Dissected legs or thoraces from three flies were used per sample and smashed in 75 µl 2x Laemmli loading buffer ( 100 mM Tris [pH 6 . 8] , 20% glycerol , 4% SDS , 0 . 2 M DTT ) . Samples were stored at −80°C until analysis . 7 . 5 µl of this lysate were loaded per lane . Blue pre-stained protein standard ( 11–190 kDa ) ( New England Biolabs ) was used . Protein was transferred to nitrocellulose membrane ( GE Healthcare ) . Membrane was blocked in 5% milk in TBST ( TBS + 0 . 1% Tween-20 ) . The following primary antibodies were used: anti-phospho ( Ser505 ) -AKT ( Cell Signal Technology ( CST ) 4054 , 1:1 , 000 ) , anti-AKT ( CST 4691 , 1:1 , 000 ) , anti-phospho ( Thr172 ) -AMPKα ( CST 2535 , 1:1 , 000 ) , anti-phospho ( Thr389 ) -p70 S6 kinase ( CST 9206 , 1:1 , 000 ) , anti-GFP ( CST 2956 , 1:1 , 000 ) , anti-phospho-p44/42 MAPK ( Erk1/2 ) ( CST 4370 , 1:1 , 000 ) and anti-α-tubulin ( clone 12G10 , Developmental Studies Hybridoma Bank , used as an unpurified supernatant at 1:3 , 000; used as a loading control for all blots ) . Primary antibodies were diluted in TBST containing 5% BSA and incubated over night at 4°C . Secondary antibodies were HRP anti-rabbit IgG ( CST 7074 , 1:5 , 000 ) and HRP anti-mouse IgG ( CST 7076 , 1:5 , 000 ) . Proteins were detected with Supersignal West Pico Chemiluminescent Substrate ( Thermo Scientific ) or Supersignal West Femto Chemiluminescent Substrate ( Thermo Scientific ) using a LAS-3000 Imager ( Fujifilm ) . Bands were quantified by densitometry using Image J . Quantifications reflect all experiments performed; representative blots from single experiments are shown . Groups of 10 flies were used per sample . After CO2 anaesthesia the flies were placed in 100 µl of ice-cold chloroform:methanol ( 3:1 ) . Samples were centrifuged for 3 min at 13 , 000 rpm at 4°C , and then flies were smashed with pestles followed by another centrifugation step . A set of standards were prepared using lard ( Sainsbury’s ) in chloroform:methanol ( 3:1 ) for quantification . Samples and standards were loaded onto a silica gel glass plate ( Millipore ) , and a solvent mix of hexane:ethyl ether ( 4:1 ) was prepared as mobile phase . Once the solvent front reached the top of the plate , the plate was dried and stained with an oxidising staining reagent containing ceric ammonium heptamolybdate ( CAM ) ( Sigma Aldrich ) . For visualization of the oxidised bands , plates were baked at 80°C for 20 min . Baked plates were imaged with a scanner and triglyceride bands were quantified by densitometry according to the measured standards using Image J . 5–7-day-old male flies , kept at 29°C , were used for the analysis . Flies were starved for 1 hr on 1% agar supplemented with 2% phosphate buffered saline ( PBS ) at 29°C before being manually smashed in 75 μl TE + 0 . 1% Triton X-100 ( Sigma Aldrich ) . three flies per sample were used . All samples were incubated at 75°C for 20 min and stored at −80°C . Samples were thawed prior to measurement and incubated at 65°C for 5 min to inactivate fly enzymes . A total of 10 μl per sample was loaded for different measurements into flat-bottom 96-well tissue culture plates . Each fly sample was measured four times , first diluted in water for calculation of background fly absorbance , second with glucose reagent ( Sentinel Diagnostics ) for the measurement of free glucose , third with glucose reagent plus trehalase ( Sigma Aldrich ) for trehalose measurement , and fourth with glucose reagent plus amyloglucosidase ( Sigma Aldrich ) for glycogen measurement . Plates were then incubated at 37°C for 1 hr before reading with a microplate reader ( biochrom ) at 492 nm . Quantities of glucose , trehalose and glycogen were calculated according to measured standards . Respiration in flies was measured using a stop-flow gas-exchange system ( Q-Box RP1LP Low Range Respirometer , Qubit Systems , Ontario , Canada , K7M 3L5 ) . Ten flies from each genotype were put into an airtight glass tube and supplied with our standard fly food via a modified pipette tip . Each tube was provided with CO2-free air while the ‘spent’ air was concurrently flushed through the system and analysed for its CO2 and O2 content . All vials with flies were normalized to a control vial with food but no flies inside . In this way , evolved CO2 per chamber and consumed O2 per chamber were measured for each tube every ~44 min ( the time required to go through each of the vials in sequence ) . For flow cytometric analysis of plasmatocytes , 90 flies per sample per genotype were anaesthetized and mechanically dissociated through a 100 µm mesh with 2 mM EDTA in PBS ( FACS buffer ) . The cell suspension is spun down and the resulting cell pellet was resuspended in 5 ml FACS buffer and again rinsed through a 100 µm mesh in a new tube . This washing step was repeated twice . Afterwards the cells were resuspended in 500 µl 2 mM EDTA and Fixable Viability Dye 780 ( ebioscience #65-0865-18 , 1:1000 ) . Samples were acquired on a FACS Canto II ( BD Biosciences ) and analyzed with FlowJo analysis software . For real-time quantitative PCR , TLCs , MFI quantification , western blot quantifications and colorimetric measurements for glucose , trehalose and glycogen levels an unpaired t-test or one-way ANOVA was used to calculate statistical significance , as noted in the figure legends . Respirometer data was analysed with a Mann-Whitney test . Lifespan/Survival assays , where analysed with the Log-Rank and Wilcoxon test . Stars indicate statistical significance as followed: p<0 . 05 , p<0 . 01 and **p<0 . 001 . All statistical tests were performed with Excel or GraphPad Prism software . All replicates are biological . No outliers were omitted , and all replicates are included in quantitations ( including in cases where a single representative experiment is shown ) . Flies were allocated into experimental groups according to their genotypes . Masking was not used . For survival experiments , typically , the 50% of flies that eclosed first from a given cross were used for an experiment . For smaller-scale experiments , flies were selected randomly from those of a given age and genotype .	The immune system helps animals fend off infections , but it also has a role in controlling the body’s metabolism – that is , the chemical reactions that sustain life . For instance , in fruit flies , high-fat diets can trigger the immune system , which results in cells becoming resistant to the hormone insulin and not being able to process sugar properly; this in turn leads to problems in sugar levels and shorter lifespans . This mechanism involves the release of an immune signal called unpaired-3 ( upd3 ) , which then binds to a receptor known as dome . Yet , it was unclear how exactly the immune system and metabolism work together , and whether their interactions are also important in flies on a normal diet . To investigate , Kierdorf et al . stopped the activity of the dome receptor in the muscles of healthy flies . This led to an increase in the activity of the enzyme AKT , a protein critical to relay insulin-type signals inside the cell . As a result , insulin signaling was hyperactivated in the tissue , leading to decreased muscle function , unhealthy changes in how energy was stored and spent , and ultimately , a shorter life for the insects . Further experiments also identified blood cells called plasmatocytes ( the flies’ equivalent of certain human immune cells ) as a key source of the upd signal . The findings by Kierdorf et al . shed a light on the fact that , even in healthy animals , complex interactions are required between the immune system and the metabolism . Further investigations will reveal if other body parts besides muscles rely on similar connections .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "immunology", "and", "inflammation" ]	2020	Muscle function and homeostasis require cytokine inhibition of AKT activity in Drosophila
TRPV1 channels are gated by a variety of thermal , chemical , and mechanical stimuli . We used optical recording of Ca2+ influx through TRPV1 to measure activity and mobility of single TRPV1 molecules in isolated dorsal root ganglion neurons and cell lines . The opening of single TRPV1 channels produced sparklets , representing localized regions of elevated Ca2+ . Unlike sparklets reported for L-type Ca2+ channels , TRPV4 channels , and AchR channels , TRPV1 channels diffused laterally in the plasma membrane as they gated . Mobility was highly variable from channel-to-channel and , to a smaller extent , from cell to cell . Most surprisingly , we found that mobility decreased upon channel activation by capsaicin , but only in the presence of extracellular Ca2+ . We propose that decreased mobility of open TRPV1 could act as a diffusion trap to concentrate channels in cell regions with high activity . The transient receptor potential vanilloid type 1 ( TRPV1 ) ion channel expressed in peripheral sensory neurons may be one of the most heavily regulated ion channels known . Its gating is multimodal , regulated by extracellular protons , phosphorylation by PKC , phosphorylation by PKA , temperature , fatty acids such as lysophosphatidic acid and anandamide , phosphoinositide signaling lipids , eicosanoids , and capsaicin , the pungent extract of hot chili peppers ( Caterina et al . , 1997; Zygmunt et al . , 1999; Premkumar and Ahern , 2000; Bhave et al . , 2002; Hu et al . , 2002; Pomonis et al . , 2003; Klein et al . , 2008; Ufret-Vincenty et al . , 2011; Nieto-Posadas et al . , 2012 ) . Furthermore , it is also subject to indirect regulation by G-protein coupled receptors and intracellular Ca2+ and its trafficking can be regulated by receptor tyrosine kinases ( Chuang et al . , 2001; Zhang et al . , 2005 , 2008; Stein et al . , 2006 ) . TRPV1 channels are highly permeable to Ca2+ , with PNa:PCa estimated to be at least 1:10 ( Caterina et al . , 1997 ) . Indeed , Ca2+ imaging at the whole-cell level was used in the seminal tour-de-force that first identified TRPV1 as the capsaicin-activated ion channel ( Caterina et al . , 1997 ) . The high permeability of TRPV1 to Ca2+ has also been a useful tool in high-throughput screening of regulatory compounds and led to the identification of a family of toxins , first purified from spiders , that act as potent activators of TRPV1 ( and have enhanced the survival of the spiders ) ( Caterina et al . , 1997 ) . In addition , Ca2+ influx through TRPV1 desensitizes sensory neurons ( Cholewinski et al . , 1993; Koplas et al . , 1997; Rosenbaum et al . , 2004 ) . Although multiple pathways are likely involved in neuronal desensitization , depletion of the signaling lipid phosphoinositide 4 , 5-bisphosphate ( PI ( 4 , 5 ) P2 ) via Ca2+-mediated activation of phospholipase C appears to contribute to desensitization of TRPV1 during periods of high channel activity ( Stein et al . , 2006; Lukacs et al . , 2007 ) . Optical recording of localized Ca2+ influx through plasma membrane ion channels can be achieved using a combination of Ca2+-sensitive fluorescent dyes and non-fluorescent Ca2+ chelators loaded into cells via a whole-cell patch pipette . When Ca2+-permeable channels open , localized Ca2+ influx produces a fluorescent ‘sparklet’ in the cytosol proximal to the active channel ( Wang et al . , 2001 ) . The presence of the nonfluorescent Ca2+ chelator in the cell acts as a sink for the excess Ca2+ , enhancing the localization of the source of the influx ( Navedo et al . , 2005 ) . Optical approaches have been used to record the activity of L-type Ca2+ channels in urinary bladder smooth muscle ( Sidaway and Teramoto , 2014 ) , arterial smooth muscle ( Navedo et al . , 2006; Amberg et al . , 2007; Navedo et al . , 2010; Tajada et al . , 2013 ) , ventricular myocytes ( Wang et al . , 2001; Zhou et al . , 2009 ) , and mammalian cell lines ( Gulia et al . , 2013 ) . More recently , sparklets due to TRPV4 channels have been reported in arterial smooth muscle ( Mercado et al . , 2014 ) and vascular endothelium ( Bagher et al . , 2012; Sonkusare et al . , 2012 ) . Two aspects of sparklets reported from L-type Ca2+ channels and TRPV4 channels are remarkable . First , multiple channels were typically clustered at the sparklet sites . Second , the sparklets remained stationary throughout the observation period . Thus , some mechanism ( s ) for clustering channels must be operating in these cells . Whether the clustering mechanism ( s ) and the mechanism ( s ) eliminating diffusion of the clusters are related is unknown . Most importantly , whether any Ca2+-permeable channels have the capability to gate ( open and close ) as they diffuse laterally in the plasma membrane of a cell has not previously been addressed . It should be noted that the muscle nicotinic aceytylcholine receptors ( AChR ) expressed in Xenopus oocytes have also been studied by optical recording , and the fluorescence signals emanating from these channels did not indicate channel clustering at the fluorescence sites ( Demuro and Parker , 2005 ) . Nevertheless , the authors did find that all fluorescence Ca2+ signals from AChR maintained a constant position for the duration of the optical recordings . Regulation of mobility in the plasma membrane has been identified as a key element in signaling for the Orai family of Ca2+-release activated channels ( CRAC ) . Orai channels diffuse throughout the plasma membrane in resting cells , but in response to the emptying of Ca2+ from the endoplasmic reticulum ( ER ) they cluster at sites in the surface membrane that juxtapose to the ER ( Lioudyno et al . , 2008; Penna et al . , 2008 ) . The interaction of Orai channels with the ER-resident protein STIM1 reduces Orai mobility , acting as a sort of diffusion trap to localize Orai channels to these sites as well as directly gating Ca2+ influx through the Orai pore ( Yeromin et al . , 2006; Zhang et al . , 2006; Wu et al . , 2014 ) . Although the diffusion trap mechanism has not yet been proposed for other types of ion channels , the addition of regulated mobility to a cell's toolkit for controlling its functions represents a powerful means of increasing the spatial and temporal specificity of cell signaling . In the present study we asked whether the mobility of TRPV1 might be regulated and whether any such regulation might be coupled to channel activity . We took advantage of the high Ca2+ permeability of TRPV1 to record Ca2+ sparklets that reflected the influx of Ca2+ through open TRPV1 channels in response to capsaicin in isolated mouse dorsal root ganglion neurons and in immortalized mammalian cell lines . Whole-cell voltage clamp was used to both minimize the signal due to voltage-gated Ca2+ channels and to introduce a combination of fluorescent and nonfluorescent Ca2+ chelators . In contrast to L-type Ca2+ channels , TRPV4 channels and AChR previously studied , Ca2+ sparklets from TRPV1 channels in isolated dorsal root ganglion neurons and cultured cell lines were observed with a wide spectrum of mobilities , from immobile to near the diffusion coefficient of a lipid moving through a bilayer . To our knowledge , the opening and closing of a channel as it diffuses through the plasma membrane has not been previously observed . Using TRPV1 fused to GFP expressed in HEK293T/17 cells , we established that each sparklet represented the activity of one and only one TRPV1 channel . Remarkably , we found that the mobility of TRPV1 , whether measured using sparklets or by tracking GFP-fused channels , decreased with activity in an extracellular Ca2+-dependent manner . We propose that activity-dependent regulation of TRPV1 mobility could , like for Orai , lead to clustering of channels in regions of the cell where they are most needed . To determine the activity of individual TRPV1 channels , we used whole-cell patch clamp of isolated mouse dorsal root ganglion neurons . TRPV1-expressing nociceptors were easily distinguished as the small-diameter ( ∼50 µm ) cells in a 12–24 hr culture . As previously shown , these cells have large , capsaicin-activated currents when studied with whole-cell patch clamp as well as when studied using inside-out patch clamp ( Stein et al . , 2006 ) . The fluorescent Ca2+ indicator Fluo-4 ( 200 μM ) and the nonfluorescent Ca2+ chelator EGTA ( 10 mM ) were included in the patch pipette , and cells were held at a potential of −40 mV . This potential was chosen empirically as the potential that minimized Ca2+ influx in the absence of capsaicin . In addition to the EGTA included in the intracellular pipette solution , 1 μM thapsigargin was included in the bath solution to deplete the endoplasmic reticulum Ca2+ stores ( see ‘Materials and methods’ ) . These solutions allowed us to attribute the capsaicin-induced rise in intracellular Ca2+ to influx through TRPV1 channels in the plasma membrane . Very little fluorescence was observed upon dialysis of Fluo-4/ EGTA solution into the isolated dorsal root ganglion neurons via the patch pipette . However , subsequent perfusion of 100 nM capsaicin into the bath generated localized bursts of fluorescence , termed sparklets ( Figure 1A and Video 1 ) . Sparklets were observed with physiological concentrations of extracellular Ca2+ ( 2 mM ) . Out of 12 cells that responded to capsaicin and showed a rise in global fluorescence due to Ca2+ influx , nine cells had easily distinguished local micro-domains of Ca2+ influx that we classified as sparklets . 10 . 7554/eLife . 03819 . 003Figure 1 . Capsaicin induced sparklets in whole-cell voltage clamped dorsal root ganglion cells loaded with Fluo-4 . ( A ) Images of immobile sparklet ( white circle ) in dorsal root ganglion cell at time points when inactive ( 1 ) and active ( 2 ) with corresponding intensity trace below ( see Video 1 ) . ( B ) Moving sparklet recorded for 5 . 6 s in dorsal root ganglion cell ( see Video 2 ) . Inset is of sparklet trace magnified fivefold . Inset scale bar is 1 µm . All other scale bars are 5 µm . ( C ) Mean fluorescence intensity ( ΔF¯/F0 ) and ( D ) open probability ( Popen ) of dorsal root ganglion sparklets ( N = 5; 3 cells ) in Figure 1—figure supplement 1 A–E . Cut-off for ‘open’ is when intensity ( ΔF/F0 ) exceeds 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 00310 . 7554/eLife . 03819 . 004Figure 1—figure supplement 1 . All points histogram of immobile sparklets in dorsal root ganglion cells ( N = 5 ) . ( A ) Histogram of sparklet intensity trace shown in inset of Figure 1A ( corresponding Video 1 ) . ( B , C ) Histograms of two separate sparklets observed in a cell ( see Video 3 ) . ( D , E ) Histograms of two additional , separate sparklets in one cell . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 00410 . 7554/eLife . 03819 . 005Figure 1—figure supplement 2 . Dose-response relation for activation TRPV1 channels in isolated dorsal root ganglion neurons by capsaicin . Points are the means from four cells , and the currents for each cell were normalized by the current measured at 300 nM capsaicin . Error bars are standard error of the mean . Data is fitted to a Hill curve with Hill coefficient of 2 . 2 . Blue arrow indicates the data point for 100 nM capsaicin . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 00510 . 7554/eLife . 03819 . 006Figure 1—figure supplement 3 . Increase in fluorescence intensity due to Ca2+ influx into F11 cell transfected with TRPV1-tagRFP , with Fluo-4 introduced via the whole-cell patch pipette . The two images of the F11 cell are a projection of bright pixels in a set of images before capsaicin was added ( left ) followed by a projection of bright pixels in a set of images after capsaicin was added ( right ) ( see Video 5 ) . Scale bars are 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 00610 . 7554/eLife . 03819 . 007Video 1 . Immobile sparklet ( white circle ) in dorsal root ganglion cell with fluorescence recording on the right elicited by capsaicin ( 100 nM ) . Scale bar is 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 007 Fluorescence intensity analysis of TRPV1 sparklets in isolated dorsal root ganglion neurons revealed two states: a dark state presumed to correspond to a closed channel state and a bright state presumed to correspond an open channel state ( Figure 1A and Video 1 ) . All-points amplitude histograms of the fluorescence intensity revealed that sparklet-to-sparklet variability in amplitude of the capsaicin-induced fluorescence was low ( Figure 1—figure supplement 1 ) . However , this amplitude is likely determined primarily by the channel's relative position along the z-axis within the TIRF evanescent excitation field as well as the concentrations and properties of Ca2+ buffers used . Thus , we did not interpret the stereotyped amplitude in terms of Ca2+ concentration at the sparklet sites . Interestingly , the two-state intensity traces ( Figure 1A and Figure 1—figure supplement 1 ) resemble those of single-channel activity measured with electrophysiology ( e . g . , Liu et al . , 2004; Rosenbaum et al . , 2004 ) , albeit with lower temporal resolution . These data are consistent with each sparklet representing a single gating TRPV1 channel or a cluster of TRPV1 channels gating cooperatively . We distinguish between these possibilities below . The all-points histograms of sparklet amplitude shown in Figure 1—figure supplement 1 could be used to calculate the open probability for each sparklet , shown in Figure 1D . We compared these open probability values to those measured from the whole population of channels in these isolated dorsal root ganglion cells using whole-cell patch clamp . Examining the dose–response relation for activation by capsaicin ( Figure 1—figure supplement 2 ) shows that the current recorded in the presence of 100 nM capsaicin normalized to the current at a saturating capsaicin concentration was 0 . 84 ± 0 . 05 ( n = 4 ) . This fractional activation value is slightly greater than the open probability values plotted in Figure 1D , for which the mean was 0 . 7 ± 0 . 1 ( N = 5 ) . This small discrepancy is expected , because the open probability of TRPV1 in the presence of a saturating capsaicin concentration is somewhat less than 1 . 0 ( Liu et al . , 2004 ) . These data indicate that the sensitivity of single TRPV1 channels as calculated from optical recordings of sparklets is similar to that of the cell population measured using whole-cell patch clamp . The fluorescence intensity of the sparklet shown in Figure 1A was straight forward to quantify because it was immobile , that is it did not move . However , many of the capsaicin-activated sparklets we observed were dynamic , that is they appeared to move laterally in the plane of the membrane ( Figure 1B ) . In order to follow the high mobility of these sparklets , we recorded their movement at a camera frame-rate of 33 fps . The inset of Figure 1B was made by tracking the indicated sparklet with the Spot Tracker 2D plugin for ImageJ ( see ‘Materials and methods’ ) , and it illustrates the significant mobility of the sparklet site over the course of 5 . 6 s ( see also Video 2; further examples of mobile sparklets in Video 3 ) . To our knowledge , simultaneous mobility in the plasma membrane and gating has not been previously reported for any ion channel , and are explored further below . 10 . 7554/eLife . 03819 . 008Video 2 . Moving sparklet observed in whole cell patch clamped dorsal root ganglion cell after addition of capsaicin ( 100 nM ) . Fluo-4 is loaded into cell as an indicator for Ca2+ influx . Track of sparklet is super-imposed in red . Scale bar is 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 00810 . 7554/eLife . 03819 . 009Video 3 . Numerous mobile sparklets in whole cell patch clamped dorsal root ganglion cell after addition of capsaicin ( 100 nM ) . Scale bar is 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 009 The heterogeneity of TRPV1 mobilities , with some sparklets immobile and others dynamic , indicates that the population of TRPV1 channels was heterogeneous in some aspect of channel structure or function . To determine the mechanisms involved in controlling TRPV1 mobility we measured the sparklet trajectories , as shown in Figure 1B , inset . Two factors , however , limited our ability to fully characterize these trajectories . The first factor , which is a general limitation of our approach in any cell type , was that our sparklet recording time was limited by the properties of the Ca2+-sensitive dye used . We required a dye sensitive enough to show the Ca2+ flux through just a single channel . The appearance of individual sparklets when capsaicin was added became difficult to discern over time as the fluorescence across the cell footprint increased . The increase in fluorescence ‘background’ was due to the rapid , global increase in intracellular Ca2+ that occurs over the course of our experiments . Although the Ca2+ chelator EGTA ( 10 mM ) in our intracellular solution extended the useful recording time by acting as a Ca2+ sink , a rise in global Ca2+ due to continued influx of Ca2+ through TRPV1 ultimately obscured the sparklet signals once intracellular Ca2+ levels exceeded the upper limit of the Ca2+ sensitive dye's optimal range ( Fluo-4: 1 μM Ca2+; Life Technologies , Carlsbad , CA ) . The second factor limiting our interpretation of sparklet activity in dorsal root ganglion neurons are the tube-like structures prevalent within the footprint of isolated neurons that brightened markedly when capsaicin was added . These tube-like structures were observed even when the SERCA inhibitor thapsigargin was present ( Video 3 ) , indicating that they are likely neurite-like processes underneath the cell footprint , rather than endoplasmic reticulum within the cell body . The processes hampered further characterization of capsaicin-activated sparklet activity in isolated dorsal root ganglion cells because in the confined space of the processes , global Ca2+ levels dominate sparklet signals very rapidly ( Video 3 ) . We attempted to circumvent this issue by conducting experiments very soon after cell isolation , before processes could form . However , cells studied soon after isolation , before significant processes had formed , showed footprints that were irregularly-adhered to the coverglass , and thus were not evenly within the evanescent TIRF field . Transient transfection of TRPV1 into the F11 cell line , a hybridoma formed between dorsal root ganglion cells and neuroblastoma cells ( Stein et al . , 2006 ) , were examined as an alternative to primary cell culture . In experiments using the F11 line with the channel expressed as a fluorescent fusion protein , TRPV1-tagRFP , we did observe moving sparklets ( Video 4 ) . However , processes running under the footprint of F11 cells ( Figure 1—figure supplement 3 and Video 5 ) interfered with sparklet characterization as they did in isolated dorsal root ganglion neurons . To study sparklet mobility we therefore turned to expressing tagRFP- and GFP-tagged TRPV1 ( TRPV1-tagRFP and TRPV1-GFP ) in transiently transfected HEK293T/17 cells , in which the neurite-like processes were not observed . 10 . 7554/eLife . 03819 . 010Video 4 . Immobile and mobile sparklets in TRPV1-tagRFP transfected F11 cell after addition of capsaicin ( 100 nM ) . Arrow indicates the site where a moving sparklet appears in a process at 7 . 75 s . This video is an exception in that the frame rate is 20 Hz . Scale bar is 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 01010 . 7554/eLife . 03819 . 011Video 5 . Sparklets occurring in TRPV1-tagRFP transfected F11 cell after addition of capsaicin ( 100 nM ) . Pinpoint accuracy of sparklets is complicated by the uneven cell footprint within the TIRF evanescent field . Scale bar is 5 µm . See Figure 1—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 011 We examined whether HEK293T/17 cells would prove suitable for quantifying TRPV1 mobility and activity . TRPV1 channels expressed robustly in these cells , which proved a problem as single tagRFP-tagged channels could not easily be distinguished and therefore nor would individual sparklets . However , we found that performing our experiments within 12–18 hr of transfection yielded the sparse density of channels required . We found that , like isolated dorsal root ganglion neurons , TRPV1-tagRFP and TRPV1-GFP in HEK293T/17 cells showed mobile and immobile capsaicin-activated sparklets ( Video 6 ) . Sparklets were not observed in the absence of capsaicin nor in the absence of extracellular Ca2+ ( Figure 2A ) . As shown in Figure 2B , C ( see also Video 7 ) , fluorescence from capsaicin-activated sparklets in HEK293T/17 cells had the signature two-amplitude intensities also observed in isolated dorsal root ganglion neurons ( Figure 1—figure supplement 1 ) . For these experiments we used Fluo-5F ( 200 μM ) together with 10 mM EGTA in the pipette solution and 20 mM Ca2+ in the extracellular buffer to closely mimic the experiments of Navedo et al . ( 2005 ) , who report on immobile sparklets from L-type Ca2+ channels ( see ‘Materials and methods’ ) . Interestingly , the amplitude of fluorescence intensity in the sparklets was comparable in isolated dorsal root ganglion neurons and HEK293T/17 cells ( compare Figure 2D to Figure 1C ) . However , differences in the dye ( Fluo-5F vs Fluo-4 ) and extracellular Ca2+ concentrations ( 20 mM vs 2 mM ) make it difficult to draw any conclusions from this similarity . The open probability for sparklets in HEK293T/17 cells was significantly less than for sparklets in isolated dorsal root ganglion neurons ( Figure 2E ) . This difference is consistent with the higher K1/2 for activation of TRPV1 expressed in HEK293T/17 cells ( Collins and Gordon , 2013 ) . Interestingly , capsaicin-independent increases in local Ca2+ were observed in untransfected as well as TRPV1-transfected HEK293T/17 cells . These could be easily distinguished from sparklets due to TRPV1 as their open times were very brief and their amplitude was spikey ( Figure 2—figure supplement 1 ) , rather than showing two states like TRPV1 ( Figure 2B , C ) . These TRPV1-independent Ca2+ events were excluded from the analysis reported here . 10 . 7554/eLife . 03819 . 012Video 6 . Immobile and mobile sparklets in TRPV1-GFP transfected HEK293T/17 cell after addition of capsaicin ( 100 nM ) . TRPV1-GFP fluorescence is observed in the first part of the video , and fluorescent sparklets become the dominant signal in second half of video with both mobile ( bottom middle ) and immobile sparklets present . Scale bar is 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 01210 . 7554/eLife . 03819 . 013Figure 2 . Fluorescence intensity traces of immobile sparklet sites in cells loaded with Fluo-5F and extracellular buffer containing 20 mM Ca2+ and voltage-clamped as described in ‘Materials and methods’ . ( A ) Without Ca2+ in the extracellular buffer very little green fluorescence was observed in the capsaicin-treated ( 100 nM ) HEK293T/17 cell that was transiently transfected with TRPV1-tagRFP . The image in the top panel was acquired after a 30 s exposure to capsaicin . Addition of 20 mM Ca2+ to the extracellular buffer increased the fluorescence of the cell footprint dramatically and individual sparklet sites could be observed ( indicated with white arrowheads in bottom panel ) . TIRF images have matching lookup table ( LUT ) values and have not been otherwise altered . Scale bar is 5 µm . ( B ) Representative trace of sparklet from TRPV1-tagRFP expressing HEK293T/17 cell . ( C ) All points histogram of trace in ( B ) . ( D ) Mean fluorescence intensity ( ΔF/F0 ) and ( E ) open probability ( Popen ) of dorsal root ganglion sparklets ( Blue; same data as in Figure 1C , D ) and TRPV1-tagRFP sparklets in HEK293T/17 cells ( green; N = 4 ) . Cut-off for ‘open’ is when intensity ( ΔF/F0 ) exceeded 0 . 05 . See Video 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 01310 . 7554/eLife . 03819 . 014Figure 2—figure supplement 1 . Fluorescence intensity traces of TRPV1-independent Ca2+ events in HEK293T/17 cells loaded with Fluo-5F , with extracellular buffer containing 20 mM Ca2+ , and studied with whole-cell voltage clamp as described in ‘Materials and methods’ . ( A ) Representative trace of Ca2+ events in an untransfected cell . ( B ) All points histogram of trace in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 01410 . 7554/eLife . 03819 . 015Video 7 . Immobile sparklet used to analyze TRPV1-tagRFP sparklet intensity levels ( See Figure 2B ) . Scale bar is 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 015 The two-state fluorescence intensity we observed with TIRF imaging of capsaicin-induced Ca2+ sparklets in dorsal root ganglion neurons and HEK293T/17 cells suggested that each sparklet was due to a single TRPV1 channel . To test this hypothesis , we used photobleaching to count the number of subunits per feature , using GFP instead of tagRFP due to its superior spectroscopic properties and because we did not need to avoid spectral overlap with the Ca2+-sensitive dye in these experiments . TRPV1 is a homomeric tetramer structurally similar to the voltage-gated superfamily of ion channels that includes voltage-gated K+ channels ( Liao et al . , 2013 ) . Each TRPV1-GFP channel should thus have four copies of GFP per functional channel . If each fluorescent spot represents one TRPV1-GFP channel , four photobleaching steps should be observed . However , if each spot represents several TRPV1-GFP channels , >4 photobleaching steps should be observed . Because measuring the GFP fluorescence intensity of mobile TRPV1 channels proved difficult , paraformaldehyde-fixed cells were used in these photobleaching experiments . As shown in Figure 3A , using a high-intensity laser to image the fluorescence from individual TRPV1-GFP features produced sudden decreases in fluorescence intensity , each of which is presumed to represent photobleaching of a single GFP ( Ulbrich and Isacoff , 2007 ) . We observed from one to four photobleaching steps across the population of TRPV1-GFP and plotted a histogram of the steps ( total number of sites with bleaching steps = 384 ) , shown in Figure 3B . We then fitted the data with a binomial model given a fixed number of subunits , n , and the probability that each GFP monomer was able to fluoresce at the time the sample was first illuminated , p , as a free parameter ( see ‘Material and methods’ ) . The best fit was obtained using n = 4 and p = 0 . 45 ( Figure 3B ) , and fits with n = 5 , as would be the case if TRPV1 were pentameric , or n = 8 , if TRPV1 tetrameric channels formed pairs , were both rejected by χ2 goodness-of-fit tests ( Figure 3C and legend ) . Although p = 0 . 45 produced by the fit with n = 4 is low with respect to this value reported by others ( Ulbrich and Isacoff , 2007; Ji et al . , 2008; Zhang et al . , 2009 ) , our experiments were performed in fixed cells , a preparation known to partially disrupt fluorescent protein fluorescence ( Chalfie and Kain , 1998 ) . Together with the two-state fluorescence intensity of capsaicin-induced sparklets , these data indicate that each TRPV1-GFP feature represented one channel , with four subunits , rather than two or more channels , with eight or more subunits . 10 . 7554/eLife . 03819 . 016Figure 3 . Photobleachstep analysis of TRPV1-GFP in fixed HEK293T/17 cells . ( A ) Fluorescence time traces of six representative TRPV1-GFP features . Arrowheads indicate individual photobleach steps ( B ) Histogram of TRPV1-GFP bleach steps in fixed HEK293T/17 cells . TRPV1-GFP fluorescence is characterized as having 1–4 bleaching steps ( blue ) . A zero-truncated binomial distribution fit by maximum likelihood estimate with n = 4 to give a probability of 0 . 45 ( red ) ( χ2 goodness-of-fit is 0 . 1712 and the model is acceptable ) . ( C ) Fits to zero-truncated binomial distributions with n = 5 ( cyan ) or n = 8 ( green ) by maximum likelihood estimate gave probabilities of 0 . 35 and 0 . 21 , respectively . Chi-squared goodness-of-fit tests were done with n = 5 and n = 8 models , yielding χ2 values of 4 . 0717 and 16 . 386 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 016 Unlike sparklet experiments , in which the duration of recording was limited by the properties of the Ca2+ chelators used , TRPV1-GFP experiments could proceed considerably longer . To further understand mechanisms underlying the spectrum of sparklet mobilities we observed , we performed a large survey of TRPV1-GFP mobilities by tracking diffusion of the channels in the plasma membrane . We acquired 15 s videos of TRPV1-GFP channels moving laterally in the plasma membrane of living cells and used the particle tracking algorithm u-track , which simplifies tracking of features in a large sample size and minimizes user bias in track identification ( Jaqaman et al . , 2008 ) , to follow channel movements . Channel trajectories for the first 6 s of a video are shown projected onto the last fluorescent image of the cell in Figure 4A ( see Video 8 ) . The mean squared displacements ( MSD ) of selected tracks labeled as 1 , 2 , and 3 are presented in Figure 4B . Each of the three tracks exhibited a different MSD vs time slope , which was characterized by an effective diffusion coefficient , Deff , as a measure of mobility . These data were then used to generate a histogram of effective diffusion coefficients for all the TRPV1-GFP channel tracks observed in the full 15 s video ( Figure 4C ) . The large variability in channel mobilities , with a 20-fold range of Deff ( 0 . 01–0 . 5 μm2/s ) , is consistent with a heterogeneous population within any given cell . Interestingly , the distribution of mobilities between cells also varied , but only by about a factor of two ( mean Deff range: 0 . 029 to 0 . 056 μm2/s ) ( Figure 4D and Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 03819 . 017Figure 4 . Distribution of TRPV1-GFP mobility in HEK293T/17 cell ( not voltage-clamped; see ‘Materials and methods’ ) . ( A ) TRPV1-GFP channel tracks in a 200 frame video ( 33 frames per second ) . Tracks are identified with the u-track algorithm ( see ‘Materials and methods’ ) and are overlaid on the final image of the series ( See Video 8 ) . Extent of cell body is shown in grey . The inset is an isolated track from the bottom center of the larger panel labeled ‘3’ . ( B ) Mean square displacements ( MSD ) at different time lags of three tracks identified as 1 ( blue ) , 2 ( green ) , and 3 ( red ) in ( A ) . ( C ) A Histogram of effective diffusion coefficients derived from tracks in the full 15 s video for the cell shown in ( A ) . Tracks identified as 1 , 2 , and 3 in ( A ) are indicated over their assigned bins . ( D ) Mean effective diffusion coefficients calculated from TRPV1-GFP tracks in HEK293T/17 cells ( N = 5 ) . Scale bar of image in ( A ) is 5 µm and inset scale bar is 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 01710 . 7554/eLife . 03819 . 018Figure 4—figure supplement 1 . Distribution of effective diffusion coefficient ( Deff ) in different cells ( A–D ) . Each event accounts for the effective diffusion coefficient ( Deff ) calculated from a time lag of 1–10 frames in a single TRPV1-GFP track ( See ‘Materials and methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 01810 . 7554/eLife . 03819 . 019Video 8 . Video of TRPV1-GFP tracks in transfected HEK293T/17 cell ( See Figure 4 ) . Scale bar is 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 019 To establish a one-to-one correspondence between capsaicin-induced sparklets and mobile TRPV1-GFP , we implemented an optical method to coincidentally detect sparklets while simultaneously tracking the channel movement with a GFP label . Figure 5A ( red trace ) shows the trajectory of a typical TRPV1-GFP channel that emitted sparklets along its track . In Figure 5B we plot the fluorescence intensity of the moving TRPV1-GFP channel as it increased intermittently to higher levels ( see also Videos 9 and 10 ) . 10 . 7554/eLife . 03819 . 020Figure 5 . TRPV1-GFP and sparklet observed in a moving channel in a voltage-clamped HEK293T/17 cell . ( A ) Track of TRPV1-GFP with increased sparklet intensity . Scale bar is 1 μm . ( B ) Fluorescence intensity trace of mobile TRPV1-GFP channel in ( A ) with sustained higher levels of fluorescence due to sparklet activity . The background image in ( A ) corresponds to the time annotated with an arrowhead in panel ( B ) . ( C ) TRPV1-GFP and sparklet positions are shown as filled circles . Red circles are positions assigned to TRPV1-GFP and cyan circles are sparklet positions . Each circle's radius is 1/10 of the standard deviation attributed to the 2-D bivariate Gaussian fitted to the fluorescence feature at that position . Scale bar is 0 . 5 μm . See Videos 9 and 10 . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 02010 . 7554/eLife . 03819 . 021Video 9 . The TRPV1-GFP channel in the left video can be followed as a GFP feature as of 0 . 0 s with intermittent sparklet activity and at 6 . 33 s is a sparklet feature for the majority of the time . Scale is 6 . 25 pixels per 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 02110 . 7554/eLife . 03819 . 022Video 10 . Video with expanded view of Figure 5A and Video 9 . This video shows the highly dynamic pool of TRPV1-GFP features before capsaicin ( 100 nM ) opens the channels and the GFP signal is overwhelmed by sparklet activity and background Ca2+ levels . Box indicates the approximate location of the ROI used for Figure 5 and Video 9 . Scale bar is 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 022 To compare the confidence with which the sparklets and TRPV1-GFP fluorescence could be localized , we measured the uncertainty of the precision as an average of the standard deviations of a 2D bivariate Gaussian fitted to the fluorescence feature in each step of the trajectory shown in Figure 5A . We used circles representing the size of the uncertainty ( 1/10th scale ) to indicate both which points were observed as sparklets ( Figure 5C , cyan circles ) and which as GFP ( Figure 5C , red circles ) . The uncertainty for all the points ( GFP and sparklet ) ranged from 0 . 13 μm to 0 . 33 μm , allowing us to determine the positions of the fluorescent feature along the full track length . Two interpretations of these data stand out . First , the co-localization between TRPV1-GFP ( Figure 5C , red circles ) and the Ca2+ sparklets ( Figure 5C , cyan circles ) confirmed that mobile sparklets co-diffuse with mobile TRPV1-GFP channels . Second , the higher fluorescence intensity of sparklets compared to TRPV1-GFP did not impede our ability to localize sparklets as compared to TRPV1-GFP . The simplest interpretation of our data is that sparklets represent Ca2+ flux through the labeled TRPV1-GFP as it gates . The mobility of actively gating TRPV1 is somewhat surprising , as Ca2+ sparklets from TRPV4 were previously reported to be immobile ( Mercado et al . , 2014 ) as have sparklets attributed to L-type Ca2+ channels ( Navedo et al . , 2005; Navedo et al . , 2006 ) and AchR channels ( Demuro and Parker , 2005 ) . In addition , in a search of the literature we did not find any reports of simultaneous observations of ion channel activity and mobility in living cells . However , the mobility of capsaicin-activated TRPV1 sparklets in three different cell types ( isolated dorsal root ganglion neurons , F11 cells , and HEK293T/17 cells ) raises the question of whether the mobility of other types of active channels should be reexamined . Does the heterogeneity of effective diffusion coefficients measured for the population of TRPV1-GFP reflect a functional heterogeneity of the channels ? Upon examining a large number of mobile TRPV1-GFP channels under various conditions , we noticed that capsaicin appeared to decrease their mobility . We therefore hypothesized that channel mobility might be regulated as a function of channel activity . Because our experiments produced direct measurements of activity and mobility of single TRPV1 molecules , we could determine whether mobility changed upon channel activation . To test the hypothesis that TRPV1 activity may regulate its mobility , we examined whether the addition of capsaicin to HEK293T/17 cells transiently transfected with TRPV1-GFP produced a change in channel mobility . In control experiments with nominally Ca2+-free extracellular buffer and in which the cells were not subject to voltage clamp , we observed a trend of increasing MSD ( Figure 6A ) . Although we did not study this effect further , we speculate it was due to photodamage by the excitation light . We did , however , design further experiments to take the observation-induced increase in apparent mobility into consideration . By expressing the data as a ratio of the change in MSD ( RΔMSD , see ‘Materials and methods’ and Figure 6—figure supplement 1 ) before addition of capsaicin to that after the addition of capsaicin for TRPV1-GFP trajectories within the same cell , we were able to correct for the observation-induced effect on TRPV1-GFP mobility . 10 . 7554/eLife . 03819 . 023Figure 6 . Mobility change in TRPV1-GFP channels after capsaicin treatment in cells not subject to voltage clamp . ( A ) HEK293T/17 cells expressing low numbers of TRPV1-GFP are initially imaged for 15 s and then a second 15 s video follows perfusion with a 1 μM capsaicin solution ( described in ‘Materials and methods’ ) . All three treatment conditions begin in HBSS ( containing Ca2+ ) with no capsaicin . Average MSD ( time lag of 90 msec ) for all tracks shows no significant change in mobility from before to after capsaicin . ( B ) Using the same source data , we characterize mobility with the MSD difference ratio ( RΔMSD , described in ‘Materials and methods’ and Figure 6—figure supplement 1 ) . With the switch from no capsaicin to 1 μM capsaicin but maintaining the Ca2+ concentration at 1 . 8 mM , the RΔMSD drops a significant amount from the initial condition ( N = 5 , blue data points ) as determined by a one-tailed paired t-test ( p < 0 . 01 ) . This significant drop in RΔMSD from the initial video to the second video is not observed under conditions where the second solution contains no capsaicin and no added Ca2+ ( N = 6 cells , green data points; n . s . : not significant ) nor is it seen when the second solution contains capsaicin ( 1 μM ) but no added Ca2+ ( N = 6 cells , red data points ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 02310 . 7554/eLife . 03819 . 024Figure 6—figure supplement 1 . Flow diagram detailing MSD difference ratio ( RΔMSD ) calculation from top to bottom . Videos taken initially or after treatment ( beginning 50 s after end of first video ) are split into two parts . Each part has the MSD calculated for each trajectory with the time lag set to 90 or 120 msec ( 3 or 4 frames , respectively for Figure 6 or Figure 6—figure supplement 2 ) before averaging these to acquire a mean for the set of trajectories , giving MSD1 , MSD2 , MSD3 , and MSD4 . The MSD difference ratio is expressed as the difference between the first part MSD and the second part MSD ( i . e . MSD2-MSD1 in initial video ) divided by the first part ( MSD1 in initial video ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 02410 . 7554/eLife . 03819 . 025Figure 6—figure supplement 2 . Mobility change in MSD calculated for 120 msec with TRPV1-GFP channels after capsaicin treatment . Using the same source data as in Figure 6 but with the time lag set to 120 msec ( 4 frames ) , we characterized mobility with the MSD difference ratio ( RΔMSD , described in ‘Materials and methods’ and Figure 6—figure supplement 1 ) . With the switch from no capsaicin to 1 μM capsaicin but maintaining the Ca2+ concentration at 1 . 8 mM , the RΔMSD drops a significant amount from the initial condition ( N = 5 , blue data points ) as determined by a one-tailed paired t-test ( p < 0 . 05 ) . This significant drop in RΔMSD from the initial video to the later video is not observed under conditions where the second solution contains no capsaicin and no added Ca2+ ( N = 6 cells , green data points; n . s . : not significant ) nor is it seen when the second solution contains capsaicin ( 1 μM ) but no Ca2+ ( N = 6 cells , red data points ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 025 In cells treated with 1 μM capsaicin and remaining in a Ca2+-containing solution ( but , as in control experiments , not voltage clamped ) , the RΔMSD of TRPV1-GFP in a given cell decreased upon the addition of capsaicin , indicating that the TRPV1 channels exhibited a decrease in mobility as they opened in the presence of Ca2+ ( Figure 6B , blue ) . In control experiments that switched from a Ca2+-containing solution to a nominally Ca2+-free solution without capsaicin ( Figure 6B , green ) or to a nominally Ca2+-free solution with capsaicin ( Figure 6B , red ) no significant change between RΔMSD in our before and after videos was observed . This difference in RΔMSD between before and after capsaicin treatment while in the presence of Ca2+ persisted out to 120 msec ( Figure 6—figure supplement 2 ) . Thus , the decrease in RΔMSD required both channel activation ( i . e . , capsaicin ) and Ca2+ influx . Our data indicate that the mobility of TRPV1-GFP decreased in a capsaicin- and Ca2+-dependent manner . Regulation of channel mobility by its activity was accessible because: 1 ) experiments were not truncated by saturation of a Ca2+-sensitive dye; 2 ) the recording of hundreds of channels per cell was possible when tracking mobility with GFP , whereas a much lower channel density was necessary for studying sparklets , 3 ) the HEK293T/17 cells were not voltage-clamped , allowing higher throughput imaging; and , most importantly , 4 ) unlike the Ca2+ dyes used to image sparklets , the fluorophore used for tracking the channels , GFP , could be imaged even when the channels were closed and even in the absence of extracellular Ca2+ . The differences in experimental conditions between these experiments and those in which we measured the mobility of sparklets raised the question of whether an activity-dependent decrease in mobility would be observed for capsaicin-activated sparklets . Sparklets could be imaged only when channels were open and conditions were permissive of Ca2+ influx . Thus , the mobility of sparklets due to TRPV1 could not be measured before the addition of capsaicin to the bath . We therefore designed experiments to look at the trend in mobility for individual sparklets that took advantage of the relatively slow perfusion time of our chamber ( estimated flow of five-times chamber volume in 30 s ) using voltage-clamped HEK293T/17 cells as described in ‘Materials and methods’ . The open probability of the channels would be expected to increase from near zero in the absence of capsaicin to ∼0 . 3–0 . 4 when the bath solution had equilibrated . By measuring the time dependence of mobility under these conditions , we accumulated a set of trajectories on mobile and active TRPV1-GFP or TRPV1-tagRFP channels in which the concentration of capsaicin , and thus their open probability , increased over time . Our hypothesis predicts that the mobility of these sparklets will decrease over the course of the experiment . We found that 13 out of 16 sparklets analyzed during a period when they remained open and observable showed a capsaicin-dependent decrease in mobility ( Figure 7 ) . The decreased mobility of these sparklets is illustrated by the trajectory shown in Figure 7A ( top ) . The red , blue , and green segments of the track represent the location of the sparklet during three consecutive epochs of equal duration . The first , red epoch was one of high mobility . The second and third epochs , represented in blue and green , respectively , showed markedly reduced mobility . Plotting the displacements as a function of time for this track ( Figure 7A bottom ) , fitting it with a line , and calculating its slope was used to measure the trend in this track toward reduced mobility . One of the three sparklets that showed increased mobility upon activation is shown in Figure 7—figure supplement 1 . The slopes of the lines from the 16 sparklets are shown in Figure 7B , revealing a trend towards decreasing mobility ( negative slope; Figure 7B ) . In conclusion , activation of TRPV1 yielded a decrease in channel mobility , whether observed as the GFP-fusion protein or as sparklets . 10 . 7554/eLife . 03819 . 026Figure 7 . Characterization of mobile TRPV1 sparklets in voltage-clamped HEK293T/17 cells . ( A ) Trajectory of sparklet with decreasing mobility . The total steps of the trajectory are divided into three equal parts representing the first third ( red ) , the second third ( blue ) and the last third ( green ) of each track . Scale bar is 1 μm . Displacement ( see ‘Materials and methods’ ) plot of sparklet in ( A ) is shown below . The blue line is the slope used to characterize the change in displacements over time . ( B ) Slopes corresponding to the change in displacements of mobile sparklets ( N = 16 ) in density plot with the center of the distribution located below 0 . 0 . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 02610 . 7554/eLife . 03819 . 027Figure 7—figure supplement 1 . Characterization of mobile TRPV1 sparklets in voltage-clamped HEK293T/17 cells . ( Left ) A representative trajectory of sparklet with increasing mobility is shown with the same color coding for the track as in Figure 7 . Scale bar is 1 μm . Displacement ( see ‘Materials and methods’ ) plot of sparklet is shown to the right . The blue line is the slope used to characterize the change in displacements over time . DOI: http://dx . doi . org/10 . 7554/eLife . 03819 . 027 We recorded the activity of individual TRPV1 channels in live cells by imaging capsaicin-activated optical sparklets with TIRF microscopy . A fluorescent Ca2+ indicator ( Fluo-4 or Fluo-5F ) was introduced , along with the nonfluorescent Ca2+ chelater EGTA , into cells via a patch pipette , which also served to clamp the voltage at a potential empirically determined to minimize Ca2+ influx in the absence of capsaicin . Capsaicin and extracellular Ca2+ were required for sparklet formation in isolated dorsal root ganglion neurons and HEK293T/17 cells transiently transfected with TRPV1-tagRFP . TRPV1 sparklets are a new tool to address critical gaps in our knowledge concerning the function of TRPV1 in its native plasma membrane environment and opens the door to further exploration into how plasma membrane dynamics and cell signaling regulate TRPV1 function . Synchronous recording of localization and activity of native TRPV1 channels in isolated dorsal root ganglion neurons and of TRPV1-GFP or TRPV1-tagRFP channels in transiently-transfected HEK293T/17 cells revealed a surprising range of mobilities in the plasma membrane . Like the L-type Ca2+ channels ( Navedo et al . , 2005; Navedo et al . , 2006 ) , AchR channels ( Demuro and Parker , 2005 ) , and TRPV4 channels ( Mercado et al . , 2014 ) discussed above , some sparklets were immobile , whereas others diffused at a rate comparable to that expected for an individual lipid ( ∼1 µm2/s ) ( Golebiewska et al . , 2008 ) ( Figures 1 , 2 , 5 and 7 ) . To our knowledge , gating of channels as they diffuse in the plasma membrane has not been previously reported , although synchronous gating and diffusion may well represent the ‘natural’ behavior of many channels and transporters . We combined our studies of capsaicin-induced sparklets in isolated dorsal root ganglion neurons and transiently-transfected HEK293T/17 cells with imaging studies of TRPV1-GFP in HEK293T/17 cells to determine the number of TRPV1 channels underlying the sparklets . Sparklets have been used previously by Navedo et al . to report on the gating of L-type Ca2+ channels as single channels or in clusters of up to seven channels ( Navedo et al . , 2005 ) , or assemblies of TRPV4 channels in arterial myocytes ( Mercado et al . , 2014 ) . The immobile TRPV1-tagRFP sparklets allowed us to measure the fluorescence intensity of capsaicin-induced sparklets and determine the number of channels at each sparklet site . The two-state fluorescence signature of the TRPV1-tagRFP sparklets could arise from single-channel openings or from multiple channels at each sparklet site with very high gating cooperativity ( Figure 2 ) . Analysis of the photobleaching steps attributed to TRPV1-GFP channels in fixed HEK293T/17 cells provided further confirmation that the channel assembles as a tetramer of TRPV1-GFP subunits without forming higher order oligomeric assemblies ( Figure 3 ) . Importantly , laterally mobile TRPV1-GFP in live cells could be tracked via both their fluorescent protein label and their sparklet activity , and switching from a lower fluorescence emission of the fluorescent protein label to the higher intensity sparklet fluorescence within the same feature demonstrated that the mobile sparklets emanated from mobile TRPV1-GFP ( Figure 5 and Video 9 and Video 10 ) . Together , these data indicate that each capsaicin-induced sparklet is due to the activation of one , and only one , TRPV1 ion channel . This doesn't preclude higher number assemblies from forming , but these were never observed in our transiently transfected cells . Our data on dorsal root ganglion sparklets activated by capsaicin recapitulates our observations in TRPV1-GFP and TRPV1-tagRFP expressing HEK293T/17 cells ( Figure 1A and Figure 1—figure supplement 1 ) . Our investigation into the lateral mobility of TRPV1 as it is influenced by the channel's activity has barely begun to scratch the surface of what may be an important new mode of functional regulation . However , our finding that Ca2+ influx is required for the capsaicin-induced decrease in TRPV1 mobility may hint at the mechanism by which the change in mobility occurs . Changes in the cytoskeleton upon upregulation of Ca2+-sensitive effectors such as α-actinin , gelsolin or scinderin ( Rodriguez Del Castillo et al . , 1990; Witke et al . , 1993; Kinosian et al . , 1998 ) can be easily imagined to regulate membrane protein mobility . The formation of local Ca2+ domains has previously been linked to regulation of cell architecture and remodeling of the cytoskeleton . A recent study looking at the influence of TRPM7 elicited Ca2+ micro-domains in migrating fibroblasts identified this process as part of the mechanism by which the cell steers towards a chemoattractant ( Wei et al . , 2009 , 2012 ) . Thus , the involvement of cytoskeleton in the Ca2+-influx mediated regulation of TRPV1 mobility is a subject of interest for future studies . We and others have shown that TRPV1 binds to several proteins important to cell signaling , including the p85β subunit of PI3-kinase and the TrkA receptor ( Chuang et al . , 2001; Stein et al . , 2006 ) . These interactions are likely related to a physiological response facilitated by nerve growth factor ( NGF ) that increases cell surface expression of TRPV1 ( Zhang et al . , 2005; Stein et al . , 2006 ) . Biochemical and electrophysiological experiments also point to a complex formed between AKAP79/150 and TRPV1 to alter sensitization of TRPV1 in response to inflammatory agents ( Zhang et al . , 2008; Por et al . , 2010; Efendiev et al . , 2013 ) . Differential properties of a signaling complex involving TRPV1 could be determined by the mobility of the channel , which makes its activity dependent mobility a point of interest . One can imagine that the activity of TRPV1 will confine the channel to a specific region of the cell , where it may ( or may not ) be more likely to encounter its binding partners in the PI3-kinase/ TrkA receptor complex . It is further possible that the binding of TRPV1 to a complex with a consequent change to the channel's mobility also feeds back on its function . At this point , we can only speculate on the physiological role of activity-regulated changes in TRPV1 mobility . Perhaps decreased mobility in regions of locally elevated channel activity serves as a sort of diffusion trap , concentrating TRPV1 channels in parts of the cell where they are most needed . Sensory neurons with cell bodies in the dorsal root ganglia are not very polarized ( Dubin and Patapoutian , 2010 ) . Indeed , in culture without other types of neurons they do not form synapses; in their role as sensory receptors they are not postsynaptic to other cells types and appear to have lost the signaling required to form postsynaptic structures . Other mechanisms for protein targeting might thus be prominent in these types of cells . Molecular Cloning—The C-terminal GFP-labeled and tagRFP-labeled TRPV1 were made by subcloning the relevant fragments into the pCDNA3 vector , as previously described for our TRPV1-CFP ( cyan fluorescent protein ) construct ( Ufret-Vincenty et al . , 2011 ) . Cell Culture and Electrophysiology—Mouse dorsal root ganglia cells were isolated as previously described ( Stein et al . , 2006 ) . Nerve growth factor was added to the final medium at a concentration of 100 ng/ml and experiments were completed within 24 hr of sacrifice of the mouse . Whole-cell voltage clamp recordings of capsaicin dose responses in Figure 1—figure supplement 2 were done as previously described ( Stein et al . , 2006 ) . F11 cells were cultured as described previously ( Stein et al . , 2006 ) . HEK293T/17 cells ( ATCC , Manassas , VA ) were cultured in Dulbecco's Modified Eagle Medium ( DMEM; Life Technologies ) supplemented with 10% fetal bovine serum , 50 units/ml penicillin , and 50 µg/ml streptomycin at 37°C and 5% CO2 . For both cell lines , low expression levels of the TRPV1-GFP or TRPV1-tagRFP protein was obtained with transient transfection by Lipofectamine 2000 ( Life Technologies ) at suboptimal concentrations of plasmid DNA ( 250 ng DNA to 1 μl Lipofectamine in one 35 mm well ) and decreased incubation times ( 2 hr ) as compared to the manufacturer's instructions . Cells were passaged after 12 hr onto poly-L-lysine coated glass coverslips in Hank's Buffered Saline Solution ( HBSS: 140 mM NaCl , 4 mM KCl , 1 mM MgCl2 , 1 . 8 mM CaCl2 , 10 mM Hepes , 5 mM glucose at pH 7 . 4 ) and experiments were conducted 2 hr later . Imaging of TRPV1-GFP in Figure 4 , Figure 6 , Figure 6—figure supplement 2 , and Video 8 was performed in HBSS or nominally Ca2+-free solution ( identical to HBSS except MgCl2 = 2 . 0 mM and no added Ca2+ ) as indicated . Activation of TRPV1-GFP in Figure 6 experiments was accomplished with 1 μM capsaicin ( from DMSO stock ) in HBSS or nominally Ca2+-free solution . Ca2+ influx through TRPV1-GFP , TRPV1-tagRFP or native TRPV1 channels was recorded as a current in whole-cell configuration of patch clamp with the membrane potential held at −40 mV , while simultaneously observing sparklets as previously described ( Navedo et al . , 2005 ) . During sparklet experiments in dorsal root ganglion cells ( Figure 1 and Figure 1—figure supplement 1 and Videos 1–3 ) and moving sparklet experiments in HEK293T/17 cells ( Figure 5 , Figure 7 and Figure 7—figure supplement 1 , Videos 9 and 10 ) , we perfused the following bath solution ( in mM ) : 140 N-methyl-D-glucamine ( NMDG ) , 5 CsCl , 1 MgCl2 , 10 glucose , 10 HEPES , and 2 CaCl2 adjusted to pH 7 . 4 , which is close to a physiological concentration of Ca2+ . In experiments quantifying the immobile sparklet intensities of TRPV1-tagRFP in HEK293T/17 cultured cells ( Figure 2 and Figure 2—figure supplement 1 , Video 7 ) , the bath solution was changed to 20 mM CaCl2 and 120 mM NMDG to increase fluorescence signal ( see reason given below for Fluo-5F use ) . In Figure 2A a nominally Ca2+-free NMDG solution was used as previously described with 120 mM NMDG but with no added Ca2+ . Pipettes were filled with a solution composed of ( in mM ) 110 Cs-aspartate , 20 CsCl , 1 MgCl2 , 5 MgATP , 10 HEPES , 10 EGTA , and either 0 . 2 Fluo-5F ( immobile sparklets in HEK293T/17 cells; Figure 2 , Figure 2—figure supplement 1 , Video 7 ) or Fluo-4 ( dorsal root ganglion sparklets: Figure 1 and Figure 1—figure supplement 1 and Videos 1–3; moving sparklets in HEK293T/17 cells: Figures 5 and 7 , Figure 7—figure supplement 1 , Videos 9 and 10 ) adjusted to pH 7 . 2 with CsOH . Both Fluo-4 and Fluo-5F were purchased from Life Technologies . Although Fluo-4 was used in dorsal root ganglion cell experiments and moving sparklet experiments in HEK293T/17 cells because of its increased affinity for Ca2+ and improved fluorescence signal in physiological solutions ( 2 mM Ca2+ ) , we used Fluo-5F dye for the immobile sparklet experiments in HEK293T/17 because we wished to more closely follow the methods of Navedo et al . ( 2005 ) , who previously studied Ca2+ sparklets from voltage-gated L-type Ca2+ channels in HEK293T/17 cells , and the cells were also allowed to equilibrate with the pipette contents for 2 min before each experiment . For F-11 cell sparklet experiments ( Figure 1—figure supplement 3 and Videos 4 and 5 ) , we used standard HBSS buffer and 0 . 2 mM Fluo-4 pipette solution , as described above . TRPV1 activation was induced by perfusion with either of the previously mentioned NMDG bath solutions supplemented with 100 nM capsaicin ( from DMSO stock ) . For some experiments , Ca2+ was depleted from intracellular stores by pre-incubating the cells for 5 min before experiments commenced in the NMDG buffer supplemented with 1 µM thapsigargin . Total internal reflection fluorescence ( TIRF ) microscopy—Fluorescence images of cellular structure in close proximity to the coverslip ( ∼100 nm ) were obtained through a Nikon Ti TIRF illumination microscope with a 100× 1 . 45 NA oil objective and a QuanEM:512SC Photometrics electron multiplying CCD camera . The 488 nm line of a 100 mW argon laser ( Melles Griot , Albuquerque , NM ) was used for all fluorophores except tagRFP , for which a 40 mW 561 nm solid state laser ( Coherent , Santa Clara , CA ) was used . Imaging of TRPV1-GFP for channel tracking ( Figures 4 and 6 , Figure 6—figure supplement 2 , Video 8 ) was done at 33 Hz at 100% laser power . Mobile sparklets were captured at 33 Hz at 100% laser power while immobile sparklet intensity experiments ( except experiments in Figure 1A and Figure 1—figure supplement 1 ) were recorded at 50 Hz at 20% laser power . All sparklet imaging began with capsaicin perfusion . Raw images acquired with NIS Elements ( Nikon Instruments Inc . , Melville , NY ) were imported into ImageJ ( http://rsbweb . nih . gov/ij/ ) for post-acquisition processing . Image Processing and Analysis—Data collected on moving TRPV1-GFP ( Figure 4 ) was processed in ImageJ by 1 ) removing background with a three pixel radius rolling ball filter , 2 ) averaging five frames together with the Windowed-sinc Filter plugin ( http://rsb . info . nih . gov/ij/plugins/windowed-sinc-filter . html ) using a spatial frequency cut-off of 0 . 2 , and 3 ) filtering the fluorescent feature with the spot enhance plug-in at 1 . 25 pixels ( Daniel Sage , spot tracker 2D , Sage et al . , 2005 ) . Detection and tracking of the channels was accomplished with the u-track software ( http://lccb . hms . harvard . edu/software . html ) in Matlab ( www . Mathworks . com , Natick , MA ) ( Jaqaman et al . , 2008 ) . Within the u-track program , the following features were selected: We chose to supply a background sample for detection purposes , the maximum gap between two segments of a track could be eight frames , the option to search for directed movement was deselected , and a maximum of three pixels movement between frames was imposed as a rule for finding tracks . Mean squared displacements [ ( MSD ) , 〈 ( R→ ( t+τ ) −R→ ( t ) ) 2〉] in Figure 4 were calculated from time averages of the x and y coordinates determined in the aforementioned u-track program . To estimate an effective diffusion , MSD values for lag times of up to 10 frames ( 1 frame = 30 msec ) were fitted with the equation for a particle diffusing in two dimensions [[〈ΔR→2 ( t ) 〉=4Dt]] ( Berg , 1993 ) . In preparation for analyzing sparklet intensity and movement observed in endogenous TRPV1 , TRPV1-tagRFP and TRPV1-GFP expressing cells , we first normalized TRPV1 sparklet signal to an average from images acquired before sparklets were observed , as previously described ( Demuro and Parker , 2005 ) . In the case of immobile sparklet intensity analysis ( Figures 1A and 2 and Figure 1—figure supplement 1 , and Figure 2—figure supplement 1 ) , all points histograms of the normalized intensity were generated in Igor Pro ( Wavemetrics , Lake Oswego , OR ) , with baseline subtraction to account for rising global Ca2+ signal in background and ( ΔF/F0 ) collected in a 12 × 12 ( 6 × 6 for dorsal root ganglion neuron experiments ) pixel region-of-interest ( ROI ) centered on each sparklet . In the case of sparklet movement analysis ( Figures 1B , 5 , 7 ) five frames were averaged in time with the frequency cut-off set to 0 . 1 using the Windowed-sinc filter plug-in ( ImageJ ) . Tracking of individual sparklets was then semi-autonomously performed with the ImageJ Spot Tracker 2D plug-in ( http://bigwww . epfl . ch/sage/soft/spottracker/ ) . In Figure 5 , the TRPV1-GFP fluorescent feature was tracked at an early stage of sparklet activity so that longer segments of the track represented contiguous positions of the GFP fluorescence followed by segments of sparklet fluorescence . Spot Tracker 2D does not provide the uncertainty for each position in the track , and we wished to compare the precision in measuring sparklet to GFP single molecule positions . We used Matlab ( www . Mathworks . com ) to calculate the standard deviations ( x- and y-direction ) of a 2D bivariate Gaussian function fit to an ROI centered on each fluoroescent peak recovered by ImageJ Spot Tracker 2D ( Schmidt et al . , 1996; Cheezum et al . , 2001 ) . A bivariate Gaussian was used because time averaging of the frames could stretch the feature by a small degree , and the uncertainty for the precision was calculated as the average of the two standard deviations . The region we ascribed to the feature ( 5 × 5 pixels ) was chosen as a compromise between isolating the feature of interest from surrounding unrelated signals and recovering the full shape of the GFP or sparklet feature . As TRPV1 channels opened , the background fluorescence levels gradually increased during the course of the video . Although most single particle detection methods rely on a background offset in their fitting function to accommodate a slowly varying and evenly distributed background level , we found that the highly variable and changing background levels of sparklet videos necessitated a background subtraction and additional fitting constraints in the tails of the 2D Gaussian fit . To suppress any effect that the background fluorescence had on our fits , an additional two layers of pixels surrounding the 5 × 5 pixel region were used to calculate a mean background level . This mean background was subtracted from the pixels in our 5 × 5 feature region , and the two layers of pixels were padded onto our feature region as zero-valued to extend the total fitted ROI to a 9 × 9 region . In Figure 7 , tracks were selected based on their contiguous sparklet fluorescence signals for ease of tracking . Fluorescence from GFP was not observed in the course of these videos due to the high background fluorescence ( for example , see Video 6 ) . This allowed us to pool displacement data from TRPV1-GFP and TRPV1-tagRFP in our analysis . Displacements of sparklet positions in Figure 7 and Figure 7—figure supplement 1 were determined for time points that are 10 steps apart ( i . e . , frame i to i + 10 ) to: 1 ) minimize the random noise contribution and 2 ) have greater confidence in the precision of the displacement calculation . Quantification used to assess the effects of capsaicin on TRPV1-GFP mobility in Figure 6 and Figure 6—figure supplement 2 was based on the trajectories acquired through the u-track program as described above with the following additional steps: 1 ) A mask to exclude regions of high channel densities is used when deemed necessary and consistently applied to all videos made on the same cell . 2 ) The MSD was calculated for time points 90 msec ( 3 frames ) apart in Figure 6 and 120 msec ( 4 frames ) apart in Figure 6—figure supplement 2 . 3 ) An MSD difference metric to compare the mobility of channels in the same cell before and after capsaicin treatment was implemented to emphasize trajectory differences due to TRPV1-GFP movements and to minimize changes over time caused by photobleaching . The number of tracks contributing to each average MSD in a video ranged from ∼80-600 . The MSD difference ratio ( RΔMSD ) reduced the impact of photobleaching on a particular cell by normalizing to the MSD measured in the first part of each video ( see Figure 6—figure supplement 1 ) . Single Molecule Photobleaching—HEK293T/17 cells , transiently transfected for low expression levels of TRPV1-GFP , were transferred to uncoated coverslips and allowed to recover in HBSS for 2 hr before being fixed with a 12 min incubation in a 4% paraformaldehyde solution prepared in HBSS . HBSS buffer was also used for the rinses and for imaging . Imaging with TIRF microscopy on the previously described Nikon instrument was performed at 100% 488 nm laser power , 10 Hz acquisition speed , and with 2 × 2 binning of the pixels using the QuantEM camera . Post-acquisition processing of the images entailed subtracting off a 7 × 7 pixel smoothed image from each frame and identifying potential bleach steps in the temporal fluorescence intensity traces of 3 × 3 ROIs centered on individual channel sites as previously described ( Demuro et al . , 2011 ) . A zero-truncated binomial distribution ( R-forge , 2008–2009 ) , ( nk ) pk ( 1−p ) n−k1− ( 1−p ) n , was fitted to the data of photobleaching steps by maximum likelihood estimates using Matlab ( www . Mathworks . com ) to determine the probability of functional GFP in fluorescent features containing up to four , five , or eight GFP molecules . The best fit probability for GFP fluorescence , p , and number of GFP sites per fluorescent feature , n , was selected using a χ2 goodness-of-fit test . For ease of interpretation the fluorescence intensity traces in Figure 3A were also filtered with a three-point running average .	Cells rely on proteins called receptors to keep them informed about what is going on around them . These receptors , which are embedded in the surface of the cell , detect and respond to specific chemical signals . It is known that receptors move around the cell surface as they search for particular chemical signals , but these movements have not been widely studied in experiments . Senning and Gordon have now investigated the movements of receptors called TRPV1 channels that can detect a chemical called capsaicin . This receptor contains an ion channel that is usually closed . However , when the receptor is activated this channel opens and allows calcium ions to enter the cell . In the experiments the receptors were tagged with a fluorescent marker , and a fluorescent calcium dye was used to indicate when the channel had been activated by capsaicin . This allowed the function of the receptors to be followed in real time . The experiments were performed on nerve cells taken from mice and in cell culture lines derived from neurons and kidney cells . Senning and Gordon showed that at first the receptors moved around freely on the surface of the cell , with the degree of mobility varying from cell to cell and also from receptor to receptor . However , when a receptor detected a capsaicin molecule and opened , it tended to move more slowly when calcium ions were present outside the cells . Further research is needed to explore the mechanism that prevents the receptor from moving . However , since calcium ions are involved in a wide range of processes in the nervous system , it is thought that this mechanism ensures that the receptors concentrate in regions of high neuronal activity .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "neuroscience" ]	2015	Activity and Ca2+ regulate the mobility of TRPV1 channels in the plasma membrane of sensory neurons
Local accumulation of oskar ( osk ) mRNA in the Drosophila oocyte determines the posterior pole of the future embryo . Two major cytoskeletal components , microtubules and actin filaments , together with a microtubule motor , kinesin-1 , and an actin motor , myosin-V , are essential for osk mRNA posterior localization . In this study , we use Staufen , an RNA-binding protein that colocalizes with osk mRNA , as a proxy for osk mRNA . We demonstrate that posterior localization of osk/Staufen is determined by competition between kinesin-1 and myosin-V . While kinesin-1 removes osk/Staufen from the cortex along microtubules , myosin-V anchors osk/Staufen at the cortex . Myosin-V wins over kinesin-1 at the posterior pole due to low microtubule density at this site , while kinesin-1 wins at anterior and lateral positions because they have high density of cortically-anchored microtubules . As a result , posterior determinants are removed from the anterior and lateral cortex but retained at the posterior pole . Thus , posterior determination of Drosophila oocytes is defined by kinesin-myosin competition , whose outcome is primarily determined by cortical microtubule density . Kinesin-1 , also known as conventional kinesin , is a major microtubule motor that transports many types of cargoes , including mRNA-containing granules towards plus-ends of microtubules ( Verhey and Hammond , 2009; Hirokawa et al . , 2009 ) . Kinesin-1 is required for the posterior accumulation of osk mRNA in Drosophila oocytes that is essential for embryo posterior determination ( Brendza et al . , 2000; Palacios and St Johnston , 2002; Krauss et al . , 2009 ) . Defects in osk mRNA localization prevent germ cell formation and abdomen specification ( Lehmann and Nüsslein-Volhard , 1986 ) . Posterior localization of osk mRNA is initially established during mid-oogenesis ( late stage 8 to stage 9 ) and maintained throughout late oogenesis ( stage 10B to stage 14 ) . Staufen , an RNA-binding protein that forms ribonucleoprotein particles ( RNPs ) with osk mRNA , is commonly used as a proxy for oskar RNA ( referred as ‘osk/Staufen’ hereafter ) ( Brendza et al . , 2000; Palacios and St Johnston , 2002; Erdélyi et al . , 1995; Shulman et al . , 2000; Lu et al . , 2016; Lu et al . , 2018; St Johnston et al . , 1991 ) . During mid-oogenesis ( stage 7–9 ) , the oocyte forms an anterior-to-posterior gradient of cortical microtubules ( Theurkauf et al . , 1992; Clark et al . , 1994; Clark et al . , 1997 ) . These microtubules are anchored at the cortex by their minus ends via the minus-end binding protein Patronin and a microtubule-actin crosslinker , Short stop ( Shot ) ( Nashchekin et al . , 2016 ) . Because Shot is excluded from the posterior cortex in a Par-1-dependent manner , more microtubule minus ends are anchored at the anterior and lateral cortex than at the posterior cortex ( Shulman et al . , 2000; Nashchekin et al . , 2016; Zimyanin et al . , 2008; Parton et al . , 2011 ) . This anterior-posterior cortical microtubule gradient is established in stage 7–8 oocytes ( Theurkauf et al . , 1992 ) , just before osk/Staufen starts accumulating at the posterior pole . The anterior-posterior gradient of cortical microtubule results in slightly more microtubule plus ends oriented towards the posterior pole ( Nashchekin et al . , 2016; Zimyanin et al . , 2008; Parton et al . , 2011; Khuc Trong et al . , 2015 ) . Kinesin-1 has been proposed to drive osk mRNA transport along these weakly biased cortical microtubules , favoring osk mRNA movement from the anterior to the posterior pole ( Brendza et al . , 2000; Palacios and St Johnston , 2002; Zimyanin et al . , 2008; Khuc Trong et al . , 2015; Nieuwburg et al . , 2017 ) . In addition to its conventional cargo-transporting activity , kinesin-1 has another essential function in reorganizing microtubules . Using its N-terminal motor domain and a second microtubule-binding site at the C-terminus of kinesin-1 heavy chain ( KHC ) , kinesin-1 slides antiparallel microtubules against each other in many cell types ( Jolly et al . , 2010; Lu et al . , 2013; Lu et al . , 2015; Winding et al . , 2016 ) . Kinesin-driven microtubule sliding together with kinesin-driven cargo transport generates the force that drives fast cytoplasmic streaming of the ooplasm in late-stage Drosophila oocytes ( Lu et al . , 2016 ) . Diffusion facilitated by streaming , rather than directed transport along microtubules , is responsible for the localization of posterior determinants during late oogenesis ( Lu et al . , 2018; Glotzer et al . , 1997; Forrest and Gavis , 2003 ) . Despite the necessity of directed transport and streaming , they are not sufficient for stable osk/Staufen accumulation at the posterior pole; an anchorage mechanism is required to counteract both diffusion and shear forces of the streaming ooplasm that can displace osk/Staufen from the posterior cap . osk mRNA anchorage has been suggested to be actin-dependent . Cortical localization of osk/Staufen is significantly reduced upon F-actin fragmentation ( Cha et al . , 2002 ) . An actin motor , myosin-V ( didum in Drosophila ) , is involved in osk mRNA cortical localization ( Krauss et al . , 2009; Sinsimer et al . , 2013 ) . However , myosin-V is uniformly localized at the oocyte cortex ( Krauss et al . , 2009 ) ; it is not clear why a uniformly distributed cortical anchor attaches osk/Staufen only at the posterior pole . Genetic data suggest that kinesin-1 activity antagonizes the function of myosin-V as a cortical anchor ( Krauss et al . , 2009 ) . In several other biological systems including neurons and pigment cells ( melanophores ) , competition between long-distance transport along microtubules by kinesins and anchorage/local transport by myosin-V can direct the distribution of cellular cargoes ( Rogers and Gelfand , 1998; Wu et al . , 1998; Kapitein et al . , 2013; van Bergeijk et al . , 2015; Pathak et al . , 2010; Bridgman , 1999 ) . In the oocyte , kinesin-1 has been proposed to actively remove osk mRNA from the cortex ( cortical exclusion ) by transporting it along cortical-anchored microtubules towards plus-ends of microtubules in the cytoplasm ( Cha et al . , 2002 ) . This cortical exclusion function of kinesin-1 against myosin-V may explain osk/Staufen specific anchorage at the posterior pole . In this study , we propose and directly test a unifying model that explains how osk/Staufen is transported in the oocyte and localized at the posterior pole . According to the model , myosin-V directly competes with kinesin-1 to achieve the correct posterior localization of osk/Staufen . Specifically , kinesin-1 , walking along microtubules , removes osk/Staufen from the cortex , while myosin-V , walking along actin filaments that are randomly oriented in the cortical network , dynamically anchors osk/Staufen and counteracts kinesin-driven cortical clearance . Kinesin-1 wins this competition at the anterior and lateral regions where cortical microtubule density is higher; myosin-V wins over kinesin-1 at the posterior pole where microtubule tracks are less abundant . We genetically modulate the motor activities of either kinesin-1 or myosin-V , and find that a proper balance between kinesin-1 and myosin-V activities is essential for osk/Staufen posterior localization . Using optogenetic tools we show that cortical microtubule density indeed controls osk/Staufen localization . Furthermore , using a synthetic complex that contains both myosin V and kinesin-1 motor domains , we demonstrate that these two motors are not only necessary but also sufficient for posterior localization in the oocyte . Thus , for the first time , we directly demonstrate that direct competition between an actin motor myosin-V and a microtubule motor kinesin-1 controls posterior determinant localization in Drosophila oocytes and that the outcome of this competition is determined by the local density of cortical microtubules . If Staufen localization is defined by kinesin-myosin competition , it can be disrupted in a predictable way by modulating the activity of kinesin-1 . Knockdown of KHC by RNAi leads to uniform cortical localization of Staufen in the oocyte ( Figure 1—figure supplement 1A–1D; Brendza et al . , 2000; Palacios and St Johnston , 2002; Lu et al . , 2018; Cha et al . , 2002 ) . This indicates that kinesin-1 activity is essential for Staufen cortical exclusion from the anterior and lateral cortex . To increase kinesin-1 activity in the oocyte , we employed a previously characterized constitutively active KHC mutant , KhcΔHinge2 , which lacks the flexible hinge region ( residues 521–642 ) . As a result , it is unable to fold into an auto-inhibited conformation , resulting in constitutive activation ( Figure 1A; Barlan et al . , 2013; Kelliher et al . , 2018 ) . Staufen staining of heterozygotes carrying one copy of KhcΔHinge2 exhibits a more loosened posterior Staufen cap in stage 9 oocytes ( Figure 1—figure supplement 1E , G and I ) ) , demonstrating that constitutively active kinesin-1 is dominant , and sufficient to partially mislocalize Staufen from the posterior crescent . Next , we examined Staufen localization in KhcΔHinge2 homozygous mutants . As KhcΔHinge2 homozygotes do not survive to adulthood ( Kelliher et al . , 2018 ) , KhcΔHinge2 homozygous germline clones are induced by heat shock-driven flippase ( hs-FLP ) . Within these KhcΔHinge2 homozygous germline clones , Staufen is mislocalized in stage 9 oocytes . Typically , Staufen forms a large aggregate in the middle of the oocyte and a small loose residual cap near the posterior cortex ( Figure 1B–E ) . The endogenous deletion of Khc hinge2 region shows a stronger phenotype of Staufen mislocalization than the ectopic expression of KhcΔHinge2 in the Khc null background from a previous study ( Williams et al . , 2014 ) . Intriguingly , later in development ( at stage 10B ) normal Staufen distribution is recovered with the restoration of the posterior cap and clearing of the central cytoplasmic aggregate ( Figure 1F–H; Figure 1—figure supplement 1F , H and J ) . Previous studies suggest that Osk protein , translated at the posterior pole , functions in a positive feedback mechanism for osk/Staufen accumulation in streaming oocytes ( Lu et al . , 2018; Vanzo and Ephrussi , 2002 ) . We postulate that enough Osk protein translates from the residual cap , initiating the positive feedback loop , while ooplasmic streaming circulates mislocalized osk/Staufen particles to the posterior cap , enhancing Osk localization , resulting the restoration of the posterior crescent . Together , our data show that osk/Staufen posterior localization is sensitive to kinesin-1 activity level . If osk/Staufen localization is indeed controlled by kinesin-myosin competition , a change of myosin-V activity would be predicted to disrupt osk/Staufen posterior localization . Previous studies reveal that inhibition of myosin-V by overexpression of the C-terminal myosin-V globular tail ( GT ) causes Staufen mislocalization to the center of the oocyte ( Krauss et al . , 2009; Lu et al . , 2018 ) , which is in agreement with our hypothesis that myosin-V anchors osk/Staufen at the posterior cortex . In addition , our hypothesis predicts that increased myosin-V activity alters the outcome of the competition between these two motors , and results in ectopic localization of osk/Staufen at the anterior-lateral cortex . Therefore , we decided to disrupt the regulation of myosin-V activity . Similar to kinesin-1 , myosin-V is auto-inhibited through its C-terminal cargo-binding GT domain binding to the motor and inhibiting its motor ATPase activity ( Thirumurugan et al . , 2006; Donovan and Bretscher , 2015; Li et al . , 2008; Figure 2A ) . The C-terminus of the coiled-coil 1 region is essential for auto-inhibition as it docks the cargo-binding tail for binding to the motor domain , stabilizing the closed inactive conformation ( Figure 2A; Donovan and Bretscher , 2015; Li et al . , 2008; Li et al . , 2006 ) . Deletion of the C-terminus of the coiled-coil 1 region of mouse myosin Va abolishes the inhibitory function of the GT domain and leads to constitutive activity of myosin-Va ( Li et al . , 2006 ) . Based on the homology between Drosophila myosin-V and mouse myosin Va , and coiled-coil region predictions , we deleted the 98 residues at the C-terminus of coiled-coil 1 region to create MyoVΔ1017-1114 ( Figure 2A ) . We purified this myosin-V deletion mutant ( Figure 2—figure supplement 1A; Wang et al . , 2000; Tóth et al . , 2005 ) to perform actin gliding assays of myosin-V variants . MyoVΔ1017-1114 shows a similar velocity in actin gliding assays compared to wild-type full-length myosin-V ( Figure 2B ) , indicating that the deletion does not disrupt myosin-V motor function . Next , we examined whether this deletion leads to constitutive activity by measuring the steady-state ATPase activity of these variants . Wild-type myosin-V displays a low ATPase activity in the absence of Ca2+ , as it adopts a closed inactive conformation ( Li et al . , 2006; Wang et al . , 2004 ) , but it is significantly stimulated by Ca2+ due to release from the auto-inhibited conformation ( Wang et al . , 2004; Figure 2C ) . In contrast , MyoVΔ1017-1114 displays an equivalent and high level of ATPase activity with and without Ca2+ , suggesting the deletion mutant is constitutively active ( Figure 2D ) . Furthermore , we performed negative stain electron microscopy on wild-type myosin-V and myosin-VΔ1017-1114 . Most wild-type myosin-V molecules at a low salt concentration adopt a closed conformation ( Figure 2E and I; Figure 2—figure supplement 1B ) , while in high salt concentration they display an open conformation ( Figure 2F ) . The closed conformation of wild-type myosin-V observed in low salt closely resembles that previously demonstrated in mammalian myosin Va , indicating that the structural basis for autoinhibition is conserved ( Thirumurugan et al . , 2006; Liu et al . , 2006 ) . In contrast , myosin-VΔ1017-1114 displays an open conformation in both low and high salt conditions ( Figure 2G–H and J ) . Measurement of head-junction-head angles shows that the Δ1017–1114 deletion results in a more open conformation than wild-type myosin-V ( Figure 2K ) ( see more details in Materials and methods ‘Electron Microscopy’ ) . Collectively , the data show that the Δ1017–1114 deletion creates a constitutively active myosin-V . In order to eliminate interference between the constitutively active mutant and endogenous myosin-V , we generated germline clones of a myosin-V loss-of-function mutant , didum[88] , in which Myosin-V level is dramatically reduced ( Krauss et al . , 2009 ) . We expressed MyoVΔ1017-1114 in didum[88] germline clones and find that Staufen staining in these oocytes becomes less restricted at the posterior pole and spreads to the lateral cortex ( Figure 2L–N ) . The effect is more pronounced at the anterior half of the lateral cortex ( Figure 2M–N ) . Thus , in agreement with the proposed role of myosin-V as an anchor for osk/Staufen at the actin cortex , constitutively active myosin-V expands the domain of cortically anchored osk/Staufen beyond the posterior pole . Together , these results indicate that osk/Staufen cortical anchorage is sensitive to myosin-V activity level . The results above demonstrate that the proper balance of kinesin-1 and myosin-V activities is essential for osk/Staufen localization at the posterior pole , consistent with the competition model . In other biological systems , such as pigment cells , local enrichment of F-actin favors myosin-V-dependent anchoring over microtubule-dependent transport ( Wu et al . , 1997 ) . However , within the oocyte , neither actin filaments nor myosin-V are concentrated at the posterior pole ( Krauss et al . , 2009; Lu et al . , 2016; Figure 2—figure supplement 1C–1F ) . In contrast , the microtubule network in the oocyte displays a clear anterior-posterior gradient , with the lowest density of cortical microtubules at the posterior cortex ( Clark et al . , 1997; Nashchekin et al . , 2016; Zimyanin et al . , 2008 ) . This anterior-posterior ( A-P ) microtubule gradient is revealed by microtubule labeling with germline-specific expressed EMTB-3XTagRFP ( Faire et al . , 1999; Figure 3A ) , and a GFP-labeled endogenous MAP , Jupiter-GFP ( Morin et al . , 2001; Karpova et al . , 2006; Figure 3—figure supplement 1A ) . This gradient is a result of reduced accumulation of the microtubule minus-end binding protein , Patronin , at the posterior pole ( Figure 3B–C; Nashchekin et al . , 2016 ) . We hypothesize that the cortical microtubule density is the key factor that determines the outcome of the kinesin-myosin competition . To test this hypothesis , we first used genetic tools to disrupt the microtubule gradient in oocytes . Germline overexpression of a depolymerizing kinesin , kinesin 13 ( Klp10A in Drosophila ) ( Goshima and Vale , 2005; Mennella et al . , 2005 ) , leads to complete depolymerization of microtubules in the oocyte as well in the nurse cells , while the somatic follicle cells in these ovaries still have intact microtubules ( Figure 3E–E’ , compared to Figure 3D–D’; Figure 3—figure supplement 1B–C’ and E ) . This lack of microtubules in the oocyte results in a failure of Staufen enrichment at the posterior pole; instead it is uniformly localized along the entire oocyte cortex ( Figure 3G–H and J ) . Conversely , we increased the density of cortical microtubules in the oocyte by overexpressing the microtubule minus-end binding protein , Patronin . Patronin stabilizes microtubules and prevents minus-end depolymerization by Klp10A ( Goodwin and Vale , 2010 ) . Overexpressing Patronin significantly increases microtubule density in the oocyte ( Figure 3F–F’; Figure 3—figure supplement 1D–1E ) . This increase in microtubule density causes Staufen to be excluded from the entire cortex . Instead , it forms aggregates in the center of the oocyte where most microtubules plus-ends are directed ( Figure 3G , I and K ) . Together , these data demonstrate that osk/Staufen localization is controlled by the microtubule density in the oocyte . Decreased density of cortical microtubules favors myosin-V resulting in uniform cortical localization of osk/Staufen , whereas increased microtubule density favors kinesin-1 resulting in exclusion of osk/Staufen from the cortex . Having established that the global manipulations of the microtubule gradient disrupt the posterior localization of osk/Staufen , we decided to directly test whether a local decrease of the cortical microtubule density is sufficient for cortical accumulation of Staufen . We employed optogenetic tools to locally recruit kinesin-13/Klp10A , which depolymerizes MTs ( Figure 3E ) , to the actin cortex to manipulate the local microtubule density in the oocyte ( Figure 4A ) . Using an improved light inducible dimer ( iLID ) ( Guntas et al . , 2015; Adikes et al . , 2018 ) , GFP-tagged Klp10A-SspB is recruited to the actin-rich cortex within seconds upon global blue light exposure ( 488 nm ) ( Figure 4 —Video 1 ) . We combined the microtubule labeling EMTB-3XTagRFP with Klp10A local recruitment , and found that locally recruiting Klp10A to the lateral cortex significantly decreases the EMTB-3XTagRFP signal ( Figure 4 —Videos 2 and 3; Figure 4—figure supplement 1 ) . We then combined Klp10A recruitment with RFP-tagged Staufen in late stage 8 oocytes prior to the formation of a compact posterior cap of Staufen with a pool of free Staufen particles remaining in the cytoplasm that are potentially available for dynamic anchorage . We locally recruited Klp10A to the lateral cortex , an area of the oocyte that normally has high microtubule density ( Figure 3A ) and no cortical Staufen ( Figure 3G ) . Recruitment of Klp10A-SspB to the lateral cortex leads to a dramatic accumulation of RFP-Staufen in the stimulated area ( Figure 4B–D ) . Over the time of stimulation , the RFP-Staufen progressively accumulated at the cortex ( Figure 4C and E ) . Remarkably , this ectopic recruitment of Staufen to the region of reduced microtubule density is reversible; the Staufen signal decreased to control levels after removal of blue light . Furthermore , this ectopic recruitment is repeatable . We are able to induce the recruitment either at a single spot of an oocyte ( Figure 4 — Videos 4 and 5 ) or at different spots ( Figure 4 —Video 6 ) . In contrast , in the samples expressing LifeAct-SsrA only ( No Klp10A-SspB ) , no increase of RFP-Staufen is observed after the identical light treatment , indicating the effect of Staufen-accumulation is specific to Klp10A recruitment ( Figure 4D ) . To ensure that the local recruitment effect we observed is explained by Klp10A-dependent microtubule depolymerization ( instead of Klp10A directly recruiting Staufen ) , we treated the flies with colcemid to globally depolymerize microtubules . After colcemid treatment , no visible microtubules remain at the cortex ( Figure 4—figure supplement 2 ) . Under these conditions , Klp10A recruitment at the actin cortex does not cause Staufen recruitment ( Figure 4D ) . Collectively , these results demonstrate that local microtubule density is the key determinant of osk/Staufen localization by tipping the balance of kinesin-myosin competition . Lower microtubule density favors cortical anchorage by myosin-V , while higher microtubule density favors cortical exclusion by kinesin-1 . We further tested whether the competition between kinesin-1 and myosin-V is sufficient to drive posterior accumulation of key polarity determinants in the oocyte . In order to avoid any effects of cargo interactions , we designed a minimal artificial system and used a rapalog-dependent dimerization system ( Kapitein et al . , 2013; DeRose et al . , 2013 ) to induce complex formation between constitutively active dimers of kinesin-1 heads ( KHC576 ) ( Ally et al . , 2009 ) and myosin-V heads ( MyoVHMM ) ( Tóth et al . , 2005; Rosenfeld and Lee Sweeney , 2004; Figure 5A ) . In the absence of rapalog , KHC576-TagRFP is distributed mostly in cytoplasm , consistent with the fact that the plus-ends of the cortical microtubules are directed away from the cortex ( Figure 5B and C–C’ ) . GFP-tagged MyoVHMM in these oocytes is uniformly localized at the actin cortex , consistent with the fact that it is bound to F-actin ( Figure 5C ) . Upon addition of rapalog , the distribution of KHC576 changes dramatically; it mostly accumulates at the posterior pole , resembling the localization of osk/Staufen ( Figure 5B and D–D’ ) . Some cortical localization of MyoVHMM remains at the anterior and lateral cortex , likely because it was expressed in the oocyte in excess of KHC576 . As a control , we repeated the rapalog treatment after microtubule depolymerization by colcemid . In the absence of microtubules , the dimerized KHC576-MyoVHMM shows no accumulation at the posterior pole ( Figure 5B ) . In summary , our rapalog experiments reveal that a synthetic protein complex containing just two active motors , kinesin-1 and myosin-V , exhibits microtubule-dependent posterior localization in Drosophila oocytes . This dimerized complex is removed by kinesin-1 from the anterior-lateral cortex that has more cortical microtubules and trapped by myosin-V at the posterior cortex where the cortical microtubule density is much lower ( Figure 5E ) . The cortical exclusion model was first proposed after uniform cortical localization of osk mRNA was observed in the kinesin-null oocytes ( Cha et al . , 2002 ) . In agreement with this model , we show that constitutively active kinesin-1 causes osk/Staufen mislocalization in the cytoplasm of the oocyte ( Figure 1 ) , whereas reducing microtubule density at the lateral cortex leads to ectopic accumulation of Staufen at the cortex ( Figure 4 ) . These data support the model that kinesin-driven cortical exclusion along cortically-attached microtubules plays an essential role in restricting osk/Staufen to the posterior pole . Previously , several groups have proposed that kinesin-1 transports osk/Staufen particles along slightly biased cortical microtubules , resulting in net movement of osk/Staufen from the anterior side to the posterior pole in stage 8–9 oocytes ( Palacios and St Johnston , 2002; Lu et al . , 2018; Zimyanin et al . , 2008; Parton et al . , 2011 ) . In fact , kinesin-driven cortical exclusion and kinesin-driven transport towards the posterior pole are not mutually exclusive; they describe the same event of osk/Staufen movement . Within the oocyte , cortical microtubules are anchored to the cortex by their minus-ends while their plus-ends face towards the cytoplasm . Due to the anterior-posterior gradient of cortical microtubule density , more microtubule plus ends are oriented towards the posterior pole ( Nashchekin et al . , 2016; Zimyanin et al . , 2008; Parton et al . , 2011; Khuc Trong et al . , 2015 ) . Thus , kinesin-1-driven transport along microtubules is a prerequisite for kinesin-1-driven cortical exclusion . Cortical exclusion of osk/Staufen by kinesin-1 results in biased transport of osk/Staufen towards the posterior pole ( Figure 5E ) . This kinesin-myosin competition model is suggested by genetic interaction data from a previous study . Specifically , increasing KHC dosage enhances osk/Staufen mislocalization phenotypes in myosin-V loss-of-function mutants , while reducing KHC dosage by half partially suppresses myosin loss-of-function phenotypes ( Krauss et al . , 2009 ) . Furthermore , double MyoV and Khc mutant clones have diffuse cytoplasmic localization of osk mRNA , compared to uniform cortical localization of osk mRNA in Khc single mutant clones ( Krauss et al . , 2009 ) . These data strongly imply that in the absence of kinesin-1 , myosin-V promiscuously anchors osk/Staufen everywhere in the cortex . In this study , we manipulate the activity of either kinesin-1 or myosin-V , and find that proper balance between the activities of these two motors is critical for correct osk/Staufen localization , supporting the model in which kinesin-myosin competition is key to the correct posterior determination in the Drosophila oocyte . The competition between kinesin-1 and myosin-V we described here is not the first example of such a mechanism for cargo transport and localization . For instance , myosin-V opposes microtubule-dependent transport and provides a dynamic anchor for melanosomes ( Rogers and Gelfand , 1998; Wu et al . , 1998; Wu et al . , 1997 ) , peroxisomes ( Kapitein et al . , 2013; van Bergeijk et al . , 2015 ) , recycling endosomes ( van Bergeijk et al . , 2015 ) , mitochondria ( Pathak et al . , 2010 ) , and synaptic vesicles ( Bridgman , 1999 ) at sites of local accumulation of F-actin . At these sites , the abundance of F-actin tracks provides an upper hand for myosin-V to win the tug-of-wars over microtubule motors . Kinesin-myosin competition appears to be an evolutionarily conserved mechanism to allow flexible refinement and/or error correction , as motors constantly undergo reversible binding and releasing activity on cytoskeletal filaments . In the oocytes , the machinery responsible for osk/Staufen localization contains the same basic building blocks; however , unlike the other systems , the outcome of the competition is not determined by actin filament density , as F-actin density is uniform along the oocyte cortex ( Figure 2—figure supplement 1 ) . Instead , the outcome of this competition is decided by abundance of microtubule tracks . Higher microtubule density at the anterior and lateral cortex favors kinesin-mediated cortical removal of osk/Staufen , while scarcity of microtubule tracks at the posterior pole favors myosin-V-dependent anchoring . To confirm this model , we used optogenetic tools to recruit a microtubule-depolymerizing kinesin , kinesin-13/Klp10A , to actin cortex , and thus locally modulate cortical microtubule density . Locally decreasing cortical microtubule density causes ectopic accumulation of Staufen at the cortex . The loss of microtubules prevents kinesin-driven cortical exclusion , which allows myosin-V to win the competition and form a patch of cortically localized Staufen . This recruitment of Staufen is reversible and repeatable , indicating this kinesin-myosin competition is continuous , and the outcome of this never-ending battle is decided by the local microtubule concentration . Previously , synthetic motor domains of a plus-end motor , kinesin-1 , and a minus-end motor , kinesin-14/Nod , were used to label overall microtubule polarity in Drosophila oocytes and neurons ( Kin:βGal and Nod:βGal ) ( Clark et al . , 1994; Clark et al . , 1997 ) . As myosin-V is essential for osk/Staufen localization , in this study we expressed two synthetic motor constructs ( KHC576 and MyoVHMM ) in the oocyte , and induced their dimerization using a rapalog-dependent dimerization system . Dimerized motors accumulate at the posterior pole , highly resembling the osk/Staufen localization ( Figure 5 ) . This posterior accumulation is dependent on the anterior-posterior microtubule gradient; dimerized motors fail to localize at the posterior pole after microtubule depolymerization ( Figure 5 ) . Together , using dimerized synthetic motors , we demonstrate that direct competition ( without any cargo binding ) between a microtubule motor , kinesin-1 , and an actin motor , myosin-V , is sufficient for initial posterior localization in a Drosophila oocyte . In summary , we have elucidated the anchorage mechanism for initial posterior localization of osk/Staufen during mid-oogenesis . Kinesin-1 competes with myosin-V to control osk/Staufen localization . The outcome of this kinesin-myosin competition is primarily determined by cortical microtubule density . Higher microtubule density at anterior-lateral cortex allows kinesin-1 to win and cortically exclude osk/Staufen , while lower microtubule density at posterior pole favors myosin-V-mediated anchorage at the cortex . Together , two cytoskeletal components ( microtubules and F-actin ) and two molecular motors ( kinesin-1 and myosin-V ) govern the posterior determination for future Drosophila embryos . pFB1-MyoV ( ∆1017–1114 aa ) -FLAG and pUASp-MyoV ( ∆1017–1114 aa ) -FLAG: Full-length myosin-V heavy chain ( 1–1792 residues ) with C-terminal FLAG was inserted into pFB1 vector by BamHI ( 5’ ) /XbaI ( 3’ ) . pFB1-MyoV∆1017–1114 aa deletion was generated by replacing the region of pFB1-MyoV . FL with synthesized oligos by PmeI ( 5’ ) /NsiI ( 3’ ) . MyoV∆1017–1114-FLAG was then subcloned into pUASp-attB vector by BamHI ( 5’ ) /XbaI ( 3’ ) . pUASp-Klp10A-GFP-SSpB: Klp10A CDS was amplified by PCR from Klp10A cDNA clone ( LD29208 , DGRC ) and inserted into pUbi-GFP-SspB construct by NheI/AgeI . Klp10A-GFP-SspB fragment was then amplified by PCR and inserted into pUASp by KpnI ( 5’ ) /SpeI ( 3’ ) . pUASp-LifeAct-HA-SsrA: LifeAct and HA-SsrA were inserted into pUASp vector by KpnI ( 5’ ) /SpeI ( 3’ ) and BamHI ( 5’ ) /XbaI ( 3’ ) , respectively . pUASp-EMTB-3XTagRFP: TagRFP was amplified by PCR and 3 copies of TagRFP were cloned into pUASp-attB vector by SpeI ( 5’ ) /BamHI ( 3’ ) using In-Fusion cloning ( Takara ) ; human Ensconsin MT binding domain ( 18–283 aa ) was amplified by PCR and inserted into pUASp-attB-3XTagRFP by NotI ( 5’ ) /SpeI ( 3’ ) . pUASp-KHC576-TagRFP-FKBP: KHC576 ( 1–576 aa ) , TagRFP , FKBP were amplified by PCR and inserted into pMT-A by EcoRI ( 5’ ) /NotI ( 3’ ) , NotI ( 5’ ) /XhoI ( 3’ ) , and XhoI ( 5’ ) /XbaI ( 3’ ) , respectively . KHC576-TagRFP-FKBP was then subcloned into pUASp by KpnI ( 5’ ) /XbaI ( 3’ ) . pUASp-MyoVHMM-GFP-FRB: MyoVHMM ( 1–1100 aa ) , GFP and FRB were amplified by PCR and inserted into pMT-A by SpeI ( 5’ ) /NotI ( 3’ ) , NotI ( 5’ ) /XhoI ( 3’ ) and XhoI ( 5’ ) /XbaI ( 3’ ) , respectively . MyoVHMM-GFP-FRB was then subcloned into pUASp by SpeI ( 5’ ) /XbaI ( 3’ ) . All these pUASp plasmids were sent to BestGene for transposase-mediated P-element insertions , except for the MyoV ( ∆1017-1114aa ) -FLAG for the PhiC31-mediated integration at the attP64 site ( 3R , 89B9 ) . Fly stocks and crosses were kept on standard cornmeal food supplemented with dry active yeast at room temperature ( ~24–25°C ) , except for the optogenetic flies were raised and kept at 18 ~ 19°C . The following fly stocks were used in this study: KhcΔHinge2 ( II , Dr . Jill C . Wildonger , University of Wisconsin-Madison Kelliher et al . , 2018 ) ; nos-Gal4-VP16 ( III Van Doren et al . , 1998; Lu et al . , 2012 ) ; mat αtubGal4-VP16[V2H] ( II , Bloomington Drosophila Stock Center #7062 ) ; mat αtubGal4-VP16[V37] ( III , Bloomington Drosophila Stock Center #7063 ) ; hs-FLP[12] ( X , Bloomington Drosophila Stock Center #1929 Chou and Perrimon , 1996 ) ; FRTG13 ( II , Bloomington Drosophila stock center # 1956 ) ; FRTG13 ubi-GFP . nls ( II , Bloomington Drosophila Stock Center # 5826 ) ; FRT42B didum88 ( II ) , UASp-MyoV . FL-GFP ( III ) ( from Dr . Anne Ephrussi , EMBL Krauss et al . , 2009 ) ; vasa-Gal4 ( II , from Dr . Yukiko Yamashita , University of Michigan ) ; vasa-Gal4 ( III , from Dr . Allan Spradling , Carnegie Institution for Science DeLuca and Spradling , 2018 ) , mat αtub-RFP-Staufen ( III , from Dr . St Johnson , Parton et al . , 2011 ) ; UASp-GFP-Patronin ( II; from Dr . Uri Abdu , Ben Gurion University ) ; Jupiter-GFP ( ZCL2183 ) ( Morin et al . , 2001 ) ; UASp-tdEOS2-αtub84B ( II Lu et al . , 2013 ) . The following fly stocks were generated in this study: UASp-MyoV ( ∆1017–1114 aa ) -FLAG ( attP64 , III ) , UASp-Klp10A-GFP-SspB ( II ) , UASp-HA-LIfeAct-SsrA ( II ) , UASp-EMTB-3XTagRFP ( III ) , UASp-KHC576-TagRFP-FKBP ( III ) , and UASp-MyoVHMM-GFP-FRB ( II ) . A combined double germline-specific Gal4 driver line ( mat αtub-Gal4-VP16[V2H]; nos-Gal4-VP16 ) was used to drive both the RNAi and overexpression , while mat αtubGal4-VP16[V2H]/+; nosGal4-VP16/+ was used as control . A standard recombination protocol was performed between FRTG13 and KhcΔHinge2 . These FRTG13 KhcΔHinge2/CyO virgin female flies were crossed with males carrying hs-flp[12]/y; FRTG13 ubi-GFP . nls/CyO . From these crosses , young pupae at day 7 and day 8 AEL ( after egg laying ) were subjected to heat shock at 37°C for 2 hr . Non CyO F1 females were collected 3–4 day after heat shock and fattened with dry active yeast overnight before dissection for Staufen staining . vasa-Gal4 ( III ) and FRTG13 ubi-GFP . nls were combined to generate yw; FRTG13 ubi-GFP . nls; vasa-Gal4 . UASp-MyoV ( ∆1017–1114 aa ) ( attP64 ) and FRT42B didum88 were combined to generate yw; FRT42B didum88/CyO; UASp-MyoV ( ∆1017–1114 aa ) ( attP64 ) . yw , hs-Flp[12]; sna[Sco]/CyO virgin female flies were crossed with males of yw; FRT42B didum88/CyO; UASp-MyoV ( ∆1017–1114 aa ) ( attP64 ) to generate males of yw , hs-Flp[12]; FRT42B didum88/CyO; UASp-MyoV ( ∆1017–1114 aa ) /+ . The males were then crossed with virgin females of yw; FRTG13 ubi-GFP . nls; vasa-Gal4 . 1st~2nd instar larvae from the cross at day 5 and day 6 AEL ( after egg laying ) were subjected to heat shock at 37°C water bath for 2 hr . Non CyO non-yellow body color ( UASp-MyoV ∆1017–1114 aa inserted attP64 carries a y+ marker ) F1 adult females were collected after heat shock and fattened with dry active yeast overnight before dissection for Staufen staining . A standard fixation and staining protocol was used ( Lu et al . , 2016; Lu et al . , 2012 ) . Samples were incubated with mouse anti-Staufen antibody ( 1:50 , a gift from Dr . Chris Q . Doe , University of Oregon ) at 4°C overnight , washed with PBTB ( 1XPBS + 0 . 1% Triton X-100 + 0 . 2% BSA ) five times for 10 min each time , incubated with TRITC-conjugated anti-mouse secondary antibody ( Jackson ImmunoResearch Laboratories , Inc ) at 1:100 at room temperature ( 24 ~ 25°C ) for 4 hr , and washed with PBTB five times for 10 min each before mounting . Samples were imaged either on a Nikon A1plus scanning confocal microscopy with a GaAsP detector and a 40 × 1 . 30 N . A . oil lens using Galvano scanning , or on a Nikon Eclipse U2000 inverted stand with a Yokogawa CSU10 spinning-disk confocal head and a 40 × 1 . 30 NA lens using an Evolve EMCCD , both controlled by Nikon Elements software . Images were acquired every 0 . 5 µm/step in z stacks . A 5 µm maximum intensity z projection was generated in each sample . The plot profile was either generated along a 3 . 6 µm-wide line delineating the oocyte cortex ( starting and ending at the area where the oocyte meets the nurse cells , Figure 1—figure supplement 1C ) , a 2 . 5 µm-wide line delineating the oocyte cortex ( starting and ending in the middle of the boundary between nurse cells and the oocyte , Figure 2N ) , or an 8 µm-wide line along the anterior–posterior axis ( starting the boundary between nurse cells and oocytes , and ending at the posterior-most follicle cells , Figure 1D ) . We normalized the distance in the plot profiles using a custom MatLab program ( Lu et al . , 2018 ) . We also normalized fluorescence intensity ( maximum and minimum ) in Figures 1 , 2 , 3 and 5; Figure 1—figure supplement 1; Figure 2—figure supplement 1; but not in Figure 4E; Lu , 2020a and Lu , 2020b ) . Both MatLab codes are now available on GitHub ( copies archived at https://github . com/elifesciences-publications/Wen-Lu ) ; normalize both fluorescence intensity and distance ( Lu , 2020a ) ; normalize distance only ( Lu , 2020b ) . cDNAs encoding for myosin-V wild-type heavy chain and myosin-V heavy chain deletion ( ∆1017–1114 aa ) were inserted into a modified pFastBac1 vector which contains a FLAG-tag sequence to the C-terminus . Transposition and the generation of recombinant baculovirus were performed following manufacturer’s protocols ( Thermo Fisher Scientific ) . For protein production , Sf9 insect cells were co-infected with recombinant baculovirus encoding for myosin-V heavy chains , Drosophila calmodulin and Drosophila cytoplasmic myosin light chain ( Mlc-c ) . The protein complex was purified via FLAG affinity chromatography ( Sigma ) as described ( Billington et al . , 2013; Melli et al . , 2018 ) with minor modifications . Eluted proteins were dialyzed overnight against high salt buffer containing 10 mM MOPS , pH 7 . 2 , 500 mM NaCl , 0 . 1 mM EGTA , 2 mM MgCl2 and 1 mM DTT . Myosins were further concentrated by low speed centrifugation ( 4000 X g , 15 min ) with Amicon filter units ( Millipore Sigma ) and flash frozen with liquid nitrogen for future use . Flow chambers were constructed using a 1% w/v nitrocellulose coated #1 . 5 coverslip . Buffers were based on a motility buffer ( MB ) consisting of 20 mM MOPS ( pH 7 . 4 ) , 5 mM MgCl2 , 0 . 1 mM EGTA . Myosin was introduced in high salt buffer ( 0 . 3 mg/ml myosin in MB + 500 mM NaCl , 1 mM DTT - 1 min incubation ) . The surface was blocked with BSA ( 1 mg/ml in MB + 500 mM NaCl , 1 mM DTT - 1 min ) . Inactive myosin heads were blocked with unlabeled F-actin ( 1 µM rabbit skeletal muscle F-actin in MB + 50 mM NaCl , 1 mM ATP - 1 min ) . The chamber was washed using MB + 50 mM NaCl , 1 mM ATP followed by MB + 50 mM NaCl . 10 nM rhodamine phalloidin-labeled actin was added and allowed to land for 1 min before starting adding the final assay buffer ( MB + 50 mM NaCl , 1 mM ATP , 50 mM DTT , 2 . 5 mg/ml glucose , 100 µg/ml glucose oxidase , 40 µg/ml catalase , 0 . 1% Methylcellulose ) . Movies were acquired using a Nikon Eclipse Ti-E microscope ( temperature maintained at 25°C ) at 10 frames per second . Movies were subsequently downsampled 10 fold , allowing for sufficient movement between frames to avoid tracking errors . Motility was quantified using the FAST program ( Aksel et al . , 2015 ) . A tolerance filter of 33% was used to exclude intermittently moving filaments and a minimum velocity filter of 20 nm/s was used to exclude stuck filaments . For each myosin , data from three separate chambers were collected and 3 fields of view were analyzed from each chamber . Steady-state ATPase activities were measured in Cary 60 spectrophotometer ( Agilent Technologies ) at 25°C in buffers containing 10 mM MOPS , pH 7 . 2 , 1 mM ATP , 50 mM NaCl , 2 mM MaCl2 , and either 0 . 1 mM EGTA or 0 . 1 mM free CaCl2 . The buffers also contained 1 µM exogenous calmodulin and an NADH-coupled , ATP-regenerating system including 40 unites/ml lactate dehydrogenase , 200 units/ml pyruvate kinase , 200 µM NADH and 1 mM phosphoenolpyruvate . The rate of ATP hydrolysis was measured from the decrease in absorbance at 340 nm caused by the oxidation of NADH . Myosins were diluted to a concentration of 50 nM in 10 mM MOPS ( pH 7 . 0 ) , 50 mM NaCl ( or 500 mM to visualize the extended conformation ) , 2 mM MgCl2 , 0 . 1 mM EGTA and 0 . 1 mM ATP where indicated . To stabilize the inhibited conformation prior to making grids , samples were crosslinked using 0 . 1% glutaraldehyde for 1 min at room temperature and the reaction was quenched by adding Tris-HCl to a final concentration of 100 mM . Samples were applied to carbon-coated copper grids and stained with 1% uranyl acetate . Micrographs were recorded on a JEOL 1200EXII microscope using an AMT XR-60 CCD camera at a nominal magnification of 60000x . Image processing was performed using SPIDER software as described previously ( Burgess et al . , 2004 ) . An initial dataset of 2255 particles ( WT myosin V ) was aligned and classified into 100 classes using K-means classification . Classes ( 44/100 ) which most clearly demonstrated the inhibited conformation were selected and those particles ( n = 962 ) were realigned and classified into 40 classes . Angles measurements were carried out by manual selection of head pairs and corresponding lever-lever junctions in raw images . The selection coordinates were then use to determine the head-junction-head angles for each molecule ( n = 176 molecules for each myosin ) . Molecules in which the two heads were superimposed due to being in a side orientation were not selected . Ovaries were dissected and gently teased apart in Brinkley Renaturing Buffer 80 ( BRB80 ) , pH 6 . 8 ( 80 mM piperazine-N , N’-bis ( 2-ethanesulfonic acid ) [PIPES] ) , 1 mM MgCl2 , 1 mM EGTA ) ; permeabilized and extracted in BRB80+1% Triton X-100 at 25°C for 1 hr ( without agitation ) ; fixed in 1XPEM ( 100 mM PIPES pH 6 . 9 , 2 mM EGTA and 1 mM MgCl2 ) +0 . 1%Triton X-100+0 . 5% Glutaraldehyde for 20 min on the rotator; briefly washed with 1XPBS three times; quenched in NaBH4+1XPBS on rotator for 20 min; washed with 1XPBS briefly; washed with 1XPBTB ( 1XPBS+0 . 1% Triton X-100+0 . 2%BSA ) five times; blocked with 5% NGS ( normal goat serum ) in PBTB for 1 hr; stained with the 647-conjugated anti-tubulin nanobody ( a gift from Dr . Helge Ewers , Freie Universität Berlin Mikhaylova et al . , 2015 ) 1:50 at 4C overnight; washed with PBTB five times before mounting . Images of nanobody tubulin staining were either imaged on Nikon A1plus scanning confocal microscopy with a GaAsP detector , and a 40 × 1 . 30 N . A . oil lens using Galvano scanning ( Figure 3 ) , or Nikon Eclipse U2000 inverted stand with a Yokogawa CSU10 spinning-disk confocal head and a 40 × 1 . 30 NA lens using an Evolve EMCCD ( Figure 3—figure supplement 1 ) , both controlled by Nikon Elements software . A single focal plane image was used to present the microtubule distribution . All the nanobody images were identically imaged and processed . Microtubules ( only in the oocyte area , not including nurse cells or follicle cells ) were recognized using an ImageJ plugins ( Curvetracing , Steger’s algoristhm , developed in Cell Biology group of Utrecht University by Jalmar Teeuw and Eugene Katrukha ) and the length of all traced lines was measured . The total length of all traced microtubules lines divided by oocyte area was used as an indicator of the microtubule density in each sample . Ovaries from the stock of Jupiter-GFP ( ZCl2183 ) were dissected and gently teased apart in BRB80 Buffer , pH 6 . 8; permeabilized and extracted in BRB80+1% Triton X-100 at 25°C for 1 hr ( without agitation ) ; fixed in 1XPEM ( 100 mM PIPES pH 6 . 9 , 2 mM EGTA and 1 mM MgCl2 ) +0 . 1%Triton X-100+0 . 5% Glutaraldehyde for 20 min; briefly washed with 1XPBS three times; quenched in NaBH4+1XPBS on rotator for 20 min; washed with 1XPBS briefly; washed with 1XPBTB ( 1XPBS+0 . 1% TritonX-100+0 . 2% BSA ) five times before mounting . Samples were imaged using Nikon A1plus scanning confocal microscopy with a GaAsP detector , and a 40 × 1 . 30 N . A . oil lens using Galvano scanning , controlled by Nikon Elements software . Images were acquired every 0 . 5 µm/step in z stacks and a 2 . 5 µm Maximum intensity projection was used to present the microtubule distribution . Ovaries from flies expression tdEOS-αtub under maternal αtub-Gal4-VP16 [V37] ( mat αtub > tdEOS-αtub ) were dissected and gently teased apart in BRB80 Buffer , pH 6 . 8 ) ; permeabilized and extracted in BRB80+1% Triton X-100 at 25°C for 1 hr ( without agitation ) ; fixed in 1XPBS +0 . 1%Triton X-100+4% EM-grade formaldehyde for 20 min on the rotator; briefly washed with 1XPBS five times before mounting . Samples were imaged using Nikon A1plus scanning confocal microscopy with a GaAsP detector , and a 40 × 1 . 30 N . A . oil lens using Galvano scanning , controlled by Nikon Elements software . Images were acquired every 0 . 5 µm/step in z stacks and a 2 . 5 µm Maximum intensity projection was used to present the microtubule distribution . Young newly-eclosed females flies were first fatten with dry active yeast for 18–24 hr ( ~10–15 females + 5 males in a vial ) , and then starved for 8 hr . A yeast paste scrambled with 200 μM colcemid or DMSO ( control ) was then provided to the flies . After 13–14 hr of feeding , flies were dissected for examination . Flies of following were used in the optogenetic recruiting experiments: ( 1 ) yw; UASp-LifeAct-HA-SsrA/UASp-Klp10A-GFP-SspB; nos-Gal4-VP16/UASp-EMTB-3XTagRFP; ( 2 ) yw; UASp-LifeAct-HA-SsrA/+; nos-Gal4-VP16/UASp-EMTB-3XTagRFP; ( 3 ) yw; UASp-LifeAct-HA-SsrA/UASp-Klp10A-GFP-SspB; nos-Gal4-VP16/mat αtub-RFP-Staufen; ( 4 ) yw; UASp-LifeAct-HA-SsrA /+; nos-Gal4-VP16/mat αtub-RFP-Staufen . Flies were crossed and raised at 18°C and protected from light to ensure normal oogenesis . Flies were fattened with dry active yeast for 48 hr at 18°C and then dissected in Halocarbon oil 700 ( Sigma-Aldrich ) as previously described ( Lu et al . , 2016; Lu et al . , 2018 ) . Dissected samples were protected from light and kept in complete darkness for at least 20 min before imaging . Samples were imaged using Nikon A1plus scanning confocal microscopy with a GaAsP detector , and a 40 × 1 . 30 N . A . oil lens using Galvano scanning , controlled by Nikon Elements software . For local recruitment , a 15 μm x 15 μm ROI was stimulated by 488 nm laser at 0 . 06% power ( 488 nm laser power: 15 mW ) for 15 scans at 1 frame per second speed per scan for each recruitment step . For RFP-Staufen imaging , after each recruitment , 561 nm laser were used to image for three frames ( 2 . 12 s intervals ) ; the stimulation and imaging procedures were repeated 30 times for total imaging time of 637 s in each samples . For EMTB-3XTagRFP imaging , the recruitment and waiting cycles were identically to the RFP-Staufen imaging; 561 nm laser were used to image once after two recruitment cycles to reduce photobleaching . Fluorescence intensity in the stimulated region of RFP-Staufen and EMTB-3XTagRFP were measured by ‘Time measurement’ in Nikon Elements , and ImageJ ROI measurement , respectively . Young newly-eclosed female flies ( yw; UASp-MyoVHMM-GFP-FRB/vasa-Gal4; UASp-KHC576-TagRFP-FKBP/+ ) were fatten with dry active yeast for 18–24 hr ( ~10–15 females + 5 males in a vial ) and then dissected in Xpress insect medium ( Lonza ) . The dissected ovaries were then carefully teased apart using G271/2 syringe needles in a glass bottom dish to loosen up the ovarioles . 10 μM a/c heterodimerizer ( Clontech , cat#635057 ) was added to the medium ( the same volume of EtOH was added to the control group ) . After 1 hr of a/c heterodimerizer incubation , the ovarioles were fixed with 4% EM-grade methanol-free formaldehyde diluted in PBT ( 1 × PBS and 0 . 1% Triton X-100 ) for 20 min on a nutator , and washed five times with PBTB for 10 min each before mounting . Samples were imaged on Nikon Eclipse U2000 inverted stand with a Yokogawa CSU10 spinning-disk confocal head and a 40 × 1 . 30 NA lens using an Evolve EMCCD controlled by Nikon Elements software . Images were acquired every 0 . 5 µm/step in Z stacks .	One of the most fundamental steps of embryonic development is deciding which end of the body should be the head , and which should be the tail . Known as 'axis specification' , this process depends on the location of genetic material called mRNAs . In fruit flies , for example , the tail-end of the embryo accumulates an mRNA called oskar . If this mRNA is missing , the embryo will not develop an abdomen . The build-up of oskar mRNA happens before the egg is even fertilized and depends on two types of scaffold proteins in the egg cell called microtubules and microfilaments . These scaffolds act like ‘train tracks’ in the cell and have associated protein motors , which work a bit like trains , carrying cargo as they travel up and down along the scaffolds . For microtubules , one of the motors is a protein called kinesin-1 , whereas for microfilaments , the motors are called myosins . Most microtubules in the egg cell are pointing away from the membrane , while microfilament tracks form a dense network of randomly oriented filaments just underneath the membrane . It was already known that kinesin-1 and a myosin called myosin-V are important for localizing oskar mRNA to the posterior of the egg . However , it was not clear why the mRNA only builds up in that area . To find out , Lu et al . used a probe to track oskar mRNA , while genetically manipulating each of the motors so that their ability to transport cargo changed . Modulating the balance of activity between the two motors revealed that kinesin-1 and myosin-V engage in a tug-of-war inside the egg: myosin-V tries to keep oskar mRNA underneath the membrane of the cell , while kinesin-1 tries to pull it away from the membrane along microtubules . The winner of this molecular battle depends on the number of microtubule tracks available in the local area of the cell . In most parts of the cell , there are abundant microtubules , so kinesin-1 wins and pulls oskar mRNA away from the membrane . But at the posterior end of the cell there are fewer microtubules , so myosin-V wins , allowing oskar mRNA to localize in this area . Artificially 'shaving' some microtubules in a local area immediately changed the outcome of this tug-of-war creating a build-up of oskar mRNA in the 'shaved' patch . This is the first time a molecular tug-of-war has been shown in an egg cell , but in other types of cell , such as neurons and pigment cells , myosins compete with kinesins to position other molecular cargoes . Understanding these processes more clearly sheds light not only on embryo development , but also on cell biology in general .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2020	Competition between kinesin-1 and myosin-V defines Drosophila posterior determination
The integrated stress response ( ISR ) attenuates the rate of protein synthesis while inducing expression of stress proteins in cells . Various insults activate kinases that phosphorylate the GTPase eIF2 leading to inhibition of its exchange factor eIF2B . Vanishing White Matter ( VWM ) is a neurological disease caused by eIF2B mutations that , like phosphorylated eIF2 , reduce its activity . We show that introduction of a human VWM mutation into mice leads to persistent ISR induction in the central nervous system . ISR activation precedes myelin loss and development of motor deficits . Remarkably , long-term treatment with a small molecule eIF2B activator , 2BAct , prevents all measures of pathology and normalizes the transcriptome and proteome of VWM mice . 2BAct stimulates the remaining activity of mutant eIF2B complex in vivo , abrogating the maladaptive stress response . Thus , 2BAct-like molecules may provide a promising therapeutic approach for VWM and provide relief from chronic ISR induction in a variety of disease contexts . eIF2B is the guanine nucleotide exchange factor ( GEF ) for the GTPase and initiation factor eIF2 and modulation of its activity is central to regulation of protein synthesis rates in all eukaryotic cells . GTP-bound eIF2 associates with the initiator methionyl tRNA and this ternary complex ( eIF2-GTP-Met-tRNAi ) delivers the first amino acid to the ribosome . GTP is hydrolyzed and eIF2-GDP is released , requiring reactivation by eIF2B to enable a new round of protein synthesis ( Hinnebusch and Lorsch , 2012 ) . Four stress-responsive kinases ( PERK , HRI , GCN2 and PKR ) that detect diverse insults converge on phosphorylation of eIF2 at serine 51 of the α subunit ( eIF2α ) . Phosphorylation converts eIF2 from a substrate of eIF2B into its competitive inhibitor , triggering the integrated stress response ( ISR ) , which reduces translation initiation events and decreases global protein synthesis ( Krishnamoorthy et al . , 2001; Yang and Hinnebusch , 1996 ) . Concomitantly , the ISR induces the translation of a small set of mRNAs with special sequences in their 5’ untranslated regions , including the transcription factor ATF4 ( Harding et al . , 2000; Watatani et al . , 2008 ) . ATF4 triggers a stress-induced transcriptional program that is thought to promote cell survival during mild or acute conditions but can contribute to pathological changes under severe or chronic insults ( Pakos-Zebrucka et al . , 2016 ) . Vanishing White Matter ( VWM; OMIM 603896 ) is a rare , autosomal recessive leukodystrophy that is driven by mutations in eIF2B ( Leegwater et al . , 2001; van der Knaap et al . , 2002 ) . The disease is characterized by myelin loss , progressive neurological symptoms such as ataxia , spasticity , cognitive deterioration and ultimately , death ( Schiffmann et al . , 1994; van der Knaap et al . , 1997 ) . Age of VWM onset is variable and predictive of disease progression , ranging from severe prenatal/infantile onset leading to death in months ( as in the case of ‘Cree leukoencephalopathy’ ) to slower-progressing adult onset presentations ( Fogli et al . , 2002; Hamilton et al . , 2018; van der Knaap et al . , 2006 ) . Because eIF2B is essential , VWM mutations are restricted to partial loss-of-function in any of the five subunits of the decameric complex ( Fogli et al . , 2004; Horzinski et al . , 2009; Li et al . , 2004; Liu et al . , 2011 ) . Nearly 200 different mutations have been catalogued to date in the Human Gene Mutation Database , which occur as homozygotes or compound heterozygotes with a different mutation in each allele of the same subunit . Reduction of eIF2B activity is analogous to its inhibition by phosphorylated eIF2α , thus it is congruent and compelling that ISR activation has been observed in VWM patient post-mortem samples ( van der Voorn et al . , 2005; van Kollenburg et al . , 2006 ) . We previously showed that a range of VWM mutations destabilize the eIF2B decamer , leading to compromised GEF activity in both recombinant complexes and endogenous protein from cell lysates ( Wong et al . , 2018 ) . We demonstrated that the small molecule eIF2B activator ISRIB ( for ISR inhibitor ) stabilized the decameric form of both wild-type ( WT ) and VWM mutant complexes , boosting their intrinsic activity . Notably , ISRIB bridges the symmetric dimer interface of the eIF2B central core , acting as a molecular stapler ( Tsai et al . , 2018; Zyryanova et al . , 2018 ) . In addition , we showed that ISRIB attenuated ISR activation and restored protein synthesis in cells carrying VWM mutations ( Wong et al . , 2018 ) . Although we found that ISRIB rescued the stability and activity of VWM mutant eIF2B in vitro , the ability of this class of molecules to prevent pathology in vivo remained an unanswered question . Knock-in mouse models of human VWM mutations have been characterized , and the severe mutations recapitulate key disease phenotypes such as progressive loss of white matter with concomitant manifestation of motor deficits ( Dooves et al . , 2016; Geva et al . , 2010 ) . Here , we generate an improved tool molecule and demonstrate that sustained eIF2B activation blocks maladaptive induction of the ISR and prevents all evaluated disease phenotypes in a VWM mouse model . To interrogate efficacy in vivo , we sought a small molecule eIF2B activator with improved solubility and pharmacokinetics relative to ISRIB . To that end , we synthesized 2BAct , which has a differentiated bicyclo[1 . 1 . 1]pentyl core and , unlike ISRIB , is no longer symmetric ( Figure 1A ) . 2BAct is a highly selective eIF2B activator and exhibited similar potency to ISRIB in a cell-based reporter assay that measures ISR attenuation ( Figure 1—figure supplement 1A and Supplementary file 1A ) . 2BAct is able to penetrate the central nervous system ( CNS ) ( unbound brain/plasma ratio ~0 . 3 ) and also demonstrated dose-dependent oral bioavailability using an aqueous-based vehicle ( Supplementary file 1B ) . Additionally , 2BAct is well-suited for formulation in diet , enabling long-term dosing without effects on body weight in WT mice ( Figure 1—figure supplement 1B ) . The molecule was well-tolerated in the animal studies described here , and did not elicit any relevant effects in a rat cardiovascular ( CV ) safety study; however , significant anomalies were observed in a dog CV model . This CV safety liability makes this particular molecule unsuitable for human dosing . We generated a previously described mouse model of VWM that harbors the severe ‘Cree leukoencephalopathy’ mutation , Eif2b5R191H/R191H ( hereafter referred to as R191H; Figure 1—figure supplement 2 ) ( Dooves et al . , 2016 ) . The homologous human Eif2b5R195H mutation causes an infantile-onset , rapidly progressing form of VWM ( Black et al . , 1988 ) . R191H mice recapitulate many aspects of the human disease , such as spontaneous myelin loss , progressive ataxia , motor skill deficits , and shortened lifespan ( Dooves et al . , 2016 ) . We selected this severe disease allele for in vivo studies as pharmacological efficacy in this model is a stringent test for eIF2B activators and , mechanistically , should generalize to milder mutations as seen in vitro ( Wong et al . , 2018 ) . We confirmed that primary fibroblast lysates from R191H embryos had lower GEF activity than WT lysates , and that 2BAct enhanced this activity threefold ( EC50 = 7 . 3 nM; Figure 1—figure supplement 1C–D ) . To test efficacy , we undertook a 21-week blinded treatment study with 2BAct and measured intermediate and terminal phenotypes associated with disease progression in R191H mice ( Figure 1B ) . 2BAct was administered orally by providing mice with the compound incorporated in rodent meal . This dosing regimen provided unbound brain exposures 15-fold above the in vitro EC50 at the end of the study , ensuring saturating coverage of the target . At the initiation of the study ( 6–11 weeks old ) , WT and R191H males had similar body weights ( Figure 1C and Figure 1—figure supplement 3A ) . However , 1 week later , a significant difference emerged and continued to grow as R191H mice failed to gain weight ( WT gain = 0 . 5 g/week vs . R191H gain = 0 . 08 g/week; p<10−4 ) . Remarkably , the body weight of 2BAct-treated R191H males caught up to WT mice 2 weeks after beginning dosing ( 8–13 weeks old ) , at which point their rate of weight gain equalized ( WT gain = 0 . 5 g/week vs . R191H gain = 0 . 55 g/week; p = 0 . 24; Figure 1C ) . Similar results were observed in female mice ( Figure 1—figure supplement 4A ) . As lack of weight gain appeared to be the first overt phenotype , rapid normalization by 2BAct was a promising prognostic sign of efficacious target engagement . Longitudinal characterization revealed that R191H mice developed progressive , age-dependent strength and motor coordination deficits . From 8 to 19 weeks of age , R191H animals were not significantly different from WT in their performance on an inverted grid test of neuromuscular function . At 23 weeks , R191H mice showed a trend towards shorter hang times and at 26 weeks , this decrease was highly significant in both sexes ( Figure 1—figure supplement 3B ) . In a beam-crossing test of balance and motor coordination , R191H mice were not significantly different from WT littermates at 8–19 weeks of age ( Figure 1—figure supplement 3C–D ) . However , at 23 weeks of age , beam-crossing time was significantly increased in mutant animals , and they exhibited more foot slips and falls from the beam while crossing . The deficit in both parameters was exacerbated at 26 weeks , and some R191H animals completely failed to cross the beam within the trial cutoff of 30 s ( Figure 1—figure supplement 3C ) . These results are consistent with the original description of the R191H model ( Dooves et al . , 2016 ) . With baseline performance measured , we assessed the effect of 2BAct treatment on R191H motor skills . Placebo-treated R191H males had significantly reduced inverted grid hang times at both tested time points , as well as more coordination errors and time spent crossing the balance beam ( Figure 1D–F ) . By contrast , 2BAct-treated R191H males were indistinguishable from WT in both assays . Full normalization was similarly observed in female animals ( Figure 1—figure supplement 4B–D ) . Together , these data show that treatment of R191H animals with 2BAct prevented the progressive deterioration of motor function caused by VWM . VWM is a leukoencephalopathy defined by progressive loss of myelin . In patients with advanced disease , an almost complete loss of cerebral white matter is observed ( van der Knaap et al . , 1997 ) . Similarly , in a previous characterization of the R191H mouse model , perturbed myelination and myelin vacuolization were noted in the brain beginning at 4–5 months of age ( Dooves et al . , 2016; Klok et al . , 2018 ) . Given the dramatic rescue of behavioral phenotypes by 2BAct , we performed immunohistochemical analysis to examine its effects on myelin and accompanying pathologies at the end of the treatment . We focused on two heavily myelinated regions , the corpus callosum and the spinal cord . Notably , severe spinal cord pathology has recently been reported in both VWM patients and R191H mice ( Leferink et al . , 2018 ) . As anticipated , R191H animals showed a clear reduction in myelin by Luxol Fast Blue staining of both regions ( 33% reduction in corpus callosum , p<10−4; 58% reduction in cervical/thoracic spinal cord , p<10−4; Figure 2A–D ) . Strikingly , 2BAct treatment maintained myelin levels at 91% and 85% of WT in the corpus callosum and spinal cord , respectively . Staining for myelin basic protein ( MBP ) , an alternative measure of myelin content , corroborated these results in both regions ( Figure 2—figure supplement 1A–D ) . As astrocytes have been implicated in the pathogenesis of VWM , we also stained for the astrocyte marker GFAP ( Dietrich et al . , 2005; Dooves et al . , 2016 ) . We found a significant increase in GFAP in both regions of placebo-treated R191H mice , which was fully normalized by 2BAct treatment ( Figure 2A–B and Figure 2—figure supplement 1C–D ) . Consistent with the literature , we noted a significant increase in Olig2 in R191H spinal cord ( Figure 2—figure supplement 1C–D ) . Olig2 is a marker of the oligodendrocyte lineage , and its increase could indicate an attempt to compensate for the myelin loss ( Bugiani et al . , 2011 ) . Additionally , we observed signs of reactive microglia in the placebo-treated R191H samples , as evidenced by a 15-fold increase in Iba-1 staining ( Figure 2C–D ) . ATF3 , an ISR target induced in the spinal cord during injury and inflammation , was also significantly increased ( Dominguez et al . , 2010; Hossain-Ibrahim et al . , 2006 ) . 2BAct treatment fully normalized all four of these markers ( Figure 2C–D and Figure 2—figure supplement 1C–D ) . In an analysis of younger mice , we observed no significant differences in myelin , Iba-1 or ATF3 between R191H and WT spinal cord at the start of treatment ( 2 months of age; Figure 2—figure supplement 2 ) . The time course of pathology was consistent with the emergence of motor deficits , and reflected the degenerative nature of VWM . Together , these results demonstrate that dosing with 2BAct before onset of histological signs prevents CNS pathology in a mouse model of VWM . In all eukaryotic systems , ATF4 protein expression is regulated by the level of ternary complex in cells ( Harding et al . , 2000; Mueller and Hinnebusch , 1986; Vattem and Wek , 2004 ) . We postulated that the decrease in eIF2B GEF activity brought about by the Eif2b5R191H mutation would reduce levels of ternary complex , leading to upregulated ATF4 translation in R191H mice . In support of this , ISR activation has been reported in patient VWM postmortem samples ( van der Voorn et al . , 2005; van Kollenburg et al . , 2006 ) . To evaluate ISR activity , we measured the expression of 15 transcripts previously identified as ATF4 target genes at three different time points ( 2 . 5 , 5 , and 7 months ) during the lifespan and development of pathology . The ISR was robustly and consistently induced at these time points in the cerebellum , forebrain , midbrain and hindbrain of R191H animals ( Figure 3A and Figure 3—figure supplement 1A ) . Significant upregulation of all targets except Gadd34 , Slc1a5 and Gadd45a was evident in cerebellum at 2 . 5 months; at later timepoints , these three transcripts also became significantly induced . The ISR signature was similar across all brain regions with Atf5 , Eif4ebp1 and Trib3 being the most upregulated transcripts in the panel . Interestingly , we did not observe ISR induction in R191H mice at postnatal day 14 ( Figure 3—figure supplement 1B ) . Thus , the ISR is activated sometime between 2 and 8 weeks of age through an as-yet unknown mechanism . Activation of this stress response preceded the appearance of pathology ( myelin loss , gliosis and motor deficits ) in VWM mice . In addition to changes in transcript levels , we confirmed translational induction of ATF4 as well as the increase in protein levels of the negative regulator of cap-mediated mRNA translation EIF4EBP1 by Western blot analysis of 7-month-old R191H cerebellum lysates ( Figure 3B ) . As expected for an ISR induced by eIF2B dysfunction rather than external stressors ( see schematic in Figure 3C ) , we did not detect an increase in eIF2α phosphorylation in R191H brains . We observed greater ISR induction in the spinal cord compared to the cerebellum ( Figure 3D , compare red and brown points ) , with the transcription factor Atf3 showing an additional 10-fold increase . The greater extent of myelin loss in the spinal cord , as well as astrocyte and microglial activation , suggests exacerbation of the phenotype due to increased ISR activation . Notably , treatment with 2BAct abolished ISR induction in both regions ( Figure 3B , D ) . The striking attenuating effect of this molecule on the ISR is consistent with full rescue of GEF activity in vivo . To determine whether our results would extend to other VWM mutations , we examined a second mouse model of VWM bearing an Eif2b5R132H/R132H mutation , which corresponds to the disease-causing human Eif2b5R136H allele ( Geva et al . , 2010 ) . This model has a normal lifespan and exhibits very mild phenotypes in comparison to R191H mice . Nevertheless , we detected significant upregulation of two ISR targets , Atf5 and Eif4ebp1 , that was blocked by 4 weeks of treatment with 2BAct ( Figure 3—figure supplement 2 ) . Even though patients carry the eIF2B mutation ( s ) in all cell types , VWM manifests as a CNS disease with the exception of ovarian failure in late-onset female patients . In extremely rare and severe cases , renal dysplasia and hepatomegaly have also been recently reported ( Hamilton et al . , 2018 ) . To evaluate the impact of the R191H mutation on other tissues , we interrogated ISR target expression in various peripheral organs . Upregulation of ATF4 targets was not detected in skeletal muscle , liver , kidney or ovaries ( Figure 3—figure supplement 3 ) , demonstrating that the CNS is particularly sensitive to a reduction in eIF2B function . Because our targeted RNA panel consisted of only 15 genes , we turned to RNA-seq in order to comprehensively profile the transcriptional changes that take place in the VWM brain . We analyzed cerebellum from WT and R191H mice at 2 , 5 and 7 months of age in order to assess potential changes in R191H mice as they develop pathology . We confirmed the upregulation of ISR target genes beginning as early as 2 months of age , the magnitude of which was sustained at 5 and 7 months ( Figure 4—figure supplement 1A–C ) . By contrast , and as expected for a disease driven by dysfunction in eIF2B , we did not observe changes in expression for known downstream targets of the parallel IRE1α or ATF6-dependent branches of the unfolded protein response at any time point ( Figure 4—figure supplement 1A–C ) . In order to identify additional classes of transcripts that distinguish R191H from WT , we performed singular value decomposition ( SVD ) analysis on the dataset from 2-month-old mice . We focused on genes with the largest positive and negative loadings on the first eigengene ( i . e . the first singular vector from SVD [Alter et al . , 2000] ) , which separated the samples by genotype ( Figure 4A–B ) . The first class consisted of 473 genes with increased expression ( positive loadings ) in R191H mice . Unsurprisingly , GO-term enrichment analysis revealed categories that contained many known ATF4-dependent targets involved in amino acid metabolism and tRNA aminoacylation ( Supplementary file 1C ) ( Adamson et al . , 2016; Han et al . , 2013 ) . A second class comprised 600 genes with reduced expression ( negative loadings ) . These genes were not restricted to expression in a specific cell type , but GO-term enrichment analysis revealed categories related to myelination and lipid metabolism ( Supplementary file 1C ) , suggesting an effect on glial cells such as astrocytes and oligodendrocytes . A gene signature of perturbed myelin maintenance is detectable as early as 2 months , preceding evidence of myelin loss by histological analysis . A heatmap of the 50 genes with the largest absolute loadings on the first eigengene revealed that the expression of both classes persisted and was consistent as the animals aged ( Figure 4C ) . Our targeted analysis revealed complete abrogation of the ISR by 2BAct treatment ( Figure 3D ) . To test whether 2BAct could also normalize the broad downregulation of transcripts related to CNS function , we performed RNA-seq on cerebellum from 2 . 5-month-old WT and R191H mice treated for only 4 weeks . Remarkably , both upregulated and downregulated classes of transcripts were normalized in 2BAct-treated R191H mice ( Figure 4D ) . Clustering of samples based on the top 50 genes from the previous analysis confirmed that 2BAct-treated R191H mice were indistinguishable from WT ( Figure 4E ) . Thus , 2BAct normalized the defective expression of glial and myelination-related genes . Moreover , 2BAct treatment of WT mice did not significantly alter gene expression compared to placebo ( Figure 4—figure supplement 2A–B ) . Together , these data demonstrated that 2BAct treatment normalizes the aberrant transcriptional landscape of VWM mice without eliciting spurious gene expression changes in WT mice . The robust induction of ISR targets in the brain of VWM mice raised the question of whether all CNS cell types are uniformly affected , or whether a subpopulation of cells is particularly susceptible to a decrease in eIF2B function . To address this , we performed single cell RNA-seq ( scRNA-seq ) on two brain regions , forebrain and cerebellum , of 2 . 5-month-old WT and R191H mice . Unbiased clustering of single cells from each region identified 13 and 10 clusters in the forebrain and cerebellum , respectively , that were subsequently assigned cell type identities using CNS cell type gene markers obtained from bulk RNA-seq data ( Figure 5A , Figure 5—figure supplement 1 and Figure 5—figure supplement 2A ) ( Koirala and Corfas , 2010; Zhang et al . , 2014 ) . Cells from both WT and R191H tissues were represented in each cluster , demonstrating that transcriptionally defined cell types are not influenced by genotype at this early time point in disease progression . Next , we generated an unbiased , tissue-independent ISR target signature by using the top 50 upregulated genes from our bulk RNA-seq analysis of R191H cerebellum as input for Clustering by Inferred Co-Expression ( CLIC ) analysis ( Li et al . , 2017 ) . CLIC identified 18 of our input genes as coherently co-expressed across 1774 diverse mouse microarray datasets , and expanded this co-expression module to a final signature comprising 95 genes ( Supplementary file 1D ) . Using the CLIC-derived signature , we first assessed ISR expression in WT cells only . Strikingly , two astrocyte clusters in the forebrain and Bergmann glia in the cerebellum showed significant enrichment of the ISR signature ( q = 0 . 004 , 0 . 008 and 0 . 004 , respectively ) , indicating that these cell types have higher basal expression of ISR targets ( Figure 5B and Figure 5—figure supplement 2B ) . By contrast , the ISR signature was insignificant ( using a threshold of q = 0 . 05 ) or even negatively enriched in other WT CNS cell types . Next , we investigated the source of upregulated ISR expression in R191H compared to WT . In the forebrain , the two astrocyte clusters with an enriched ISR signature in WT showed significant further upregulation in R191H tissue ( q < 10−3 for both clusters; Figure 5C ) . In the cerebellum , Bergmann glia also showed the most significant enrichment of the ISR signature in R191H tissue ( q = 0 . 0004; Figure 5—figure supplement 2C ) . Astrocytes , and more recently Bergmann glia , have long been implicated as the affected cell type in VWM based on histology and ex vivo analyses ( Bugiani et al . , 2011; Dietrich et al . , 2005; Dooves et al . , 2016; Dooves et al . , 2018; Geva et al . , 2010; Leferink et al . , 2018; Wong et al . , 2000 ) . For the first time , our work provides in vivo evidence that in VWM mice , astrocytes exhibit a molecular signature of an early maladaptive ISR . Interestingly , the comparison of R191H to WT cerebellum also revealed significant enrichment of the ISR signature in other non-neuronal cell types , including myelinating oligodendrocytes , an unassigned cell type , and endothelial cells ( q = 0 . 001 , 0 . 005 and 0 . 02 , respectively; Figure 5—figure supplement 2C ) . We were unable to confidently assign an identity to the unknown cell type , but its expression profile is consistent with a non-neuronal lineage ( cluster four in Figure 5—figure supplement 1B ) . Our findings held true when we repeated the analysis using a manually curated list of ISR target genes ( Figure 5—figure supplement 3 and Supplementary file 1D ) . Collectively , the results of our unbiased analysis suggest that in R191H brain , other cell types beyond astrocytes upregulate the ISR and may contribute to pathology . Because eIF2B is essential for translation initiation and the Eif2b5R191H mutation reduces its GEF activity , we wondered how well the changes observed by RNA-seq would correlate with changes in the proteome . To address this , we performed tandem mass tag mass spectrometry ( TMT-MS ) on cerebellum samples at the end of the 2BAct treatment study . We discovered 42 proteins that increased >1 . 5 fold in abundance , and 19 proteins that decreased >1 . 5 fold in abundance in placebo-treated R191H vs . WT at a posterior probability >90% ( Figure 6A ) . Of the upregulated proteins , 21/42 were transcriptionally upregulated in the RNA-seq experiment . The remaining half did not meet an abundance threshold in the RNA-seq analysis . Among these were known ISR targets such as amino acid transporters ( SLC1A4 and SLC7A3 ) and metabolic enzymes ( CTH and PYCR1 ) . However , TMT-MS did not detect ATF4 or some of its low-abundance targets ( e . g . the transcription factors ATF3 , ATF5 and CHOP ) . Nevertheless , the presence of the other ISR targets in this set drove the enrichment of ISR-associated pathways in GO-term analysis ( Figure 6B ) . The good agreement between RNA-seq and TMT-MS data , as well as the ability to detect targets present in one but not the other , highlight their utility as complementary approaches . Of the downregulated proteins , 13/19 were transcriptionally downregulated in the RNA-seq experiment , one did not change transcriptionally and five did not meet the RNA-seq abundance threshold . We did not identify significant enrichment of gene sets in the downregulated proteins , likely due to the small number of targets that met our cutoff criteria . Remarkably , 14/19 of these targets are most highly expressed in Bergmann glia or astrocytes , and 9/19 are highly expressed in oligodendroyctes or oligodendrocyte precursor cells ( Tabula Muris Consortium et al . , 2018 ) . One interesting example that falls into both categories is the fatty-acid binding protein FABP7 , the downregulation of which is again suggestive of dysregulation of lipid metabolism ( Kipp et al . , 2011; Kurtz et al . , 1994 ) . Our proteomic data are consistent with the downregulated GO categories observed in RNA-seq , as well as the ISR activation seen in astrocytes and myelinating oligodendrocytes by scRNA-seq . Together , they implicate glial cells as a potential source of dysregulation . Importantly , all downregulated proteins and 40/42 upregulated proteins were normalized by 2BAct treatment ( Figure 6A right panel and Figure 6C ) . Interestingly , we discovered that levels of all five eIF2B subunits were reduced 15–35% in R191H cerebellum compared to WT ( Figure 6D ) . This decrease occurred at the protein level , as no changes were observed in transcript abundance by RNA-seq . We had previously observed destabilization of eIF2B in HEK293T cells , wherein a VWM mutation in one member of the complex caused a reduction in itself as well as the other subunits ( Wong et al . , 2018 ) . The finding that long-term treatment with 2BAct does not rescue eIF2B complex levels in vivo suggests that the normalization of all measured endpoints in R191H mice ( Table 1 ) is due to boosting the GEF activity of the remaining mutant complex to functionally normal levels . Introduction of a severe VWM mutation into mice led to spontaneous loss of myelin and motor deficits that reproduce key aspects of the human disease . The impaired GEF activity of mutant eIF2B underlies the observed translational upregulation of the transcription factor ATF4 and in turn , induction of a maladaptive ISR program that precedes behavioral pathology . Our data demonstrate that a subset of astrocytes possess a basal ISR in normal mice , which is further activated when eIF2B is mutated . Notably , we also detected ISR upregulation in other non-neuronal cell populations , including myelinating oligodendrocytes . Histological analysis has shown that white matter astrocytes , but not grey matter astrocytes , are dysmorphic and a subset of oligodendrocytes are characterized as foamy in VWM patients ( Hata et al . , 2014; Wong et al . , 2000 ) . Moreover , Bergman glia , astrocytes found in the cerebellum , and Muller glia , astrocytes found in the retina , were shown to be severely affected in patients and in VWM mouse models ( Dooves et al . , 2016; Dooves et al . , 2018 ) . In agreement with histology , scRNA-seq of two brain regions identified astrocyte subtypes with exacerbated ISR induction in R191H . As cells in the brain are highly interconnected , cellular and metabolic dysfunction in astrocytes could lead to malfunction in other cell types such as the myelinating oligodendrocytes . Whether ISR activation in the non-astrocytic cells is cell-autonomous or triggered by signals from astrocytes remains to be explored . Among the ATF4 targets revealed by both transcriptomics and proteomics are solute transporters ( SLC1A4 , SLC1A5 , SLC3A2 , SLC7A1 , SLC7A3 , SLC7A5 , SLC7A11 ) and metabolic enzymes ( ASNS , CTH , CBS , PLPP4 , PHGDH , PSAT1 , PSPH , SHMT2 and MTHFD2 ) , and their upregulation is likely to disrupt cellular functions in the CNS . Interestingly , ATF4 is chronically induced in various mouse models of mitochondrial dysfunction , and mutations in human mitochondrial proteins can lead to loss of myelin or leukoencephalopathies ( Carvalho , 2013; Dogan et al . , 2014; Huang et al . , 2013; Mendes et al . , 2018; Moisoi et al . , 2009; Pereira et al . , 2017; Quirós et al . , 2017; Taylor et al . , 2014 ) . An important question for future investigation is whether these diseases are also characterized by a chronic ISR that may be protective or maladaptive , and which cell types are susceptible to its induction . The induction of Eif4ebp1 and Sesn2 , two negative regulators of the mTOR signaling pathway , suggests that impairment of translation in VWM may extend beyond the direct effect of crippled eIF2B activity ( Budanov and Karin , 2008; Wolfson et al . , 2016; Yanagiya et al . , 2012 ) . We speculate that inhibition of the eIF2 and mTOR pathways could lead to a dual brake on protein synthesis . If protein synthesis is globally reduced by chronic ISR activation in oligodendrocytes , this could directly disrupt maintenance of the myelin sheath . The ISR signature of VWM mice is unique as it originates downstream of the canonical sensor , that is eIF2α phosphorylation by stress-responsive kinases . Thus , it provides insight into the transcriptional and proteomic changes driven by the response in vivo in the absence of exogenously applied pleiotropic stressors . In contrast to ISR induction via stress-responsive kinases , the negative feedback loop elicited by both the constitutive CREP-containing and the ISR-inducible GADD34-containing eIF2α phosphatases is not effective in attenuating the response in VWM . Therefore , it constitutes a ‘locked-on’ ISR ( Figure 3C ) . However , the system can still respond to further stress , as seen in VWM patients wherein provoking factors ( e . g . febrile infections or head trauma ) exacerbate the disease and lead to poorer outcomes ( Hamilton et al . , 2018 ) . 2BAct had a normalizing effect on the transcriptome , proteome , myelin content , microglial activation , body weight and motor function of R191H VWM animals . Long-term administration of 2BAct starting at ~8 weeks of age is a preventative treatment paradigm . Beginning treatment at later time points would likely result in ISR attenuation , as we predict that eIF2B activation by 2BAct is age-independent . However , the degree of pathological amelioration in a therapeutic mode would likely depend on the extent of neuronal damage and myelin perturbation in the nervous system at the start of treatment . The accrued damage in turn would depend on the severity of the eIF2B mutant allele and the time elapsed since disease onset . The remyelination capacity of humans and mice is likely to differ significantly , thus the therapeutic value of eIF2B activators in VWM is best interrogated in the clinic . The small molecule 2BAct has cardiovascular liabilities at a minimal efficacious dose and is not suitable for human dosing . Therefore , therapeutic testing of this class of molecules awaits the generation of a suitable candidate . Nevertheless , ISRIB has shown beneficial effects in mouse models when administered either acutely or for short periods of time by intraperitoneal delivery ( Chou et al . , 2017; Halliday et al . , 2015; Li et al . , 2018; Nguyen et al . , 2018; Sidrauski et al . , 2013; Wang et al . , 2018 ) . The improved in vivo properties of 2BAct make it an ideal molecule to interrogate the efficacy of this mechanism of action in rodent disease models that require long-term dosing , including those characterized by chronic activation of any of the four eIF2α kinases . We and others previously demonstrated that stabilizing the decameric complex can boost the intrinsic GEF activity of both WT and various VWM mutant eIF2B complexes ( Tsai et al . , 2018; Wong et al . , 2018 ) . To date , all tested VWM mutations have responded to eIF2B activators in vitro . Here , we show that increased GEF activity can compensate in vivo under conditions where a severe mutation leads to both decreased levels of eIF2B complex and reduced intrinsic activity . Furthermore , 2BAct also shut off the ISR in vivo in mice bearing an Eif2b5R132H/R132H VWM mutation . Although it is possible to introduce artificial mutations into eIF2B that interfere with the compound binding site , no VWM mutations are known to exist in this region ( Sekine et al . , 2015; Tsai et al . , 2018; Wong et al . , 2018 ) . Based on this , we anticipate that 2BAct-like compounds may be broadly efficacious across the range of mutations identified in the VWM patient population . Finally , our results demonstrate that eIF2B activation is a sound strategy for shutting off the ISR in vivo . This raises the possibility that various diseases exhibiting a maladaptive ISR could be responsive to this mechanism of action . To a solution of 5- ( difluoromethyl ) pyrazine-2-carboxylic acid ( 20 . 3 g , 117 mmol ) and tert-butyl ( 3-aminobicyclo[1 . 1 . 1]pentan-1-yl ) carbamate ( 22 . 0 g , 111 mmol ) in N , N-dimethylformamide ( 400 mL ) at ambient temperature was added triethylamine ( 61 . 9 mL , 444 mmol ) . The 2- ( 3H-[1 , 2 , 3] triazolo[4 , 5-b]pyridin-3-yl ) - 1 , 1 , 3 , 3-tetramethylisouronium hexafluorophosphate ( V ) ( 44 . 3 g , 117 mmol ) was added portion wise over 60 min and the mixture was allowed to stir at ambient temperature for 23 hr . The mixture was quenched with saturated , aqueous NH4Cl ( 75 mL ) and water ( 30 mL ) and diluted with EtOAc ( 100 mL ) . The layers were separated and the aqueous layer was extracted with EtOAc ( 2 × 20 mL ) and CH2Cl2 ( 2 × 30 mL ) . The combined organic extracts were dried over anhydrous Na2SO4 , filtered , concentrated under reduced pressure . The residue was crystallized from EtOH ( Et2O wash ) to give some solid product . The mother liquor was concentrated under reduced pressure and purified via column chromatography ( SiO2 , 75% EtOAc/heptanes ) to give additional solids . All solids were combined to give tert-butyl ( 3- ( 5- ( difluoromethyl ) pyrazine-2-carboxamido ) bicyclo[1 . 1 . 1] pentan-1-yl ) carbamate ( 38 g , 107 mmol , 97% yield ) . To a solution of tert-butyl ( 3- ( 5- ( difluoromethyl ) pyrazine-2-carboxamido ) bicyclo[1 . 1 . 1]pentan-1-yl ) carbamate ( 38 . 0 g , 107 mmol ) in CH2Cl2 ( 350 mL ) at 0°C was added trifluoroacetic acid ( 132 mL , 1720 mmol ) dropwise over 1 hr . This mixture was allowed to warm to ambient temperature and was stirred for 2 hr then was concentrated under reduced pressure and azeotroped with toluene to give N- ( 3-aminobicyclo[1 . 1 . 1]pentan-1-yl ) −5- ( difluoromethyl ) pyrazine-2-carboxamide , trifluoroacetic acid ( 40 . 0 g , 109 mmol , quantitative yield ) . To a solution of N- ( 3-aminobicyclo[1 . 1 . 1]pentan-1-yl ) −5- ( difluoromethyl ) pyrazine-2-carboxamide , trifluoroacetic acid ( 39 . 4 g , 107 mmol ) and 2- ( 4-chloro-3-fluorophenoxy ) acetic acid ( 24 . 1 g , 118 mmol ) in N , N-dimethylformamide ( 400 mL ) was added triethylamine ( 59 . 7 mL , 428 mmol ) . The mixture was cooled to 0°C then 2- ( 3H-[1 , 2 , 3] triazolo[4 , 5-b]pyridin-3-yl ) −1 , 1 , 3 , 3-tetramethylisouroniumhexafluorophosphate ( V ) ( HATU , 44 . 8 g , 118 mmol ) was added portionwise over 30 min and the mixture was allowed to stir at ambient temperature for 16 hr . The mixture was quenched with saturated , aqueous NH4Cl ( 100 mL ) and water ( 50 mL ) and diluted with EtOAc ( 200 mL ) . The layers were separated and the aqueous layer was extracted with EtOAc ( 2 × 50 mL ) and CH2Cl2 ( 2 × 50 mL ) . The combined organic extracts were dried over anhydrous Na2SO4 , filtered and concentrated under reduced pressure . The solids were crystallized from EtOAc/heptanes to give some product . The mother liquor was concentrated under reduced pressure and purified via column chromatography ( SiO2 , 75% EtOAc/heptanes ) to give additional solids . All solids were combined and recrystallized again using charcoal to remove color ( material in boiling EtOAc and hot filtered ) to give product which was dried under vacuum at 45°C for 2 days to give 2BAct ( 39 g , 88 mmol , 83% yield ) as a white solid . An aqueous suspension of 2BAct was prepared by suspending the drug in 0 . 5% hydroxypropyl methylcellulose ( HPMC; Hypromellose 2910 , 4000 mPa . s; Spectrum Chemical Manufacturing Corp , NJ ) in water . The suspending vehicle was first prepared by adding 5 g of HPMC to 500 mL of miliQ water heated to 60°C . This mixture was allowed to stir until all of HPMC was dispersed . This solution was then transferred to a volumetric flask with two additional rinses of the original container . Sufficient quantity of water was then added to prepare 1 L of vehicle and allowed to stir overnight to obtain a clear suspension . The vehicle was kept refrigerated and allowed to come to room temperature before each use . Fresh vehicle was prepared every month . For preparation of the aqueous suspension of 2BAct , the compound was weighed into an appropriately sized mortar and levigated with a pestle using a small amount of the vehicle . This was then collected into an appropriately sized glass vial , previously marked with a q . s . line . The mortar was rinsed five times , adding each rinse into the glass vial . Additional vehicle was added to the glass vial until q . s . line was reached and entire suspension mixed by vortexing for 10 s . Six- to eight-week-old CD1 male mice were dosed with 2BAct at 1 mg/kg or 30 mg/kg orally at a dosing volume of 10 mL/kg . For dosing , 2BAct was micronized and suspended in 0 . 5% hydroxypropyl methylcellulose ( HPMC ) ( see Microsuspension preparation above ) . Blood was drawn into EDTA charged capillary tubes via the tail vein at the following timepoints: 0 . 25 , 0 . 5 , 1 , 3 , 6 , 9 , 12 and 24 hr ( N = 3 measurements per timepoint , mice bled at each timepoint , and combined in pairs for extraction ) . Blood was centrifuged at 3000 rpm and plasma harvested . Plasma samples and standards were extracted by protein precipitation with acetonitrile containing internal standards . The supernatant was diluted with 0 . 1% formic acid in water before injection into an HPLC-MS/MS system for separation and quantitation . The analytes were separated from matrix components using reverse phase chromatography ( 30 × 2 . 1 mm , 5 µm Fortis Pace C18 ) using gradient elution at a flow rate of 0 . 8 mL/min . The tandem mass spectrometry analysis was carried out on SCIEX triple quadrupole mass spectrometer with an electrospray ionization interface , in positive ion mode . Data acquisition and evaluation were performed using Analyst software ( SCIEX ) . 2BAct was administered orally by providing mice with the compound incorporated in rodent meal ( 2014 , Teklad Global 14% Protein Rodent Maintenance Diet; Envigo , WI ) . For this , the compound was weighed , added to a mortar with small amount of powdered meal , and ground with a pestle until homogenous . This was further mixed with additional powdered meal in HDPE bottles by either geometric mixing with hand agitation or using a Turbula mixer ( Glen Mills Inc . , NJ ) set at 48 rpm for 15 min or contract manufactured at Envigo to achieve a 2BAct concentration of 300 ppm ( 300 µg 2BAct/g of meal ) . Teklad 2014 without added compound was offered as the placebo diet . The Eif2b5R191H/R191H knock-in mutant mouse model was generated in the background strain C57BL/6J as a service by genOway ( Lyon , France ) . Briefly , a targeting vector was designed against the Eif2b5 locus to simultaneously insert: ( 1 ) a Flp-excisable neomycin resistance cassette between exons 2 and 3; ( 2 ) a CGC - > CAC codon substitution in exon 4 ( changing residue Arg191 to His ) ; ( 3 ) loxP sites flanking exons 3 and 7 ( Figure 1—figure supplement 2 ) . Successful homologous recombination in ES cells was verified by PCR and Southern Blotting . Chimeras were generated by blastocyst injection , which were then mated to WT C57BL/6J mice to identify F1 heterozygous Eif2b5+/R191H;FRT-neo ( flox ) progeny . The neomycin resistance cassette was removed by mating of heterozygous mice to Flp deleter mice . The resulting Eif2b5+/R191H ( flox ) mice were used as colony founders . Experiments were performed using homozygous mutant mice and their WT littermates as controls . The Eif2b5R132H/R132H mouse model was generated in a similar manner . Pregnant female C57BL/6J mice ( WT and Eif2b5R191H ( flox ) /R191H ( flox ) ) were sacrificed 13 days after discovery of a post-mating plug . Embryos were dissected out of the uterine horn and separated for individual processing . The head and blood-containing organs of each embryo were removed . The remaining tissue was finely minced with a razor blade and triturated using a pipet . Cells were dissociated with 0 . 25% trypsin-EDTA for 30 min at 37°C . Dissociated cells were washed once in DMEM +10% FBS+1X antibiotic-antimycotic solution , then plated onto 10 cm dishes in fresh medium . Cells were expanded for two passages before freezing for storage . Cells from each embryo were treated as separate , independent lines . Cerebellum lysates were prepared in RIPA buffer ( Sigma #R0278 ) +protease/phosphatase inhibitors ( Pierce #A32959 ) . Tissues were lysed in a Qiagen TissueLyser II for 2 × 2 min intervals at 30 Hz . Lysates were incubated on ice for ten minutes and centrifuged ( 21 , 000 x g , 10 min , 4°C ) to remove cellular debris . Protein concentrations were determined using a Pierce BCA assay ( Thermo #23227 ) and adjusted to 2 mg/mL using RIPA buffer . Lysates were aliquoted , flash-frozen and stored at −80°C . For western blots , samples were run on Mini-PROTEAN TGX 4–20% gradient gels ( Bio-Rad #4561096 ) and transferred using Trans-Blot Turbo Mini-PVDF Transfer packs ( Bio-Rad #1704156 ) on a Trans-Blot Turbo apparatus . Membranes were blocked with 5% milk in TBS-T and incubated overnight with primary antibody in the same blocking buffer at 4°C . After three washes of 15 min each in TBS-T , HRP-conjugated secondary antibodies were applied for 1 hr . Membranes were washed in TBS-T as before . Advansta WesternBright chemiluminescent substrate was applied to the membranes and images were obtained on a Bio-Rad ChemiDoc MP imaging system in signal accumulation mode . The experiment was performed as previously described ( Wong et al . , 2018 ) . Briefly , HEK293T cells expressing an ATF4-luciferase reporter ( Sidrauski et al . , 2013 ) were seeded into 96-well plates and treated with 100 nM thapsigargin for 7 hr to induce ER stress . Cells were co-treated with 2BAct or ISRIB in dose response . Luminescence was measured using ONE-Glo Luciferase assay reagent ( Promega ) and a Molecular Devices SpectraMax i3x plate reader . Data were analyzed in Prism ( GraphPad Software ) . The experiment was performed as previously described ( Wong et al . , 2018 ) . Briefly , Bodipy-FL-GDP-loaded eIF2 was used as a substrate for lysates generated from WT and R191H MEFs . The assay was performed in 384-well plates . In a final assay volume of 10 µL/well , the following conditions were kept constant: 25 nM Bodipy-FL-GDP-loaded eIF2 , 3 nM phospho-eIF2 , 0 . 1 mM GDP , 1 mg/mL BSA , 0 . 1 mg/mL MEF lysate . 2BAct was dispensed from a 1 mM stock . For each run , triplicate measurements were made for each concentration of 2BAct . Reactions were read on a SpectraMax i3x plate reader using the following instrument parameters: plate temperature = 25°C; excitation wavelength = 485 nm ( 15 nm width ) ; emission wavelength = 535 nm ( 25 nm width ) ; read duration = 30 mins at 45 s intervals . Data were analyzed in Prism . GDP release half-lives were calculated by fitting single-exponential decay curves . EC50s were calculated by fitting log ( inhibitor ) vs response curves . Mice were allowed to habituate to our facilities for at least 1 week prior to the start of experiments . Mice were housed on a 12-hr light/dark cycle ( lights on at 6:00 , lights off at 18:00 ) in a temperature- and humidity-controlled environment ( 22 ± 1°C , 60–70% humidity ) . Experimental procedures took place during the illuminated phase of the cycle . Male mice were individually housed due to aggression . Female mice were housed as shipped from the vendor in groups of 1–3 animals/cage . Animals had free access to food and water . AbbVie is committed to the internationally accepted standard of the 3Rs ( Reduction , Refinement , Replacement ) and adhering to the highest standards of animal welfare in the company’s research and development programs . Animal studies were approved by AbbVie’s Institutional Animal Care and Use Committee or Ethics Committee . Animal studies were conducted in an AAALAC-accredited program where veterinary care and oversight was provided to ensure appropriate animal care . Body weights were measured weekly for mice throughout the study . For experiments , mice were pseudo-randomly balanced between treatment groups by date of birth . Analysis of groups was performed in a blinded fashion . Due to the number of animals required for the 2BAct treatment study , mice were binned into age cohorts spanning 5 weeks each . Power analysis was not done prior to beginning the study as experiments were animal-limited . Post-hoc analysis demonstrated that the sample sizes used were sufficient to detect an effect size of 1 . 1 at 90% power and alpha = 0 . 05 ( two-sided t-test ) . A 100 cm long x 2 . 5 cm diameter PVC pipe , suspended 30 cm above a padded landing area , was used as the balance beam . A length of 80 cm was marked on the pipe as the distance the animals were required to cross . The pipe was placed in a dimly lit room , with a spotlight suspended above the starting end to serve as the aversive condition . A darkened enclosure with bedding was placed at the opposite end to promote a more agreeable condition . A video camera was set above the starting end to record foot slips as the animal progressed across the pipe . Animals were habituated to the experimental room for at least one hour prior to testing . To begin an experiment , each subject was placed on the starting end of the beam and allowed a cutoff time of 30 s to cross . Elapsed time was only recorded while the subject actively moved towards the finish line; if it turned around or stopped to groom , the timer was paused . If a fall occurred , the subject was restarted at the beginning of the beam . Each subject was given three attempts to complete the balance beam . If a subject failed all three attempts , it was assigned a time of 30 s and a value of 3 falls . The number of foot slips was quantified from the video recording only if the animal completed the task . The balance beam was cleaned and dried between each subject . Due to the non-normal distribution of the data , they were analyzed using a Mann-Whitney ( non-parametric ) test in Prism . Reported p-values are adjusted for multiple comparisons by the Bonferroni method . Animals were habituated to the experimental room for 30–60 mins prior to testing . A grid screen measuring 20 cm x 25 cm with a mesh density of 9 squares/cm2 was elevated 45 cm above a cage with bedding . Each subject was placed head oriented downward in the middle of the grid screen . When it was determined that the subject had proper grip on the screen , it was inverted 180° . The hang time ( duration a subject held on to the screen without falling ) was recorded , up to a cutoff time of 60 s . Any subject that was able to climb onto the top of the screen was assigned a time of 60 s . The grid was cleaned and dried between trial days . Due to the non-normal distribution of the data , they were analyzed using a Mann-Whitney ( non-parametric ) test in Prism . Reported p-values are adjusted for multiple comparisons by the Bonferroni method . Mice were deeply anesthetized with CO2 and rapidly fixed by transcardiac perfusion with normal saline followed by 10% formalin . Brains and spinal cords were excised and allowed to post-fix in formalin for an additional 2 hr . Samples were processed for paraffin embedding , sectioned at 6 µm , and mounted on adhesive-coated slides . Immunohistochemical staining was carried out using the antibodies described in the Key Resources Table and species-appropriate polymer detection systems ( ImmPRESS HRP , Vector Laboratories ) , and developed with 3 , 3'-Diaminobenzidine ( DAB ) followed by hematoxylin counterstaining . For GFAP , DAB development was enhanced with nickel . Myelin staining by Luxol Fast Blue was carried out by the Klüver-Barrera method ( Kluver and Barrera , 1953 ) . Sections were dehydrated through successive ethanol solutions , cleared in xylene , and coverslipped using xylene-based mounting media . To examine histopathological endpoints in the brain , sections from two coronal levels of the corpus callosum were examined beginning at approximately Bregma + 0 . 7 and+1 . 0 mm . For the spinal cord , coronal sections from cervical and thoracic levels were examined . Image capture and analysis was achieved using either a 3DHISTECH Pannoramic 250 Digital Slide Scanner ( 20X , Thermo Fisher Scientific ) and HALO image analysis software ( version 2 . 1 . 1637 . 26 , Indica Labs ) , or a BX-51 microscope fitted with a DP80 camera ( 10X ) and cellSens image analysis software ( Olympus ) . The same parameters for microscopy and image analysis were uniformly applied to all images for each endpoint . For the spinal cord , the mean area fraction from a single section from both the cervical and thoracic levels served as the value for each subject to normalize to the different size of the anatomical levels . For the corpus callosum , the mean area of positive staining from a single section at each coronal level served as the value for each subject . Data were analyzed by one-way ( for single time point experiments ) or two-way ANOVA ( for multiple time point experiments ) . Reported p-values are adjusted for multiple comparisons by the Holm-Sidak method . Total RNA was isolated from ~20 mg pieces of each frozen tissue , with a minimum of one extraction performed from each sample . All steps were performed on dry ice prior to the addition of 350 µL of RTL Buffer + 40 µM DTT . Samples were homogenized at 4°C using a Qiagen TissueLyser II ( Retsch , Castleford , UK ) for 2 × 2 min intervals at 30 Hz with the addition of one 5 mm stainless steel bead ( Qiagen catalog # 69965 ) . This was followed by incubation in a Vortemp 56 ( Labnet International , Edison , NJ ) at 65°C with shaking ( 300 rpm ) . The samples were then centrifuged at 15 , 000 x g for 5 min . RNA extraction was performed using the RNeasy Mini kit ( Qiagen , Germany ) according to manufacturer’s instructions , eluting in 30 µL nuclease-free H2O . RNA concentration and purity were determined spectrophotometrically . To measure multiplexed transcript expression levels , we designed a custom QuantiGene Plex Panel ( Flagella et al . , 2006 ) ( Thermo Fisher Scientific , Waltham , MA ) . Either purified RNA or crude lysates ( ~15–20 mg of frozen tissue , homogenized using the QuantiGene 2 . 0 Sample Processing Kit ) were used since comparable results were achieved in a head-to-head comparison with a set of samples ( data not shown ) . RNA or tissue homogenates were then subjected to the QuantiGene assay following manufacturer’s instructions . Briefly , this involved: ( 1 ) capturing target RNAs to corresponding genes on specified beads through an overnight hybridization; ( 2 ) a second hybridization to capture the biotinylated label probes for signal amplification; and ( 3 ) a final hybridization to capture the SAPE reagent , which emits a fluorescent signal from each bead set . Signal intensities for each bead set , proportional to the number of captured target RNA molecules , were read on a Luminex Flexmap 3D ( Luminex , Northbrook , IL ) . To control for differences in total RNA input between reactions , background-subtracted output data were normalized to the mean of two housekeeping genes , RPL13A and RPL19 . The final data are presented as fold-change for each gene relative to the mean of the corresponding WT samples . Data were analyzed in Prism . Student’s t-test was performed for the panel of genes comparing R191H to WT for each time point and brain region . Reported p-values are adjusted for multiple comparisons by the Holm-Sidak method . RNA-seq libraries were prepared using purified RNA isolated as described above . RNA quality and concentration were assayed using a Fragment Analyzer instrument . RNA-seq libraries were prepared using the TruSeq Stranded Total RNA kit paired with the Ribo-Zero rRNA removal kit ( Illumina , San Diego , CA ) . Libraries were sequenced on an Illumina HiSeq 4000 instrument . For the experiment comparing different ages , N = 3 males/genotype/time point . For the 4-week 2BAct treatment experiment , N = 3 females/condition . RNA-seq library mapping and estimation of expression levels were computed as follows . Reads were mapped with STAR aligner ( Dobin et al . , 2013 ) , version 2 . 5 . 3a , to the mm10 reference mouse genome and the Gencode vM12 primary assembly annotation ( Mudge and Harrow , 2015 ) , to which non-redundant UCSC transcripts were added , using the two-round read-mapping approach . This means that following a first read-mapping round of each library , the splice-junction coordinates reported by STAR , across all libraries , are fed as input to the second round of read mapping . The parameters used in both read-mapping rounds are: outSAMprimaryFlag =‘AllBestScore’ , outFilterMultimapNmax =‘10’ , outFilterMismatchNoverLmax =‘0 . 05’ , outFilterIntronMotifs =‘RemoveNoncanonical’ . Following read-mapping , transcript and gene expression levels were estimated using MMSEQ ( Turro et al . , 2011 ) . Transcripts and genes which were minimally distinguishable according to the read data were collapsed using the mmcollapse utility of MMDIFF ( Turro et al . , 2014 ) , and the Fragments Per Kilobase Million ( FPKM ) expression units were converted to Transcript Per Million ( TPM ) units . In order to test for differential expression between the R191H and WT samples , for each of the three time points , we used MMDIFF with a design matrix corresponding to two groups without extraneous variables . In order to test whether R191H relative to the WT fold-change varied between each of the two consecutive time points ( i . e . , 5 months vs . 2 months , and 7 months vs . 5 months ) , we used MMDIFF with the following design matrix testing whether the fold-change between groups A and B ( e . g: R191H and WT at 5 months ) is different from the fold-change between groups C and D ( e . g: R191H and WT at 2 months ) : # M 0 , 0 , 0 , 0 , 0 , 0; # C 0 0 , 0 0 , 0 0 , 0 1 , 0 1 , 0 1; # P0 1; # P1 1 0 0 , 0 1 0 , 0 0 1 . Genes were ranked in descending order according to the Bayes factor and the posterior probability , reported by MMDIFF . Singular value decomposition analyses were carried out using R’s SVD function . The right singular vectors of the decomposition , referred to as eigengenes , are used to capture the canonical gene expression patters in the data set ( Alter et al . , 2000 ) . We applied a threshold on the absolute gene loading values defined as two-fold the mean of all absolute values . For GO-term enrichment , analyses were carried out using the online tool at geneontology . org ( Ashburner et al . , 2000; The Gene Ontology Consortium , 2017 ) . Samples were clustered using the heatmap . 2 function from the gplots package ( R Core Team , 2018; Warnes et al . , 2005 ) . Forebrain and cerebellum were dissected from 2-month-old female mice . Tissues were dissociated by incubation in 2 mg/mL papain solution ( BrainBits , Springfield , IL ) for 30 min at 37°C with agitation , followed by trituration . To remove large debris , suspensions were successively passed through 100 μm and 40 μm strainers . Cells were pelleted by centrifugation at 280 x g , and pellets were subjected to the Miltenyi Debris Removal protocol with Debris Removal solution according to manufacturer’s instructions ( Miltenyi Biotec , Bergisch Gladbach , Germany ) . Following this , cells were re-filtered through a 40 μm Flowmi cell strainer ( Belart , Wayne , NJ ) . Single cell RNA-seq libraries were created using the Chromium Single Cell 3′ Library and Gel Bead Kit v2 and associated consumables ( 10X Genomics , Pleasanton , CA ) . ~7000 cells per sample were loaded onto the Chromium microfluidic chip for an expected recovery of 4000 cells per sample . Manufacturer’s instructions were followed for library preparation , and cDNAs were amplified for 12 cycles . Samples were sequenced on an Illumina HiSeq 4000 . scRNA-seq fastq files were demultiplexed to their respective barcodes using the 10X Genomics Cell Ranger mkfastq utility . Unique Molecular Identifier ( UMI ) counts were generated for each barcode using the Cell Ranger count utility , with the mm10 reference mouse genome used for mapping reads . For each sample , barcodes that were not likely to represent captured cells were filtered out by detecting the first local minimum above two in a distribution of log10 ( UMIs ) . To identify cell clusters , the four samples were concatenated and we defined genes with variable expression dispersion using Seurat ( Butler et al . , 2018 ) , which resulted in 2622 genes . PCA was performed on these genes using the rsvd R package to reduce the dimensionality of the data , retaining the 50 PCs explaining the highest amount of variation . We then used Seurat’s methodology to build a Shared Nearest Neighbor ( SNN ) graph of these cell-embedding data , first generating a K-Nearest Neighbor ( KNN ) graph using K = min ( 750 , #cells-1 ) and a Jaccard distance cutoff of 1/15 . The SNN graph was then used as input to the Louvain algorithm , implemented in the ModularityOptimizer software ( Waltman and van Eck , 2013 ) . Since this implementation uses a resolution parameter that strongly affects the number of clusters , we searched the 0 . 05–1 . 225 range of this parameter , using the mean unifiability isolability clustering metric as our maximization parameter . This process was initially done on all cells in our data , and subsequently repeated for each cluster individually , in an iterative manner where convergence was defined as not being able to break down a cluster into sub-clusters . This resulted in 10 and 13 clusters in the cerebellum and forebrain , respectively . In order to obtain gene markers for each cluster , we ran a differential expression test implemented in Seurat ( using a likelihood ratio test between a model that assumes that a gene’s expression values of two compared clusters values were sampled from two distributions , versus a null model which assumes they were sampled from a single distribution ) . A marker gene of a given cluster was defined as a gene which was found to be significantly overexpressed ( adjusted p-value<0 . 05 ) in that cluster compared to all other clusters . In order to assign cell identities to our identified clusters , we utilized published bulk RNA-seq gene expression data obtained by cell sorting of major brain cell types ( Koirala and Corfas , 2010; Zhang et al . , 2014 ) . For each gene that is a cell-type marker in the list compiled from the published data and is expressed in a cluster , we computed its mean expression level across all cells of the cluster and multiplied it by the fraction of cells it was captured in . We then summed these gene scores across all markers of the cell type that are expressed in the cluster of interest and divide that sum by the number of markers of that cell type to account for possible ascertainment biases ( as some cell types may have many more markers than others ) . Finally , we scaled the cell-type scores of each cluster to sum up to one and hence reflect probabilities . Whenever a probability was higher than 0 . 5 , we assigned the identity of that cluster to be of the high probability cell type . In order to detect whether ISR activation is cell-type-specific in the WT samples , for each gene in each cluster we performed a differential expression analysis between that cluster and all other clusters as described above . Since the clustering analysis using both WT and R191H samples ( described above ) found that transcriptionally defined clusters are not influenced by genotype , we used these cluster assignments for the WT , as independent clustering of the WT data alone was not powered enough to obtain the same clustering as when performed on both genotypes . Subsequently , for each cluster we performed a Gene Set Enrichment Analysis ( GSEA ) , using the fgsea R package , with our ISR gene lists – one derived from the CLIC analysis ( Li et al . , 2017 ) and a manually curated list of ATF4 targets ( Supplementary file 1D ) , as the gene sets . In order to detect whether ISR activation is cell type-specific between WT and R191H , for each cluster in the combined WT and R191H data , for each gene in each cluster we performed a differential expression analysis contrasting the cells that correspond to the two genotypes . We subsequently performed the ISR GSEA for each cluster . Frozen brain samples were lysed with 0 . 9 mL 50 mM HEPES , 75 mM NaCl , 3% SDS , pH 8 . 5 + protease/phosphatase inhibitors and homogenized using a Qiagen TissueLyser II for 8 cycles at 20 Hz . Samples were then sonicated for 5 min . Lysates were centrifuged ( 16 , 000 x g , 5 min ) to remove cellular debris . Proteins were reduced with 5 mM dithiothreitol ( 56°C , 30 min ) and alkylated with 15 mM iodoacetamide ( room temperature , 30 min ) . Excess iodoacetamide was quenched with 5 mM dithiothreitol ( room temperature , 30 min ) . Proteins were precipitated by sequential addition of 4 volumes methanol , 1 vol chloroform and three volumes water , with vigorous vortexing in between . Protein pellets were washed with methanol , air-dried , and resuspended in 50 mM HEPES , 8 M urea , pH 8 . 5 . The samples were then diluted to 4 M urea and digested with Lys-C protease ( 25°C , 15 hr; WAKO Chemicals , Richmond , VA ) . The urea concentration was then diluted to 1 M and the samples were digested with trypsin ( 37°C , 6 hr; Promega , Madison , WI ) . After protein digestion , samples were acidified with 10% trifluoroacetic acid and desalted using C18 solid-phase extraction columns ( SepPak ) . Samples were eluted with 40% acetonitrile/0 . 5% acetic acid followed by 80% acetonitrile/0 . 5% acetic acid , and dried overnight under vacuum at 30°C . Peptide concentrations were measured using a Pierce BCA assay , split into 50 µg aliquots , and dried under vacuum . Dried peptides were resuspended in 50 µL 200 mM HEPES/30% anhydrous acetonitrile . 5 mg TMT reagents ( Thermo Fisher Scientific ) were dissolved in 250 µL anhydrous acetonitrile , of which 10 µL was added to peptides . TMT reagent 131 was reserved for the ‘bridge’ sample and the other TMT reagents ( 126 , 127 c , 127 n , 128 c , 128 n , 129 c , 129 n , 130 c and 130 n ) were used to label the individual samples . Following incubation at room temperature for 1 hr , the reaction was quenched with 200 mM HEPES/5% hydroxylamine to a final concentration of 0 . 3% ( v/v ) . TMT-labeled amples were acidified with 50 µL 1% trifluoroacetic acid and pooled into 11-plex TMT samples at equal ratios , desalted with SepPak , and dried under vacuum . Pooled TMT-labeled peptides were fractionated using high pH RP-HPLC . Samples were resuspended in 5% formic acid/5% acetonitrile and fractionated over a ZORBAX extended C18 column ( Agilent , 4 . 6 mm ID x 250 mm length , 5 µm particles ) . Peptides were separated on a 75 min linear gradient from 5% to 35% acetonitrile in 10 mM ammonium bicarbonate at a flow rate of 0 . 5 mL/min on an Agilent 1260 Infinity pump ( Agilent Technologies , Waldbronn , Germany ) . The samples were fractionated into a total of 96 fractions , and then consolidated into 12 as described previously ( Edwards and Haas , 2016 ) . Samples were dried under vacuum and reconstituted in 5% formic acid/4% acetonitrile for LC-MS/MS processing . Peptides were analyzed on an Orbitrap Fusion Lumos mass spectrometer ( Thermo Fisher Scientific ) coupled to an Easy-nLC ( Thermo Fisher Scientific ) . Peptides were separated on a microcapillary column ( 100 µm ID x 25 cm length , filled in-house with Maccel C18 AQ resin , 1 . 8 µm , 120 A; Sepax Technologies ) . The total LC-MS run length for each sample was 180 min comprising a 165 min gradient from 6–30% acetonitrile in 0 . 125% formic acid . The flow rate was 300 nL/min and the column was heated to 60°C . Data-dependent acquisition mode was used for mass spectrometry data collection . A high resolution MS1 scan in the Orbitrap ( m/z range = 500–1200 , resolution = 60 , 000; AGC = 5×105; max injection time = 100 ms; RF for S-lens = 30 ) was collected , from which the top 10 precursors were selected for MS2 analysis followed by MS3 analysis . For MS2 spectra , ions were isolated using a 0 . 5 m/z window using the mass filter . The MS2 scan was performed in the quadrupole ion trap ( CID , AGC = 1×104 , normalized collision energy = 30% , max injection time = 35 ms ) and the MS3 scan was analyzed in the Orbitrap ( HCD , resolution = 60 , 000; max AGC = 5×104; max injection time = 250 ms; normalized collision energy = 50 ) . For TMT reporter ion quantification , up to six fragment ions from each MS2 spectrum were selected for MS3 analysis using synchronous precursor selection . Mass spectrometry data were processed using an in-house software pipeline as previously described ( Huttlin et al . , 2010 ) . Briefly , raw files were converted to mzXML files and searched against a composite mouse Uniprot database ( downloaded on 9 May 2017 ) containing sequences in forward and reverse orientations using the Sequest algorithm . Database searching matched MS/MS spectra with fully tryptic peptides from this composite dataset with a precursor ion tolerance of 20 ppm and a product ion tolerance of 0 . 6 Da . Carbamidomethylation of cysteine residues ( +57 . 02146 Da ) and TMT tags of peptide N-termini ( +229 . 162932 Da ) were set as static modifications . Oxidation of methionines ( +15 . 99492 Da ) was set as a variable modification . Linear discriminant analysis was used to filter peptide spectral matches to a 1% FDR ( false discovery rate ) . Non-unique peptides that matched to multiple proteins were assigned to proteins that contained the largest number of matched redundant peptides sequences . Quantification of TMT reporter ion intensities was performed by extracting the most intense ion within a 0 . 003 m/z window at the predicted m/z value for each reporter ion . TMT spectra were used for quantification when the sum of the signal-to-noise for all the reporter ions was greater than 200 and the isolation specificity was greater than 0 . 75 ( Ting et al . , 2011 ) . Peptide-level data from each 11-plex was column-normalized to have equivalent geometric means and protein level estimates were obtained by fitting a previously described compositional Bayesian model ( O'Brien et al . , 2018 ) . For each protein , log2 fold-changes are estimated as the posterior means from the model . Variances and 95% credible intervals from the posterior distributions are also reported .	Cells must be able to respond to their changing environment in order to survive . When cells encounter particularly unfavorable conditions , they often react by activating a so-called ‘stress’ response . A group of proteins collectively known as eIF2B helps to regulate this response . In a severe neurological condition called Vanishing White Matter ( VWM ) , the genes that produce the eIF2B proteins contain mutations that make eIF2B less active . As a result , certain cells in people with VWM are always stressed . Six years ago , researchers discovered a molecule that boosts the activity of eIF2B . In 2018 , they found that it also works on various mutant forms of eIF2B found in VWM . The molecule had so far only been tested in biochemical laboratory experiments . Now , Wong et al . – including some of the researchers involved in the 2018 study – have tested whether an improved version of the molecule treats VWM in mice . The trial treatment successfully halted all signs of the disease in the mice . The molecule blunted the persistent stress response of the cells in the brain and spinal cord , primarily in a cell type that is severely affected by the human form of VWM . Cells in other parts of the body were spared . Overall , the results of the experiments suggest that an eIF2B activator may prove to be an effective treatment for VWM in humans . It could similarly be used to treat other conditions that activate this abnormal cell stress response . The molecule Wong et al . used is not suitable for use in humans , so work is continuing to find a suitable variant .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "neuroscience" ]	2019	eIF2B activator prevents neurological defects caused by a chronic integrated stress response
Natural sounds can be characterised by their spectral content and temporal modulation , but how the brain is organized to analyse these two critical sound dimensions remains uncertain . Using functional magnetic resonance imaging , we demonstrate a topographical representation of amplitude modulation rate in the auditory cortex of awake macaques . The representation of this temporal dimension is organized in approximately concentric bands of equal rates across the superior temporal plane in both hemispheres , progressing from high rates in the posterior core to low rates in the anterior core and lateral belt cortex . In A1 the resulting gradient of modulation rate runs approximately perpendicular to the axis of the tonotopic gradient , suggesting an orthogonal organisation of spectral and temporal sound dimensions . In auditory belt areas this relationship is more complex . The data suggest a continuous representation of modulation rate across several physiological areas , in contradistinction to a separate representation of frequency within each area . Frequency structure ( spectral composition ) and temporal modulation rate are fundamental dimensions of natural sounds . The topographical representation of frequency ( tonotopy ) is a well-established organisational principle of the auditory system . In mammals , tonotopy is established in the receptor organ , the cochlea , and maintained as a systematic spatial separation of different frequencies in different areas of the ascending auditory pathway , and in the auditory cortex . While temporal modulation is recognised as an essential perceptual component of communication sounds such as human speech and animal vocalisations ( Rosen , 1992; Drullman et al . , 1994; Shannon et al . , 1995; Wang , 2000; Chi et al . , 2005; Elliott and Theunissen , 2009 ) , its representation in the auditory system is poorly understood . In contrast to sound frequency , amplitude modulation rate is not spatially organised in the cochlea but represented in the temporal dynamics of neuronal firing patterns . However , a considerable proportion of neurons in the auditory brainstem and cortex show tuning to amplitude modulation rates ( reviewed in Joris et al . , 2004 ) . It has been proposed that temporal information in sound is extracted by amplitude modulation filter banks ( Dau et al . , 1997a , 1997b ) that are physiologically instantiated in the midbrain ( Rees and Langner , 2005 ) . Studies in rodents ( Langner et al . , 2002 ) , cats ( Schreiner and Langner , 1988 ) and primates ( Baumann et al . , 2011 ) have shown that at the stage of the inferior colliculus amplitude modulation rate and frequency of sound are represented in approximately orthogonal topographical maps . Whether the spatial organisation of amplitude modulation rate is preserved in the auditory cortex remains is debated . Data from gerbils ( Schulze et al . , 2002 ) and cats ( Langner et al . , 2009 ) suggest a topographical map for temporal modulation rates in the auditory cortex . In cats , orthogonal gradients for modulation rate and frequency have been shown , similar to those in the inferior colliculus . However , it is not clear how such an organisation might be preserved across the multiple auditory fields of primate cortex where different fields show different orientations of the tonotopic gradient . While earlier fMRI studies in humans ( Giraud et al . , 2000; Schonwiesner and Zatorre , 2009; Overath et al . , 2012 ) reported robust responses to a range of amplitude-modulated sounds , but no systematic organisation of rate , two more recent studies suggested an orthogonal relationship of frequency and rate in areas homologous to non-human primate auditory core ( Herdener et al . , 2013 ) and beyond ( Barton et al . , 2012 ) . Electrophysiology studies in non-human primates have shown tuning of individual neurons to different modulation rates and suggest a tendency for neurons in primary fields to prefer faster rates than neurons in field higher up the hierarchy ( Bieser and Muller-Preuss , 1996; Liang et al . , 2002 ) : see also ( Joris et al . , 2004 ) . However , no clear topographical organisation of modulation rate across different auditory fields has been demonstrated in non-human primates . In the current study we mapped the blood oxygen level dependent ( BOLD ) response to a wide range of amplitude modulation rates from 0 . 5–512 Hz applied to a broad-band noise carrier in the auditory cortex of macaque monkeys using functional magnetic resonance imaging ( fMRI ) . This range of modulation rates covers preferred rates for cortical neurons ( Joris et al . , 2004 ) . We investigated whether the preference for specific amplitude modulation rates in neuronal ensembles is systematically represented in the auditory cortex and , if so , how such an organisation is arranged relative to the tonotopic gradients across auditory fields . The data reveal a topographic organisation of amplitude modulation rate in the macaque auditory cortex arranged in concentric iso-rate bands that are mirror-symmetric across both hemispheres with a preference for the highest rates in the postero-medial auditory cortex at the medial border of A1 and for the lowest rates in lateral and anterior fields . This organisation results in a modulation rate gradient running approximately orthogonal to the tonotopic gradients in the auditory core fields , A1 and R . In a first experiment we recorded the BOLD response to amplitude-modulated broad-band noise at six different rates ( 0 . 5 , 2 , 8 , 32 , 128 , 512 Hz , see also [Baumann et al . , 2011] ) across the auditory cortex of three monkeys . We generated two different maps to reveal the spatial organisation of modulation rate by projecting the data of the acquired volumes onto the cortical surface derived from the anatomical scans . In a first map ( contrast map ) we contrasted the response strength of the lower rate bands vs the higher rate bands to reveal the gradual change of preference for higher to lower modulation rates across the auditory cortex ( Figure 1 , top panels ) . The data corresponding to the two highest rates ( 512 Hz , 128 Hz ) and lowest rates ( 2 Hz , 0 . 5 Hz ) were combined before contrasting , while the intermediate rates were ignored ( see also Baumann et al . , 2011 ) . In a second map ( best-rate map ) , for each position in the auditory cortex , we represented the rate band that showed the strongest response . For the best-rate maps , all six individual rates were mapped ( Figure 1 , bottom panels ) . The contrast maps reveal a topography for different amplitude modulation rates with preferences for high rates consistently clustered in the postero-medial auditory cortex with the maxima at the medial border of the primary field A1 in both hemispheres of all tested animals . Preferences for low rates were located lateral , anterior and to some degree posterior to the high-rate clusters . In the areas with the maximal preference for low or high rates , the contrast between low and high rates was statistically significant ( p < 0 . 05 ) in all animals and hemispheres , corrected for multiple comparisons with family-wise error ( FWE ) correction over the recorded volume . The best-modulation-rate maps for the six tested rates ( Figure 1 , bottom ) confirmed the systematic organisation of the response pattern with a topographic representation of rate arranged in approximately concentric frequency bands starting with high rates in the postero-medial auditory cortex ( at least a few voxels showed a best modulation rate of 128 Hz postero-medially in 5 of 6 hemispheres ) and progressing anterior , lateral and in some cases posterior to lower rates ( see also the schemata in Figure 4 ) . The highest rate ( 512 Hz ) was hardly represented in the best-rate maps . 10 . 7554/eLife . 03256 . 003Figure 1 . Representation of amplitude modulation rates in the auditory cortex . Top panels: map showing contrast of low vs high rates ( rate contrast map ) projected on rendered surfaces of the superior temporal planes in three animals ( M1-3 ) . Green arrows indicate mean gradient direction of the contrast in auditory field A1 ( gradient directions derived from 2D regression; see also Table 1 ) . Green circles indicate the position of the centre of mass of A1 . Bottom panels: map of preferred response to different rates ( Best-rate map ) . A; anterior , P; posterior , L; left , R; right . DOI: http://dx . doi . org/10 . 7554/eLife . 03256 . 003 In a second experiment , we recorded the BOLD response to bandpass noise in three different frequency bands ( 0 . 5–1 kHz , 2–4 kHz , 8–16 kHz ) in the same animals used in the first experiment . Based on these data , we generated contrast maps ( Figure 2 , top panels ) and best frequency maps ( Figure 2 , bottom panels ) for each animal similar to experiment 1 . These maps confirm the well-established tonotopic pattern with multiple reversals of the frequency gradient that serve as a basis for the delineation of auditory fields in the primate auditory cortex ( Morel et al . , 1993; Kosaki et al . , 1997; Petkov et al . , 2006; Bendor and Wang , 2008; Baumann et al . , 2010 ) . In a further experiment , we confirmed these findings using a ‘phase-encoded’ design ( Sereno et al . , 1995; Joly et al . , 2014 ) in animal M1 and M2 ( Figure 3 ) . Based on these frequency maps , we identified auditory fields according to Hackett ( 2011 ) , using procedures described in Petkov et al . ( 2006 ) and Baumann et al . ( 2010 ) , adapted by the use of T1/T2 weighted MRI data to inform on the location of the core/belt border ( Joly et al . , 2014 ) . 10 . 7554/eLife . 03256 . 004Figure 2 . Representation of spectral frequency in the auditory cortex . Top panels: map of contrast of low vs high frequency band ( frequency-contrast map ) projected on rendered surfaces of the superior temporal planes in three animals ( M1-3 ) . Green arrows indicate mean gradient direction of the contrast in auditory field A1 . Green circles indicate the position of the centre of mass of A1 . Bottom panels: map of preferred response to different frequency bands ( Best-frequency map ) . A; anterior , P; posterior , L; left , R; right . DOI: http://dx . doi . org/10 . 7554/eLife . 03256 . 00410 . 7554/eLife . 03256 . 005Figure 3 . Representation of spectral frequency in the auditory cortex derived from ‘phase-encoded’ experiment . Top panels: map of preferred response to nine different frequencies between 0 . 5–8 KHz projected on rendered surfaces of the superior temporal planes in two animals ( M1 , M2 ) . Approximate border between A1 and R is marked by grey bars . A; anterior , P; posterior , L; left , R; right . DOI: http://dx . doi . org/10 . 7554/eLife . 03256 . 005 Here , we describe organisational features of these maps ( summarised in Figure 4 , right ) which are well in line with previous studies ( reviewed in Baumann et al . , 2013 ) . The anterior-posterior low to high gradient that defines the auditory core field A1 is particularly clear in the contrast maps ( Figure 2 , top panels ) . In line with previous studies in macaques , this gradient has generally a lateral-to-medial component in addition to the main anterior-posterior direction ( Morel et al . , 1993; Kosaki et al . , 1997; Baumann et al . , 2010 ) . Similar to the situation in humans , the low frequency area that defines the border between fields A1 and R is typically wider on the lateral side compared to the medial side . The anterior high frequency area , which forms the border between the auditory fields R and RT , is typically located on the medial side of the superior temporal plane , in the depth of the circular sulcus ( see also Figure 4 , Baumann et al . , 2013 ) . The resulting anterio-medial direction of the main frequency gradient in R forms , in combination with gradient direction in A1 , an inward inflection in the gradient axis across A1 and R , which has previously been reported in Morel et al . ( 1993 ) and Kosaki et al . ( 1997 ) in macaques , Bendor and Wang ( 2008 ) in marmosets and described in ( Jones , 2003; Hackett , 2011; Baumann et al . , 2013 ) . This directional change of the frequency gradient axis across the core fields , which is also obvious from the different mean angles of the gradients in A1 and R ( Table 1 ) , is of particular relevance for the relationship of temporal and spectral gradients in these fields ( see Figure 4 and ‘Discussion’ ) . 10 . 7554/eLife . 03256 . 006Figure 4 . Schematic representation of amplitude-modulation-rate organisation in macaque auditory cortex . Model of modulation rate organisation in context of functional-field borders and frequency reversals ( left side ) . Schematic organisation of tonotopy with indication of main gradients for tonotopy and modulation rate in selected functional fields ( right side ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03256 . 00610 . 7554/eLife . 03256 . 007Table 1 . Directions and relative orientations of amplitude modulation rate and frequency gradients in selected auditory fieldsDOI: http://dx . doi . org/10 . 7554/eLife . 03256 . 007AnimalHemis-phereFieldRel . angle ( α ) ( degrees ) Frequency ( degrees ) R2p-valueRate ( degrees ) R2p-valueN ( vertices ) M1LA11371620 . 796<1E-16250 . 636<1E-1684R731550 . 546<1E-16820 . 696<1E-16101M2L1201750 . 927<1E-16560 . 884<1E-16154R1331600 . 803<1E-16660 . 750<1E-16103M3L1211640 . 824<1E-16430 . 699<1E-16156R1281750 . 722<1E-16470 . 675<1E-1687Average118 . 7165 . 20 . 7753 . 20 . 72114 . 2Std dev23 . 38 . 20 . 1319 . 70 . 0932 . 5M1LR81380 . 1061 . 40E-021200 . 782<1E-1688R140230 . 3912 . 89E-111170 . 600<1E-16107M2L51290 . 697<1E-16800 . 4202 . 00E-09102R103440 . 4412 . 50E-101470 . 660<1E-1693M3L180810 . 4761 . 40E-101000 . 2554 . 20E-0687R14890 . 6114 . 20E-151570 . 691<1E-1673Average117 . 237 . 30 . 45120 . 20 . 5791 . 7Std dev47 . 624 . 60 . 2028 . 70 . 2012 . 1M1LCL161130 . 5772 . 80E-041480 . 5871 . 70E-0633R1131090 . 5163 . 30E-071380 . 2641 . 90E-0344M2L163420 . 5633 . 30E-071550 . 5822 . 00E-1240R102140 . 5525 . 70E-101160 . 559<1E-1658M3L168270 . 7953 . 30E-161650 . 7512 . 60E-1448R15820 . 7064 . 20E-111560 . 7684 . 20E-1342Average144 . 234 . 50 . 62146 . 30 . 5944 . 2Std dev28 . 839 . 00 . 1117 . 40 . 188 . 4Main gradient directions ( relative to anterior-posterior axis ) and the resulting relative angle ( α ) between the orientations of the amplitude modulation rate ( Rate ) and spectral frequency ( Frequency ) gradients in auditory fields A1 , R and CL are listed for two hemispheres ( L , R ) in three animals ( M1-3 ) . Additionally , R2 values , p-values and number of data points ( n ) from the respective 2D regression analysis are included . The systematic , topographic representation of amplitude modulation rate we demonstrate extends over multiple fields to form a wider concentric organisation across the entire auditory cortex in both hemispheres . If this organization is overlaid with the spectral pattern derived from experiment 2 , we notice that in the auditory core areas ( A1 , R ) the gradients for modulation rate and frequency lie approximately orthogonal to one another ( illustrated in schemata of Figure 4 ) . This arrangement is most obvious in A1 in individual animals , but the relationship generally holds for R and the adjacent lateral belt fields as well . This occurs as a consequence of the change in the direction of the tonotopic axis between A1 and R such that the tonotopic gradient follows approximately the iso- amplitude-modulation rate lines . In other words , the tonotopic axis runs anteriorly and medially in area R whilst the rate gradient runs anteriorly and laterally to preserve the orthogonal relationship . In this arrangement , the frequency reversals form spokes in the concentric organisation of the amplitude modulation rate ( Figures 1 , 2 , 4 left ) . However , rostral of R and caudal of A1 the orthogonality of spectral and temporal features breaks down and the gradients become more collinear and less obvious ( Figures 1 , 2 , 4 right ) . In order to quantify these observations we calculated gradient directions and relative orientations based on the contrast maps for the topographies of modulation rate and frequency in the auditory core fields A1 and R using a two dimensional regression analysis as described in Baumann et al . ( 2011 ) . We also calculated the relative gradients for the caudio-lateral field CL , representative for extra-core areas where the orthogonal direction breaks down . The auditory fields have been delineated based on the tonotopy reversals as described in Baumann et al . ( 2010 ) and Petkov et al . ( 2006 ) . The results are summarised in Table 1 and gradient directions in A1 are indicated with green arrows in Figures 1 , 2 overlaid on the contrast maps with a circle marking the centre of mass of field A1 . The gradient directions in the individual hemispheres generally follow the schemata in Figure 4 . All the calculated fields show a clear gradient . In A1 , the gradient directions for rate and frequency clearly cross each other but deviate somewhat from perfect orthogonality with an average angle of 118 . 7 ± 22 . 3° . The calculated directions in field R are more variable , but also show an average relative angle for rate and frequency gradients of about 120° ( 117 . 2 ± 47 . 6° ) . In contrast , relative angles in the postero-lateral field CL are much closer to anti-parallel with 4 of 6 hemispheres showing a relative angle around 160° and higher . Due to their sparse , non-parametric nature , the best rate/frequency maps are less suited for gradient analysis . Furthermore , in some belt fields that feature a single best frequency , no gradient can be specified . However , for comparison , we provided a respective gradient analysis of the best rate/frequency for the fields with a defined gradient in the supplementary methods ( Supplementary file 1 ) and the gradients for A1 are highlighted on the respective maps in Figures 1 , 2 . In most cases , the calculated gradient directions in core areas differ little from the analysis based on the contrast maps . The resulting relative angles are also similar with means closer to 110° in core areas ( 108 . 5 ± 42 . 1° for A1 , 112 . 0 ± 55 . 1° for R ) . However , the correlation values ( r2 ) , p values and the variance across animals and hemispheres are clearly worse , which can be attributed to the sparse nature of the data . This is particularly true for the posterior belt field CL . Previous studies that investigated the representation of amplitude-modulation rate ( Bieser and Muller-Preuss , 1996; Giraud et al . , 2000; Bendor and Wang , 2008; Schonwiesner and Zatorre , 2009; Overath et al . , 2012 ) , have not reported a topographical organisation of this temporal dimension in the auditory cortex . We can only speculate why such an organisation has not been reported in these previous studies . Electrophysiology studies in other primate species have highlighted a tendency for differential responses across different auditory fields with more primary areas preferring faster rates or showing shorter latencies and areas higher up the hierarchy preferring lower rates or showing longer latencies ( Bieser and Muller-Preuss , 1996; Bendor and Wang , 2008 ) . These results have been interpreted in terms of a posterior-to-anterior high-to-low rate gradient ( Bendor and Wang , 2008 ) or a core-to-belt high-to-low rate gradient ( Camalier et al . , 2012 ) . Given our results , both schemes have some merit , but they do not capture the full complexity of the organisation for temporal rates in concentric bands . Furthermore , the maps for amplitude-modulation rate reported in this study ( Figure 1 ) indicate that in addition to differential responses between fields a clear gradient can be observed within the fields , particularly in the auditory core . While earlier human fMRI studies ( Giraud et al . , 2000; Schonwiesner and Zatorre , 2009; Overath et al . , 2012 ) did not show a clear topographical representation of amplitude modulation rate , two recent studies suggest a topographical gradient for this sound dimension ( Barton et al . , 2012; Herdener et al . , 2013 ) . Furthermore , both studies reported an orthogonal relationship of temporal rate and frequency in some auditory areas . The representation of modulation rates reported in Barton et al . ( 2012 ) in the human auditory cortex resembles the data from the current study with a maximal preference of high rates in the postero-medial cortex surrounded by areas with a preference for lower rates . However , the interpretation of the topographical pattern of this study differs to ours , in proposing a concentric organisation for frequency representation overlapping with an angular representation of modulation rate in contrast to our interpretation of a concentric organisation for modulation rate and an angular representation of frequency in core areas . An important result of this difference is that Barton et al . ( 2012 ) suggest an orthogonal relationship of frequency and modulation rate representation in the entire auditory cortex while in our case we find such a relationship mainly in the core areas . Furthermore , the study of Barton et al . suggested that each field contained a separate complete amplitude modulation rate gradient in addition to a separate and complete frequency gradient whilst the scheme that we demonstrate in macaques suggests that amplitude modulation rate is represented across multiple areas , none of which contain a complete map . However , a detailed comparison between the two studies is complicated by Barton et al . ( 2012 ) , use of a definition of the auditory fields that is incompatible with the definition commonly used in non-human primates . The study by Herdener et al . ( 2013 ) focused on human homologues of the auditory core areas ( hA1 , hR ) in the vicinity of the Heschl's gyrus . The reported results in these areas are consistent with ours in that they suggest orthogonal relationships of frequency and amplitude modulation rates with similar gradient directions found to those in our study . Finally , a further recent study in humans tested the representation of temporal and spectral features ( Santoro et al . , 2014 ) . The chosen approach differed from our and previous studies in that a computational analysis was applied to test different models of stimulus feature representation . Thus , the emphasis was not on mapping individual rates and frequencies or identifying stimulus gradients . Furthermore , combined spectro-temporal modulations where used as stimuli in addition to pure temporal and spectral modulations . Nevertheless , the study identified an area in postero-medial auditory cortex , just posterior of the medial Heschl's gyrus , with a preference for fast temporal rates . Regions anterior and lateral of this area showed a preference for low temporal rates ( summarised in Figure 7 of Santoro et al . , 2014 ) . Based on anatomical relationships in the auditory cortex between human and non-human primates as suggested in Baumann et al . ( 2013 ) , such a pattern for temporal rate preference is consistent with the concentric pattern of rate representation observed here . In electrophysiology , neuronal responses to modulation rate are usually divided into two different coding principles: rate coding and temporal coding ( e . g . , Lu et al . , 2001; Liang et al . , 2002; Yin et al . , 2011 ) , reviewed in Joris et al . ( 2004 ) . The rate code is a measure of the average number of spikes fired over a defined period of time to the different modulation rates , and the temporal code is a measure of how well the neuron's firing synchronises with the amplitude envelope of the stimulus at different modulation rates . The average response of neurons in the cortex shows slightly different best rates for the two measures and some neurons are properly tuned to only one of the two measures ( Joris et al . , 2004 ) . Different neurons in the auditory cortex of non-human primates have been reported with tuning to modulation rates between 2–120 Hz , measured by temporal response and between 1–250 Hz measured by rate response ( Bieser and Muller-Preuss , 1996; Liang et al . , 2002 ) , see also ( Joris et al . , 2004 ) . However , neurons responding with a synchronised temporal response are rarely tuned to rates above 64 Hz and are preferentially located in primary areas . Neurons demonstrating rate tuning make up the majority of neurons that respond to modulated sounds in-non primary auditory areas and are frequently tuned to rates at 64 Hz and above . It is not entirely clear which of the two response types are represented by the BOLD response reported in this study . While an increased average firing rate of a local sample of neurons to a certain modulation rate would certainly lead to an increased BOLD response , an increased synchronisation within the same sample of neurons would probably have a similar effect . This is supported by a simulation study which suggested that both types of firing patterns would influence the BOLD response in similar way ( Chawla et al . , 1999 ) . In our study , the range of amplitude-modulation rates represented in the BOLD response and the pattern of areas that respond to specific rates is better matched by response properties reported for single neurons responding with a rate code in non-human primates ( Bieser and Muller-Preuss , 1996; Liang et al . , 2002; Bendor and Wang , 2008 ) . Nevertheless , it is also possible that the measured BOLD signal is a response to a combination of rate- and temporal-coding neurons . The responses to different amplitude-modulation rates reported in this study are in line with results from previous electrophysiological studies in other non-human primates ( Bieser and Muller-Preuss , 1996; Liang et al . , 2002 ) in the range of the preferred rates ( 2–128 Hz ) as well as the preferred rates for the different fields . The current study showed the highest preference in A1 for a rate of 32 Hz and to a lesser extent to 8 Hz while similar electrophysiology studies in monkeys additionally highlighted 16 Hz , a rate not used in this study . Human studies ( Giraud et al . , 2000; Harms and Melcher , 2002; Overath et al . , 2012 ) showed slightly lower preferred rates between 2–8 Hz for primary areas while rates above 64 Hz where hardly represented . These species-specific values are consistent with psychophysical comparisons between humans and macaques in an amplitude modulation discrimination task showing average peak sensitivities at lower rates in humans ( 10–60 Hz ) than in macaques ( 30–120 Hz ) ( O'Connor et al . , 2010 ) or even a low-pass function in humans ( Viemeister , 1979 ) . We demonstrated a topographic map for a specific temporal feature of sound , the rate of amplitude modulation . Temporal variation of sound can also be characterised by other means such as frequency modulation . Furthermore , differences in response latencies in areas at a similar hierarchy level were used as an indicator of the temporal resolution of the local circuits ( Langner et al . , 1987 , 2002 ) . Various studies show however that different temporal response measures to different temporal stimulus features are highly correlated . Liang et al . ( 2002 ) , for example , shows a high correlation of single neuron responses in the auditory cortex of non-human primates to amplitude-modulation rate and frequency-modulation rate . Studies in the inferior colliculus showed good correlation of the representation of amplitude-modulation rate and the response latency of neurons in the inferior colliculus ( Langner et al . , 1987 , 2002 ) . Furthermore , studies that recorded response latencies across different auditory fields , showed a tendency for longer latencies in rostral fields ( Bendor and Wang , 2008; Camalier et al . , 2012 ) and belt fields ( Camalier et al . , 2012 ) , consistent with the preference for slower modulation rates we found in these areas . This suggests that preferred amplitude-modulation rate is representative of the processing , or integration , time windows which characterise the time scale over which temporal features are integrated in a particular area or circuit . Here , we demonstrate a systematic , topographical representation of modulation rate in the auditory cortex of a non-human primate , organised , in parts , orthogonally to the established gradient for frequency . The systematic concentric organisation of amplitude-modulation rate that we demonstrate here , could provide an anatomical basis for the analysis of different modulation rates in separate modulation filterbanks ( Dau et al . , 1997a , 1997b ) . The superposition of maps of modulation rate and frequency also occurs in the inferior colliculus ( Baumann et al . , 2011 ) where some higher modulation rates are represented than in the cortex , whilst in the cortex the modulation rates mapped decrease with greater distance from A1 . The mapping of distinct dimensions of sound , amplitude-modulation rate and frequency , as distinct vectors in an anatomical space , is analogous to the mapping of polar angle and eccentricity that occurs at all processing levels of the visual system . Representation in the auditory cortex differs , however , in that a complete mapping of modulation rate does not occur within each cortical area , in contrast to the multiple and complete representations of spectral frequency that these areas contain .	The arrival of sound waves at the ear causes the fluid inside a part of the ear known as the cochlea to vibrate . These vibrations are detected by tiny hair cells and transformed into electrical signals , which travel along the auditory nerve to the brain . After processing in the brainstem and other regions deep within the brain , the signals reach a region called the auditory cortex , where they undergo further processing . The cells in the cochlea that respond to sounds of similar frequencies are grouped together , forming what is known as a tonotopic map . This also happens in the auditory cortex . However , the temporal properties of sounds—such as how quickly the volume of a sound changes over time—are represented differently . In the cochlea these properties are instead encoded by the rate at which the cochlear nerve fibres ‘fire’ ( that is , the rate at which they generate electrical signals ) . However , it is not clear how the temporal properties of sound waves are represented in auditory cortex . Baumann et al . have now addressed this question by scanning the brains of three awake macaque monkeys as the animals listened to bursts of white noise with varying properties . This revealed that just as neurons that respond to sounds of similar frequencies are grouped together within auditory cortex , so too are neurons that respond to sounds with similar temporal properties . When these temporal preferences are plotted on a map of auditory cortex , they form a series of concentric rings lying at right angles to the frequency map in certain areas . Recent brain imaging studies in humans have also suggested the existence of a ‘temporal map’ . Further experiments are now required to determine exactly how neurons within the auditory cortex encode the temporal characteristics of sounds .	[ "Abstract", "Introduction", "Results", "Discussion" ]	[ "neuroscience" ]	2015	The topography of frequency and time representation in primate auditory cortices
Genetic and environmental factors , such as metals , interact to determine neurological traits . We reasoned that interactomes of molecules handling metals in neurons should include novel metal homeostasis pathways . We focused on copper and its transporter ATP7A because ATP7A null mutations cause neurodegeneration . We performed ATP7A immunoaffinity chromatography and identified 541 proteins co-isolating with ATP7A . The ATP7A interactome concentrated gene products implicated in neurodegeneration and neurodevelopmental disorders , including subunits of the Golgi-localized conserved oligomeric Golgi ( COG ) complex . COG null cells possess altered content and subcellular localization of ATP7A and CTR1 ( SLC31A1 ) , the transporter required for copper uptake , as well as decreased total cellular copper , and impaired copper-dependent metabolic responses . Changes in the expression of ATP7A and COG subunits in Drosophila neurons altered synapse development in larvae and copper-induced mortality of adult flies . We conclude that the ATP7A interactome encompasses a novel COG-dependent mechanism to specify neuronal development and survival . Copper , iron , and manganese act as micronutrients , yet in excess behave as environmental neurotoxicants ( Bush , 2013; Perl and Olanow , 2007; Wright and Metals , 2007; Madsen and Gitlin , 2007 ) . These metals are handled by diverse membrane transporters and dedicated chaperones that deliver metals to cellular compartments to act as cofactors in essential enzymatic reactions while also maintaining cellular levels of these metals within a narrow range ( Robinson and Winge , 2010 ) . Mutations in these metal membrane transport and chaperoning mechanisms cause neurodegenerative disorders in both children and adults ( Madsen and Gitlin , 2007; Kaler , 2011 ) . In addition , metal exposure modulates the severity of neurological disease by yet unknown metal-sensitive mechanisms ( Sparks and Schreurs , 2003 ) . Thus , genetic and environmental factors , such as metals , converge to impinge on neurodegeneration and neurodevelopmental phenotypes . Cellular copper content is regulated by two chief transporters , ATP7A and CTR1 ( SLC31A1 ) ( Madsen and Gitlin , 2007; Kaler , 2011 ) . Cellular copper homeostasis is maintained by virtue of ATP7A and CTR1 subcellular localization and metal transport topology . ATP7A resides at the Golgi apparatus where it sequesters copper topologically into the Golgi lumen and away from the cytoplasm . In contrast , CTR1 localizes to the plasma membrane where it transports copper into the cytoplasm from the extracellular milieu ( Kaler , 2011; Lutsenko et al . , 2007; Polishchuk and Lutsenko , 2013 ) . The subcellular localization of ATP7A and CTR1 is modulated by copper-dependent vesicle traffic . After an extracellular copper challenge , ATP7A translocates from the Golgi complex to the plasma membrane where it extrudes copper out of cells while CTR1 undergoes copper-dependent endocytosis , thus down-regulating CTR1-dependent copper transport across the plasma membrane ( Kaler , 2011; Lutsenko et al . , 2007; Polishchuk and Lutsenko , 2013 ) . Of these two transporters , ATP7A is associated with neurological phenotypes in vertebrates ( Kaler , 2011; Menkes , 1999; Menkes et al . , 1962; Kennerson et al . , 2010; Tümer , 2013; Kaler et al . , 1994 ) , justifying it as our choice to test the hypothesis that interactomes of molecules specialized in the handling of metals encompass novel metal homeostasis mechanisms capable of modulating the expression of neurological traits . ATP7A loss-of-function genetic defects cause three neurological diseases , including Menkes disease ( OMIM 309400 ) , occipital horn syndrome ( OMIM 304150 ) , and X-linked distal spinal muscular atrophy type 3 ( OMIM 300489 ) . Menkes causing mutations systemically deplete organisms and cells of copper due to defective gut copper uptake , which in turn reduces the activity of cuproenzymes involved in neurotransmitter , neuropeptide , melanin , mitochondrial respiration , and extracellular matrix synthesis ( Kaler , 2011; Lutsenko et al . , 2007; Menkes , 1999; Menkes et al . , 1962; Kennerson et al . , 2010; Tümer , 2013; Kaler et al . , 1994 ) . Menkes disease is dominated by early childhood neurodegeneration whose mechanisms remain poorly understood ( Zlatic et al . , 2015 ) , thus making ATP7A an ideal candidate to identify metal homeostasis mechanisms capable of modulating neuropathology phenotypes . Here , we defined the ATP7A interactome which was enriched in products implicated in neurodegeneration and neurodevelopmental disorders , including the conserved oligomeric Golgi ( COG ) complex . COG is a Golgi vesicle tether necessary for intra Golgi vesicular traffic and the retention of Golgi-localized enzymes at the Golgi complex , such as glycosyltransferases ( Climer et al . , 2015; Willett et al . , 2013; Miller and Ungar , 2012; Ungar et al . , 2002; Kranz et al . , 2007 ) . COG genetic defects cause disruption of the Golgi complex function characterized by the loss of Golgi-resident enzymes and alterations in the glycosylation of membrane proteins traversing the exocytic route . These COG mutations cause a systemic disorder that includes cerebral atrophy , developmental delay , hypotonia , ataxia and epilepsy ( Climer et al . , 2015; Willett et al . , 2013; Miller and Ungar , 2012; Ungar et al . , 2002; Kranz et al . , 2007 ) . A requirement for the COG complex in cellular copper homeostasis and ATP7A-dependent metal buffering mechanisms was previously unrecognized and it has become the focus of our attention . In order to identify ATP7A-dependent and copper-sensitive mechanisms , we developed an unbiased approach to comprehensively define the ATP7A interactome in neuroblastoma cells ( Figure 1 ) . We used human SH-SY5Y neuroblastoma cells to perform immunoaffinity chromatography isolation of cross-linked ATP7A complexes . ATP7A was isolated with a monospecific monoclonal antibody that robustly recognizes a band in SH-SY5Y cells whose immunoreactivity is abolished in ATP7A-null cells ( Figure 1A , lanes 1–4 and Figure 1B ) . Like other cells , SH-SY5Y ATP7A surface content increased after copper addition as determined by surface biotinylation followed by streptavidin precipitation and immunoblot against ATP7A . Transferrin receptor , a copper-insensitive membrane protein , did not increase at the surface upon copper addition ( Figure 1 , compare lanes 1’ and 4’ ) . These results validate our choice of cells and antibody to define the ATP7A interactome . 10 . 7554/eLife . 24722 . 003Figure 1 . Isolation of ATP7A interactome . ( A ) In SH-SY5Y neuroblastoma cells , the addition of increasing amounts of copper leads to an increase of ATP7A at the cell surface as measured by surface biotinylation followed by streptavidin pull-down ( lanes 1’−4’ ) , while the total levels of ATP7A remain unchanged ( lanes 1–4 ) . Transferrin receptor shows consistent surface expression regardless of copper addition ( lanes 1’−4’ ) . The cytosolic chaperone Hsp90 was used to assess the selectivity of streptavidin pulldowns ( B ) The monoclonal ATP7A antibody used in these studies recognizes a single band by immunoblot . This band is missing in ATP7A null human Menkes fibroblasts ( lane 2 ) . ( C1–C4 ) Experimental designs to isolate ATP7A interactomes . ATP7A immunoaffinity chromatography was performed in two cell types , SH-SY5Y cells ( C1-2 ) and human skin fibroblasts ( C3-4 ) . In the former , left , SH-SY5Y cells were incubated with either 400 μM BCS ( C1 ) , a copper chelator , or 200 μM CuCl2 for 2 hr ( C2 ) . Cells were crosslinked with DSP , cell extracts were immunoprecipitated with the monoclonal ATP7A antibody either in the absence or presence of 22 μM ATP7A antigenic peptide . The same peptide was used to elute samples , which were then analyzed by label free quantitative mass spectrometry or silver stain ( D ) . C3-4 , ATP7A immunoaffinity chromatography was performed in ATP7A-null human skin fibroblasts ( C3 ) as well as the same cells recombinantly expressing ATP7A ( C4 ) . The experiment was performed as in SH-SY5Y cells , with the exception of the ATP7A antigenic peptide being omitted for outcompetition . ( D ) Silver stain from ATP7A immunoprecipitation depicted in ( C ) except that immunocomplexes were eluted with Laemmli sample buffer . Immunoprecipitations were performed for BCS-treated ( lanes 1–3 ) and CuCl2 treated ( lanes 1’−3’ ) SH-SY5Y cell extracts . Asterisks indicate immunoglobulin G chains , and densitometry profiles show differential protein enrichment in samples immunoprecipitated with ( lanes 3 and 3’ , black traces ) and without ( lanes 2 and 2’ , blue traces ) the antigenic ATP7A peptide . Below are immunoblots performed in parallel revealing positive identification of ATP7A and known interacting partner dopamine beta hydroxylase ( DBH ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 003 Intact cells were first cross-linked with dithiobis ( succinimidyl propionate ) to identify ATP7A protein interaction partners with weak or transient interactions ( DSP , Figure 1C ) . DSP is a short arm 12 Å cross-linker , which is cell permeant and cleavable by reducing agents ( Alloza et al . , 2004; Lomant and Fairbanks , 1976; Zlatic et al . , 2010 ) . Cross-linking increases the coverage of interactome components co-purifying with membrane proteins of complex topology , such as ATP7A ( Gokhale et al . , 2012; Perez-Cornejo et al . , 2012 ) . We developed a multipronged approach to maximize ATP7A-specific interactions . First , non-selective binding to magnetic bead-ATP7A antibody complexes was determined by the addition of an excess of the antigenic ATP7A peptide recognized by the monoclonal antibody ( Figure 1C1–2 and Figure 2B ) . Second , the ATP7A antigenic peptide was used to selectively elute cross-linked ATP7A interacting proteins from magnetic beads instead of Laemmli sample buffer ( Figures 1C1–4 ) . Third , we mock isolated ATP7A from ATP7A-null human fibroblasts ( ATP7A-/- ) , along with the same cells rescued by expression of recombinant ATP7A ( Figures 1C3–4 , ATP7AR/R ) ( La Fontaine et al . , 1998 ) . Proteins isolated from ATP7A-null cells were eliminated from all experimental datasets ( Figure 1C3 ) . Fourth , we performed label free quantitative mass spectrometry and thresholded positive protein identification as a ratio >2 between samples immunoisolated in the absence and presence of the ATP7A antigenic peptide ( Figures 1C and 2A ) ( Gokhale et al . , 2016 ) . Finally , we isolated cross-linked ATP7A complexes from SH-SY5Y neuroblastoma cells preincubated in the presence of either CuCl2 or the copper chelator bathocuproine disulphonate ( Figures 1C1–2 , BCS ) . Figure 1D depicts immunoaffinity chromatography experiments with cell extracts from SH-SY5Y cells pre-treated in the absence ( Figure 1D , BCS ) or presence of copper ( Figure 1D ) . Non-specific interactors bound to ATP7A antibody-decorated magnetic beads are depicted in lanes 3 and 3’ where incubations were performed with an excess of the ATP7A antigenic peptide ( Figure 1D ) . The presence of ATP7A and one previously described interactor , dopamine beta hydroxylase ( DBH ) , was confirmed by immunoblot of isolated cross-linked complexes ( Figure 1D ) ( Gokhale et al . , 2015a ) . Protein complexes copurifying with ATP7A isolated from BCS or copper-treated cells were not evidently different ( Figure 1D , compare lanes 2 and 2’ and traces , and Figure 2C ) . Therefore , we reasoned that proteins common to BCS and copper conditions would represent stable and/or strong interactors carried by ATP7A irrespective of its subcellular location ( Figure 2C ) . 10 . 7554/eLife . 24722 . 004Figure 2 . Identification of ATP7A interactome components by mass spectrometry . ( A ) Peptide spectrum matches ( PSM ) from quantitative mass spectrometry of proteins co-isolating with ATP7A are plotted for cells incubated with copper chelator BCS ( left ) , CuCl2 ( middle ) , and peptides identified in both samples ( right ) . Blue dots represent peptides that were enriched 2-fold over negative control samples incubated with an antigenic ATP7A peptide . ( B ) ATP7A peptides identified by mass spectrometry . Peptides corresponded to the antigenic peptide sequence shown above the black line as well as other ATP7A peptides identified via mass spectrometry ( blue lines ) . ( C ) Five hundred and forty one proteins co-isolated with ATP7A , one hundred thirty four of which were present regardless of cellular copper status are listed . Three COG subunits were present in both BCS and copper-treated samples ( blue text ) , the other three subunits were found either in BSC or copper-treated samples . Curation with a dataset from the CRAPome reveals minimal overlap . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 00410 . 7554/eLife . 24722 . 005Figure 3 . The ATP7A interactome enriches gene products implicated in Golgi function and neuropathologies . ( A–B ) The gene ontology algorithm DAVID was used to analyze the ATP7A interactome using the GO Terms Cellular Component ( A ) and Biological Process ( B ) . Size of the circles increases with gene set size , and p-values are represented by colors ranging from blue ( p=3×10−11 ) to red ( p=9×10−4 ) . Lines connecting circles depict ontology categories with shared gene products . ( C–F ) The ATP7A interactome was also evaluated using ENRICHR algorithm , to characterize dataset enrichment in GO Term Biological Process ( C ) , protein complexes from the CORUM database ( D ) , and phenotypes in mice ( E ) . Significance is represented as a combined score ( z-score x -log ( p-value ) ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 00510 . 7554/eLife . 24722 . 006Figure 4 . The ATP7A interactome enriches gene products associated with nervous system diseases and mental disorders . ( A–B ) The ATP7A interactome was analyzed using the GDA bioinformatics tool , which derives disease-gene links from human databases OMIM and Genopedia . Nervous system diseases ( A ) and mental disorders ( B ) were both significantly enriched in the dataset ( p<0 . 0001 ) . Circle size represents number of gene product from the ATP7A interactome in the disease category . Shades of blue depict p values . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 006 Quantitative mass spectrometry identified five hundred and forty one positive hits co-isolating with ATP7A , as defined in Figure 1C and Supplementary files 1–3 . These five hundred and forty one proteins are the result of a curation with a dataset obtained under the same conditions but using cellular homogenates from ATP7A null human cells ( Figure 1C3 , see Supplementary files 1–3 ) . We confirmed the quality of this curation step by comparing the five hundred and forty one positive hits to a published control dataset of three hundred and thirty five proteins that spuriously bind magnetic beads ( Figure 2C , CRAPome ) ( Mellacheruvu et al . , 2013 ) . The overlap between our five hundred and forty one ATP7A interactors and the CRAPome was just thirty one proteins , a 5 . 7% overlap ( Figure 2C , CRAPome , Supplementary file 3 ) . One hundred and thirty four proteins co-isolated with ATP7A regardless of whether cells were incubated with BCS or copper ( Figure 2A and C , see Supplementary files 1–3 ) . Mass spectrometry identified ATP7A peptides corresponding primarily to the antigenic primary sequence at the C-terminal domain of ATP7A ( Figure 2B ) . The interactome includes known ATP7A interactors such as dopamine beta hydroxylase ( Gokhale et al . , 2015a ) ; subunits of the WASH complex , FAM21 and WASH1 ( Phillips-Krawczak et al . , 2015; Ryder et al . , 2013 ) ; subunits of the retromer complex , VPS26 and 35 ( Phillips-Krawczak et al . , 2015; Steinberg et al . , 2013 ) ; and components of clathrin-coated vesicles such as CLTB ( Holloway et al . , 2013; Montpetit et al . , 2008; Yi and Kaler , 2015; Hirst et al . , 2012 ) . In order to prioritize candidate proteins for functional studies , we analyzed the ATP7A interactome with gene ontology algorithms ( Figure 3 , Supplementary file 3A–B ) . Gene ontological categories were enriched in Golgi-related terms such as Golgi transport complex ( GO:0017119 , p<5E-6 ) , which contained six of the eight subunits of the Golgi localized COG tethering complex . Three of these six COG subunits were identified in the BSC and copper-challenged ATP7A interactomes ( Figure 2C , COG2 , 4 and 7 ) . Additionally , bioinformatics identified neuronal ontological terms including growth cone , neuron projection , and dendrite ( GO:004300 , 0030426 , 003042 , respectively; all p<10E-4 , Figure 3A , Supplementary file 3A–B ) . Similar bioinformatic results were obtained irrespective of the algorithm used , DAVID ( Figure 3B , see Supplementary file 3A ) or ENRICHR ( Figure 3C , see Supplementary file 3B ) ( Huang et al . , 2007 , 2009; Chen et al . , 2013 ) . The ATP7A interactome is overrepresented in biological processes related to membrane trafficking and vesicular transport , Golgi-related transport in particular . Proteins contained in these ontological terms were among the products identified in both BSC and copper-challenged ATP7A interactomes , such as the COG complex subunits ( COG2 , 4 and 7 , Figure 2C ) . In fact , the COG complex was among the most prominently enriched complexes present in the ATP7A interactome ( CORUM database , Figure 3D , combined score >7 , see Supplementary file 3B ) ( Ruepp et al . , 2010 ) . These gene ontology tools point to the Golgi-dependent and neuronal trafficking mechanisms as key components of the ATP7A interactome . We next asked what traits would associate with loss-of-function mutations in components of the ATP7A interactome using mouse phenotypic databases ( Chen et al . , 2013 ) . Analysis of gene sets describing mouse phenotypes demonstrated that gene products contained in the ATP7A interactome were significantly enriched in categories describing neuronal pathology such as abnormal brain-neuronal morphology and neurodegeneration ( Figure 3E , combined score >9 . 5 , Supplementary file 3B ) . To further our understanding of phenotypes associated with members of the ATP7A interactome , we utilized the GDA bioinformatic tool that derives disease-gene links from human databases OMIM and Genopedia ( Park et al . , 2014 ) . This analysis revealed a significant association with neurodegenerative nervous system diseases ( Figure 4A , Supplementary file 3C ) and psychiatric disorders ( Figure 4B , Supplementary file 3C ) . The ATP7A interactome enriched 62 neurodegeneration associated or causative genes ( MeSH , Medical Subject Heading C10 . 574 , p<0 . 0001 ) , which include COG2 , VPS26 , VPS35 , DBH , and NSF . Moreover , 181 gene products contained in the ATP7A interactome were associated with mental disorders ( Medical Subject Heading F03 , p<0 . 0001 , Supplementary file 3C ) . Within the mental disorder category the MeSH term neurodevelomental disorders stood out ( Medical Subject Heading F03 . 625 , p<0 . 0001 , Supplementary file 3C ) represented with 34 gene products . These analyses indicate that the ATP7A interactome enriches gene products previously implicated in neurodegenerative and neurodevelopmental disorders , suggesting that these associations could participate in the neuropathogenesis of ATP7A genetic deficiencies . We prioritized for confirmation ATP7A interactome candidates due to their association with neurodegenerative and neurodevelopmental disorders . We immunoprecipitated ATP7A from cell extracts of cross-linked SH-SY5Y neuroblastoma cells and immunoblotted for proteins of interest . We focused on proteins that were either co-enriched with ATP7A to a similar extent as the bait ( fold of enrichment log2 >6 ) , were present in the BCS and copper treated ATP7A interactomes ( Figure 2C ) , were associated with or causative of neurodegeneration ( VAC14 , strumpellin , NFKB1 , and GIGYF2/PARK11 ) as well as mental disorders ( NSF , DOCK7 , GIGYF2 , and COG complex subunits , Figure 3 ) , and/or were present in compartments where ATP7A traffics to and from . The WASH complex subunit strumpellin and clathrin heavy chain ( CHC ) served as controls for previously described ATP7A interacting proteins ( Phillips-Krawczak et al . , 2015; Ryder et al . , 2013; Holloway et al . , 2013; Montpetit et al . , 2008; Yi and Kaler , 2015; Hirst et al . , 2012 ) . The ATP7A monoclonal antibody co-immunoprecipitated strumpellin and clathrin heavy chain , and other proteins belonging to the ATP7A interactome listed above ( Figure 5A , lane 2 ) . The addition of ATP7A immunogenic peptide prevented the coprecipitation of ATP7A with these selected interactome components ( Figure 5A , lane 3 ) . To further establish specificity of our immunoprecipitation method , we immunoblotted for abundant cytosolic proteins that were identified by mass spectrometry either in the negative control samples or that fell below the threshold of enrichment established for significance . These proteins include actin , several peroxiredoxins , thioredoxin , and the ubiquitin hydrolase UCHL1 ( Figure 5B ) . None of these proteins co-isolated with ATP7A even in the presence of a crosslinker ( Figure 5B , lane 2 ) . Neither specific interactors of ATP7A ( Figure 5A ) nor proteins used as controls ( Figure 5B ) changed their association with ATP7A after pretreating cells with copper chelator ( BSC-treated cells , Figure 1C ) or copper ( data not shown ) . These results validate the ATP7A interactome and confirm that this interactome enriches neurodevelopmental and neurodegenerative gene products . 10 . 7554/eLife . 24722 . 007Figure 5 . ATP7A co-immunoprecipitates with proteins implicated in neurodegeneration and neurodevelopmental disorders . ( A–B ) ATP7A was immunoprecipitated from DSP-crosslinked SH-SY5Y neuroblastoma cell lysates . Whole cell extracts ( lanes 1 ) and immunoprecipitated samples ( lanes 2 and 3 ) were resolved by SDS-PAGE and analyzed by immunoblot . Lane 3 contains samples in which an antigenic ATP7A peptide was added during immunoprecipitation as a negative control . ( A ) ATP7A and previously characterized interactors , clathrin heavy chain ( CHC ) and strumpellin , were selectively identified , along with newly identified interactome components that were highly enriched by mass spectrometry and associated with neuropathologies . Note the presence of two COG complex subunits . ( B ) Abundant cytosolic proteins that fell below the significant fold of enrichment by mass spectrometry or were identified in the negative control samples failed to co-purify with ATP7A . None of these proteins coprecipitate with ATP7A demonstrating the specificity of the interactions depicted in A . Asterisks denote mouse IgG chains . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 007 We selected the COG complex to test the hypothesis that genetic defects in components of the ATP7A interactome impair copper homeostasis by metal transporters for three reasons . First , we find that ATP7A co-purifies with six of the eight COG complex subunits to a similar extent ( Figure 2A and C; Figure 5A ) . Secondly , our bioinformatics analysis prioritizes this complex among components of the ATP7A interactome ( Figure 3A–E ) . Finally , mutations in the human COG subunits result in neurological defects some overlapping with Menkes disease ( Figures 3E and 4 ) ( Foulquier et al . , 2006; Kodera et al . , 2015; Reynders et al . , 2009; Paesold-Burda et al . , 2009; Fung et al . , 2012; Rymen et al . , 2012; Lübbehusen et al . , 2010; Huybrechts et al . , 2012; Shaheen et al . , 2013; Wu et al . , 2004; Morava et al . , 2007; Zeevaert et al . , 2009; Foulquier et al . , 2007 ) . . We used HEK293 cells rendered null for COG subunits 1 ( COG1Δ/Δ ) or 8 ( COG8Δ/Δ ) using CRISPR-Cas9 genome editing ( Blackburn and Lupashin , 2016; Bailey Blackburn et al . , 2016 ) . These cells reproduce cellular phenotypes associated with COG complex deficiencies , such as defective Golgi-dependent glycosylation of membrane proteins and degradation of Golgi-localized proteins in lysosomal compartments ( Blackburn and Lupashin , 2016; Bailey Blackburn et al . , 2016; Shestakova et al . , 2006 ) . Both COG null cell lines exhibited decreased expression of a known COG-dependent and Golgi-localized membrane protein , GPP130 ( Figure 6A ) . In addition , the GPP130 protein was found to migrate faster during SDS-PAGE , an indication of defective GPP130 glycosylation in COG null cells ( Figure 6A ) ( Sohda et al . , 2010; Oka et al . , 2004; Zolov and Lupashin , 2005 ) . Similarly , ATP7A expression was reduced by ~25–50% ( Figure 6A–B ) , and ATP7A SDS-PAGE migration was increased in COG-null cell lines ( Figure 6A ) . 10 . 7554/eLife . 24722 . 008Figure 6 . ATP7A stability and surface expression requires the COG complex . ( A–B ) Immunoblots for ATP7A , known COG-dependent protein GPP130 , COG seven and actin were performed in wild type HEK293 cells ( 6A , lane 1 ) and cells null for COG subunit 1 ( 6A , lane 2 , COG1Δ/Δ ) or COG subunit 8 ( 6A , lane 3 , COG8Δ/Δ ) . ( C ) Surface biotinylation of the same cell types was performed with ( 6C , even lanes ) and without ( 6C , odd lanes ) the addition of 200 μM CuCl2 for two hours . Total protein extracts ( 6C , lanes 1–4 ) and surface biotinylated proteins precipitated by streptavidin beads ( 6C , lanes 5–8 ) were probed for ATP7A and CTR1 , along with streptavidin-peroxidase to compare biotinylation efficiency . ( D–F ) ATP7A and CTR1 quantitations to measure the ratio of surface to total ATP7A ( D ) , total surface ATP7A with and without copper ( E ) , and total surface CTR1 with and without copper ( F ) . D to F surface signals were corrected by the efficiency of biotinylation that was 1 . 84 ± 0 . 7 higher in COG1Δ/Δ cells ( average ± SEM ) . Surface levels of ATP7A and CTR1 for each experimental condition were compared using non parametric Kriskal Wallis test followed by pairwise Mann-Whitney U test comparisons , n = 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 008 We determined whether COG deficiency also affected the surface expression of ATP7A by surface biotinylation . Wild type , COG1Δ/Δ , and COG8Δ/Δ HEK293 cells were incubated at 37°C for two hours in the absence and presence of 200 µM CuCl2 , followed by surface biotinylation at 4°C . Biotinylated proteins were precipitated with streptavidin beads and analyzed by immunoblot with antibodies against the copper transporters ATP7A and CTR1 ( Figure 6C–F ) . The efficiency of cell surface biotinylation was at nearly two-fold higher in COG-null cells as compared to wild type cells ( Figure 6C compare odd and even lanes ) , thus we normalized all transporter surface expression levels as a ratio of the surface content to the biotinylation efficiency . Normalized surface levels of ATP7A were decreased in COG1Δ/Δ HEK293 cells as compared to controls ( Figure 6C , compare lanes 5–6 , Figure 6E ) . Copper addition to wild type HEK293 cells modestly decreased the normalized surface levels of ATP7A . In contrast , normalized ATP7A surface levels in COG1Δ/Δ HEK293 cells did not change after copper ( Figure 6C , compare lanes 5 and 7 and 6 and 8; Figure 6E ) . The mobilization of ATP7A from the surface in the presence of excess copper in HEK293 cells is in contrast with the ATP7A response in SH-SY5Y neuroblastoma cells ( Figure 1A ) . These findings demonstrate that the total and cell surface levels of ATP7A are decreased in cells with genetic defects in the COG complex . Our findings suggested that either the ATP7A trafficking to and from the plasma membrane in the presence of copper differ between wild type HEK293 and SH-SY5Y cells or , alternatively , HEK293 cells are poorly responsive to a copper challenge . We tested the latter hypothesis by asking if the normalized surface content of plasma membrane copper transporter CTR1 was sensitive to copper addition . CTR1 is the main plasma membrane transporter required for copper influx into cells ( Gupta and Lutsenko , 2009; Kuo et al . , 2001 ) . We used the well-known copper-induced endocytosis of CTR1 as a tool to assess if HEK293 cells respond to copper addition ( Figure 6C ) ( Petris et al . , 2003; Clifford et al . , 2016 ) . The CTR1 antibody recognized monomeric and oligomeric species when CTR1 was enriched in cell surface biotinylated proteins , thus we could not assess CTR1 total cellular expression ( Figure 6C compare inputs 1–2 to lanes 5–8 ) . However , we found pronounced down-regulation of normalized cell surface CTR1 in COG1Δ/Δ HEK293 cells in basal conditions ( Figure 6C compare lanes 5 and 6 , and Figure 6F ) . Addition of copper to wild type and COG1Δ/Δ HEK293 cells further exacerbated the depletion of CTR1 at the surface ( Figure 6C compare lanes 5 and 7 , lanes 6 and 8 , and Figure 6F ) , demonstrating that wild type and COG1Δ/Δ HEK293 cells are responsive to copper challenge . These results demonstrate that COG complex genetic defects decrease the surface expression of two copper transporters , ATP7A and CTR1 , thus suggesting complex copper metabolism phenotypes in COG deficiencies . We addressed whether COG genetic defects impair cellular copper uptake by directly measuring the metal content in cells . In addition , we determined secondary effects of copper imbalances by quantifying transcripts whose expression is sensitive to metals and metal pathway dysfunction . We measured total copper and zinc content in wild type , COG1Δ/Δ , and COG8Δ/Δ HEK293 cells with inductively-coupled plasma mass spectrometry . Copper was readily detectable in wild type HEK293 cells . However , copper content in COG-null HEK293 cells was below detection limit ( 1 ng/sample , Figure 7A ) . COG-null cellular copper phenotype was selective since zinc levels remained similar to wild type ( Figure 7B ) . Addition of the copper-selective chelator BCS brought down copper content in wild type cells while addition of copper chloride or the copper ionophore disulfiram increased cellular copper levels in all three genotypes ( Figure 7A ) . However , copper chloride uptake was impaired in COG null cells as compared to wild type HEK293 cells even when COG-null cells were exposed to 25 μM copper ( Figure 7A ) . Only the addition of the copper ionophore disulfiram increased the copper content at or above wild type levels in COG null HEK293 cells ( Figure 7A ) . These results demonstrate that the decreased expression of ATP7A and CTR1 observed in COG mutant cells results in selective copper deficiency . 10 . 7554/eLife . 24722 . 009Figure 7 . Copper content and expression of copper-sensitive transcripts is altered in COG deficient cells . Copper ( A ) , zinc ( B ) , and transcript ( C ) levels from wild type ( grey bars ) , COG1Δ/Δ ( light blue bars ) , and COG8Δ/Δ ( dark blue bars ) HEK293 cells were measured either by inductively coupled plasma mass spectrometry ( A-B ) or quantitative real-time PCR ( C ) . In A and B cells were incubated for 24 hr with the indicated drugs in complete media with 10%FBS . Transcripts were normalized to beta-actin mRNA . Inductively coupled plasma mass spectrometry was performed in two independent biological replicates where each determination was in triplicate . Five independent biological replicates were performed for each determination in triplicate for quantitative real-time PCR . For metal determinations , significant p-values were determined by ANOVA followed by Dunnett test . p-values associated with transcript changes were determined by non-parametric Kriskal Wallis test followed by pairwise Mann-Whitney U test comparisons; all unlabeled comparisons are not significant in ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 009 Imbalances in cellular copper metabolism modify the expression of transcripts encoding copper transporters and metallochaperones that carry this metal to distinct molecules and subcellular compartments ( Robinson and Winge , 2010 ) . We previously used copper-sensitive gene expression to document copper metabolism imbalances in mutations of an ATP7A complex interactor , the BLOC-1 complex ( Gokhale et al . , 2015a ) . We predicted that transcripts of ATP7A and/or metallochaperones should be altered in COG deficient cells as a consequence of copper dyshomeostasis ( Figure 7A ) . We focused on the metallochaperones ATOX1 , which delivers copper to ATP7A ( Strausak et al . , 2003; Voskoboinik et al . , 1999; Walker et al . , 2002; Lutsenko et al . , 1997 ) ; CCS , which carries copper to the mitochondrially localized SOD1 and that itself is imported into mitochondria ( Suzuki et al . , 2013; Wang et al . , 2013 ) ; COX17 , which is required for copper delivery to the mitochondrial cytochrome c oxidase ( Cobine et al . , 2006 ) ; and two isoforms of metallothioneins , both cysteine-rich proteins that bind metals in the cytoplasm ( Palmiter , 1998 ) . We used a qRT-PCR assay capable of detecting changes in the expression of COG1 and COG8 transcripts in COG1Δ/Δ and COG8Δ/Δ HEK293 cells ( Figure 7C ) . We normalized all transcript determinations against beta-actin mRNA . The housekeeping gene glyceraldehyde 3-phosphate dehydrogenase message showed no differences among cell genotypes when normalized to actin mRNA ( Figure 7C , GAPDH ) . In stark contrast with the ATP7A protein expression levels , both COG1Δ/Δ and COG8Δ/Δ deficiency doubled ATP7A transcript levels as compared with wild type cells . These changes in ATP7A mRNA contrasted with the expression of the ATP7A metallochaperone , ATOX1 , which remained unchanged ( Figure 7C ) . Transcripts encoding metallothioneins IA and IIB were increased only in COG8Δ/Δ ( MT1A and MT2A , Figure 7C ) . However , mRNA levels of two metallochaperones that traffic copper to mitochondria , CCS and COX17 , were significantly down-regulated in both COG1Δ/Δ and COG8Δ/Δ cells ( Figure 7C ) . CCS mRNA was reduced to 58% in both COG deficient cells whereas COX17 mRNA was decreased to 64% of the control values . These findings indicate that the expression of mitochondrial copper homeostasis factors is perturbed in COG deficient HEK293 cells . We further tested this hypothesis by measuring mRNA levels of mitochondrially localized co-factors that work in concert with COX17 to load copper into cytochrome c oxidase ( COX19 , COX23 , PET191 , and SCO1 , Figure 7C ) ( Cobine et al . , 2006 ) . Of these factors PET191/COA5 mRNA was significantly decreased to ~70% in both COG1Δ/Δ and COG8Δ/Δ HEK293 cells ( Figure 7C ) . In contrast , SCO1 and COX23 were selectively down-regulated in COG1Δ/Δ and COG8Δ/Δ cells , respectively ( Figure 7C ) . Thus , COG complex deficient cells have altered expression of at least four transcripts implicated in the delivery of copper to mitochondrial enzymes . These results suggest that COG complex deficiency cellular copper depletion leads to copper-dependent cellular and mitochondrial metabolism . We assessed the effects of increasing extracellular copper on cellular and mitochondrial metabolism with tetrazolium salts . Tetrazolium is reduced into formazan by NAD ( P ) H-dependent oxidoreductases and dehydrogenases localized to cytoplasm and mitochondria ( Berridge et al . , 2005 ) . We measured 2- ( 4 , 5-dimethyl-2-thiazolyl ) -3 , 5-diphenyl-2H-tetrazolium bromide reduction ( MTT ) in HEK293 cells either wild type , COG1Δ/Δ , COG8Δ/Δ , or carrying combined defects in COG1 and COG8 ( Figure 8 , COG1 , 8Δ/Δ ) . Cells were exposed to increasing extracellular copper concentrations for 72 hr and MTT reduction was measured . Wild type cells exposed to low copper , 3–20 µM , increased MTT reduction as compared to wild type cells incubated with media alone ( Figure 8A , grey symbols ) . In contrast , COG1Δ/Δ , COG8Δ/Δ , and COG1 , 8Δ/Δ HEK293 cells exposed to low copper failed to increase MTT metabolization ( Figure 8A , blue symbols and Figure 7A ) . Irrespective of the cell genotype , all cells experienced a decrease of MTT metabolization at copper concentrations above 50 µM , likely due to copper toxicity ( Figure 8A ) . These COG- and copper-dependent phenotypes do not reflect general cellular sensitivity to metal-based toxicants as assessed with cisplatin , a cytotoxic agent whose import into HEK293 cells does not require CTR1 copper transporter activity ( Figure 8B ) ( Bompiani et al . , 2016 ) . These results demonstrate that COG deficient cells fail to respond to extracellular copper as predicted from the defects in ATP7A and CTR1 surface transport mechanisms . 10 . 7554/eLife . 24722 . 010Figure 8 . COG null cells possess copper metabolism defects . ( A ) Wild type , COG1Δ/Δ , COG8Δ/Δ and COG1 , 8Δ/Δ HEK293 cells were incubated with CuCl2 ranging from 3–300 μM for 72 hr . Each condition was carried out in quadruplicate , and the activity of NAD ( P ) H-dependent oxidoreductases was measured by the reduction of MTT . Each dot represents the average absorbance at 595 nm ± SEM , normalized to a baseline reading ( n = 5 ) . One Way ANOVA followed by Bonferroni's All Pairs Comparisons ( B ) Wild type and COG null cells were incubated with 0 . 4–100 μM cisplatin for 72 hr and MTT reduction was measured as above ( n = 2 ) . ( C–D ) The copper ionophore disulfiram ( DSF ) was added to wild type , COG1Δ/Δ , and COG8Δ/Δ HEK293 cells for 24 hr either in concentrations ranging from 1 . 5–200 nM either in the absence ( C ) or presence ( D ) of 2 . 5 μM CuCl2; each condition was carried out in quadruplicate . Reduction of MTT by NAD ( P ) H-dependent oxidoreductases was measured and normalized to a baseline reading with no drug added . Each dot represents the average of five independent biological replicates ± SEM . Non-parametric Kriskal Wallis test followed by pairwise Mann-Whitney U test comparisons . ( E–F ) Crystal violet staining was performed in parallel to MTT analysis to measure changes in cell number . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 010 We hypothesized that if decreased cellular copper and normalized surface levels of ATP7A and CTR1 in COG deficient HEK293 cells ( Figure 6 ) prevent MTT metabolization , then direct copper delivery across the plasma membrane via a copper ionophore should revert MTT phenotypes ( Figure 8C–D ) . Disulfiram is a cell permeant copper chelation agent that inhibits copper dependent enzymes in diverse compartments including mitochondria ( Simonian et al . , 1992; Kuroda and Cuéllar , 1993; Goldstein , 1966; Gaval-Cruz and Weinshenker , 2009 ) . However , disulfiram complexed with copper increases metal cellular levels ( Cen et al . , 2004; Allensworth et al . , 2015 ) ( Figure 7A ) , and rescues respiration phenotypes in CTR1 null cells ( Schlecht et al . , 2014 ) . We incubated wild type and COG deficient HEK293 cells with disulfiram in the absence ( Figure 8C and E ) or presence of copper ( Figure 8D and F ) . Cells were incubated for 24 hr to minimize the effect of modifications in cell numbers on MTT activity readings . We controlled for cell numbers with a crystal violet colorimetric assay ( Feoktistova et al . , 2016 ) ( Figure 8E–F ) . Addition of increasing disulfiram to wild type and COG-null HEK293 cells did not affect MTT activity at low concentrations , yet disulfiram above 25 nM decreased MTT activity ( Figure 8C ) . None of these disulfiram concentrations significantly affected cell numbers ( Figure 8E ) indicating that MTT metabolization was impaired by the copper chelation activity of disulfiram at high doses . Next , we added increasing disulfiram concentrations to wild type and COG null HEK293 cells in the presence of 2 . 5 µM of copper ( Figure 8D and F ) . Disulfiram concentrations above 25 nM in the presence of added copper decreased MTT activity and cell numbers irrespective of the cell genotype ( Figure 8D and F ) . Thus , we focused on disulfiram and copper conditions that did not compromise cell numbers ( Figure 8F ) . Loading cells with copper using low concentrations of disulfiram ( <25 nM ) significantly increased MTT metabolization in COG1Δ/Δ and COG8Δ/Δ HEK293 cells as compared to wild type controls ( Figure 8D , compare gray and blue symbols ) . These results show that delivering copper with a copper ionophore to bypass copper transporter plasma membrane defects increases the metabolization of MTT in COG null cells . COG null HEK293 cells have copper-dependent cellular and metabolic phenotypes that can be rescued with a copper-loaded ionophore , disulfiram . We focused on ATP7A overexpression because it cell-autonomously decreases cellular levels of copper due to ATP7A mistargeting to the cell surface ( Hwang et al . , 2014; Lye et al . , 2011 ) . Thus , we hypothesized that phenotypes induced by genetically increasing ATP7A expression in neurons should be modulated by loss-of-function mutations in the COG complex , which decreases ATP7A expression ( Figure 6A–B ) . We first tested whether Drosophila ATP7A and COG complex subunits genetically interact to specify synapse morphology in the developing neuromuscular junction of the third instar larva ( Figure 9 ) . We overexpressed ATP7A in Drosophila neurons using the pan-neuronal elav GAL4 c155 driver ( C155 ) ( Lin and Goodman , 1994 ) . Overexpression of ATP7A reduced the cumulative synapse branch length; thus , inducing a collapse of the synapse as measured as an increased synaptic bouton density ( Figure 9A image C155>UAS-ATP7A , Figure 9B–C , column 4 . Compare 4 to control columns 1 and 2 ) . In contrast , animals carrying one copy of the null allele cog1e02840 increase cumulative synapse branch length while maintaining wild type synaptic bouton density ( Figure 9A–C , column 3 ) . As predicted by our hypothesis , overexpression of ATP7A in cog1e02840 flies restored synaptic bouton density to wild type levels ( Figure 9A and B , compare columns 4 and 5 ) . These results demonstrate that a component of the ATP7A interactome , the COG complex , genetically interact with ATP7A to specify a neurodevelopmental synapse phenotype . 10 . 7554/eLife . 24722 . 011Figure 9 . Drosophila ATP7A and COG1 genetically interact to specify Drosophila melanogaster synapse development . Third instar larvae neuromuscular junction synapses were stained with anti HRP antibodies ( A ) imaged and their morphology assessed using as parameters branch length ( B ) and bouton density ( C ) . Scoring was done blind to the animal genotype . Control animals ( C155 outcross , column 1; or UAS-ATP7A outcross , column 2 ) , animals carrying one copy of the null allele cog1e02840 ( cog1 e02840 outcrossed , column 3 ) , flies overexpressing ATP7A in neuronal cells ( c155>UAS-ATP7A; column 4 ) , and animals overexpressing ATP7A and mutant for cog1 ( C155> UAS-ATP7A x cog1e02840 , column 5 ) were analyzed . Numbers in parentheses and italics in C depict the number of animals . Statistical comparisons were performed with One Way ANOVA followed by Fisher’s Least Significant Difference Comparison . Box plots depict percentiles fifth and 95th . Box line represents sample median and diamonds sample mean and notches mark the half-width . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 011 Second , we examined whether ATP7A and COG complex subunits genetically interact to specify neurodegeneration in the Drosophila adult nervous system ( Figure 10 ) . We controlled the expression of ATP7A in adult dopaminergic neurons , a group of cells frequently used to model Parkinson’s disease in Drosophila ( Feany and Bender , 2000; Haass and Kahle , 2000; Li et al . , 2000; Yang et al . , 2003; Lin et al . , 2010 ) . We drove the expression of UAS-ATP7A selectively in dopaminergic and serotoninergic neurons with the dopa decarboxylase ( Ddc ) -GAL4 driver ( Feany and Bender , 2000 ) . We reasoned that overexpression of ATP7A , which decreases cellular levels of copper ( Hwang et al . , 2014; Lye et al . , 2011 ) , should reduce the toxicity to copper diet exposure . We previously observed a high sensitivity to copper in the diet of wild type animals ( Gokhale et al . , 2015a ) . Copper feeding progressively increased mortality in wild type male ( Figure 10A ) and female adults ( Figure 10B ) over a period of three days . Overexpression of ATP7A in adult dopaminergic neurons was sufficient to significantly protect males and female adult animals from the toxic effect of copper feed at 48 hr ( Figure 10A–B , ( Ddc>UAS-ATP7A ) ) . Similarly , mutation of the COG complex subunit cog1 protected animals from copper diet induced death ( Figure 10A–B , ( Ddc x cog1e02840 ) ) . In contrast , the mortality phenotype observed in animals overexpressing ATP7A was restored to the levels of wild type lethality by adding in trans a genetic defect in cog1 ( Figure 10A–B , ( UAS-ATP7A; Ddc x cog1e02840 ) ) . Importantly , mortality was negligible when copper was omitted from the diet fed to animals of any genotype ( Figure 10C–D ) . Our experiments demonstrate that the COG complex and ATP7A genetically interact in adult dopaminergic neurons to specify copper-dependent mortality . 10 . 7554/eLife . 24722 . 012Figure 10 . Drosophila ATP7A and COG1 genetically interact in dopaminergic neurons to specify copper-induced Drosophila melanogaster viability . Control animals ( Ddc outcross ) , animals carrying one copy of the null allele cog1e02840 ( cog1 e02840 outcrossed ) , flies overexpressing ATP7A in dopaminergic neuronal cells ( Ddc>UAS-ATP7A ) , and animals overexpressing ATP7A and mutant for cog1 ( UAS-ATP7A; Ddc x cog1e02840 ) were fed a glucose diet ( C-D ) or a glucose diet supplemented with 1 mM CuCl2 for three consecutive days . Numbers in parentheses and italics depict the number of independent experiments each one performed with at least 10 animals per genotype . Statistical comparisons were performed with Non-parametric Kriskal Wallis test followed by pairwise Mann-Whitney U test comparisons . Box plots depict percentiles fifth and 95th . Box line represents sample median and diamonds sample mean and notches mark the half-width . DOI: http://dx . doi . org/10 . 7554/eLife . 24722 . 012 We isolated and defined the ATP7A interactome to identify novel metal homeostasis mechanisms capable of modulating the expression of neurological traits . We selected candidate gene products from the ATP7A interactome based on three criteria; they strongly coenriched with ATP7A , were present in Golgi and post-Golgi compartments where ATP7A is present at steady state or traffics through after a copper challenge , and phenocopied some of the chief neurological phenotypes in Menkes disease . One such candidate that met all of these criteria , the COG complex , is associated with severe neuropathologies when mutated in humans and as we demonstrate have a severe depletion of cellular copper ( Foulquier et al . , 2006; Kodera et al . , 2015; Reynders et al . , 2009; Paesold-Burda et al . , 2009; Fung et al . , 2012; Rymen et al . , 2012; Lübbehusen et al . , 2010; Huybrechts et al . , 2012; Shaheen et al . , 2013; Wu et al . , 2004; Morava et al . , 2007; Zeevaert et al . , 2009; Foulquier et al . , 2007 ) . Here , we demonstrate that the COG complex is a hub where two copper transporters , ATP7A and CTR1 , converge to regulate cellular copper content , homeostasis , synapse development and copper-induced neurodegeneration . We compared the ATP7A interactome from cells treated in the presence of either excess copper or the copper chelator BCS ( Figure 1 ) . We reasoned that copper-dependent translocation of ATP7A from Golgi to post-Golgi compartments would reveal putative compartment-specific ATP7A interactors . However , we found that the most enriched proteins were associated with ATP7A regardless of whether cells were challenged with excess copper or not . It may be possible that we did not identify proteins that selectively associate with ATP7A in a copper-dependent manner because of our purification strategy . ATP7A peptides identified by mass spectrometry from the interactome were predominantly sequences enriched at the C-terminal domain of ATP7A , which could either be a reflection of the tertiary structure of the protein ( Figure 2 ) or an indication that copper-induced protein associations with ATP7A mask the N-terminal domain from recognition by the antibody used as a bait . We favor the latter hypothesis because the antibody used to isolate the interactome is directed to the N-terminal domain , the same domain where the copper binding sites reside in ATP7A and a domain necessary for ATP7A interaction with ATOX1 , the copper chaperone that shuttles copper to ATP7A ( Strausak et al . , 2003; Voskoboinik et al . , 1999; Walker et al . , 2002; Lutsenko et al . , 1997 ) . Consistent with this hypothesis , we did not detect ATOX1 in the ATP7A interactome although it is a robustly documented ATP7A interactor . However , our ATP7A antibody bait allowed the identification of both known and novel interacting proteins . ATP7A possesses 664 predicted and 13 experimentally tested interactors according to a comprehensive computational interactome of the human genome ( Garzón et al . , 2016 ) . Our experimentally generated ATP7A interactome includes 541 proteins , of which only 21 proteins are in common with a computational ATP7A interactome . Therefore , the majority of the ATP7A interactome components identified here consisted of uncharacterized relationships . We identified several known trafficking proteins and complexes present at the plasma membrane , endosome , or Golgi complex that are known to regulate the trafficking of ATP7A , including the WASH complex and components of clathrin coated vesicles , as well as novel factors such as VAC14 and the COG complex ( Figure 2 ) . We independently confirmed that these trafficking components co-isolated with ATP7A ( Figure 5 ) . Strikingly , these trafficking proteins and complexes also associate with diverse neuropathologies , suggesting that they may share common mechanisms with the neuropathology observed in Menkes disease ( Figures 3–4 ) ( Climer et al . , 2015; Lenk et al . , 2016; Bonifati et al . , 2003 ) . One prominent finding was the strong co-enrichment of COG complex subunits with ATP7A . We predicted that if the COG complex were responsible for trafficking of ATP7A , deletion of the complex would alter cellular copper content , subcellular localization and levels of copper transporters , the expression of copper-sensitive molecules , and cellular susceptibility to copper challenges . We found that in COG Δ/Δ cells , the total cellular and normalized surface levels of ATP7A are decreased ( Figure 6 ) , a molecular defect leading to selective copper depletion in COG null cells ( Figure 7A ) . While the non-normalized surface levels of ATP7A measured by immunoblot were comparable in wild type and COGΔ/Δ cells at steady state , we normalized by the biotinylation efficiency , which was greater in the COG null cells . We attribute this increased biotinylation efficiency to a reduction in the surface glycocalyx negative charge in COG null cells , which would increase surface biotinylation efficiency with the anionic sulfo-NHS-biotin reagent ( Pokrovskaya et al . , 2011 ) . When observing normalized surface levels of ATP7A after a copper challenge , we were surprised to find that , unlike in previously observed cell types including Caco-2 , CHO cells , and human fibroblasts , ATP7A did not increase at the surface of wild type HEK cells ( Nyasae et al . , 2007; Petris and Mercer , 1999 ) . We were unable to assess if this was due to an accelerated rate of endocytosis of ATP7A after copper addition , or a small pool of ATP7A at the Golgi unable to sustain increased ATP7A surface expression . However , we were able to document copper dependent membrane traffic of CTR1 in the same cells ( Gupta and Lutsenko , 2009; Kuo et al . , 2001; Petris et al . , 2003; Clifford et al . , 2016 ) , thus excluding unresponsiveness to copper in HEK293 cells . We also assessed the total and normalized surface levels of CTR1 , a copper importer that is known to internalize in the presence of excess copper . CTR1 normalized surface levels were dramatically decreased in COGΔ/Δ cells in resting conditions , and this deficiency was exacerbated by the addition of copper , leaving no measurable CTR1 protein on the surface . This led us to hypothesize that COG deletion is perturbing not only copper regulation by ATP7A , but copper homeostasis in other cellular compartments that require the entry of copper from the plasma membrane via CTR1 . We decided to use copper-sensitive metabolism of MTT to determine whether the effects of COG deletion mutations affecting ATP7A and CTR1 surface expression extended beyond ATP7A-dependent Golgi trafficked enzymes , such as DBH , LOX and PAM ( Kaler , 2011; Lutsenko et al . , 2007 ) . Cytochrome c oxidase , though not directly downstream of ATP7A , requires copper as a cofactor , and thus mitochondrial function is impaired by decreased activity of cytochrome c oxidase when copper delivery to mitochondria is impaired ( Leary et al . , 2004 , 2007 ) . Our hypothesis that copper homeostasis is impaired in COGΔ/Δ cells is supported by reports describing that RNAi or mutation of COG subunits alters copper content in mammalian cells and yeast , respectively ( Schlecht et al . , 2014; Malinouski et al . , 2014 ) and by our data that levels of copper and copper sensitive transcripts encoding proteins present in or transferring copper to different subcellular compartments are altered in COG null cells ( Figure 7 ) . Copper regulatory genes are known to undergo coordinated regulation in response to changing copper levels , and we found that multiple copper sensitive transcripts are decreased in COGΔ/Δ cells ( Barresi et al . , 2016 ) . Of particular interest are two metallochaperones that deliver copper to mitochondria , CCS and COX17 , whose message levels are reduced in COG null cells ( Robinson and Winge , 2010; Cobine et al . , 2006 ) . We found that MTT metabolism is impaired in COG null cells due to defects in copper cellular homeostasis . These data suggest the existence of a mechanistic link between a Golgi-dependent metal buffering and copper-dependent oxidoreductases localized outside the Golgi complex . Several lines of evidence further support a model of a multicompartment regulation of copper homeostasis . Genetic evidence indicates that the expression of CTR1 is modulated by the mitochondrial metallochaperone SCO1 ( Hlynialuk et al . , 2015 ) , down-regulation of mitochondrial components including complex I subunits ( NDUFB4 , NDUFA1 ) and the mitochondrial transporter SLC25A41 lead to global cellular changes in copper ( Malinouski et al . , 2014 ) , and the elimination of ATP7A rescues mitochondrial metallochaperone mutant phenotypes ( Malinouski et al . , 2014; Leary et al . , 2013 ) . Along with the neurodegeneration characteristic of Menkes disease , disruptions in copper homeostasis have been implicated in several other prevalent neurodegenerative diseases , including Parkinson’s and Alzheimer’s disease ( Greenough et al . , 2016; Davies et al . , 2016 ) . There is evidence that therapeutic modulation of copper levels may be effective in alleviating symptoms of or delaying the onset of Parkinson’s and Alzheimer’s disease , yet there are gaps in our understanding of the regulation of copper homeostasis , particularly in the brain ( Sparks and Schreurs , 2003; Brewer , 2015; Rose et al . , 2011 ) . Polymorphisms in genes encoding the COG complex subunits COG4 , 6 , and 8 associate with neurodegenerative and neurobehavioral phenotypes in humans ( Li et al . , 2008; Scharf et al . , 2013; Kendler et al . , 2011; Li et al . , 2012 ) . Our bioinformatic studies also indicate that the COG2 subunit is one of 42 gene products present in the ATP7A interactome associated with tauopathies and Alzheimer’s disease; one of fourteen genes products associated with Parkinson’s disease; and one of the 49 gene products in the ATP7A interactome associated with neurocognitive disorders , such as dementia ( Figure 3 ) . Deletions of COG complex subunits lead to type II congenital disorders of glycosylation ( CDG II ) and are strongly associated with cerebral atrophy , developmental delay , hypotonia , ataxia and epilepsy ( Foulquier et al . , 2006; Kodera et al . , 2015; Reynders et al . , 2009; Paesold-Burda et al . , 2009; Fung et al . , 2012; Rymen et al . , 2012; Lübbehusen et al . , 2010; Huybrechts et al . , 2012; Shaheen et al . , 2013; Wu et al . , 2004; Morava et al . , 2007; Zeevaert et al . , 2009; Foulquier et al . , 2007 ) . Since COG mutations drastically decreased the cellular copper levels ( Figure 7A ) , it is possible that COG neurological phenotypes reflect in part decreased neuronal copper much like is the case in Menkes disease . Mutations in cog1 prevent toxicity of dietary copper in Drosophila ( Figure 10 ) supporting this concept of decrease copper content in COG deficient neurons . Similarities between Menkes and congenital disorders of glycosylation are suggested by neurological phenotypic overlap between these disorders . Common phenotypes include developmental delay , seizures , hypotonia and cerebral atrophy ( Kaler , 2011; Menkes , 1999; Menkes et al . , 1962; Kennerson et al . , 2010; Tümer , 2013; Kaler et al . , 1994 ) . We genetically tested whether phenotypes caused by ATP7A dosage increase could be modulated by genetic defects in COG using the Drosophila developing neuromuscular synapse and adult dopaminergic neuron ( Figures 9–10 ) . Phenotypes due to overexpression of ATP7A in developing synapses and adult neurons could be reverted to wild type by adding a mutation in cog1 . Inspired by the effects of COG deficiency in human ATP7A ( Figure 6A–B ) , we interpret this Drosophila rescue results as a reduction in the levels of overexpressed ATP7A in fly neurons due to a ATP7A down-regulation effect of the cog1 mutant allele . This interpretation assumes cell autonomous effects of the cog1 mutation on neuronal overexpressed ATP7A , and it does not consider possible non-cell autonomous contributions of the cog1 genomic defect . Our findings suggest that part of the neurological phenotypes in COG genetic defects may be due to impaired copper content and metabolism as we report here . While we initially focused on the role of the COG complex in ATP7A trafficking , it is clear that copper-related phenotypes downstream of COG extend beyond cuproenzymes that traverse the Golgi complex . Our strategy to identify an ATP7A interactome has revealed novel ATP7A interacting partners and may be a fruitful way to further explore networks related to other copper regulatory proteins . Our finding that the COG complex , a Golgi localized tether is implicated in mitochondrial copper homeostasis indicates that essential mineral homeostasis , such as copper , results from coordinated multicompartment metabolite sensing and response mechanisms . SH-SY5Y ( ATCC Cat# CRL-2266 , RRID:CVCL_0019 ) and HEK293T ( ATCC Cat# CRL-3216 , RRID:CVCL_0063 ) cells were cultured in Dulbecco’s modified Eagle’s medium ( DMEM ) supplemented with 10% fetal bovine serum ( FBS ) and 100 μg/ml penicillin and streptomycin ( Hyclone ) at 37°C in 10% CO2 . HEK293T cells deficient for COG subunits were generated as described in ( Blackburn and Lupashin , 2016 ) . Two lines of Menkes deficient fibroblasts were used: one in conjunction with a rescue line expressing recombinant ATP7A ( described in La Fontaine et al . , 1998 ) and the other in conjunction with a familial control ( Coriell GM01981 and GM01983 ) . The former was cultured in DMEM supplemented with 10% FBS and 100 μg/ml penicillin and streptomycin at 37°C in 5% CO2 , while the latter were cultured in minimum essential media ( MEM ) ( Thermo Fisher 11095080 ) supplemented with 15% FBS and 100 μg/ml penicillin and streptomycin at 37°C in 5% CO2 . Antibodies and primers can be found in Supplementary file 1 . Cell were tested for mycoplasma infection using the MycoAlert Mycoplasma Detection Kit from Lonza . None of the cell lines used in this study belong to the Database of Cross-Contaminated or Misidentified Cell Lines defined by the International Cell Line Authentication Committee . To assess interactions of ATP7A in the presence and absence of copper , we performed cross-linking in intact cells with dithiobis ( succinimidylpropionate ) ( DSP ) followed by immunoprecipitation as previously described but with the following modifications ( Gokhale et al . , 2015a ) . Briefly , SH-SY5Y neuroblastoma cells or ATP7A−/− and ATP7AR/R human fibroblasts were washed with ice-cold PBS with MgCl2 and CaCl2 ( PBS/Mg/Ca ) . 200 μM Copper chloride or 400 μM BCS diluted in PBS/Mg/Ca buffer was then added to the respective plates and they were placed back in the 37°C incubator for 2 hr . The plates were then placed on ice , rinsed twice with PBS/Mg/Ca , and incubated with 10 mM DSP ( Thermo Scientific 22585 ) , diluted in PBS for 2 hr on ice . Tris , pH 7 . 4 , was added to the cells for 15 min to quench the DSP reaction . The cells were then rinsed twice with PBS and lysed in buffer A ( 150 mM NaCl , 10 mM HEPES , 1 mM EGTA , and 0 . 1 mM MgCl2 , pH 7 . 4 ) with 0 . 5% Triton X-100 and Complete anti-protease ( catalog #11245200 , Roche ) , followed by incubation for 30 min on ice . Cells were scraped from the dish , and cell homogenates were centrifuged at 16 , 100 × g for 10 min . The clarified supernatant was recovered , and at least 500 μg of protein extract was applied to 30 μl Dynal magnetic beads ( catalog #110 . 31 , Invitrogen ) coated with 5 μl ATP7A antibody ( NeuroMab Cat No . 75–142 , RRID:AB_10672736 ) , and incubated for 2 hr at 4°C . In some cases , immunoprecipitations were done in the presence of the antigenic ATP7A peptide as a control . The peptide ( sequence VSLEEKNATIIYDPKLQTPK , custom made by Biosynthesis ) was prepared in 10 mM MOPS and used at 22 μM . The beads were then washed 4–6 times with buffer A with 0 . 1% Triton X-100 . Proteins were eluted from the beads with sample buffer . Samples were resolved by SDS-PAGE and contents analyzed by immunoblot or silver stain . In the case of the large-scale proteomic analysis , proteins eluted from the beads were combined and concentrated by TCA precipitation . Cells were washed three times with cold phosphate-based saline and detached by gentle squirting . Cell pellets were heated in a continuous cycle from 35°C to 95°C for 30 min with 2% nitric acid/H2O2 . Copper and zinc were analyzed using inductively-coupled plasma mass spectrometry . Samples were quantified with an 8-point calibration using indium as an internal standard . The limits of detection were 1 ng/sample for copper and 0 . 1 ng/sample for zinc . Gene list to disease associations were performed with GDA algorithm ( http://gda . cs . tufts . edu/; Park et al . , 2014 ) . We performed gene ontology analysis with ENRICH ( http://amp . pharm . mssm . edu/Enrichr/ ) ( Chen et al . , 2013 ) , and Database Annotation , Visualization and Integrated Discovery ( DAVID , https://david . ncifcrf . gov ) ( Huang et al . , 2007 , 2009 ) . Cytoscape with Enrichment Map plugin for visualizing DAVID outputs was used in order to depict integrations between GO terms associated with the ATP7A interactome as described ( Shannon et al . , 2003; Merico et al . , 2010 ) . Charts were narrowed down in order to simplify Cytoscape representations by eliminating broad GO terms . All Bioinformatic analyses are presented in Supplementary file 3A—B Cells were collected and seeded in a 96-wells plate at a density of 10 × 103 cells/well in DMEM +10% FBS+100 U/ml penicillin , 100 μg/ml streptomycin and incubated overnight . On day 2 , cells were treated with serial dilutions of the appropriate drug ( 0 . 4–100 μM cisplatin , 3–300 μM copper chloride , or 1 . 5–200 nM disulfiram ) . After incubation at 37°C for either 24 or 72 hr depending on the experiment , 20 μl 5 mg/ml ( 3- ( 4 , 5-Dimethylthiazol-2-yl ) -2 , 5-Diphenyltetrazolium Bromide ) ( MTT ) ( Life Technologies M6496 ) was added to each well , and the plates were incubated for an additional 3 . 5 hr at 37°C . MTT was aspirated , 150 μl DMSO was added and cells were agitated on an orbital shaker for 15 min . The absorbance was read at 595 nm using a microplate reader . Each condition was carried out in quadruplicate , and MTT absorbance was expressed as percentage absorbance of untreated cells . Cells were plated and treated with the appropriate drug as described for MTT experiments above . After incubation with drug , cells were washed once with PBS-Ca-Mg . Cells were then fixed for five minutes in 65% MeOH , followed by a fifteen minute fixation with 100% MeOH . The methanol was removed and wells were allowed to dry completely , after which 100 μl 0 . 1% w/v crystal violet in H2O was added to each well for five minutes . Plate was washed 3–5 times with H2O , and crystal violet was solubilized with 2% sodium deoxycholate for 10 min . The absorbance was read at 595 nm using a microplate reader . Each condition was carried out in quadruplicate , and crystal violet absorbance was expressed as percentage absorbance of untreated cells ( Feoktistova et al . , 2016 ) . Plates of ~75% confluent HEK293T cells were incubated for 2 hr at 37°C in the absence or presence of 200 μM copper chloride and then moved to an ice bath . Plates were then washed two times with ice-cold PBS-Ca-Mg ( 0 . 1 mM CaCl2 and 1 mM MgCl2 in PBS ) . All solutions used were ice-cold when applied to cells , and cells were maintained on an ice bath throughout . A biotin labeling solution of 0 . 5 mg/ml Sulfo-NHS-Biotin ( Thermo Scientific 21217 ) in PBS-Ca-Mg was applied for 15 min , aspirated , and fresh labeling solution was applied for an additional 15 min . The labeling solution was then aspirated and cells were washed three times with 1 mg/mL glycine in PBS-Ca-Mg followed by one wash with plain PBS . Cells were collected from plates and incubated for 30 min in buffer A with 0 . 5% Triton X-100 , supplemented with Complete antiprotease . Lysates were spun at 16 , 100 × g for 15 min . The supernatant was recovered and diluted to 1 mg/ml . A small volume ( 50 μl of NeutrAvidin-coated agarose bead slurry ( Thermo Scientific 29200 ) was prewashed two times in buffer A with 0 . 1% Triton X-100 , and 500 μg of cell lysate was incubated with the beads for 2 hr with end-over-end rotation at 4°C . Beads were then washed five times in buffer A with 0 . 1% Triton X-100 for 5 min with end-over-end rotation at 4°C . Proteins were eluted from the beads by boiling in Laemmli sample buffer at 75°C for 5 min , and samples were analyzed by SDS–PAGE followed by immunoblot . qRT-PCR Quantiative rt-PCR was performed as described in Gokhale et al . ( 2015a ) and Larimore et al . ( 2013 ) ) . RNA from HEK293T cells was TRIzol-extracted ( Invitrogen ) , and isolated RNA was reverse transcribed into cDNA using SuperScript III first strand synthesis ( Invitrogen ) . PCR amplifications were performed on a LightCycler480 real time plate reader using LightCycler 480 SYBR Green reagents ( Roche ) . The following strains were obtained from the Bloomington Drosophila Stock Center , Bloomington , Illinois: w[1118] . w[1118]; PBac{w[+mC]=RB}CG4848[e02840]/TM6B , Tb[1]; w[1118]; w[1118]; P{w[+mC]=Ddc-GAL4 . L}Lmpt[4 . 36] and C155 GAL4 animals were from Bloomington ( P{GawB}elavC155 , Fly Base ID FBti0002575 ) . w[1118]; UAS-ATP7-wt was a gift of Richard Burke , Monash University , Australia . Flies were reared on standard Molasses Food ( Genesee Scientific ) at 25C in 12 hr:12 hr light:dark cycle . Copper feeding toxicity experiments were performed as described using 5% glucose as control feed or 5% glucose supplemented with 1 mM CuCl2 ( Gokhale et al . , 2015a ) . Larval dissections , immunohistochemistry , and confocal microscopy were performed as described previously ( Gokhale et al . , 2016 , 2015b; Mullin et al . , 2015 ) . Wandering third-instar female larvae were dissected in normal HL3 , fixed in 4% paraformaldehyde for 1 hr , and stained with HRP–FITC conjugated antibody for 2 hr at room temperature ( 1:500 ) . An inverted 510 Zeiss LSM microscope was used for confocal imaging of synapses at muscle 6/7 on either the second or third segment . Bouton counts and branch length were calculated using FIJI ( Schindelin et al . , 2012 ) with experimenters blind to the genotype of the animal . Experimental conditions were compared using Synergy Kaleida-Graph , version 4 . 1 . 3 ( Reading , PA ) or Aabel NG2 v5 × 64 by Gigawiz as specified in each figure .	People need a source of copper in their diet because this nutrient is used to produce the pigment in hair and skin , the connective tissue in tendons and ligaments , and some of the small molecules that allow brain cells to communicate . There is an ideal range of copper that allows cells to carry out these processes . Both too much and too little copper can have negative effects on health , particularly related to how the brain works . Cells contain multiple proteins that bind to copper and transport it wherever it is needed . People with mutations that mean they lack one of these copper transporters , ATP7A , often have serious damage to their nervous system that cannot be explained by the current understanding of how this protein works . Comstra et al . set out to establish a comprehensive list of proteins that interact with ATP7A to better understand how this transporter works and how it is regulated . The search revealed that ATP7A interacts with hundreds of proteins present in different compartments within cells , many of which had not previously been associated with balancing copper levels in cells and the body . Like ATP7A , many of these proteins ( or the protein complexes that contain them ) are known to affect nerves and brain activity when they are mutated . Next , Comstra et al . engineered human cells grown in the laboratory to lack one of the protein complexes that interacts with ATP7A , the COG complex . Cells without this protein complex had 50% less ATP7A than normal human cells and very low levels of copper too . These mutant cells also had problems generating the energy that they need , because the structures in cells that provide them with energy – the mitochondria – were impaired; adding copper to the cells improved the activity of their mitochondria . Mutations in the COG complex cause the brain to develop abnormally , and the finding that deleting the COG complex from cells causes copper deficiency now helps to explain why . Further characterization of the proteins that interact with ATP7A and the COG complex will contribute to our understanding of how cells regulate copper and how copper levels affect the brain .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "neuroscience" ]	2017	The interactome of the copper transporter ATP7A belongs to a network of neurodevelopmental and neurodegeneration factors
The importance of genome replication has inspired detailed crystallographic studies of enzymatic DNA/RNA polymerization . In contrast , the mechanism of nonenzymatic polymerization is less well understood , despite its critical role in the origin of life . Here we report the direct observation of nonenzymatic RNA primer extension through time-resolved crystallography . We soaked crystals of an RNA primer-template-dGMP complex with guanosine-5′-phosphoro-2-aminoimidazolide for increasing times . At early times we see the activated ribonucleotides bound to the template , followed by formation of the imidazolium-bridged dinucleotide intermediate . At later times , we see a new phosphodiester bond forming between the primer and the incoming nucleotide . The intermediate is pre-organized because of the constraints of base-pairing with the template and hydrogen bonding between the imidazole amino group and both flanking phosphates . Our results provide atomic-resolution insight into the mechanism of nonenzymatic primer extension , and set the stage for further structural dissection and optimization of the RNA copying process . Many lines of evidence point to a central role for RNA in the emergence of life on Earth ( Sutherland , 2016; Wachowius et al . , 2017 ) . As part of the RNA world hypothesis , the earliest forms of life must have replicated their simple RNA genomes without the aid of complex enzymes ( Orgel , 2004 ) . A key part of this process would have been the nonenzymatic template directed polymerization of activated RNA nucleotides . The resulting complementary strand could then act as a template to produce a copy of the original sequence . Previous studies have suggested that the local conformation at the reaction site greatly impacts the nonenzymatic RNA polymerization reaction . For example , a wide range of nucleic acid backbones , such as peptide ( Schmidt et al . , 1997 ) or threose nucleic acids ( Heuberger and Switzer , 2006 ) , can template the polymerization of activated RNA monomers , but the fastest reactions occur on A-form structured RNA or LNA templates ( Zielinski et al . , 2000; Schrum et al . , 2009 ) . These results indicate that the A-form conformation of the template provides a structure conducive for the polymerization reaction . Downstream helper oligomers , commonly utilized to improve the yield of primer extension reactions , are also thought to improve the reaction in part by stabilizing the A-form structure of an otherwise disordered ssRNA template ( Vogel et al . , 2005; Prywes et al . , 2016 ) . Structural aspects of the template near the reaction site may underlie many of the characteristics of nonenzymatic RNA polymerization , such as the rate and fidelity of polymerization ( Leu et al . , 2013 ) . A better atomic-level understanding of the structural factors that promote nonenzymatic RNA polymerization could help to optimize the reaction for generalized copying of sequences , thus leading to a better understanding of the conditions required for RNA replication during the origin of life . Crystallography is a powerful tool for gaining atomic-level understanding of the mechanism of nonenzymatic RNA polymerization . However , the widely used nucleoside-5′-phosphoro-imidazolide substrates are too labile to use directly in crystallographic studies . We have instead utilized stable phosphonate analogs , co-crystallized with a primer template duplex , to help model the structure of the reaction site ( Zhang et al . , 2016; Zhang et al . , 2018 ) . These studies indicate that pseudo-activated guanosine monomers bind the template through both Watson-Crick and non-canonical modes of hydrogen bonding , with the phosphonate-leaving group mimic highly disordered . In contrast , the symmetrical 5′−5′ triphosphate-linked dinucleotide GpppG , which is an analogue of the imidazolium-bridged dinucleotide intermediate , was observed in a crystal structure to bind the template through two consecutive Watson-Crick base-pairs , with the 3′-hydroxyl of the primer positioned for in-line attack ( Zhang et al . , 2017 ) . This observation suggested that the high reactivity of the imidazolium-bridged dinucleotide intermediate could be due to a favorable preorganized geometry in the template bound state . Consistent with this model , the superiority of the newly identified nucleotide-5′-phosphoro-2-aminoimidazolide ( 2-AIpN ) monomers ( Li et al . , 2017 ) , is thought to be due to the greater stability of the corresponding imidazolium-bridged dinucleotide intermediate ( Np-AI-pN ) ( Walton and Szostak , 2017 ) . In addition , our recent study of the mechanism by which downstream helper oligonucleotides enhance the rate of nonenzymatic primer extension ( Zhang et al . , 2018 ) showed that the helper oligomers pre-organize the reaction site and decrease the distance between the 3′-hydroxyl of the primer and the phosphate of the adjacent nucleotide . To fully understand the mechanism of nonenzymatic template-directed primer extension , static 3D structures of RNA primer-template complexes co-crystallized with substrate and intermediate analogues , or even with the actual substrates and intermediates , are not sufficient to depict the dynamic progress of the reaction over time . A molecular movie of actual substrates reacting on the template , captured through the methods of crystallography , would provide greater insight . The approach of time-resolved X-ray crystallography has been greatly developed to study the dynamic processes of macromolecules and small molecules , by utilizing X-ray probe pulses in the pump-probe scheme ( Ki et al . , 2017 ) . The technique of time-resolved serial femtosecond crystallography , as well as the conventional Laue-type polychromatic time-resolved crystallography ( Ren et al . , 1999 ) , has been widely applied to probe the structural transitions of proteins in both the solid crystal phase and in solution . Another category of time-resolved X-ray crystallography depends on the diffusion of small molecules into protein crystals . For the study of enzymes , it has been found that catalytic activity is sometimes preserved in crystallized proteins , which allows a structural analysis of the mechanism through time-resolved X-ray crystallography ( Hajdu et al . , 2000 ) . In such experiments , the catalytic process can be triggered by diffusion-based reaction initiation techniques . At various time points after the initiation of the reaction , the crystal is placed in liquid nitrogen to freeze the reaction . This technique has provided deep insight into the polymerization reaction catalyzed by DNA polymerase η . In addition to capturing intermediate steps in phosphodiester bond formation , these structural studies provided evidence for a previously unknown third Mg2+ ion that catalyzes the rate-limiting step of the reaction ( Nakamura et al . , 2012; Gao and Yang , 2016 ) . Inspired by the above studies , we conducted a time-resolved analysis of nonenzymatic RNA polymerization by soaking chemically activated substrates into crystals containing a primer-template duplex . Here , we present the crystal structures of the template-bound substrate 2-AIpG ( Figure 1A ) and the reaction intermediate Gp-AI-pG ( Figure 1B ) . These structures corroborate our previous studies using stable analogs and uncover new features of the template-bound substrates . By obtaining crystal structures at various times after the initiation of the reaction , we were able to capture a sequence of molecular ‘snapshots’ that follow the progress of the nonenzymatic RNA polymerization reaction . This reaction sequence consists of monomers binding to the template in multiple conformations , followed by formation of the imidazolium-bridged dinucleotide intermediate , and finally new phosphodiester bond formation between the primer and the adjacent nucleotide of the intermediate . These experiments strongly suggest that the observed conformation of the template-bound dinucleotide intermediate Gp-AI-pG is pre-organized so as to favor the primer extension reaction . The approach described herein should be broadly useful for answering additional mechanistic questions regarding nonenzymatic RNA polymerization . In order to follow nonenzymatic primer extension by crystallography , we first prepared crystals of an RNA primer-template complex in which unactivated monomers were bound to the template region . These crystals were used in crystal soaking experiments to exchange the unactivated monomers with activated monomers or the reaction intermediate . To facilitate comparison with previously obtained structures , we used the same partially self-complementary RNA 14-mer oligonucleotide that we used in our structural studies of the stable substrate analogues guanosine 5′- ( 3-methyl-1H-pyrazol-4-yl ) phosphonate ( PZG ) and guanosine 5′- ( 4-methylimidazolyl ) phosphonate ( ICG ) , and the stable intermediate analogue GpppG ( Zhang et al . , 2016; Zhang et al . , 2017; Zhang et al . , 2018 ) . Our attempts to crystallize a native RNA-monomer complex led to a highly disordered flanking template regions with no evidence for bound monomers . We therefore chose to use the oligoribonucleotide 5′-CCCGACUUAAGUCG-3′ which contains four locked nucleic acid ( LNA ) residues , denoted by italics . The LNA modification locks the sugar into the C3′-endo A-form conformation characteristic of a canonical RNA duplex . The two-nucleotide template overhang is composed of 5-methyl cytidine LNA to rigidify the backbone and facilitate crystallization . This duplex was co-crystallized with 2′-deoxyguanosine-5′-monophosphate ( dGMP ) , to bind the template before soak-in experiments ( Figure 1C ) . The use of dGMP allows unambiguous observation of the replacement of the initially bound deoxynucleotides with activated ribonucleotides during crystal soaking experiments , based upon the appearance of the electron density of the 2′-hydroxyl . Since the pyrazole and imidazole groups of monomer analogues are often disordered , the imidazole groups of the 2-AIpG would not be a reliable indicator of molecular exchange . Several other modified monomers , including 8-bromoguanosine , guanosine-3′-monophosphate and guanosine 3′ , 5′-cyclic monophosphate , were tested for crystallization with the RNA duplex without success , possibly due to steric perturbation of molecular packing . At both ends of the crystallized duplex , the two dGMP monomers bound to the -CC template in a well defined conformation ( Figure 2A ) , exclusively through Watson-Crick base-pairing ( Figure 2B ) . This result contrasts with our previous study of the RNA monomer analogues PZG and ICG , which displayed a variety of binding modes in addition to Watson-Crick base-pairing ( Zhang et al . , 2016; Zhang et al . , 2018 ) . The sugar and phosphate of the first bound dGMP ( in the +1 position adjacent to the 3′ end of the primer ) are well ordered , with the sugar in a C3′-endo A-form conformation . At the +2 position , the electron density clearly indicates the sugar is in the C3′-exo B-form conformation , but the phosphate is disordered . Previous crystal structures of DNA-RNA hybrid duplexes have shown DNA sugar puckers in both C3′-endo and C3′-exo conformations ( Horton and Finzel , 1996 ) . In contrast , our previous measurement using transferred nuclear Overhauser effect spectroscopy ( TrNOESY ) indicated that DNA monomers bound to an RNA template maintained a C2′-endo conformation ( Zhang et al . , 2012 ) . This difference may result from the LNA template used for crystallization , sequence differences or crystal packing forces . In the crystal structure of the RNA duplex with dGMP , we also clearly observed electron density near the N7 of the first template-bound dGMP that appears to represent a Mg2+ ion based on its octahedrally coordinated waters ( Figure 2C ) ; Mg2+ was present in the crystallization buffer . Two water molecules coordinated to the presumed Mg2+ are located within hydrogen bonding distance of N7 and O6 of the primer guanosine ( distances 2 . 8 Å and 2 . 6 Å ) . A third water molecule coordinated to the presumed Mg2+ is within hydrogen bonding distance of O6 of the first bound monomer , and is also hydrogen bonded to a fourth water molecule that appears to be hydrogen bonded to N7 of the second bound monomer ( distances 3 . 0 Å , 2 . 7 Å and 2 . 8 Å ) . These proposed Mg2+-mediated electrostatic and hydrogen-bonding interactions bridge the two bound monomers and the primer , and thus potentially stabilize the template-bound structure of dGMP . To confirm the proposed Mg2+-guanosine interaction , we replaced Mg2+ with Sr2+ in our crystallization buffer , to increase the electron density of the bound metal cation . We obtained crystals of the same RNA duplex in 20 mM Sr2+ with the RNA nucleotide guanosine-5′-monophosphate bound to the template region , with the same overall structure as the other complexes . In this structure , much stronger electron density was present near N7 and O6 of the first template-bound GMP , indicating the presence of Sr2+ ( Figure 2—figure supplement 1 ) . This Sr2+ ion binds in a similar manner as the proposed Mg2+ ion observed in the structure of template-bound dGMP , supporting our identification of the interaction between N7 of dGMP in the +1 position with Mg2+ . Similar water-mediated metal-guanine interactions have previously been structurally identified in the deep major groove of A-form nucleic acid duplexes ( Robinson et al . , 2000; Ennifar et al . , 2003 ) . Importantly , we observed a large channel in between the slip-stacked RNA duplexes that could allow activated substrates to reach the template by diffusion ( Figure 2D ) . As in our previous studies , the self-complementary RNA duplex complexed with dGMP crystallized with hexagonal symmetry ( space group P3121 ) . The individual RNA-monomer double helices are slip-stacked with one another end-to-end , and groups of three adjacent duplexes form a triangular prism-like structure ( Figure 2D , side view ) . Groups of three of these triangular prisms are in turn arranged in a triangular lattice , with a large central channel with a diameter of approximately 30 Å ( Figure 2D , top view ) . This channel should allow mononucleotides or dinucleotides to diffuse into the crystal and bind the template in place of dGMP . Having obtained crystals of the RNA-dGMP complex , we sought to replace the bound dGMP residues with purified imidazolium-bridged diguanosine intermediate ( Gp-AI-pG ) . We soaked the crystals in buffer containing Gp-AI-pG for various times to allow the intermediate to diffuse into the crystal . No Mg2+ was included in either the crystallization or soak buffers , to improve the solubility and stability of the Gp-AI-pG . The Gp-AI-pG molecule was observed by crystal structure determination to bind the template in place of dGMP , and optimal structures were obtained after 4 hr of soaking the crystals in the exchange buffer . In the conditions used for crystal soaking , Gp-AI-pG decays with a half-life of 10 hr as observed by 31P NMR , indicating that the majority of Gp-AI-pG remains intact during this experiment ( Figure 3—figure supplement 1 ) . In the crystal structures , Gp-AI-pG binds the -CC template at both ends of the RNA duplex exclusively through Watson-Crick base-pairing ( Figure 3A ) , with hydrogen bond distances of 2 . 8 to 3 . 3 Å . Similar to our previous analysis of the analog GpppG , the Gp-AI-pG intermediate is well-ordered and displays only the C3′-endo sugar conformation . The 4 . 6 Å distance between the 3′-hydroxyl and the phosphate of the nucleotide in the +1 position is shorter than the corresponding 6 to 6 . 5 Å distance for the 2′-hydroxyl , consistent with the observed 3′ regioselectivity of polymerization using 2-methyl and 2-aminoimidazole activated monomers ( Inoue and Orgel , 1981; Giurgiu et al . , 2017 ) . The two guanosine-5′-phosphate residues of Gp-AI-pG are linked by the 2-aminoimidazolium bridge . The electron density corresponding to the phosphate-aminoimidazole-phosphate bridge suggests that its structure is stabilized by hydrogen bonds between the 2-amino group of the imidazolium moiety and the non-bridging oxygen atoms of the flanking phosphates ( Figure 3B–C ) . Unfortunately , the observed electron density does not clearly indicate whether the 2-amino group of the imidazolium bridge is pointing toward the major or minor groove of the duplex . The Gp-AI-pG intermediate is modeled in the two different orientations in Figure 3B–C . It may be that the two orientations are randomly distributed throughout the crystal . The orientation of the 2-aminoimidazolium group affects the distances and angle of attack between the primer 2′- or 3′-hydroxyl and the P-N bond to be broken during phosphodiester bond formation . For the 3′-hydroxyl , the distance to the phosphorous atom is 4 . 6 Å in both orientations , but the angle of attack changes from 170° when the 2-amino points toward the minor groove , to 132° when it points towards the major groove . For the 2′-hydroxyl , the angle decreases from 148° to 129° but the distance increases from 6 . 0 to 6 . 5 Å when the 2-amino group points towards the major groove . These results suggest that rotation of the 2-aminoimidazolium group between the two phosphates may affect the rate and regioselectivity of nonenzymatic RNA polymerization . A comparison between template-bound dGMP and Gp-AI-pG reveals similar overall structures ( Table 1 ) . A key difference is the decreased distance between the 3′-hydroxyl of the primer and the phosphate in the +1 position , from 6 . 5 Å for bound dGMP to 4 . 6 Å for template-bound Gp-AI-pG . This decreased distance is likely due to the constraint imposed by the imidazolium-bridge of Gp-AI-pG . No monovalent metal ions or water molecules are observed close to the N7 positions of Gp-AI-pG or the 3′ G of the primer . Encouraged by our success at soaking the reactive Gp-AI-pG intermediate into preformed crystals of the RNA duplex , we asked whether soaking the activated monomer 2-AIpG into the crystal in the presence of Mg2+ might lead to formation of the intermediate , possibly followed by reaction with the primer in the crystal . We therefore grew crystals of the RNA duplex in buffer at pH 7 . 0 containing dGMP and 20 mM Mg2+ . The RNA-dGMP crystals were then transferred to a new drop of the same pH 7 . 0 crystallization buffer , containing 20 mM 2-AIpG and 20 mM Mg2+ . Crystal soaking was terminated by equilibrating the crystal in cryo-protectant ( 35% 2-methyl-2 , 4-pentanediol , MPD ) for 2 min followed by storage in liquid nitrogen prior to analysis . In the crystallization buffer , only 8% of the 2-AIpG decayed in 3 hr as determined by 31P NMR , indicating that the majority of the monomer remains activated during these experiments ( Figure 4—figure supplement 1 ) . After 5 min of soaking in exchange buffer containing 2-AIpG , the dGMP monomers co-crystallized with the RNA duplex had already been replaced by 2-AIpG monomers ( Figure 4A ) . The rapid diffusion and exchange of the monomer 2-AIpG relative to the dimer Gp-AI-pG is clear from the appearance of the electron density of the 2′-hydroxyl in these structures , which cannot come from dGMP . The faster diffusion of the 2-AIpG monomer into the crystal is likely due to its smaller size . At this early time point , the 2-AIpG monomer at the +1 position is bound through Watson-Crick base pairing , but the monomer at the +2 position is bound through a non-canonical interaction . This atypical hydrogen bonding pattern has been observed in our previous studies of analogues for the guanosine-5′-phosphoro-2-methylimidazole monomer ( Zhang et al . , 2016 ) . However , the 2-AIpG monomers in both the +1 and +2 positions are Watson-Crick base paired in the crystal structure obtained after 15 min of soaking the crystal ( Figure 4B ) . These results further suggest that a variety of interactions are possible between monomers and the template during primer extension reactions . It is interesting to note that the distance between the primer 3′-hydroxyl and the 5′-phosphate of the incoming monomer increases when both monomers are bound to the template in Watson-Crick conformation ( Table 1 ) . In addition , we again observe electron density , likely corresponding to Mg2+ and coordinated water molecules , that bridges from N7 of the 3′-G of the primer to N7 of the 2-AIpG monomer in the +1 position . Structures obtained after continued incubation of the crystals with activated monomer 2-AIpG reveal the gradual appearance of the imidazolium-bridged dinucleotide Gp-AI-pG . While the 2-aminoimidazole groups are largely disordered and the imidazolium bridge is not observed in the structures obtained after 5 and 15 min of soaking , new density corresponding to formation of the imidazolium bridge clearly emerges after 30 min to 1 hr of soaking ( Figure 4C–D ) . In addition , the ~6 . 2 Å ( 15 min ) distance between the two phosphorus atoms of separate monomers decreases to 5 . 2 Å , very close to the 5 . 1 Å distance observed for the template-bound structure of Gp-AI-pG ( Table 1 ) . Similar to the structure obtained by soak-in of Gp-AI-pG ( Figure 3 ) , these structures show the dinucleotide intermediate bound to the template through two Watson-Crick base pairs and an imidazolium bridge that is ordered through hydrogen bonds between the 2-amino group and the non-bridging oxygen atoms of the phosphates . These results suggest that the 2-AIpG monomers are reacting inside the crystal to form the dinucleotide intermediate Gp-AI-pG . When the crystal was incubated in a solution of 2-AIpG for times over 1 hr , we began to observe the appearance of a new phosphodiester bond between the primer and the adjacent G residue ( Figure 4E–G ) . At times from 1 to 3 hr , new electron density appears between the primer 3′-hydroxyl and the 5′-phosphate of the monomer in the +1 position . The structures at 1 . 5 hr and 2 hr represent the average of a distribution of chemical states , and are refined as a mixture of template-bound Gp-AI-pG and the primer +1 extension product with a 2-AIpG monomer in the +2 binding site of the template . The ratio of these two states changes over time as phosphodiester bond formation proceeds to near completion by 3 hr . The structure at 3 hr clearly indicates a new well-ordered phosphodiester bond , with a 2-AIpG monomer bound to the template in the +2 position through Watson-Crick base pairing . The distance between the N7 atoms of the primer guanosine and the incoming guanosine remains the same throughout the reaction in the crystal ( Table 1 ) , suggesting that the base pairs and base stacking features of the RNA-substrate duplex are not perturbed during the course of the reaction . Only local conformational changes associated with formation and then loss of the imidazolium-bridge and formation of the new phosphodiester bond are observed . These results are consistent with a recent analysis indicating that the predominant mechanism of primer extension by 2-aminoimidazole activated monomers is through the imidazolium-bridged dinucleotide intermediate ( Walton and Szostak , 2017 ) . However , the observed decline in density for the imidazolium bridge in our current structures may also be due in part to hydrolysis of Gp-AI-pG . The mechanism of template-directed nonenzymatic RNA polymerization has recently been shown to involve an imidazolium-bridged dinucleotide intermediate ( Walton and Szostak , 2016; Kervio et al . , 2016 , also noted the high reactivity of this molecule ) . A major question is why the primer reacts so much more rapidly with the template-bound intermediate than with template-bound monomers . Our previous investigation of GpppG , a stable analog of the Gp-AI-pG intermediate , confirmed that a 5′ , 5′ linked dinucleotide could bind the template through two Watson-Crick base pairs and suggested that the true intermediate would be pre-organized for phosphodiester bond formation ( Zhang et al . , 2017 ) . In contrast , the leaving group is highly disordered in our previously determined structures of stable phosphonate analogues of activated monomer substrates bound to the same template . Our current studies using the actual substrate 2-AIpG and intermediate Gp-AI-pG strongly support the model that structural pre-organization contributes to the reactivity of the covalent intermediate . The distance from the primer 3′-hydroxyl to the adjacent phosphate decreases from 6 . 5 Å for the monomer 2-AIpG , to 4 . 6 Å for the intermediate Gp-AI-pG . In addition , hydrogen bonding between the 2-amino group of the imidazolium bridge and the flanking non-bridging phosphate oxygen atoms likely helps to stabilize the conformation of the imidazolium bridge and pre-organize the template-bound intermediate for reaction with the primer . In contrast , the 2-aminoimidazole group of 2-AIpG was disordered in our structures , suggesting that the formation of a covalent link between two activated nucleotides is required to organize the correct conformation of the leaving group . We note that the hydrogen bonds of the 2-aminoimidazolium bridge would be absent for other imidazole groups such as 2-methylimidazole or 2-methylaminoimidazole . In addition , bulkier substitutions on the imidazole ring may distort the structure of the intermediate through steric clashes with the non-bridging oxygen atoms of the flanking phosphates . These considerations may help to explain the superiority of 2-aminoimidazole activation relative to other imidazole groups ( Li et al . , 2017 ) . Our observation of phosphodiester bond formation in the crystal strongly suggests the relevance of these structures for template-directed nonenzymatic RNA polymerization . However , the rate of the primer extension reaction is much slower in the crystal than in solution . In the crystal structure of template-bound Gp-AI-pG , the distance between the 3′-hydroxyl and the 5′-phosphate of the incoming nucleotide is quite long at ~4 . 6 Å . Clearly , significant conformational changes must occur before reaction with the primer , and these changes may be constrained by molecular packing . Such constraints may explain the slow rate of reaction in the crystal . In our recently determined structures in which GpppG is sandwiched between the primer and a downstream helper , the O3′-P distance is reduced , consistent with the enhanced reaction rate observed in the presence of a downstream helper oligonucleotide . If the Gp-AI-pG intermediate can be assembled in between the primer and a downstream oligonucleotide in crystal soaking experiments , and a decreased O3′-P distance is observed , primer extension in the crystal will require smaller conformational changes and may proceed more rapidly . The strategy of soaking reactive substrates into a pre-crystallized RNA duplex followed by time-resolved X-ray crystallography could also be applied to address other aspects of the reaction mechanism and outcome . For example , a mismatch at the +1 position results in decreased 3′−5′ regioselectivity , but the structural basis of this observation remains unclear ( Giurgiu et al . , 2017 ) . Many important aspects of phosphodiester bond formation between the primer and template-bound Gp-AI-pG remain unclear from our current series of time-resolved X-ray crystal structures . Foremost among these unanswered questions is the role of the catalytic metal ion in catalysis of the primer extension reaction . The primer extension reaction we observed in the crystal was Mg2+ catalyzed , since the reaction was observed only when the monomer 2-AIpG was soaked into the crystal in the presence of Mg2+ . In contrast , when Gp-AI-pG was soaked into the crystal without Mg2+ , no reaction was observed . In our structures , the only Mg2+ interactions observed involved bridging the N7 atoms of the primer and the monomer at the +1 position . These interactions between consecutive G nucleotides in the primer and template may help to properly orient the G monomers , and may help to explain the historically faster nonenzymatic template directed primer extension observed with G compared with the other nucleotides . However , Mg2+ has long been thought to have a direct catalytic role through activation of the primer 3′-hydroxyl nucleophile and possibly also through interactions with the monomer phosphate . Although we did not observe any bound Mg2+ near the site of phosphodiester formation , this is not surprising due to the very weak binding of the catalytic metal ion . Strategies to decrease the O3′-P distance and to further pre-organize the reaction center , for example through the use of a downstream helper oligonucleotide and/or the synthesis of transition state analogues , may allow visualization of the catalytic metal ion in future studies . The oligonucleotide 5′-CCCGACUUAAGUCG-3′ was synthesized in-house by standard solid-phase techniques using native and locked nucleoside phosphoramidites from Exiqon Inc . , followed by HPLC purification . C denotes LNA 5-methylcytidine residues and G denotes LNA guanosine residues , while A , C , G and U denote unmodified RNA residues . For the annealing step prior to crystallization , the RNA duplex ( 1 mM ) was mixed with an equal volume of dGMP ( 50 mM ) and heated to 80o C for 2 min before being slowly cooled to room temperature . All crystals were grown at 18o C . The Nucleic Acid Mini Screen Kit , Natrix High Throughput Kit and Index High Throughput Kit ( Hampton Research , Aliso Viejo , CA ) were used for screening crystallization conditions by the hanging drop or sitting drop vapor diffusion methods . Optimal crystals grew in a crystallization buffer of 5% v/v ( +/- ) −2-methyl-2 , 4-pentanediol , 20 mM sodium cacodylate pH 7 . 0 , 6 mM spermine tetrahydrochloride , 40 mM sodium chloride , and 20 mM magnesium chloride . Crystals used for soak-in of Gp-AI-pG were obtained under identical conditions except that magnesium chloride was omitted . For crystallization of the RNA duplex with GMP , 20 mM SrCl2 was added in place of MgCl2 . To soak 2-AIpG or Gp-AI-pG into crystals of the RNA duplex with dGMP , the preformed crystals were first stabilized in a drop containing crystallization buffer for 20 min . Then , the crystals were transferred to a drop containing crystallization buffer containing 20 mM Gp-AI-pG or 2-AIpG and incubated for various amounts of time to allow diffusion of the molecules into the crystal . As with the crystallization procedure , Mg2+ was omitted for soaking Gp-AI-pG into the crystal . After incubation with the activated substrates , the crystal was dipped into cryo-protectant ( 35% MPD ) for 2 min and then placed in liquid nitrogen . All crystal diffraction data were collected under a stream of nitrogen at 99 K . The data sets were obtained at the SIBYLS beam lines 821 and 822 at the Advanced Light Source , Lawrence Berkeley National Laboratory . The distance from the detector to the crystal was set between 200–300 mm , and the collecting wavelength was set to 1 Å . The crystals were exposed for 1 s per image with one degree oscillations , and 180 images were taken for each data set . These data were processed using HKL2000 ( Otwinowski and Minor , 1997 ) . All the structures were solved by molecular replacement ( McCoy et al . , 2007 ) . The RNA models were from our previously reported structure ( PDB code 5DHC ) . All the structures were refined using Refmac ( Murshudov et al . , 1997 ) . The refinement protocol included simulated annealing , refinement , restrained B-factor refinement , and bulk solvent correction . During refinement , the topologies and parameters for locked nucleic acids and for the ligand 2-AIpG or Gp-AI-pG were constructed and applied . After several cycles of refinement , the water and metal atoms were added . Data collection , phasing , and refinement statistics of the determined structures are listed in Supplementary file 1 . The structures at 1 . 5 and 2 hr were refined as mixtures of the intermediate state and product state . 2-AIpG and Gp-AI-pG were synthesized and purified as previously described ( Li et al . , 2017; Zhang et al . , 2017 ) . To characterize the stability of these substrates during the crystal soaking procedure , samples of 2-AIpG and Gp-AI-pG were prepared in crystallization buffer and monitored by 31P NMR . These samples were placed in a Shigemi tube with a co-axial insert containing D2O for locking the NMR signal . All NMR spectra were taken on a Varian INOVA 400 MHz NMR spectrometer at 25°C and analyzed by integration of peaks using MestReNova software . 31P NMR signals are referenced to internal trimethyl phosphate added after the experiments . The NMR spectra were shown in Figure 3—figure supplement 1 and Figure 4—figure supplement 1 .	Enzymes speed up chemical reactions that are essential to life . Most enzymes are proteins , but some are molecules of ribonucleic acid or RNA . Like DNA , RNA is made from a chain of building blocks called nucleotides . In modern organisms , protein-based enzymes build RNAs by linking nucleotides together , while the building blocks of proteins are linked by an RNA-based enzyme at the heart of a structure called a ribosome . The earliest life on Earth most likely relied only on RNA-based enzymes , but during the emergence of life , scientists believe that RNA molecules must have replicated spontaneously before dedicated RNA-based enzymes had evolved . How RNA could replicate without enzymes has been a puzzle for decades . Recently , scientists discovered a previously unsuspected chemical intermediate that forms during the process , and hypothesized that this molecule’s special structure is what enables the chemical reaction that adds new nucleotides to a growing strand of RNA . To test this hypothesis , Zhang et al . diffused free RNA nucleotides into a crystalized complex containing template strands of RNA attached to short pieces of RNA called primers , which kick-start replication . Then , the crystals were frozen at various intervals and viewed using X-rays . This allowed Zhang et al . to observe the structural changes that occurred over time as the compounds reacted . The approach first revealed that the free nucleotides had paired with complementary nucleotides on the RNA template strands . Then , pairs of free nucleotides reacted with each other to form the intermediate . Finally , the intermediate reacted with the primer , forming a new bond that connects the RNA primer to one of the nucleotides of the intermediate , while the other nucleotide of the intermediate was released as a free nucleotide . This experiment confirms that the specific structure of the intermediate molecule promotes RNA replication without help from enzymes . These findings will benefit chemists and biologists who study how RNA evolves and replicates . Future research building upon this work will deepen scientific understanding of the environmental conditions that were required for life to appear on Earth .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology" ]	2018	Crystallographic observation of nonenzymatic RNA primer extension
Visual projection neurons ( VPNs ) provide an anatomical connection between early visual processing and higher brain regions . Here we characterize lobula columnar ( LC ) cells , a class of Drosophila VPNs that project to distinct central brain structures called optic glomeruli . We anatomically describe 22 different LC types and show that , for several types , optogenetic activation in freely moving flies evokes specific behaviors . The activation phenotypes of two LC types closely resemble natural avoidance behaviors triggered by a visual loom . In vivo two-photon calcium imaging reveals that these LC types respond to looming stimuli , while another type does not , but instead responds to the motion of a small object . Activation of LC neurons on only one side of the brain can result in attractive or aversive turning behaviors depending on the cell type . Our results indicate that LC neurons convey information on the presence and location of visual features relevant for specific behaviors . Many animals use vision to guide their interactions with the environment . Doing so requires their visual systems to extract information about the presence of ethologically relevant visual features from diverse and dynamic sensory landscapes . Most organisms with elaborated nervous systems compartmentalize this task; in vertebrates and insects , for example , visual processing begins in specialized brain regions of similar general structure , called , respectively , the retina and the optic lobe ( Sanes and Zipursky , 2010 ) . The signals computed in these early visual areas are then conveyed to different higher order brain regions by visual projection neurons ( VPNs ) ; ultimately these signals must be passed on to the neural circuits that control behaviors . While VPNs are anatomically diverse and not necessarily closely related , their unique position as output channels of early visual centers makes these neurons attractive entry points for circuit-level analyses of visual processing . Studies of such neurons , for example of retinal ganglion cells in vertebrates and lobula plate tangential cells in insects , have provided insights into both the computations performed by the early visual system and the visual information that is available to higher brain regions ( Borst , 2014; Gollisch and Meister , 2010 ) . However , the relationship between signals encoded by the VPNs and visual behaviors has been difficult to systematically explore in any animal . Compared to photoreceptor neurons , which primarily respond to local luminance changes , VPNs can show much more specialized responses , some of which have been interpreted as encoding visual features directly relevant for specific behaviors , for example the presence of prey ( Lettvin et al . , 1959 ) or predators ( Zhang et al . , 2012 ) . Here we present anatomical , behavioral and physiological analyses of Lobula Columnar ( LC ) neurons in Drosophila that support such a role for this class of VPNs . In flies , visual information is first processed in the optic lobes , which are comprised of four neuropils called the lamina , medulla , lobula and lobula plate ( Fischbach and Dittrich , 1989; Meinertzhagen and Hanson , 1993 ) . Each neuropil has a repetitive structure of several hundred retinotopically-arranged columns that support the parallel processing of visual signals from different points in space . Neurons projecting out of the optic lobes originate in the medulla , lobula and lobula plate with the majority from the latter two , deeper neuropil layers . The response properties of several lobula plate VPNs have been characterized in great detail , mainly through studies in larger flies ( Borst et al . , 2010; Krapp et al . , 1998 ) . These lobula plate tangential cells ( LPTCs ) show strongly directionally selective responses to a variety of motion stimuli and some LPTCs have been proposed to function as matched filters for the complex optic flow patterns associated with a fly’s movements ( Krapp et al . , 1998 ) . Many recent advances have also revealed key components of the upstream circuitry that provides LPTCs with their direction-selective response properties ( reviewed in Borst [2014] ) . However , visual processing of stimuli other than wide-field motion is generally much less well understood . For example , flies respond to the movement , shape and position of objects ( Card and Dickinson , 2008b; Coen et al . , 2016; Egelhaaf , 1985a , 1985b , 1985c; Ernst and Heisenberg , 1999; Götz , 1980; Liu et al . , 2006; Maimon et al . , 2008; Ofstad et al . , 2011; Reichardt and Wenking , 1969; Robie et al . , 2010; Tang et al . , 2004 ) or to different wavelengths of light ( Gao et al . , 2008; Karuppudurai et al . , 2014; Melnattur et al . , 2014 ) . The neural substrates of these behaviors are largely unidentified but have often been proposed to involve neurons in the lobula . A major role of the lobula in visual processing is also indicated by anatomy , since the vast majority of medulla outputs project to the lobula ( Fischbach and Dittrich , 1989 ) . Of the different classes of VPNs found in the lobula , one group , the LC neurons ( Fischbach and Dittrich , 1989; Otsuna and Ito , 2006; Strausfeld and Okamura , 2007 ) , has received particular attention . The LC neurons are the most numerous VPNs of the lobula and , as we show here , can be divided into over twenty distinct types . Anatomical characteristics of typical LC neurons are illustrated in Figure 1 . Each LC type comprises multiple neurons of similar morphology whose individual dendritic arbors spread across only part of the array of lobula columns but which , with a few exceptions , cover the entire visual field as a population ( Fischbach and Dittrich , 1989; Otsuna and Ito , 2006; Strausfeld and Okamura , 2007 ) ( Figure 1C , F , K ) . These anatomical properties contrast with those of LPTCs whose large dendritic arbors vary across lobula plate columns but are sufficiently stereotyped to be individually identifiable across animals . In addition to their dendritic arrangements , the second anatomical hallmark of most LC neurons is the convergence of their axons onto cell-type specific target regions in the central brain ( Otsuna and Ito , 2006; Strausfeld and Okamura , 2007 ) . These target regions represent distinct neuropil structures often referred to as ‘optic glomeruli’ ( Strausfeld , 1976; Strausfeld and Okamura , 2007 ) ( Figure 1A , B , D , E , G–I ) , named by analogy to the similarly shaped subunits of the antenna lobes , the olfactory glomeruli ( Laissue et al . , 1999 ) . Similar to olfactory glomeruli , the optic glomeruli are enriched for synaptic sites and can be directly visualized with general neuropil markers ( Figure 1D , E ) ( Ito et al . , 2014 ) . The most prominent of these optic glomeruli are found in the ventrolateral central brain , specifically the posterior ventrolateral protocerebrum ( PVLP ) and the posterior lateral protocerebrum ( PLP ) ( Figure 1B ) . In addition , a small dorsal brain region with a glomerular structure , the anterior optic tubercle ( AOTu ) ( Figure 1B ) , is considered a specialized optic glomerulus and also receives LC cell inputs ( Otsuna and Ito , 2006; Strausfeld and Okamura , 2007 ) . 10 . 7554/eLife . 21022 . 003Figure 1 . Introduction to lobula columnar ( LC ) neurons . Schematics ( A–C ) and confocal images ( D–K ) show the lobula and adjacent parts of the visual system . ( A , D , G , H ) Horizontal sections . ( B , E , I , J ) Anterior views . ( C , F , K ) Cross-section views of the lobula . Some subregions of the optic lobe ( Me , Medulla; Lp , Lobula plate; Lo , Lobula ) and central brain ( AOTu , Anterior Optic Tubercle; PVLP , Posterior Ventrolateral Protocerebrum; PLP , Posterior Lateral Protocerebrum ) are indicated in selected panels . Dendrites of individual LC neurons ( red and blue cells in the schematics and red , blue and green cells in J , K ) span only part of the visual field . As populations , the neurons of a given LC cell type cover most or all of the lobula ( F ) , though LC neurons with regionally restricted lobula arbors also exist ( e . g . LC14 ( Otsuna and Ito , 2006 ) , see Figure 1—figure supplement 1 ) . LC neurons receive feed forward visual inputs from photoreceptors in the retina via a series of optic lobe interneurons ( a few lamina neurons , in brown , and transmedullary neurons [Tm] , in green , are illustrated as examples in [A] ) . This places LC neurons at least 2–3 synapses downstream of the photoreceptors . The majority of LC neurons projects to distinct target regions in the central brain called optic glomeruli; some of these are illustrated in ( A ) and ( B ) and also visible as distinct structures in the anti-Brp pattern in the images in ( D ) and ( E ) . Most optic glomeruli are located in the PVLP and the adjacent more posterior PLP . The more dorsal AOTu ( illustrated in [B] ) is considered a specialized optic glomerulus . For a more detailed map of optic glomeruli see Figure 3 . Confocal images show either populations of neurons ( D–I ) or individual cells labeled using Multicolor FlpOut ( MCFO ) ( Nern et al . , 2015 ) ( J , K ) . LC cell types shown are LC17 ( D , G , H ) and LC16 ( E , F , I–K ) . Population labeling ( D–I ) was with split-GAL4 driven expression of a membrane marker ( green; myr::smFLAG , using pJFRC225-5XUAS-IVS-myr::smFLAG in VK00005 ) with a presynaptic marker also shown [magenta; synaptotagmin-HA , using pJFRC51-3XUAS-IVS-syt::smHA in su ( Hw ) attP1] in ( D , E ) and by itself in ( H ) . A neuropil marker ( anti-Brp ) is included in grey in ( D–F , K ) and neuropil regions are also in grey in the schematics . Images in ( D , E , G–J ) were generated using brains that were computationally aligned to a template brain using the anti-Brp pattern as reference . The anti-Brp pattern in ( D , E ) is that of the standard brain used for alignment . Images in ( D–K ) show projection images of different views of three-dimensional image stacks; these were generated in either Fiji ( http://fiji . sc/ ) ( D , E , G–J ) or Vaa3D ( Peng et al . , 2010 ) ( F , K ) . Scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 00310 . 7554/eLife . 21022 . 004Figure 1—figure supplement 1 . Examples of additional LC cell types and similar neurons . These cell types were not selected for further analyses in this study since they did not project to glomerular target regions in the ventrolateral central brain , covered only part of the lobula as cell populations or are VPNs of a different optic lobe region . ( A , B ) LC14 cells , which project to the contralateral lobula , have been previously described ( Hassan et al . , 2000; Otsuna and Ito , 2006 ) . LC14 is the name used by Otsuna and Ito; some closely related cells , here named LC14b , project to the contralateral medulla in addition to the lobula ( Hassan et al . , 2000; Langen et al . , 2013 ) . LC19 ( C ) and LC23 ( D ) are newly identified cell types that project to the contralateral ( LC19 ) or both the ipsi- and contralateral ( LC23 ) central brain . LC19 and LC23 neurons have cell bodies in the cell body rind between optic lobe and central brain near the dorsal tip of the lobula neuropile . LC19 arbors in the lobula are located mainly in layers Lo5 and Lo6 , those of LC23 neurons primarily in layers Lo2 and Lo4 ( see Figure 5—figure supplement 1 for details of lobula stratification ) . LC19 processes are present across the entire lobula; those of LC23 appear to cover only the anterior , not the posterior lobula . Images show projection views ( generated in Vaa3D ) of MCFO labeled neurons together with the anti-Brp neuropil marker ( grey ) . ( E ) An example of a columnar lobula plate VPN ( LPC1 ) also identified by others ( Panser et al . , 2016 ) . LPC1 cells have cell bodies in the lobula plate cortex . Their processes in the lobula plate are mainly in lobula plate layer Lp2 . ( A ) , ( B ) and ( C ) show multiple cells; ( D ) and ( E ) segmented single cells . Scale bar represents 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 004 LC neurons have been proposed to encode behaviorally relevant visual features ( Strausfeld and Okamura , 2007 ) , a conjecture that is largely based on the striking anatomical reorganization associated with the convergence of their axons into glomeruli . While the dendrites of different LC neuron types are arranged in overlapping retinotopic arrays in the lobula , their synaptic termini in the central brain are grouped into discrete glomeruli , most of which receive input from a single LC neuron type ( Otsuna and Ito , 2006; Strausfeld and Okamura , 2007 ) . The retinotopic structure of the lobula and other early visual neuropils is thought to facilitate the extraction of visual features from the spatiotemporal patterns of the activity of visual interneurons ( ultimately going back to patterns of photoreceptor activity ) . By contrast , the organization of LC target regions into separate glomeruli , which is similar to the projection pattern of olfactory receptor neurons in the antennal lobe , suggests that different cell types already encode distinct features ( such as the presence of specific odorants in the olfactory system ) and that spatial information is of secondary importance . Though recent studies have begun to explore functional properties of LC neurons ( Aptekar et al . , 2015; Mu et al . , 2012 ) , the hypothesis that LC neurons are feature responsive cells remains largely untested and little is known about LC neuron function in general . One limitation has been experimental access to defined LC types: although several types of LC neurons have been described in Drosophila ( Fischbach and Dittrich , 1989; Otsuna and Ito , 2006 ) , specific genetic reagents for the study of these neurons were largely lacking and inputs to several prominent optic glomeruli had not been identified , though candidates for new LC neuron types are , for example , recognizable in images of fly brain clonal units ( Ito et al . , 2013 ) . In parallel to our work , two recently published independent studies ( Costa et al . , 2016; Panser et al . , 2016 ) have made use of existing image data ( such as ( Chiang et al . , 2011; Jenett et al . , 2012 ) ) to reveal additional VPN pathways from the lobula to optic glomeruli . Costa et al . report the identification of several new LC neuron types as one of several examples of the application of a new computational method that groups similar neurons using aligned brain images . Their work illustrates the potential power of these computational tools but places less emphasis on the description or interpretation of specific findings regarding VPN neuroanatomy . Panser et al . use a different computational method to identify GAL4 driver lines from the Janelia and Vienna Tiles collections ( Jenett et al . , 2012; Kvon et al . , 2014 ) that appear to include expression in VPN inputs to optic glomeruli and use these lines to generate an anatomical map of these glomeruli . In this study , we apply a previously established genetic intersectional approach ( Aso et al . , 2014a; Tuthill et al . , 2013 ) to generate highly specific split-GAL4 ( Luan et al . , 2006; Pfeiffer et al . , 2010 ) driver lines that target different LC types . We first use these lines to provide detailed anatomical descriptions of 22 different LC types about half of which had not been previously described by Otsuna and Ito ( Otsuna and Ito , 2006 ) . For each type , we examine not only the position and shape of the target glomerulus but also many other anatomical features ( such as lateral spread and layer patterns of lobula arbors , cell body positions and cell numbers ) , significantly extending the analyses in other studies . We find that each LC type is characterized by a distinct layer distribution of dendrites in the lobula and , with a few exceptions , a unique axonal output region in the central brain . Comparison of these output regions with the pattern of optic glomeruli indicates that most prominent glomeruli are the target regions of a specific LC type . Our glomerulus map largely concurs with the results of Panser et al . but is based on higher resolution images that allow us to better separate adjacent glomeruli and to define the target regions of two additional LC cell types . We also show that another LC type can be subdivided into four anatomically and genetically defined subtypes . Independent of the overall anatomical transformation associated with the convergence of LC neuron axons into glomeruli ( see above ) , LC neuron axons of a given type might either retain or discard retinotopy within their target glomerulus . Potential retinotopy within optic glomeruli has not been examined in detail and images with sparse labeling of LC neurons have been interpreted as arguing either for or against such axonal retinotopy ( Otsuna and Ito , 2006; Panser et al . , 2016 ) . To further explore possible retinotopy within individual glomeruli , we used multicolor stochastic labeling of individual LC neurons . In general , we did not observe detectable retinotopy of LC neuron axons within a glomerulus . However , we did identify a few exceptions; in particular , axonal projections of LC neurons to the AOTu retain retinotopy for azimuthal positions , suggesting a specialized role of the AOTu in the processing of spatial information . Stochastic labeling also allowed us to examine additional anatomical features of LC cells such as the dendritic arbor size and shape that cannot be observed at the population level . We also used our new driver lines to explore behaviors associated with LC neuron activity , by examining the response of freely behaving flies to optogenetic depolarization of individual LC types . In several cases such activation triggers distinct , highly penetrant behavioral responses that resemble natural , visually guided behaviors . In particular , using high-speed videography we show that two of these evoked behaviors , flight-initiating jumping ( takeoff ) and backward walking , resemble natural avoidance behaviors that can be elicited by a looming visual stimulus . Moreover , the two LC types whose activation evokes these avoidance behaviors respond to looming stimuli as assayed by two-photon calcium imaging . The encoding of looming is not a feature of all LC types , as we found a third LC type to be selectively responsive to small object motion . Finally , we present evidence that activation of LC neurons on only one side of the brain can induce attractive or aversive turning behaviors , depending on the cell type . Taken together , our anatomical and functional data suggest that each LC type conveys information about the presence and at least general location of a behaviorally relevant visual feature . Although details of our data suggest further downstream integration of signals from different LC types , the highly penetrant phenotypes we observe with activation of some LC types are consistent with a simple model for the initiation of several behaviors . To study and further identify LC neurons , we screened collections of GAL4 driver lines ( Jenett et al . , 2012; Kvon et al . , 2014 ) ( Barry J . Dickson , personal communication ) . We searched for cell types that consisted of many similar cells which as populations covered the entire array of visual columns in the lobula and whose axonal projections converged onto single glomerulus-like regions in the ipsilateral central brain ( Figure 1 ) . Other types of lobula VPNs were also identified in the screen but will not be further characterized here . These included a number of additional LC-like cell types which were excluded here because of the different structure or location of their target regions or because their combined dendrites appeared to be restricted to lobula subregions corresponding to only part of the visual field . Some examples of such cells , which include the previously described LC14 ( Hassan et al . , 2000; Otsuna and Ito , 2006 ) , are shown in Figure 1—figure supplement 1 . In addition to neurons having dendrites in the lobula , we identified columnar VPNs associated with other optic neuropils that also had glomerular target regions ( see Figure 1—figure supplement 1 for an example ) but we excluded them from further analysis . For the cell types that met our criteria , we used the split-GAL4 intersectional approach ( Luan et al . , 2006; Pfeiffer et al . , 2010 ) , where GAL4 activity is restricted to the overlap in the expression patterns of two GAL4 driver lines , to generate driver lines with predominant or exclusive expression in individual LC types . In combination , the split-GAL4 driver lines reported here have expression in 22 different types of LC neurons . Seven of these LC types ( LC4 , LC6 , LC9-LC13 ) have been previously described ( Fischbach and Dittrich , 1989; Otsuna and Ito , 2006 ) . For consistency , we named new LC types by extending a previously used numbering scheme ( Otsuna and Ito , 2006 ) and coordinated these names with another group that also found , and very recently reported ( Panser et al . , 2016 ) , several of the new LC neurons described here; except for the cell types shown in Figure 1—figure supplement 1 ( LC14 , LC19 , LC23 ) , the gaps in the sequence of LC cell type names ( LC1-LC3 , LC5 , LC7 and LC8 ) are due to the naming scheme and do not correspond to known LC types not covered in this study . Anatomical characteristics of the different LC neuron types are described below; for genotypes of the split-GAL4 driver lines see Materials and methods . Overall expression patterns of the main split-GAL4 lines used in this study can be found in Figure 2 and Figure 2—figure supplement 1 . Expression patterns of some additional lines for these same cell types are shown in Figure 9—figure supplement 1 and Figure 10—figure supplement 1 . Confocal stacks of all lines can be downloaded from www . janelia . org/split-GAL4 . Details of which split-GAL4 driver lines and other transgenes were used in individual experiments are provided in the Materials and methods and in Supplementary file 1B , D . 10 . 7554/eLife . 21022 . 005Figure 2 . Expression patterns of LC neuron split-GAL4 lines . Split-GAL4 driven expression of 20xUAS-CsChrimson-mVenus ( insertion in attP18; visualized using anti-GFP antibody labeling; green ) and a neuropil marker ( anti-Brp , magenta ) are shown . Genotypes are identical to those used in the behavioral experiments in Figure 8 . Some adjustments of brightness and contrast of individual samples were made . For each driver line ( but not across different lines ) , adjustments and microscope settings were identical for the brain shown in this Figure and the corresponding ventral nerve cord ( VNC ) shown in Figure 2—figure supplement 1 . The images in this and other figures with CsChrimson expression patterns are representative of 2–5 brains and 2–3 VNCs imaged for each split-GAL4 line . Scale bar represents 50 µm . Original confocal stacks are available from www . janelia . org/split-GAL4 . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 00510 . 7554/eLife . 21022 . 006Figure 2—figure supplement 1 . VNC expression patterns of the corresponding brains shown in Figure 2 . Imaging parameters and brightness or contrast adjustments were identical for each brain/VNC pair . Scale bar represents 50 µm . Original confocal stacks are available from www . janelia . org/split-GAL4 . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 006 We first present detailed anatomical studies of the LC cell types labeled by our split-GAL4 driver lines ( Figures 1–7 ) . We then focus on LC neuron function ( Figures 8–13 ) . To anatomically characterize LC neuron types and to confirm the identity of the LC neurons labeled by each split-GAL4 driver line , we examined cell shape ( Figures 2 and 3; examples shown in Figure 1D , E , G , I ) and the location and shape of target regions in the central brain ( Figure 3; visualized with a presynaptic marker [HA-tagged synaptotagmin; syt-smHA]; Figure 1H shows an example ) for each LC cell population . We also carried out stochastic labeling experiments to reveal the morphology of individual cells ( illustrated in Figure 1J , K ) and to explore the arrangements of LC neuron axon terminals relative to the retinotopic positions of their dendrites ( Figure 4 ) . Several stereotyped morphological features revealed in these experiments support the classification of LC neurons into the cell types described here: the shape and location of the target glomerulus as well as the size , shape and layer pattern of the lobula arbors; the approximate position of cell bodies; and the path followed by the axons . The anatomical features of LC neurons are summarized in a table in Supplementary file 1A . This table also includes estimates of the approximate number of cells per LC type for most LC types . We describe in detail the distribution and structure of axonal target regions of LC neurons in the central brain ( Figure 3 ) and the layer pattern , size and shape of their dendrites ( Figures 5 , 6 and 7 ) as these features provide information about the potential synaptic partners of each LC type . 10 . 7554/eLife . 21022 . 007Figure 3 . LC neuron terminals in the central brain are organized into distinct neuropil structures . ( A ) Illustration of the projection patterns of 12 LC cell types that project to major optic glomeruli in the PVLP ( or in the PVLP/PLP boundary region ) . Image is a substack maximum intensity projection of a composite image stack generated from 12 computationally aligned image stacks ( one for each cell type ) . Images were manually segmented to exclude background and some off-target cell types . Unedited pre-alignment stacks are available from www . janelia . org/split-GAL4 . For details of genotypes see Supplementary file 1D . ( B , C ) Target regions of the LC neurons shown in ( A ) match the optic glomeruli pattern in the PVLP . Target regions of different LC cell types were labeled by split-GAL4 driven expression of a presynaptic marker ( pJFRC51-3XUAS-IVS-syt::smHA in su ( Hw ) attP1 , detected using anti-HA antibody labeling ) . Images for different cell types were edited and combined as described above . The anti-Brp pattern of the standard brain used for alignment is shown in grey . ( C ) Pattern of optic glomeruli revealed by anti-Brp labeling . Image is the same as the anti-Brp channel of the overlay shown in ( B ) . Note the close correspondence of the presynaptic terminals of LC cell populations and optic glomeruli . Asterisk marks a large synapse rich ( based on anti-Brp labeling ) glomerular structure in the PLP that appears to be the target of several columnar VPNs that were not included here since we considered them to be primarily associated with the lobula plate , not the lobula ( though some have lobula branches ) . As an example , one cell of one such type ( LPC1 ) is shown in Figure 1—figure supplement 1E . LPC1 was also identified by ( Panser et al . , 2016 ) . ( D–F ) Overlays generated as in ( A ) and ( B ) showing the projection patterns ( D ) and target regions ( E , F ) of additional LC neurons with terminals in the posterior PVLP and in the PLP . LC24 , LC25 and LC26 projected to locations close to those of LC15 , LC6 and LC16 but slightly more posterior and might also slightly overlap with each other . In particular , LC25 was unusual in that its terminals spread along the surface of , and perhaps partly overlapped with , the LC15 target region and other adjacent glomeruli ( E and Figure 3—figure supplement 2 ) . Similarly , part of the boundary of the LC22 target region , as visualized by synaptic marker expression in LC22 cells , was less well defined than the boundaries of most other glomeruli ( F ) . The LC22 glomerulus also appears to overlap with the target region of a second columnar VPN: stochastic labeling experiments revealed an additional LPLC-like cell type ( distinct from LPLC1 and LPLC2 and tentatively named LPLC4 in agreement with [Panser et al . , 2016] ) that projects to the same approximate location as LC22 . While we have yet to generate specific split-GAL4 drivers for this additional LPLC cell type and therefore did not further characterize these neurons here , their overlap with LC22 terminals was directly confirmed by co-labeling of single cells of both types in the same specimen ( Figure 3—figure supplement 1 ) . ( G–I ) LC10 neurons project to the large subunit of the AOTu . Overlays ( generated as described above ) showing cell shapes ( G ) and presynaptic sites ( H ) of LC10 cells labeled by two different split-GAL4 driver lines . LC10 neurons showed two distinct general projection paths with axons entering the large subunit of the AOTu from both dorsally and ventrally ( magenta cells ) or only from ventrally ( green cells ) , in agreement with previous findings suggesting the existence of LC10 subtypes ( Otsuna and Ito , 2006 ) . ( I ) Anti-Brp reference pattern alone . The LC10 terminals are in a distinct large subcompartment ( white bracket ) of the AOTu . The AOTu also includes smaller subunits adjacent to this LC10 target region . Scale bars represent 30 µm ( A , G , I ) , 20 µm ( B , E ) or 50 µm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 00710 . 7554/eLife . 21022 . 008Figure 3—figure supplement 1 . A second type of columnar VPN projects to the LC22 target region . Main image and inset show two views along approximately orthogonal axes of MCFO labeled LC22 ( green ) and LPLC4 ( magenta ) cells generated using a GAL4 driver line ( R11C10 ) with expression in both cell types . The layer pattern of LPLC4 dendrites is different from both LPLC1 and LPLC2 ( see Figure 5 and Figure 7—figure supplement 2 ) . Images show reoriented substack projection views generated using Vaa3D . Scale bar represents 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 00810 . 7554/eLife . 21022 . 009Figure 3—figure supplement 2 . Presynaptic marker expression in individual LC cell types . Single confocal sections are shown . Images were rotated to show similar views . The anti-Brp reference pattern ( grey ) and the presynaptic marker ( syt-smHA , magenta ) are shown . The white appearance of the LC12 glomerulus is due to cross-talk between different labeling channels that occurred in some samples ( see Materials and methods ) . Genotypes are the same as in the images used for the overlays in Figure 3 . Many LC types also show some syt-smHA labeling in the lobula; examples of this are shown in Figure 5—figure supplement 2 and additional cases included in Supplementary file 1A . Scale bar represents 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 00910 . 7554/eLife . 21022 . 010Figure 4 . Multicolor stochastic labeling reveals differences in the arrangement of the terminal arbors of different LC types within their target glomerulus . ( A–C ) LC16 . ( A ) Position of the dendrites of three LC16 cells in the lobula in a layer cross-section view . Each cell occupies a distinct position along the long ( DV ) and short ( AP ) axes of the lobula ( dotted white lines were added to facilitate comparison of the positions of different cells ) . By contrast , within the target glomerulus ( B , C; two roughly orthogonal views are shown ) the arbors of the same cells are intermingled without an obvious correlation to the retinotopic pattern in the lobula . Most other LC types appeared similar to LC16; additional examples are shown in Figure 4—figure supplement 1 . ( D–G ) LC9 . Most LC9 cells have presynaptic arbors that spread through more than one half of the glomerulus but do not fill it completely ( F , G ) . Comparison of the approximate positions of arbors of individual LC9 cells in the glomerulus ( F , G ) with the AP positions of their dendrites in the lobula ( D , E ) suggests some retinotopy within the glomerulus , though with very low spatial resolution: For example , out of 57 examined LC9 cells , all of the cells ( 27/27 ) with axon terminals in the anterior-ventral tip of the glomerulus ( such as the green cell in [F] and both cells in [G] ) also had dendrites in the posterior half of the lobula while this was the case for only a few of the remaining LC9 cells ( 6/30 ) . White dotted lines in ( F ) and ( G ) indicate the approximate boundaries of the LC9 glomerulus . ( H–M ) LC10 . The relative order of LC10a ( H–J ) and LC10d ( K–M ) terminals along the DV axis of the AOTu ( shown in two orthogonal views , I , J for LC10a and L , M for LC10d ) matches the order along the AP axis of the lobula ( H for LC10a and K for LC10d ) . Individual cells were labeled using MCFO . Reconstructed views of reoriented subtstacks generated in Vaa3D are shown . For both LC10a and LC10d , similar results were observed for LC10 cells from five optic lobes , each with at least three labeled cells . Analyses of MCFO-labeled LC10b ( 36 cells from 18 brains ) and LC10c ( 33 cells from 17 brains ) single cells also showed an approximate correspondence between AP positions of dendrites in the lobula and DV positions of axonal arbors in the AOTu . LC10b cells also showed considerable variation in their lateral-medial spread within the medial zone of the AOTu but further analyses will be required to explore possible correlations between these differences and arbor positions in the lobula . Scale bars represent 20 µm ( A , D , E , H , K ) or 10 µm ( C , F , G , I , L ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01010 . 7554/eLife . 21022 . 011Figure 4—figure supplement 1 . Terminal arbor arrangements of additional LC cell types . ( A–C ) LC6 , ( D–F ) LC11 , ( G–I ) LC13 , ( J–L ) LC12 , ( M–O ) LC18 , ( P–R ) LPLC2 . Substantial overlap of the arbors of co-labeled single cells in their target regions was also observed for the remaining LC cell types ( at least two brains with two or more labeled cells were examined for each cell type ) . While there are occasional differences between the spread of single cells of the same type within their target glomerulus ( e . g . the LC6 cells in ( B ) do not all extend to the lateral tip of the glomerulus ) , we did not find obvious correlations between such differences and these cells’ retinotopic positions in the lobula . Lobula cross-section views and two orthogonal views of the target glomerulus are displayed as in Figure 4 . Scale bars represent 20 µm ( A ) or 10 µm ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01110 . 7554/eLife . 21022 . 012Figure 5 . Layer specific arborizations of LC neurons in the lobula . ( A ) Anti-Brp neuropil marker shows bands of different intensity in the lobula that can serve as approximate markers of layer boundaries . The image is a maximum intensity projection through 10 adjacent sections ( 0 . 38 µm spacing ) of the reference channel of the standard brain used for alignments . Approximate layer boundaries are indicated . Layer boundaries were defined by the positions of known cell types and closely match the anti-Brp pattern ( see Figure 5—figure supplement 1 ) . Lo5 was divided into two sublayers based on the anti-Brp pattern . Further subdivisions of strata based on the positions of arbors of different cell types would be possible but were not applied here . ( B , C ) Layer patterns of LC6 ( B ) and LC12 ( C ) . Split-GAL4 expression of a membrane marker is shown in green . Images were aligned to a standard brain using the anti-Brp pattern ( shown in grey ) . Images are maximum intensity projections through the same series of sections of brains aligned to the same template as in ( A ) . Approximate layer boundaries are marked with white lines . ( D ) Layer patterns of the remaining LC cell types ( except LC10 neurons ) . Projections were generated as in ( B , C ) but are shown without the anti-Brp pattern . All layers , but only a portion of the lobula is shown . Schematics in ( B–F ) indicate innervated layers as filled circles; black circles represent more extensive arborizations than grey circles . Note that these simplified schematics do not capture some details of the layer patterns ( such as sublayer patterns ) . An additional description of layer patterns can be found in Supplementary file 1A . ( E ) Single cell layer patterns are consistent with layer patterns seen at the population level . LC4 and LC15 are shown as examples . Additional single cell images can be found in Figure 7—figure supplement 1 . ( F ) Layer patterns of LC10 subtypes . LC10b and LC10d cells have similar layer patterns but differ in other aspects such as arbor size ( LC10b arbors in the lobula are larger ) . Additional examples of MCFO labeled LC10 cells of different subtypes can be found in Figure 7—figure supplement 1 and Figure 10—figure supplement 2 . Images in ( E , F ) were manually matched to the layer markers using the anti-Brp pattern . The scale of some images was slightly adjusted to compensate for varying depth of the lobula . Scale bars represent 20 µm ( A , also applies to B and C ) or 10 µm ( D , E; scale bar in E also applies to F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01210 . 7554/eLife . 21022 . 013Figure 5—figure supplement 1 . Layer positions of arbors of known cell types . Lobula terminals of MCFO labeled T5 , Tm9 , Tm4 , Tm3 and Tm20 cells ( Fischbach and Dittrich , 1989 ) are shown . Anti-Brp pattern is in grey . Layer boundaries are marked with white lines . Golgi studies ( Fischbach and Dittrich , 1989 ) have described T5 and Tm9 in Lo1 , Tm4 and Tm3 proximal terminals in Lo4 near the boundary of Lo5 , and Tm20 terminals in Lo5 near the boundary of Lo6 . Tm4 and Tm3 also have apparent presynaptic varicosities in Lo2 ( both ) and Lo1 ( Tm4 ) . These cell types therefore can be used to identify the approximate boundaries of all six lobula strata . Further division of Lo5 into two sublayers is suggested by the anti-Brp pattern and supported by the differential presence of arbors of several neuronal cell types in these sublayers ( compare for example , LC20 and LC25 in Figure 5D ) . Images show reoriented substack projection views generated using Vaa3D . Scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01310 . 7554/eLife . 21022 . 014Figure 5—figure supplement 2 . Potential presynaptic sites of LC neurons in the lobula . In addition to the target glomerulus in the central brain ( see Figure 3 and Figure 3—figure supplement 2 ) , split-GAL4 driven pJFRC51-3XUAS-IVS-syt::smHA in su ( Hw ) attP1 expression also resulted in labeling in the lobula in several LC cell types . Single confocal sections with anti-Brp in grey , syt-smHA in magenta and/or a membrane marker ( pJFRC225-5XUAS-IVS-myr::smFLAG in VK00005 ) in green are shown for each cell type as indicated in ( A ) . Although the interpretation of such fainter syt-smHA labeling patterns is not always straightforward , they strongly suggest that at least some LC neurons ( including cell types not shown in this Figure ) also have presynaptic sites within the lobula . Additional details on syt-HA labeling in the lobula are provided in Supplementary file 1A . Strong syt-smHA labeling in layer Lo6 in ( E ) appears to be mainly due to LC10b cells ( compare with ( C ) for LC10d cells alone ) , though a lobula intrinsic neuron type weakly labeled by the split-GAL4 driver line used might also contribute to the pattern . Scale bar represents 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01410 . 7554/eLife . 21022 . 015Figure 6 . Column spread of LC neurons and other cell types in the lobula in cross-section views . Cells were labeled using MCFO . Cross-section views of the lobula were generated using Vaa3D . The AP and DV axes of the lobula are indicated . Anti-Brp reference marker is shown in grey . LC neuron arbor sizes and shapes in the lobula are diverse across different cell types but similar within each type . LC cell types shown are LC11 ( A ) , LC10a ( B ) and LC22 ( C ) . The remaining LC cell types are shown in Figure 6—figure supplement 1 . MCFO labeled cells of a columnar medulla neuron type ( Tm3 ) present in each of ~750 visual columns ( D ) and a lobula tangential cell ( LT ) that spans the entire lobula ( E ) are shown for comparison . All LC arbors are multicolumnar with estimated sizes from about 10 ( LC10a ) to over 60 ( LC11 ) visual columns . LC22 cells were similar in size to LC11 along the long ( DV ) but not the short ( AP ) axis of the lobula . ( F ) Schematic summary of the column spread of different cell types in the lobula . Scale bar represents 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01510 . 7554/eLife . 21022 . 016Figure 6—figure supplement 1 . Layer cross-section views of lobula arbors of the LC neuron types not shown in Figure 6 . Cross-section views of LC cells of the indicated types were generated as described in the Figure 6 legend . Asterisks in the LC10b and LC21 panels mark two MCFO labeled cells ( LC10c and LC11 , respectively ) that do not belong to the indicated cell type . LO and LP in the LPLC panels indicate cross-section views of the lobula and lobula plate , respectively . Scale bar in the LC4 panel represents 20 µm; images of other cell types are shown at very similar scale . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01610 . 7554/eLife . 21022 . 017Figure 7 . Single cell shapes of LC neurons . Maximum intensity projection images of MCFO labeled single cells were manually segmented to exclude other labeled cells or background signal and converted to inverted grayscale images . Cells are shown in a similar orientation ( with dorsal approximately up and lateral approximately to the left ) and at the same scale . Scale bar represents 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01710 . 7554/eLife . 21022 . 018Figure 7—figure supplement 1 . Layer patterns of single cells of 22 LC neuron types . Images are reoriented views ( generated using Vaa3D ) of MCFO labeled LC neurons . Anti-Brp neuropil marker is in grey . For each cell type , two different views , along the AP and along the DV axes of the lobula , respectively , are shown . The two panels for a cell type can either show the same specimen in two views or two different specimens . Scale bars in the LC4 panels represent 20 µm; images of other cell types are shown at very similar scale . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01810 . 7554/eLife . 21022 . 019Figure 7—figure supplement 2 . Single cell labeling of LPLC arbors in lobula and lobula plate . ( A ) LPLC1 . ( B ) LPLC2 . ( C ) LPLC4 . Numbers indicate the four LP layers ( Fischbach and Dittrich , 1989 ) . Similar to lobula layers , the LP layers can be identified by changes in the intensity of anti-Brp labeling: Each LP layer shows a band of strong anti-Brp signal flanked by weaker labeling . All three LPLC cell types have processes in all lobula plate layers and several layers of the lobula but differ in details of arbor structure and layer patterns . For example , lobula layers of LPLC4 ( mainly Lo2 , Lo4 and Lo6 ) are different from both LPLC1 ( Lo2-Lo4 and Lo5B ) and LPLC2 ( Lo4 and Lo5B ) . The presence of processes in multiple lobula plate layers , as shown by all LPLC types , could support responses to stimuli that combine different directions of motion , such as a visual loom ( Schilling and Borst , 2015 ) . Scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 01910 . 7554/eLife . 21022 . 020Figure 8 . Optogenetic activation of LC neurons induces distinct behavioral responses that differ between LC cell types . ( A ) A representative video image of group of freely walking flies in the circular arena assay . ( B ) Representative video images of a freely behaving fly on a small glass platform in the single-fly assay . A side view ( upper part of the panel ) and a bottom view ( lower part ) of the fly are simultaneously recorded on a single high-speed video camera with the aid of two small prisms ( see Materials and methods ) . ( C , D ) The results of CsChrimson activation of different LC cell types in the arena ( C ) and the single-fly ( D ) assays are summarized in a grayscale intensity map . Each column represents a distinct behavior and each row represents a different split-GAL4 driver line with a predominant expression in the indicated cell type . Shading represents the behavioral penetrance ( percentage of trials or flies of a specific genotype in which a given behavior was observed , see Supplementary file 1B for the names of the GAL4 lines used and penetrance values ) . In both assays , the occurrences of reaching and jumping behaviors were annotated manually , while locomotor behaviors including forward walking , backward walking and turning were determined based on velocity and angular speed derived from automated fly tracking ( see Figure 8—figure supplement 1 , and Materials and methods ) . For locomotor behaviors we set a conservative threshold of two standard deviations away from the mean to be considered an activation phenotype ( Figure 8—figure supplement 1E , F ) . In this way , we determined the behavioral penetrance for five phenotypes – reaching , jumping , forward walking , backward walking and turning - in both the arena and the single-fly assays . In the single-fly assay , the high jumping penetrance of four LC cell types ( LC4 , LC6 , LPLC1 and LPLC2 ) resulted in too few flies ( <= 12 ) available for analyses of walking and turning behaviors ( indicated with a ‘\’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02010 . 7554/eLife . 21022 . 021Figure 8—figure supplement 1 . Quantification of locomotor behaviors and determination of behavioral penetrance . ( A ) A representative tracking of a fly with LC16 activation in the arena assay . Arrows indicate the start and end positions . Fly’s trajectory is denoted in red ( during stimulation , 1 s ) or in white ( before and after stimulation , 1 s each ) . ( B ) Time-series plot of velocity and angular speed of pBDPGAL4U control ( black , trial count = 109 , fly count = 23 ) and LC16 activated ( red , trial count = 65 , fly count = 14 ) flies . Solid and dashed lines indicate median and inner-quartile range , respectively . Grey rectangles in this and other panels signify optogenetic stimulation with red light . ( C , D ) Notched box plots showing the distribution of mean velocity ( C ) and angular speed ( D ) , before , during , and after optogenetic stimulation , of pBDPGAL4U control ( black ) and LC16 activated ( red ) flies . The increased translational velocity of LC16 flies compared to controls prior to CsChrimson stimulation appears to be due to reduced walking of the control flies after repeated red light stimulation: The data in ( B–D ) are averages from multiple trials ( trials 1 to trial 5 ) with the maximum light stimulus from experiments with increasing intensities ( i . e . the same flies were stimulated repeatedly within an experiment with a series of light intensities; see Materials and methods ) . To examine whether the apparent higher basal locomotion of LC16 flies during pre-stimulation light off was a response to prior stimulations , we compared trial 1 data to trial 5 data from the experiments in Figure 9 , because in these experiments only the maximum light stimulus was used ( data for different LC16 split- GAL4 lines and for the different control flies shown in Figure 9 were pooled ) . Data for trial 1 showed no pre-stimulus difference between LC16 and control flies ( p=0 . 88 , by Wilcox rank sum test ) , while locomotion was reduced for the controls in trial 5 ( p<0 . 01 ) . ( E , F ) Jitter plots showing the distribution of mean velocity and angular speed during optogenetic stimulation of control and LC neuron driver lines in the arena ( E ) and single-fly ( F ) assays . Each dot represents an individual trial . A conservative threshold is set at the 97 . 7th percentile of the pBDPGAL4U control distribution , indicated by dashed red lines . Grey and black dots represent trials within and outside of the normal range , respectively . For walking , trials above the top red line are counted as displaying a forward walking phenotype and trials below the bottom red line are counted as displaying a backward walking phenotype . For turning , trials above the red line are counted as displaying a turning phenotype . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02110 . 7554/eLife . 21022 . 022Figure 8—figure supplement 2 . Variability of LC16 backward walking behavior across trials and across flies . In the circular arena experiments , each fly was repeatedly tested in a series of trials ( see Materials and methods ) . For most analyses , we pooled data across all flies and trials ( as shown in the left panel for walking behavior after optogenetic stimulation of different LC neuron types; these are the same data plotted as a separate data point for each trial in Figure 8—figure supplement 1 ) . To further explore sources of variability in these experiments , using LC16 activation induced backward walking as an example , we reanalyzed the data by pooling results for either all flies of the same trial ( per trial ) or all trials of the same fly ( per fly ) . This revealed different degrees of variability both across flies within the same trial and across trials for the same fly . Nonetheless , this variability does not change our conclusion that LC16 activation results in a strong backward walking response . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02210 . 7554/eLife . 21022 . 023Figure 9 . Experiments with additional LC6 and LC16 driver lines confirm the activation phenotypes of these cell types . ( A , B ) Behavioral penetrance for different controls and multiple split-GAL4 driver lines for ( A ) jumping ( flies that jumped within 200 ms of stimulation onset ) with LC6 controls based on the OL0077B driver line and ( B ) backward walking and turning with LC16 controls based on the OL0046B driver line . Each dot represents an experiment done with a different genotype: orange , LC neuron activation in blind norpA flies that also carry an LC6 ( A ) or LC16 ( B ) split-GAL4 line; green , pBDPGAL4U control in blind norpA flies; blue , flies reared on food without supplemental retinal; red , split-GAL4 DBD or AD halves; grey , genetically distinct split-GAL4 driver lines with targeted expression in LC6 ( A ) or LC16 ( B ) . Horizontal and vertical lines indicate mean and standard deviation , respectively , for the control group and split-GAL4 group . The genotypes of the driver lines , behavioral penetrance and total trial and fly counts are listed in Supplementary file 1B . Expression patterns of the split-GAL4 driver lines used are shown in Figure 9—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02310 . 7554/eLife . 21022 . 024Figure 9—figure supplement 1 . Expression patterns of multiple split-GAL4 driver lines for LC6 and LC16 . Brain ( A , B ) and VNC ( C , D ) expression patterns of split-GAL4 drivers for LC6 ( A , C ) and LC16 ( B , D ) . Split-GAL4 driven expression of 20xUAS-CsChrimson-mVenus ( insertion in attP18; visualized using anti-GFP antibody labeling; green ) and a neuropil marker ( anti-Brp , magenta ) are shown . Imaging parameters and brightness or contrast adjustments were identical within each brain/VNC pair but not across all images . OL0077B and OL0046B images are the same as those shown in Figure 2 and Figure 2—figure supplement 1 . Scale bar represents 50 µm . Original confocal stacks are available from www . janelia . org/split-GAL4 . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02410 . 7554/eLife . 21022 . 025Figure 9—figure supplement 2 . Quantification of LC16 activation induced locomotor behaviors and determination of behavioral penetrance . Jitter plots show the distribution of mean velocity and angular speed during optogenetic stimulation of multiple LC16 split-GAL4 driver lines and control lines . Color codes are the same as in Figure 9 . Penetrance was determined as in Figure 8—figure supplement 1E . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02510 . 7554/eLife . 21022 . 026Figure 10 . Reaching behavior resulting from activation of LC10 subtypes . ( A ) Representative video images of a fly exhibiting reaching behavior in the single-fly assay . Time stamps indicate milliseconds ( ms ) after the start of reaching . ( B ) Comparison of lobula layer patterns and penetrance of reaching upon optogenetic activation for 15 LC10 split-GAL4 lines . The images show reconstructed views of CsChrimson expression in the lobula ( generated using Vaa3D and manually aligned using the anti-Brp reference marker ) . Approximate layer positions are indicated on the left . Scale bar represents 10 µm . CsChrimson expression using two additional LC10 driver lines ( SS00941 and SS00942 ) resulted in unexpected uncoordinated behaviors in response to optogenetic activation that precluded analyses of reaching behavior . These lines , which are related as they only differ from each other in the insertion site of the AD hemidriver ( see Supplementary file 1B ) , were therefore excluded from further analyses . ( C ) Single cell labeling reveals subtype expression patterns of LC10 driver lines . Overall lobula expression of an LC10 split-GAL4 line ( SS02664 ) ( top panel; displayed as in B ) and examples of MCFO labeled single cells from this line ( bottom panels ) are shown . LC10 subtypes ( indicated below the images ) of single cells were assigned based on layer pattern , arbor size and shape as follows: LC10c cells mainly arborize in layer Lo5B with some processes in the adjacent layers . LC10a cells also have arbors in Lo5B , but differ from LC10c by having additional processes in the more distal layers Lo2 , Lo3 and Lo4 . LC10b and LC10d differ from LC10a and LC10c by having major arbors in Lo6 and only few processes in Lo5B . Their distal arbors reach Lo4 . LC10b cells are wider than LC10d cells and show many small varicosities ( presumably presynaptic sites; compare Figure 5—figure supplement 2E ) in Lo6 . LC10 neurons also differ in their axonal paths in the AOTu: LC10a and LC10d axons run both dorsally and ventrally , and LC10b and LC10c only ventrally ( Figure 3G; also see Figure 10—figure supplement 1 ) . Scale bars represent 10 µm . ( D ) Reaching penetrance observed upon activation of the LC10 split-GAL4 drivers and various control lines in the single-fly assay . Split-GAL4 drivers are grouped by expression patterns in the LC10 subtypes: either subtype a and/or d , or neither a nor d . Driver lines were included in the ‘a and/or d’ category if two or more cells of these types were identified in MCFO experiments ( see Figure 10—figure supplement 2 ) . Colors representing various controls and split-GAL4 driver lines are the same as those in Figure 9 . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02610 . 7554/eLife . 21022 . 027Figure 10—figure supplement 1 . Expression patterns of LC10 split-GAL4 driver lines . Brain ( A ) and VNC ( B ) expression patterns using 20xUAS-CsChrimson-mVenus in attP18 visualized with an anti-GFP antibody ( green ) and a neuropil marker ( anti-Brp , magenta ) are shown . While some driver lines also had expression in a few non-LC10 cells , none of these additional neurons was consistently present in lines with the reaching behavior . Imaging parameters and brightness or contrast adjustments were identical for each brain/VNC pair but not across all images . OL0020B and OL0023B images are the same as those shown in Figure 2 and Figure 2—figure supplement 1 . Scale bar represents 50 µm . Original confocal stacks are available from www . janelia . org/split-GAL4 . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02710 . 7554/eLife . 21022 . 028Figure 10—figure supplement 2 . Stochastic single cell labeling reveals LC10 subtype expression patterns of the LC10 split-GAL4 driver lines . ( A ) Examples ( 10 per driver line ) of MCFO-labeled LC10 cells from five different lines . Subtype classification is indicated for each cell ( white lowercase letters ) . Anti-Brp pattern ( used to facilitate comparison of layer positions across specimens ) is shown in grey; views were generated using Vaa3D . Layer patterns of these projection images were manually aligned using the anti-Brp marker . To do so , the scale of some images was slightly adjusted to compensate for the non-uniform depth of the lobula . Approximate layer positions are indicated in the bottom right panel . Scale bar represents 10 µm . ( B ) Summary of LC10 MCFO results . For each driver line , the number of MCFO-labeled LC10 cells of each subtype was determined for the indicated number of fly brain samples . Some LC10 cells that were located at an edge of the lobula neuropil or overlapped with other labeled cells and thus could not be identified were excluded from the counts . Total cells refer to the number of all LC10 cells that were assigned a subtype . Subtypes that were only seen as single labeled cell are in parentheses in the summary . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02810 . 7554/eLife . 21022 . 029Figure 11 . LC6 and LC16 activation behaviors resemble avoidance responses evoked by visual looming . ( A ) LC6 and LC16 project to adjacent , non-overlapping target glomeruli . The image was generated using a 3D image rendering software ( FluoRender ) ( Wan et al . , 2012 ) on aligned confocal images . ( B ) LC6 and LC16 have similar layer patterns in the lobula . An overlay of substack projections of aligned image stacks is shown . Anti-Brp reference marker is in grey . ( C ) Representative video images from the single-fly assay showing that a looming stimulus and LC6 activation evoke very similar coordinated behavioral sequences , which include wing elevation , middle leg extension and initiation of flight . Time stamp is set at 0 ms for the frame of takeoff . Negative and positive values are for frames before and after takeoff , respectively . ( D ) Notched box plots showing the duration of the takeoff sequence measured as the time from the first moment of wing movement to the last moment of tarsal contact with the ground after the stimulus ( Mann-Whitney test , p=0 . 29 between looming and LC6 activation , and p<0 . 001 between LC6 activation and no stimulus ) . ( E ) Representative video images from the single-fly assay showing that a looming stimulus and LC16 activation evoke very similar backward walking behaviors . Time stamp is set at 0 ms for the start of backward walking . ( F ) Total distance flies walked on the platform of the single-fly assay . Positive and negative values are for forward and backward walking , respectively ( Mann-Whitney test , p=0 . 71 between looming and LC16 activation , and p<0 . 001 between LC16 activation and no stimulus ) . Scale bars represent 50 µm ( A ) or 20 µm ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 02910 . 7554/eLife . 21022 . 030Figure 11—figure supplement 1 . Behavioral consequences of silencing LC6 and LC16 by Kir2 . 1 expression . Split-GAL4 driver lines OL0077B ( LC6 , A–B ) or OL0046B ( LC16 , C ) were crossed to flies with pJFRC49-10XUAS-IVS-eGFPKir2 . 1 in a DL strain background . pBDPGAL4U crossed to the same effector or DL flies were used as controls . ( A , B ) Looming stimuli were presented at an elevation of 45° , an azimuth of 90° , and an l/v of 20 ( as in Figure 11C , D ) . In addition , a slow stimulus ( using an l/v of 160 ) was included in ( A ) since calcium imaging suggests LC6 is tuned to slower looms ( Figure 12—figure supplement 1 ) . Confidence intervals were determined using the Clopper-Pearson method . Fly counts for each experiment from left to right are 170 , 117 , 141 , and 89 . ( B ) Notched box plots showing the duration of the takeoff sequence ( as in Figure 11C ) . ( C ) Looming stimuli were presented as in Figure 11E–F , specifically chosen to induce backward walking . Total distance flies walked on the platform of the single-fly assay is shown . Fly counts from left to right are 396 , 43 , and 40 . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 03010 . 7554/eLife . 21022 . 031Figure 12 . LC16 and LC6 , but not LC11 , respond to visual looming stimuli with robust calcium increases . ( A ) Visual stimuli evoked calcium responses of LC neurons were imaged in head-fixed flies . ( B ) The axon terminals of LC cells bundle to form cell-type specific glomeruli ( subset shown in C ) . We imaged from a single glomerulus by using spilt-GAL4 lines labeling individual cell-types ( LC16 , LC6 or LC11 ) . Representative regions for calcium imaging experiments are marked with the yellow dashed rectangles . Exemplary responses of LC16 , LC6 and LC11 to a slow dark looming disk are shown ( C; each single frame taken from the peak response of an individual fly , distinct genotypes were used to image from each glomerulus ) . ( D ) LC16 , LC6 and LC11 responses to looming visual stimuli are shown for three variants of the stimulus ( from top to bottom: dark looming disk , bright looming disk , luminance-matched ) expanding at r/v = 550 ms ( n = 5 per genotype ) . Error bars indicate mean ± SEM . Statistics were performed on mean ∆F/F during a time window in which the response peaks ( 2 s before and after the looming stimulus stops expanding ) . ( E ) As a comparison to looming stimuli , we also presented moving object stimuli that contain local motion that is distinct from looming . LC11 responds strongly to the motion of the small ( 9°x9° ) spot , but not the long bar ( 9°x72° ) moving object . The objects moved at 22 . 5°/s , starting 18° left of the visual midline and stopping 108° to the right of the midline . Statistics were performed on mean ∆F/F during the whole stimulus epoch . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 03110 . 7554/eLife . 21022 . 032Figure 12—figure supplement 1 . LC16 and LC6 are tuned to slower looming speeds . Stimulus evoked calcium responses to different loom speeds . The responses are mean ∆F/F during a time window in which the response peaks ( 2 s before and after the looming stimulus stops expanding ) . Stimuli from top to bottom rows: dark looming disk , bright looming disk , luminance-matched . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 03210 . 7554/eLife . 21022 . 033Figure 13 . Behavioral responses to unilateral LC neuron stimulation differ from bilateral activation behaviors and are directional . ( A ) Schematic illustration of a genetic method for stochastic labeling and activation of LC neurons . A ‘stop-cassette’ reporter ( pJFRC300-20XUAS-FRT>-dSTOP-FRT>-CsChrimson-mVenus in attP18 ) was used for Flp-recombinase mediated control of CsChrimson expression . This reporter/effector construct ( small schematic in the right brain hemisphere of the larval brains in the illustration ) is based on the ‘Flp-out’ design ( Struhl and Basler , 1993 ) . It contains 20 Upstream Activating Sequences ( UAS ) and a core promoter ( grey rectangle ) for GAL4-activated expression , a transcriptional terminator ( white rectangle with a red prohibition sign ) flanked by Flp-recombinase target ( FRT ) sites ( black triangles ) , and a CsChrimson-mVenus fusion gene ( green rectangle ) . Heat shock induces expression of the Flp-recombinase which can excise the transcriptional terminator , allowing expression of CsChrimson-mVenus under the control of a split-GAL4 driver ( not shown ) . By expressing a limiting amount of Flp-recombinase early in development ( first instar larval stage ) , stochastic stop-cassette excision occurs in LC precursor cells , generating adults in which most or all neurons of one LC neuron type ( determined by the split-GAL4 driver ) express CsChrimson-mVenus in either no , one , or two optic lobes . ( B ) Strong backward walking behavior requires bilateral LC16 activation . Notched box plots showing the distribution of mean velocity for bilateral LC16 activation ( trial count = 101 , fly count = 10 ) , unilateral LC16 activation ( trial count = 75 , fly count = 7 ) and no labeling controls ( trial count = 114 , fly count = 11 ) . Behavioral responses of individual flies were assayed and their brains were subsequently dissected to determine expression patterns . Unilateral LC16 activation produced far less backward walking than bilateral activation ( Mann-Whitney test , p<0 . 001 ) . ( C ) Unilateral activation of LC16 and LC10 induces aversive and attractive turning , respectively . Notched box plots showing the distribution of mean angular velocity for different labeling categories of LC16 and LC10: For LC16 , trial and fly counts are the same as in ( B ) . For LC10 , bilateral ( trial count = 91 , fly count = 8 ) , unilateral ( trial count = 268 , fly count = 24 ) and no labeling ( trial count = 281 , fly count = 22 ) . For unilateral activation , behavioral data from animals with labeling only on the left brain hemisphere were reversed and combined with those from animals with labeling only on the right brain hemisphere . In addition to flies with clear bilateral or unilateral expression , 18 flies in the LC10 stochastic activation experiments had expression in both brain halves but showed differences in the apparent number of labeled LC10 cells between the hemispheres . Because of the wide range of these labeling differences we did not include these flies in the above analysis . However , the behavioral results for this group also showed a turning bias towards the side with stronger labeling ( see Figure 13—figure supplement 1 ) , suggesting that even small differences of LC10 activation between the two hemispheres may be sufficient to induce ipsilateral turning . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 03310 . 7554/eLife . 21022 . 034Figure 13—figure supplement 1 . Turning behavior of LC10 flies with bilateral labeling that is stronger in one brain hemisphere ( trial count = 212 , fly count = 18 ) . Data for flies with uniform bilateral expression or no labeling are the same as in the main Figure and included for comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 034 We used split-GAL4 driven expression of syt-smHA to visualize presynaptic sites of individual LC neuron populations in the central brain ( Figure 1H ) . To facilitate visualization of combinations of LC cell types , we aligned data collected from confocal stacks of individual driver lines to a standard brain ( Aso et al . , 2014a ) . Together , the LC neurons characterized in this study project to 19 distinct target regions in the ipsilateral central brain ( Figure 3 , Videos 1 and 2 ) . In addition to the shape and location of these target regions , the axonal paths followed by LC neurons are also stereotyped with individual cells of the same type showing a similar projection pattern ( Figure 2 and Figure 3A , D , G ) . To compare LC neuron target areas to the position of optic glomeruli , we focused on the PVLP where several large glomeruli can be readily visualized with general markers of synaptic density ( Ito et al . , 2014 ) . LC neuron target regions in the PVLP visualized by expression of a presynaptic marker in LC neurons closely matched the glomerular neuropil pattern revealed by anti-Brp staining , which labels presynaptic active zones ( Wagh et al . , 2006 ) ( Figure 3B , C and Figure 3—figure supplement 2 , Video 1 ) . This result confirmed optic glomeruli as the target regions of these LC neurons and identified LC input neurons for each of several glomeruli whose projection neuron inputs were unknown . In particular , our data show that each of the 12 most prominent glomeruli in the PVLP ( or , in one case , the boundary region of PVLP and PLP ) ( Figure 3B , C , Video 1 ) is the target of one columnar VPN type from the lobula . Ten of these glomeruli are the targets of typical LC cell types while two glomeruli receive inputs of neurons with dendrites in both lobula and lobula plate; we refer to these LC-like cells as LPLC1 and LPLC2 . Similar cells have been described in other Diptera ( Douglass and Strausfeld , 1998 ) . Although we visually screened the expression patterns of several thousand GAL4 lines , we did not observe additional cases where the terminal fields of a VPN similarly coincided with one of these glomeruli ( as defined by anti-Brp labeling ) , suggesting that LC neurons are the primary inputs to these structures . 10 . 7554/eLife . 21022 . 035Video 1 . Composite of aligned images showing a series of sections through the ventrolateral central brain with the target regions of 18 LC ( or LPLC ) neurons visualized by presynaptic marker expression . Original images , composite image assembly and color scheme are as in Figure 3 . The anti-Brp reference pattern ( grey ) of the template brain used for alignment is shown both together with the labeled LC terminals ( left panel ) and individually ( right panel ) . Terminals of LC10 cell types in the AOTu are not included in the movie . We believe the slight overlap seen between LC12 and LC18 and between LC24 and LC26 is due to imperfect alignment , rather than true overlap . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 03510 . 7554/eLife . 21022 . 036Video 2 . Three-dimensional rendering of the target regions of LC ( or LPLC ) neurons in the ventrolateral central brain shown in Figure 3 and Video 1 . Image assembly and color scheme are as described in Figure 3 . Anti-Brp reference pattern is in grey . Terminals of LC10 cell types in the AOTu are not included . The movie was generated using Vaa3D . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 036 The target regions of most LC cell types in the PVLP were clearly distinct with no or only minimal overlap ( Figure 3B , C ) ; some glomeruli appear to overlap in the projection image in Figure 3B but occupy distinct three-dimensional positions ( see Videos 1 and 2 ) . In addition to projections to the large PVLP glomeruli , we also characterized several LC neuron types that converged onto more posterior , often smaller regions in the PVLP and PLP ( Figure 3D–F , Videos 1 and 2 ) . The target regions of these additional LC cell types appeared more variable in shape and arrangement . Finally , several LC neuron types project to the large subunit of the AOTu ( Figure 3G–I ) ; the properties of these LC10 cells are discussed further below . In sum , our results provide a high resolution map of LC neuron target regions and establish a link between the pattern of glomeruli detected with general neuronal markers and the neuron projections of specific LC cell types . With a few exceptions , primarily the AOTu ( discussed further below ) and the target regions of LC25 and LC22 , each glomerulus described here can be uniquely identified as the target region of a single LC neuron type that appears to be the major input to that glomerulus . Although other , non-columnar VPNs may also have some presynaptic sites in or near these optic glomeruli , we did not observe similarly prominent non-LC input neurons to these structures in our GAL4 line screen . A prominent feature of LC neurons is the change in the anatomical arrangement of LC neurons as they project from the lobula to the central brain . In the lobula , LC neuron dendrites form parallel retinotopic arrays , whereas their synaptic termini in the central brain are grouped into discrete glomeruli . However , it has not been examined in detail whether any retinotopy is preserved in the glomeruli; that is , whether the positions of the dendrites of individual LC neurons in the lobula and of their presynaptic terminals within their target glomerulus are correlated . To directly compare the relative positions of the dendrites of different LC cells of the same type in the lobula with the arrangement of the presynaptic arbors of the same cells , we examined samples in which several LC cells of the same type were stochastically labeled in distinct colors ( Figure 4 ) using the Multicolor FlpOut ( MCFO ) technique ( Nern et al . , 2015 ) . These experiments revealed that most individual LC neurons , with the exceptions discussed below , had branched terminals that appeared to spread throughout their target glomerulus without obvious spatial restriction to subregions or correlation between the distributions of their dendritic arbors in the lobula and their presynaptic arbors in the glomerulus . Figure 4A–C shows examples of this analysis for LC16 . Qualitatively , most other LC cell types appeared similar to LC16 in that they lacked any obvious preservation of retinotopy at the glomerulus level; each individual cell’s axonal terminal was intermingled with others , featuring pre-synaptic boutons throughout the glomerulus ( see Figure 4—figure supplement 1 for examples of additional LC cell types ) . Of course , retinotopic patterns in the synaptic connections between these LC neurons and their targets that are not apparent at the resolution examined here may exist . LC9 is an example of a cell type that appeared to retain some retinotopy , though with very low spatial resolution , at the level of the axonal terminals: terminals of single LC9 cells expanded through only part of the glomerulus and their position correlated with the approximate position of the corresponding dendrites in the lobula ( Figure 4D–G ) . The most striking example of retinotopy within a glomerulus was the arrangement of the terminals of the LC10 neurons that project to the AOTu ( Figure 4H–M ) . LC10 cells have branched terminals similar to other LC neurons but with limited dorso-ventral ( DV ) spread so that individual cells did not cover the entire AOTu along this axis ( Figure 4I , J , L , M ) . In the AOTu , the relative positions of presynaptic arborizations of LC10 cells along the DV axis of the AOTu largely matched the order of their dendrites along the anterior-posterior ( AP ) axis of the lobula ( Figure 4H , K ) , though with more overlap between arbors in the AOTu compared to the lobula . By contrast , the DV positions of LC10 dendrites in the lobula were not obviously correlated with arrangements of neuronal processes in the AOTu and the positions of different LC10 terminals largely overlapped along the lateral-medial ( LM ) and AP axes of the tubercle ( Figure 4I , J , L , M ) . LC10 cells also differed in a second key respect from LC cell types innervating most other optic glomeruli; consistent with previous reports ( Otsuna and Ito , 2006 ) , we found that the AOTu is the target of axonal projections of several anatomically distinct LC10 subtypes . Our classification of LC10 cells is mainly based on the size and layer patterns of the lobula arbors of these neurons; these differences , which are predicted to reflect different presynaptic inputs to these cells , are described in detail below ( Figure 5F , Figure 7—figure supplement 1 and Figure 10—figure supplement 2 ) . Previously reported subtype distinctions ( Otsuna and Ito , 2006 ) were based on differences of the axonal path ( dorsal or ventral ) within the AOTu , which we also observed but which do not unambiguously identify the subtypes described here , and distinct lateral-medial positions of LC10 terminals , which we did not find evidence for; all LC10 types labeled by our driver lines projected to the large medial zone ( Ito et al . , 2014 ) of the AOTu ( Figure 3G–I ) , while inputs to the lateral zone , previously interpreted as LC10C cells ( Otsuna and Ito , 2006 ) , appear to be projections from the medulla ( Otsuna et al . , 2014; Panser et al . , 2016 ) . We refer to the subtypes defined here as LC10a-LC10d ( lowercase letters ) to distinguish them from the previously proposed LC10A-LC10C ( Otsuna and Ito , 2006 ) . Costa et al . ( 2016 ) also propose several new LC10 subtypes based on computational clustering of single cell data . These putative cell types are primarily defined by having axonal arbors that are restricted to different positions along the DV axis of the AOTu with corresponding regional patterns of their dendrites along the AP axis of the lobula . Our analyses of retinotopy of LC10 neurons using subtypes specific split-GAL4 lines showed that the axonal terminals of both LC10a ( Figure 4H–J ) and LC10d ( Figure 4K–M ) subtypes are retinotopically distributed along the full DV axis of the AOTu , suggesting that different subtypes form independent retinotopic maps that each cover the entire visual field . For this reason , we believe that the clustering method of Costa et al , which relied on the central brain arbors of LC cells , not their layer patterns in the lobula , led to a misclassification of groups of cells of LC10 neurons at different retinotopic positions as distinct cell types . In agreement with this possibility , attempts to identify GAL4 lines with specific expression in such regionally restricted LC10 subtypes were largely unsuccessful ( Panser et al . , 2016 ) . By contrast , our split-GAL4 driver lines provide genetic support for the LC10 subdivisions we describe here . In sum , within the AOTu the terminals of multiple LC cell types overlap , showing clear retinotopic spatial segregation along one axis . Just as the positions of the target glomeruli of LC neurons are indicative of the spatial location of their as yet uncharacterized postsynaptic partners , the distribution of LC neuron arbors across lobula layers can provide clues to the presynaptic inputs to the LC cells . Similar to other optic lobe neuropils , the lobula has a distinctly stratified structure ( Fischbach and Dittrich , 1989; Strausfeld , 1976 ) . To compare layer patterns across samples , we used the anti-Brp reference marker ( Figure 5A ) to both directly identify lobula strata and to enable alignment of different optic lobes to a common reference . In the lobula , anti-Brp staining shows seven bands of alternating labeling intensity ( Figure 5A ) ; examination of processes of identified neurons with known positions in the lobula indicated that these anti-Brp strata largely correspond to the previously described ( Fischbach and Dittrich , 1989 ) lobula layers Lo1 to Lo6 ( Figure 5A and Figure 5—figure supplement 1 ) , with Lo5 represented by two anti-Brp bands of different intensity . Comparison of the layer patterns of different LC cell types , each visualized using a specific split-GAL4 driver line , with the anti-Brp label indicated that , for a given type , layer patterns were similar throughout the lobula ( Figure 5B , C ) but differed for cells of different types ( Figure 5B–D ) . In some cases , differences between layer patterns of different LC cell types , though consistent across samples , were small ( for example , between LC6 and LC16 , LC4 and LC12 or LC25 and LC26 ) while other pairs of LC cell types occupied , except for connecting neurites , entirely non-overlapping sets of layers ( for example , LC4/LC12/LC18 compared to LC25/LC26 ) . In principle , distinct and apparently uniform layer patterns of LC neuron populations ( as shown in Figure 5B–D ) could still consist of multiple cell types with distinct arbor stratifications that were not resolved in these images . However , for nearly all LC neuron types , we found that the layer patterns of individual cells that projected to the same target glomerulus were similar to each other and matched the overall layer patterns seen by labeling the entire cell populations using the type-specific split-GAL4 driver lines ( compare LC4 and LC15 patterns in Figure 5E and D , respectively; also see Figure 7—figure supplement 1 for examples of single cell labeling of lobula dendrites for all LC cell types ) . These results indicate that nearly all optic glomeruli receive input from a single LC neuron type that can be recognized independently by either the dendritic arbor stratification in the lobula or the axonal projection pattern . As mentioned above , an exception to this apparent one-to-one correspondence between LC cell types and target glomeruli were LC10 neurons projecting to the AOTu . The majority of our LC10-specific split-GAL4 drivers are expressed in more than one of the four LC10 subtypes . However , we identified several lines with subtype selectivity , including two lines with expression in a single subtype ( see Figure 10—figure supplement 2 ) . This provides genetic support for the treatment of LC10a , LC10b , LC10c and LC10d as distinct , though related , cell types . We found that all lobula layers contain processes of at least one LC cell type , although the number of types contributing to each layer varies widely . At one extreme , only LC4 cells have processes in layer Lo1; by contrast , all but three LC ( LC4 , LC12 and LC18 ) cell types have processes in layer Lo5 . As anatomical overlap between different cells is a necessary condition for synapses between them , the different layer patterns of LC cells can provide clues to their potential connectivity: for example , some LC cell types with Lo5 dendrites might receive inputs from Tm neurons implicated in the processing of spectral information ( Gao et al . , 2008; Karuppudurai et al . , 2014; Lin et al . , 2016 ) ( Tm20 , shown in Figure 5—figure supplement 1 , is an example ) . However , since all lobula layers contain terminals of many neurons , connectivity cannot be inferred based on layer patterns alone , as not all the neuronal types that arborize in a shared layer will be synaptic partners . In addition , the lobula arbors of some LC neurons may be not only postsynaptic but also make some presynaptic contacts; we found that for several LC cell types the presynaptic marker ( syt-smHA ) used to visualize the target regions was also detectable , though much more weakly , in one or more lobula layers ( Figure 5—figure supplement 2 , Supplementary file 1A ) , suggesting that some LC neurons are also presynaptic in the lobula . Like layer patterns , lateral dendritic spread is also a stereotyped characteristic of LC neurons with functional implications . As populations , the dendrites of each of the LC neuron types form retinotopic arrays that cover the entire lobula ( see Figure 1C , F , K ) . However , the lateral spread of individual LC neuron arbors in the lobula shows considerable cell-type specific variation ( Figure 6 and Figure 6—figure supplement 1 ) . The lateral spread of the lobula arbors of cells of all LC types covered lobula regions corresponding to several of the ~750 visual columns within each eye . The largest arbor spreads were observed for LC11 , estimated to be over 60 columns ( Figure 6A and Supplementary file 1A ) and LC25 , approximately a hundred columns for some cells ( Figure 6—figure supplement 1 and Supplementary file 1A ) . The smallest arbors were those of LC10a ( ~10 columns; Figure 6B and Supplementary file 1A ) . For comparison , some of the medulla inputs to the lobula have lobula arbors corresponding to ~1 visual column ( Figure 6D ) and some lobula tangential cells span all visual columns of the lobula as single cells ( Figure 6E ) . Since most LC cell types did not show obvious retinotopy within their target glomeruli , we propose that the distinct arbor spreads of different LC cells determine the spatial extent over which their specific response properties are computed , rather than provide retinotopic information at different spatial resolutions . Taken together , our estimates of the number of LC cells of each type and their approximate dendritic arbor spread within lobula layers suggest that the dendrites of a given LC type , with the possible exception of the small arbors of LC10a , do not show a strict tiling pattern but rather overlap ( Supplementary file 1A ) . We also directly observed overlap of co-labeled cells in MCFO experiments ( some examples can be found in Figure 6—figure supplement 1 ) for nearly all LC cell types ( Supplementary file 1A ) . Overlap between processes of cells of the same type is also common in the medulla; our previous study ( Nern et al . , 2015 ) of Dm neurons using similar methods provides detailed examples of both overlapping and strict tiling patterns of medulla neurons . The morphology of single cells also revealed many additional details of LC neuron arbor structure that appear to be stereotyped within cells of the same type ( Figure 7 , Figure 7—figure supplement 1 , Figure 7—figure supplement 2 and Supplementary file 1A ) : For example , the arbor spread , even for a single cell type , can differ between layers ( examples are LC11 , LC12 , LC17 and LC18 ) and dendritic processes of some cell types appear to point in a specific direction relative to the main neurite ( examples include LC26 and LPLC2 ) . In summary , the multicolumnar arbor span ( Figure 6 and Figure 6—figure supplement 1 ) , intricate arbor shapes ( Figure 7 , Figure 7—figure supplements 1 and 2 ) and diverse and , in most cases , multistratified layer patterns ( Figure 5 , Figure 7—figure supplements 1 and 2 ) of LC neurons , suggest that these neurons spatially integrate inputs from multiple presynaptic cells of several types . These anatomical characteristics suggest that as a whole , the population of LC cell types has the potential to encode a diversity of visual stimuli , with large differences expected between distinct types . What information do LC neurons provide to the central brain ? As discussed in the Introduction , the anatomical reorganization resulting from the convergence of LC axons into largely cell-type specific glomeruli , has led to the speculation ( Mu et al . , 2012; Strausfeld and Okamura , 2007 ) that the activity of members of a given LC cell type primarily signals the presence of a particular behaviorally relevant visual feature rather than its precise location . Thus , it is reasonable to expect that strong artificial activation of an individual LC neuronal cell type might elicit a behavioral response that could provide clues to the function of these neurons . To explore this possibility , we conducted an optogenetic activation screen , using our LC cell-type specific split-GAL4 driver lines to genetically target expression of CsChrimson , a red light-activated cation channel ( Klapoetke et al . , 2014 ) , to specific LC neuron populations . We initially focused on 20 split-GAL4 driver lines , each with expression in a different LC neuron type . For LC10 cells , we chose two driver lines that together included all four LC10 subtypes described above . Driver line expression patterns were confirmed by imaging CsChrimson-mVenus expression in flies of the same genotypes used in the behavioral experiments ( Figure 2 and Figure 2—figure supplement 1 ) . We transiently depolarized these cells with a short pulse of red light in the context of two independent , complementary behavioral assays ( Figure 8A , B ) . One assay used a circular arena ( 100 mm diameter and 3 mm high ) ( Aso et al . , 2014b; Klapoetke et al . , 2014 ) , in which groups of about 25 freely walking flies ( Figure 8A and Video 3 ) were subjected to periodic , repeated light pulses . The second was a single-fly assay ( Figure 8B , see Materials and methods ) in which flies were released individually onto a small ( 5 mm2 ) open platform where they were exposed to light activation and their response recorded in two views ( side and bottom ) at high temporal and spatial resolution , permitting analysis of behaviors with greater detail . All activation experiments were carried out under infrared light to permit video recording while minimizing potential activation of CsChrimson by ambient light . 10 . 7554/eLife . 21022 . 037Video 3 . An example of an LC neuron activation phenotype ( backward walking and turning ) in the arena assay . A representative video showing groups of freely walking flies tested in the circular arena . The video is shown at 0 . 4x actual speed and the red indicators at the corners indicate the timing of optogenetic activation ( 1 s ) . This example shows a highly penetrant backward walking and turning phenotype resulting from LC16 activation . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 037 An overview of the screen results is presented in Figure 8C , D and in Supplementary file 1B . The table in Supplementary file 1B includes some additional behavioral responses that were below the 25% penetrance threshold used in Figure 8C , D . The red light pulse used for optogenetic activation itself induced weak behavioral responses in flies from a control driver line without detectable expression in the nervous system ( the ‘enhancerless’ pBDPGAL4U , [Pfeiffer et al . , 2010] ) . These control responses ( also see Klapoetke et al . [2014] ) included a slight increase in forward walking speed or turning . However , we found that red light activation of CsChrimson in LC neurons resulted in several phenotypes that were dramatically different from these baseline responses ( Videos 3–6 ) . These phenotypes included larger increases in forward walking speed or turning activity as well as three striking behaviors rarely seen in control flies , which we refer to as jumping , reaching and backward walking . 10 . 7554/eLife . 21022 . 038Video 4 . Examples of distinct LC neuron activation phenotypes in the arena assay . One representative fly for each phenotype is shown before , during and after optogenetic stimulation ( 1 s each , 3 s in total ) . Flies’ centers of mass are tracked and their trajectories are represented with blue ( before and after stimulation ) or red ( during stimulation ) dotted lines . The video is shown at 0 . 4x actual speed . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 03810 . 7554/eLife . 21022 . 039Video 5 . Examples of distinct LC neuron activation phenotypes in the arena assay . 10 representative flies for each phenotype are shown for the duration of optogenetic stimulation ( 1 s ) . Flies’ centers of mass are tracked for the duration of stimulation and their trajectories are represented with red dotted lines . Many flies with the reaching behavior also extended one or both wings in response to CsChrimson activation , though we did not further characterize this aspect of the behavior in this study . The video is shown at 0 . 2x actual speed . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 03910 . 7554/eLife . 21022 . 040Video 6 . Examples of distinct CsChrimson activation phenotypes in the single-fly assay . 10 representative flies of each genotype are shown during the 50 ms optogenetic stimulation and for the following 450 ms . Jumping and reaching phenotypes are shown in the side view whereas forward walking , backward walking and turning phenotypes are shown in the bottom view . The pBDPGAL4U control flies are shown in both the side and bottom views . In the bottom view , a red horizontal line through flies’ center of mass is used to help visualize forward walking , backward walking and turning behaviors . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 040 We found that the relationship between LC neuron type activated and observed behavior was not one-to-one . In some cases , the same behavior was elicited from activation of different LC cell types . For example , five different cell types ( LC4 , LC6 , LC15 , LPLC1 and LPLC2 ) drove highly penetrant jumping in at least one of the two assays . In other cases , multiple behaviors were observed from activation of a single LC neuron type . For example , LPLC2 activation elicited both jumping and backward walking behaviors with about equal penetrance in the arena assay . Overall , when activated , nearly half of LC neuron types ( 10/22 ) drove one of the five behaviors we assessed in a majority of flies tested . Since we did not independently confirm CsChrimson stimulation of LC neurons in a separate assay , we cannot exclude the possibility that the absence of a clear behavioral response of some LC cell types may have been simply due to insufficient activation of these neurons under our experimental conditions . The results obtained with the two assays were largely consistent but did show some differences . For example , the jumping phenotype of the LC4 and LPLC1 lines had much higher penetrance in the single-fly assay than in the arena assay whereas the opposite was the case for the forward walking phenotype of the LC24 split-GAL4 driver ( Figure 8C , D ) . We attribute these differences to design differences between the two assays; in particular , the higher intensity and shorter duration of the red light stimulus used in the single-fly assay ( 50 ms pulse of 3 mW/mm2 compared to the 1 s pulse of 94 µW/mm2 stimulus in the arena ) and the smaller area of the single-fly assay platform , which limits assessment of walking and turning behaviors . The screen results suggest that CsChrimson-mediated acute depolarization of individual LC cell types can induce diverse , cell-type specific behavioral responses . To confirm these phenotypes and pursue a more detailed analysis , we selected three LC neuron driver lines that produced robust and highly penetrant activation phenotypes in both assays: LC6 ( jumping ) , LC10 ( reaching ) and LC16 ( backward walking ) ( Figure 8C , D , Videos 3–6 ) . We used the higher resolution , single-fly assay to further resolve details of LC6 and LC10 phenotypes . LC16 phenotypes , which also include a distinct turning component ( Figure 8—figure supplement 1A–D ) , were further examined in the larger arena assay . To confirm that the jumping , backward walking , and turning phenotypes could not be attributed to the light response or unrelated differences in genetic background , we performed further control experiments with LC6 and LC16 lines . We combined our split-GAL4 driver lines with a norpA mutation that renders the fly blind . CsChrimson activation of these lines produced the same behaviors as seen in non-norpA flies ( Figure 9 and Supplementary file 1B ) , indicating that the behavior is not triggered by the light itself ( behavioral penetrance measured in genetically blind flies and experimental lines were not significantly different , binomial test , p=0 . 06 for jumping , p=0 . 10 for backward walking and p=0 . 40 for turning ) . Additionally , parental lines that do not express CsChrimson as well as flies reared on food without supplemental retinal show little or no jumping , backward walking or turning in response to red light ( mean penetrance of 0% for jumping , 7% for backward walking and 3% for turning ) ( Figure 9 , Figure 9—figure supplement 2 and Supplementary file 1B ) . We also tested several additional , genetically different , split-GAL4 driver lines with targeted LC6 or LC16 expression ( Figure 9—figure supplement 1 ) and found that these lines produced the same , highly penetrant jumping or backward walking and turning behaviors ( Figure 9 , Figure 9—figure supplement 2 and Supplementary file 1B; 80–100% penetrance for jumping , 89–95% for backward walking and 29–36% for turning ) . These results confirm that the distinct LC6 and LC16 phenotypes observed in the screen are due to the experimental activation of these cell types . Furthermore , an independent thermogenetic activation screen of more than 2000 lines from the Janelia GAL4 collection also identified the LC6 and LC16 cell types as producing the jumping and backward walking phenotypes through correlation between behavior and anatomy ( A A Robie and K Branson , personal communication , September 2016 ) . Thus , activation of specific LC neuron types can result in highly penetrant , cell-type specific behavioral responses . Activation of LC10 can result in a reaching behavior ( Figure 10A and Video 6 ) . However , the two LC10 split-GAL4 driver lines we used in our screen produced different activation phenotypes ( Figure 8C , D and Supplementary file 1B ) . Unlike other LC cell types , LC10 can be divided into four subtypes that each has a distinct layer pattern in the lobula ( Figure 5F ) , while projecting to the same target glomerulus ( Figure 3G–I ) . These different lobula layer patterns suggest different presynaptic inputs and , as a result , different visual response properties of LC10 subtypes that may be associated with different behavioral outputs . To confirm the reaching behavior observed in the screen with additional LC10 driver lines and to look for possible correlations between subtype expression and this behavior , we assayed a panel of 15 LC10 split-GAL4 driver lines ( Figure 10; Figure 10—figure supplement 1 ) . Upon activation , eight LC10 lines showed strong reaching responses ( >60% ) and six driver lines showed medium or little reaching ( <30% ) ( Figure 10B ) . Consistent with possible functional differences between LC10 subtypes in these experiments , we found that the expression patterns of the lines with the strongest reaching behavior appeared to differ from those of the other lines by having denser processes in more distal lobula layers ( mainly ~ Lo3 and Lo4 ) ( Figure 10B ) . To determine which of the four LC10 subtypes were included in each of the 15 lines used in our behavioral assay , we visualized individual LC10 cells by MCFO labeling ( Figure 10C , Figure 10—figure supplement 2 ) . Although individual LC10 cells showed considerable morphological variability , nearly all labeled LC10 cells in these lines could be readily classified as belonging to one of the four subtypes . This allowed us to determine the presence of LC10a , b , c or d in each driver line . We found LC10a and/or LC10d cells in all lines with strong reaching phenotypes ( Figure 10C , D and Figure 10—figure supplement 2 ) . These lines included one split-GAL4 driver ( OL0019B , Figure 10D ‘a only’ ) that appeared specific for the LC10a type and another driver ( SS03822 , Figure 10D ‘d only’ ) specific for the LC10d type . Control experiments confirmed that the observed reaching behavior did not result from a simple response to light or from unrelated differences in genetic background ( Figure 10D and Supplementary file 1B ) . Thus , activation of LC10a or LC10d neurons alone is sufficient to drive highly penetrant reaching behavior . The lines with weaker reaching phenotypes showed labeling of LC10b and LC10c cells and little or no expression in the other two LC10 subtypes , further suggesting that , unlike LC10a or LC10d , LC10b and LC10c neurons only have a minor , if any , role in generating the reaching behavior . The LC10 subtypes examined above have diverse layer patterns in the lobula but project to a common target region ( Figure 5F and Figure 3G–I ) . By contrast , LC6 and LC16 are examples of LC cell types that have very similar arbor stratification in the lobula , but project to distinct optic glomeruli ( Figure 11A , B , also see Figures 3 , 5 and Video 1 and 2 ) . Despite the similarity in their lobula arbors , LC6 and LC16 show different , highly penetrant , activation-induced behaviors—jumping and backward walking , respectively ( Figure 11B and Videos 4–6 ) . The ‘jumping’ phenotype of LC6 is reminiscent of fly takeoff , a behavior that occurs both spontaneously and in response to specific visual stimuli such as a predator-mimicking loom ( Card and Dickinson , 2008b ) . To further explore this similarity , we made use of the single-fly assay that provides a platform in which optogenetically triggered behaviors , such as LC6-mediated jumping , and responses to specific visual stimuli can be directly compared in identical experimental conditions . To visually elicit takeoff behavior , we presented flies with a looming stimulus previously shown to mimic a predator’s approach and trigger fast escape responses ( von Reyn et al . , 2014 ) . We compared high-speed videos of these flies to similar recordings of the jumping phenotype resulting from optogenetic depolarization of LC6 . For the flies that took off in response to the looming stimulus at 90° azimuth ( 47/174 flies ) we observed a consistent , coordinated behavioral sequence that started with the fly beginning to elevate its wings , followed by rapid middle ( jumping ) leg extension and initiation of flight as previously described ( Figure 11C ) ( Card and Dickinson , 2008b; von Reyn et al . , 2014 ) . Flies perform the same sequence when taking off voluntarily , however the elevation of their wings is slower ( Card and Dickinson , 2008a ) , resulting in a significantly longer-duration takeoff sequence than those evoked by looming stimuli . Strikingly , the CsChrimson-mediated LC6 depolarization not only produced a takeoff with this same sequence of events ( Figure 11C ) , but the duration of the takeoff sequence was more similar to that of a looming-evoked takeoff than a voluntary one performed when no stimulus was present ( Figure 11D ) . Thus , LC6 depolarization results in a behavior very closely resembling a looming-evoked escape response . In another set of experiments , we found that backward walking is also a possible response to a specific type of looming stimulus . When a fast-approaching looming stimulus was presented in front of the fly at 0° azimuth , many of the flies that did not take off during the observation period instead responded with a backward walking behavior ( 174/200 flies ) similar to that induced by optogenetic depolarization of LC16 ( Figure 11E ) . It has been shown that backward walking can also be part of a fly’s response to a predator . For example , upon close encounter with a nymphal praying mantis , wild-caught Drosophila melanogaster flies have been reported to occasionally respond with a backward walking or ‘retreat’ behavior usually followed by turning ( Parigi et al . , 2014 ) . We found that looming and LC16 depolarization both caused flies to move backward , and they do so with similar speeds . In contrast , without a visual stimulus , flies on the platform moved less , and their average movements were in a forward direction ( Figure 11F ) . These comparisons show that LC6 and LC16 activation phenotypes resemble typical , visually guided behaviors . In view of these findings , we performed additional experiments to ask whether silencing of LC6 or LC16 reduces the amount of takeoff or backward walking in response to visual stimuli . We presented the same looming stimuli used in Figure 11 to flies that expressed the inward-rectifying potassium channel Kir2 . 1 in LC6 or LC16 neurons and to control flies ( Figure 11—figure supplement 1 ) . Neither cell type appeared to be essential for the assayed behaviors: No reduction of backward walking compared to controls was observed for LC16 ( Figure 11—figure supplement 1C ) and a reduced jump frequency when blocking LC6 was apparent only for the fast loom ( Figure 11—figure supplement 1A ) . However , no difference in takeoff sequence duration was observed ( Figure 11—figure supplement 1B ) . Because multiple LC neuron types show jumping or backward walking responses to optogenetic stimulation ( Figure 8 ) and non-LC cell types have also been reported to contribute to looming-evoked escape ( de Vries and Clandinin , 2012 ) , the simplest explanation of these results is that functional redundancy exists at the level of neurons mediating looming-evoked escape . To explore whether LC6 and LC16 encode visual features , such as looming , that are sufficient to evoke jumping and backward walking ( Figure 11 ) , we investigated the visual responses of these cell types using in vivo two-photon Ca2+ imaging from head-fixed flies ( Figure 12A ) . We measured calcium responses of single LC neuron types by imaging from the axons within each glomerulus using an imaging plane selected to obtain the largest slice through the volume of the glomerulus ( Figure 12B ) . In agreement with the anatomical results showing that retinotopy is not simply preserved in the LC6 and LC16 glomeruli ( Figure 4A–C , Figure 4—figure supplement 1A–C ) , we observed that looming stimuli evoke responses in several axons that span each glomerulus ( Figure 12C ) . We quantified the population response of these axons by integrating the calcium signals within the glomerulus region . Calcium signals integrated over each of the LC6 and LC16 glomeruli both show similar increases in response to dark looming disks ( Figure 12D , top row ) . Both cell types appear similarly tuned , responding with larger calcium increments to the slower looming speeds presented ( Figure 12—figure supplement 1 , top row ) . Both cell types are selective to dark looming stimuli as a looming disk that was brighter than the background did not elicit large responses ( Figure 12D and Figure 12—figure supplement 1 , middle row ) . Looming stimuli provide compound visual cues , so to further test for the specificity of these neurons’ responses to dark looming objects , we presented a luminance-matched stimulus that darkened over time with the same temporal profile as the dark looming disk , but lacked any coherent edge motion . This stimulus elicited moderate responses in both cell types that were significantly smaller than the dark looming disk ( Figure 12D and Figure 12—figure supplement 1 , bottom row; Mann-Whitney test , p<0 . 01 for both LC6 and LC16 ) , indicating their selectivity for stimuli with looming motion . To confirm that the similar responses of LC6 and LC16 to looming stimuli were a genuine property of these cells , we performed the identical experiments on an additional LC neuron type . We measured the responses of LC11 ( Figure 12B , C ) , that was selected because its dendrites arborize in lobula layers that are distinct from LC6 and LC16 ( Figure 5 and Figure 7—figure supplement 1 ) , and because LC11 was not found to have any strong activation phenotypes across our behavioral assays ( Figure 8C , D ) . LC11 did not show calcium response changes to any of the looming-related stimuli ( Figure 12D and Figure 12—figure supplement 1 , black trace ) . However , we observed large responses from LC11 , when we presented simpler moving stimuli that did not contain looming motion . LC11 showed large responses to a small moving object , and more moderate responses to a moving bar spanning our visual display ( Figure 12E ) . In contrast , LC6 and LC16 also showed calcium responses to the small object , but these responses were smaller than those to the loom stimuli , and much smaller than those of LC11 ( Figure 12E , Mann-Whitney test , p<0 . 01 for both LC6 and LC16 ) . Taken together , these results demonstrate that LC6 and LC16 exhibit selectivity for slow , dark looming objects , while LC11 encodes a distinct set of visual features . Our results support the idea that LC cell types respond to the presence of specific visual features in the environment and that the activation of several LC neuron types can simulate the presence of such features , triggering appropriate behavioral responses . However , during most natural visually guided behaviors , flies respond not only to the general presence of a particular visual feature but also to its location . For example , flies use visual cues to orient toward the far side of a gap when attempting to cross it ( Pick and Strauss , 2005 ) or toward another fly during courtship behavior ( Coen et al . , 2016 ) . Flies also adjust their takeoff direction relative to the azimuthal orientation of a looming stimulus ( Card and Dickinson , 2008b ) and show a range of orientation behaviors that depend on the location of objects ( Aptekar et al . , 2012; Bahl et al . , 2013; Heisenberg and Wolf , 1984; Maimon et al . , 2008; Reiser and Dickinson , 2010 ) . Such orientation behaviors must depend on neural circuits that convey the spatial location of specific visual features . Indeed , simultaneous activation of most or all LC neurons of a particular type , as in the experiments described above , is probably rather uncommon under natural conditions , and may partially explain why our activation screen revealed robust phenotypes for only half of the LC cell types . Taking LC11 as an illustrative example , if a function of this population is to signal the presence of a small moving object to downstream circuits , then it is not surprising that activating all LC11 neurons simultaneously failed to result in a coherent behavioral response . While our anatomical data suggest that most LC neurons largely discard retinotopic information , the inputs from the fly’s two eyes remain distinct . In order to provide optogenetic stimuli that , similar to many naturalistic visual stimuli , differ between the two eyes , we designed a genetic approach to activate LC neurons on only one side of the brain . By using a construct in which GAL4-driven CsChrimson expression requires the prior removal of a transcriptional terminator , we can add temporal control to the cell-type specificity provided by the split-GAL4 driver ( Figure 13A , and Materials and methods ) . If excision of the transcriptional terminator is induced for a brief time window early in development—after the precursor populations for LC cells in the left and right optic lobes have been established , but before extensive cell proliferation has occurred—animals can be readily obtained in which cells specific to that GAL4 driver on only one side of the animal are labeled . As a stochastic method , this approach also generates flies with bilateral labeling or no labeling ( Figure 13B ) providing both positive and negative controls of identical genotype and experimental history . To correlate behavioral responses with expression patterns , individual flies were tested in a modified arena ( see Materials and methods ) and then retrieved to examine labeling patterns by histology . We applied this method to two LC neuron populations , one of whose activation induces an avoidance response , LC16 , and the other an approach by reaching , LC10 . Expression patterns were scored as bilateral , unilateral ( on the left or right ) , or no labeling ( Figure 13B ) . The behaviors of flies with either bilateral or no expression were consistent with what we observed in our earlier experiments ( Figure 8 and 9 ) . Unlike bilateral activation , unilateral LC16 optogenetic depolarization only rarely resulted in backward walking ( Figure 13B ) , but instead generated strong turning responses ( Figure 13C and Video 7 ) . This turning behavior had a strong directional bias with turns predominantly away from the activated side ( Figure 13C ) , suggesting a directional avoidance response . A possible interpretation of these results is that bilateral activation might represent a large object directly ahead , resulting in backward walking , while unilateral activation would mimic an object on one side of the fly , resulting in turning towards the opposite side . Interestingly , unilateral activation of LC10 also caused a strong turning bias , but in the opposite direction; flies predominantly made turns toward the activated side and the most common LC10 activation phenotypes were now turning responses rather than reaching ( Figure 13C ) . Taken together , these data show that ( 1 ) bilateral and unilateral activation of LC neurons can have different behavioral consequences and ( 2 ) unilateral activation can result in either attractive or aversive turning behaviors , depending on the cell type . These results suggest that LC neurons can convey information on both the nature of a visual feature , by means of the differential activity of different LC cell types , and its location , by the differential activation of the same type in the two optic lobes . 10 . 7554/eLife . 21022 . 041Video 7 . Examples of phenotypes upon bilateral and unilateral activation of LC16 in the stochastic activation experiment . A representative fly for bilateral and unilateral LC16 activation is shown before , during and after optogenetic stimulation ( 1 s each , 3 s in total ) . Flies’ centers of mass are tracked and their trajectories are represented with blue ( before and after stimulation ) or red ( during stimulation ) dotted lines . The video is shown at 0 . 4x actual speed . DOI: http://dx . doi . org/10 . 7554/eLife . 21022 . 041 In this report , we present anatomical and functional studies of lobula columnar ( LC ) cells , prominent visual projection neurons from the lobula to target regions in the central brain called optic glomeruli . Comprehensive anatomical analyses of the dendritic arbors and central brain projections of LC neurons support the notion that these cells encode diverse visual stimuli , distinct for each LC cell type , and convey this information to cell-type specific downstream circuits . Precise genetic tools that target individual LC cell types allowed us to explore the behavioral consequences of optogenetic activation of these cell types . We found that activating cells of single LC neuron types was often sufficient to evoke a range of coordinated behaviors in freely behaving flies . Using two-photon calcium imaging from head-fixed flies , we showed that two LC cell types with activation phenotypes similar to avoidance responses , selectively encode visual looming , a stimulus that also evokes similar avoidance behaviors , while a third cell type responded strongly to a small moving object . These results suggest that LC cell types encode visual features that are relevant for specific behaviors . Activation of LC cells in only one brain hemisphere can result in either an attractive or repulsive directional turning response , depending on cell type . Thus which LC neuron channel is activated determines the valence of the behavior , whereas comparison across the brain by two such channels of the same type provides information about the location of relevant visual features . Anatomical properties of LC neurons have been previously described both in Drosophila and other Diptera ( Fischbach and Dittrich , 1989; Otsuna and Ito , 2006; Strausfeld and Okamura , 2007 ) . Our work extends these studies by providing a comprehensive description of LC neurons in Drosophila , including the identification of several previously unreported cell types . Further , we combine these anatomical analyses with the generation of highly specific genetic markers ( split-GAL4 lines ) for each cell type . We found that each of the 22 LC types described here has morphologically distinct dendritic arbors in the lobula with stereotyped arbor stratification , size and shape . As observed in the medulla , where synapse-level connectomics data are available for many cell types ( Takemura et al . , 2013 ) , different layer patterns and arbor shapes are likely to reflect differences in synaptic connectivity and neuronal computation . Arbors of LC neurons are found in all lobula strata , though with large differences between layers . Only LC4 ( and perhaps LPLC1 and LPLC2 ) cells are potentially postsynaptic to neurons in the most distal lobula layer , Lo1 , while other strata such as Lo4 and Lo5B include processes of more than half of the LC types . The presence of at least some LC dendrites in each lobula layer implies that all of the about 50 different interneuron types that convey visual information from the medulla , and to a lesser extent from the lobula plate , to the lobula , are potentially presynaptic to some LC cells , although a far smaller number is likely presynaptic to any single LC cell type . The predicted differences in the synaptic inputs to different LC cell types also suggest that they will differ in their responses to visual stimuli . Thus , individual LC neuron types are expected to encode specific visual stimuli , while the population of all LC cell types together should signal a wide range of behaviorally relevant visual features . The visual responses of several LC cell types measured using two-photon calcium imaging ( Figure 12 ) support the expectation that different types selectively respond to different visual features . The three LC neuron types examined preferentially responded to distinct stimuli , with either a dark looming stimulus ( LC6 and LC16 ) or a small moving object ( LC11 ) evoking the strongest measured responses . LC6 and LC16 showed stronger responses to a dark expanding disc than to related stimuli such as an expanding bright disk or a darkening stimulus that lacks the expanding motion . The reduction in the LC6 and LC16 responses when the edge motion is removed from the stimulus is precisely what is expected of loom-sensitive neurons and is reminiscent of behavioral studies in houseflies showing that darkening contrast combined with edge motion is the most effective stimulus for triggering takeoffs ( Holmqvist and Srinivasan , 1991 ) . Consistent with their similar responses in the imaging experiments , LC6 and LC16 have very similar lobula layer patterns while LC11 has a different arbor stratification indicating that LC11 receives inputs from a different set of medulla cell types than LC6 and LC16 . It is likely that the selectivity for visual stimuli observed in LC neuron responses is both a property of the stimulus selectivity of their inputs—some selectivity was seen while imaging in the dendrites of a few LC cell types ( Aptekar et al . , 2015 ) —and specific computations implemented by individual LC neuron types . In addition , cells post-synaptic to the LC cells may integrate the responses of several individual LC neurons of the same type to provide more robust detection of specific visual features . For example , while LC6 and LC16 cells as populations are strongly excited by dark looming stimuli , we currently do not know whether individual LC6 and LC16 neurons , which have dendritic extents well below the maximum size of our looming stimuli , and also well below the size known to elicit maximal behavioral responses ( von Reyn et al . , 2014 ) , show the same response properties . Our anatomical data and genetic reagents provide a starting point for the additional functional and ultra-structural studies that will be required to elucidate the circuit mechanisms that produce the response properties of these and other LC cell types . The suggestion that LC cells are feature-responsive neurons has been partly based on the apparent dramatic reduction in retinotopy between LC neuron dendrites , which have a retinotopic arrangement in the lobula , and their axons , which appear to discard this spatial information as they converge onto target glomeruli ( Mu et al . , 2012; Strausfeld and Okamura , 2007; Strausfeld et al . , 2007 ) . We extended previous analyses of LC neuron arbor convergence by directly visualizing multiple single LC cells in a glomerulus in the same fly ( Figure 4 ) . These experiments revealed no detectable retinotopy of LC cell processes in most glomeruli even at this cellular level of resolution . It is possible that the responses of individual LC cells carry information about retinotopic position; given the comparatively small size of LC dendrites ( the lateral spread of even the largest LC cells covers less than 20% of visual columns ) and the retinotopic distribution of these dendrites in the lobula it would be surprising if they did not . Such retinotopic responses could for example be relevant for those LC cell types that appear to have presynaptic sites in the lobula and are thus likely to provide input to retinotopically organized circuits . However , with the caveat that we did not examine synapse-level connectivity , for most LCs the available anatomical information appears to support the view that much retinotopic information is discarded at the glomerulus level . Consistent with this anatomical observation , the calcium imaging experiments from single LC cell types revealed visual responses to localized stimuli that could be measured throughout a cross-section of the glomerulus without clear retinotopic arrangement of the responding axons ( Figure 12B , C ) . Because of the columnar nature and apparently restricted visual field of the dendrites of LC neurons , the features computed by individual LC neurons are likely to be well defined in subregions of the eye , with perhaps downstream circuits required to integrate these locally-extracted features , as discussed above for looming . We currently have little insight into how these computations are initiated in the optic glomeruli and this remains an exciting area for future investigation . Unlike the other LC neurons , we found that LC10 , and to a lesser extent LC9 , cells retain some retinotopic information in the arrangement of their axon terminals indicating that the loss of retinotopy is not a necessary consequence of axonal convergence onto a glomerular target region . More specifically , we observed that the order of LC10 axonal terminals in the AOTu along the DV axis matches the sequence of AP positions of the corresponding dendrites in the lobula . This organization could facilitate synaptic interactions of LC10 cells corresponding to different azimuthal positions in the visual field with distinct target cells . Consistent with a possible general role of the AOTu in the processing or the relaying of retinotopic information , retinotopic responses have recently been observed in the dendrites of central complex neurons ( Seelig and Jayaraman , 2013 ) that , mainly based on work in other insects ( Pfeiffer and Homberg , 2014; Pfeiffer et al . , 2005 ) , are thought to be synaptic targets of output neurons of the lateral zone of the AOTu ( Ito et al . , 2014 ) . We found that , independent of the presence or absence of retinotopy at the glomerulus level , positional information can be extracted from the differential activity of LC cells between the two optic lobes . We directly demonstrated this capability by genetically restricting optogenetic LC neuron activation to only one optic lobe . This unilateral activation evoked directional turning responses relative to the activated brain side . Thus , LC neuron signaling appears to convey information on both different visual features and their location . This may further extend the similarities to the antennal lobes where differences in odorant receptor neuron activity between the left and right antennal lobes may contribute to odorant tracking ( Gaudry et al . , 2013 ) . We found that activation of different types of LC neurons can induce distinct behaviors including jumping , reaching , wing extension , forward walking , backward walking and turning . While specific activation phenotypes have been reported for a variety of cell types and behaviors , many of these studies have focused on command-like neurons thought to orchestrate specific motor programs ( Bidaye et al . , 2014; Flood et al . , 2013; Lima and Miesenböck , 2005; von Philipsborn et al . , 2011; von Reyn et al . , 2014 ) . By contrast , the activation phenotypes we report here result from the optogenetic stimulation of different types of related visual projection neurons . A plausible interpretation of these results is that activation of LC neurons can mimic the presence of the visual features that these neurons normally respond to and thus elicits behavioral responses associated with these fictive stimuli . This possibility is supported by several lines of evidence from our studies of LC6 and LC16 . First , optogenetic depolarization of each of these cell types evokes a specific behavioral response—backward walking for LC16 and jumping for LC6—that resembles a similar natural avoidance or escape behavior ( Parigi et al . , 2014; von Reyn et al . , 2014 ) . Second , backward walking and jumping can both also be elicited by presentation of a predator-mimicking visual loom ( Card and Dickinson , 2008b ) ( and this study ) and , third , in calcium imaging experiments both LC16 and LC6 showed a preferential response to a similar looming stimulus compared to a number of related stimuli . Although we did not explore LC10 response properties , we note that LC10-activation phenotypes also show similarities to natural behaviors: movements resembling the directed foreleg extension displayed during activation-evoked reaching occur , for example , during gap-climbing behavior ( Pick and Strauss , 2005 ) and in aggressive fly-fly interactions ( Chen et al . , 2002 ) . Overall , the LC neuron activation phenotypes we observed suggest that the encoding of visual information at the level of LC neurons is sufficiently specialized to contribute to distinct behavioral responses in a cell-type dependent fashion . However , patterns of LC neuron activation that produce more refined fictive stimuli than we employed in the current work will be required to fully explore the LC neuron behavioral repertoire . Likewise , more comprehensive physiological studies of the response properties of the LC cell types will be needed . How does LC cell activation evoke specific behavioral responses ? In the simplest scenario , LC neuron depolarization could directly activate a single postsynaptic premotor descending interneuron that would then in turn trigger the observed behavior . This appears plausible in some cases: for example , activation of LC4 neurons ( called ColA cells in larger flies ) might evoke a jumping response via activation of the Giant Fiber ( GF ) cells , a pair of large descending neurons known to be postsynaptic to ColA ( Strausfeld and Bassemir , 1983 ) and LC4 ( K von Reyn and GM Card , personal communication , September 2016 ) and which have a known role in escape behavior ( von Reyn et al . , 2014; Wyman et al . , 1984 ) . For other LC cell types , there is currently no evidence suggesting a direct connection to descending neurons . For example , candidate descending neurons for the LC16 backward walking response , the moon-walker descending interneurons ( Bidaye et al . , 2014 ) , do not have dendrites in or near the LC16 glomerulus . Responses to diverse visual stimuli , some of which may derive from LC neuron activity , have also been observed in higher order brain centers without direct connections to LC neurons such as the central complex ( Seelig and Jayaraman , 2013; Weir and Dickinson , 2015 ) . Our activation experiments also provide several indications that the signaling downstream of LC neurons is likely to be more complex; for example , activation of a single LC cell type can give rise to multiple behaviors such as reaching , wing extension and turning for LC10 , or backward walking and turning for LC16 . Changes of the spatial pattern of LC neuron activation , as in our stochastic labeling experiments , can further modify activation phenotypes . For example , unilateral LC16 activation primarily evokes turning away from the location of LC16 activation , not backward walking , suggesting that the relative differences in LC16 activity between the two eyes can guide the direction of motor output through downstream signaling . Furthermore , several different LC neuron types may contribute to the same or similar behaviors , as suggested by the jumping phenotypes of LC4 , LC6 , LC15 , LPLC1 and LPLC2 . Presumably , visual signals and other information downstream of LC neurons are integrated to select appropriate behavioral actions . Such additional processing is also suggested by the cases of neurons with overlapping response properties but distinct activation phenotypes such as LC6 and LC16 . We also note that some responses to LC neuron activation appear to be context dependent; for example , we observed reduced forward walking for several LC cell types on the platform of the single-fly assay that is much smaller than the arena used in the arena assay ( Figure 8—figure supplement 1E and F , Supplementary file 1B ) . In addition , we only examined the behavior of standing or walking flies and LC neuron signaling might have different consequences depending on the behavioral state . For example , looming stimuli can also elicit avoidance responses in flying flies ( Muijres et al . , 2014; Tammero and Dickinson , 2002 ) , but these responses differ from the takeoff or retreat behaviors of walking animals . Therefore , while LC cell activity appears to convey visual information that is specialized for sets of related behavioral responses , LC neurons do not appear to instruct a single behavioral output . The most common activation phenotypes observed in our screen were apparent avoidance responses . Furthermore , in addition to the LC cells studied here , other VPNs may also contribute to avoidance behaviors ( de Vries and Clandinin , 2012 ) . This predominance of avoidance phenotypes is perhaps not unexpected . Since escape responses have to be fast and reliably executed under many different conditions , neurons that signal features that can evoke escape may be particularly likely to show phenotypes in an activation screen . Given the importance of predator avoidance for fly survival , it appears plausible that a considerable fraction of visual output neurons might be utilized for the detection of visual threats ranging from looming to small objects ( Card , 2012; Maimon et al . , 2008 ) . Furthermore , it is likely that CsChrimson-mediated depolarization of an entire population of LC neurons is more similar to the pattern of neuronal activity induced by an imminent collision , and thus responses of many individual loom-sensitive neurons , so it is not surprising that our activation screen revealed at least two looming-sensitive neuron types . The escape-inducing neurons we identified could provide inputs to different escape response pathways , such as long- and short-mode escape ( von Reyn et al . , 2014 ) , or act as multiple inputs to the same downstream circuits . Interestingly , neurons with avoidance-like activation phenotypes project to two separate groups of adjacent glomeruli , one in the dorsal PVLP ( LC6 , LC16 and also LC15 ) and one more ventral and medial ( LC4 , LPLC1 and LPLC2 ) . This spatial organization may facilitate synaptic interactions of functionally related LC neuron types with common downstream pathways for a specific behavior . The second group is close to dendritic branches of the GF , large descending neurons required for short-mode responses in Drosophila and a postsynaptic partner of LC4/ColA ( Strausfeld and Bassemir , 1983 ) ( K von Reyn and GM Card , personal communication , September 2016 ) and possibly also the two LPLC cell types . LC6 terminals do not overlap with GF dendrites and LC6 cells may play a role in the GF-independent escape pathways that have been proposed in both Drosophila ( Fotowat et al . , 2009; von Reyn et al . , 2014 ) and housefly ( Holmqvist , 1994 ) . Parallel neuronal pathways involved in escape behaviors have been identified or postulated in both vertebrates and invertebrates ( Burrows and Rowell , 1973; Fotowat and Gabbiani , 2011; Fotowat et al . , 2011; Münch et al . , 2009; Yilmaz and Meister , 2013 ) , but a contribution of several identified visual projection neurons to such pathways , as suggested by our activation screen , has not been previously reported . Different visual output neurons with distinct tuning of their response properties to looming parameters such as speed , size , luminance change or edge detection might have evolved to ensure robust responses to avoid predators or collisions . It is , however , currently not known whether LPLC1 , LPLC2 , LC4 and LC15 are indeed sensitive to looming stimuli and if so , whether their response details differ from LC16 , LC6 and each other . Nevertheless , the identification of these neurons opens the possibility to examine the potential contribution of several visual pathways to avoidance behaviors . LC neurons are a subset of the about a hundred VPN cell types that relay the output of optic lobe circuits to targets in the central brain . Our data strongly support existing proposals for LC cell types as feature-detecting neurons , which have been mainly based on the distinct anatomical properties of LC cells ( Strausfeld and Okamura , 2007 ) . While these anatomical features distinguish LC neurons from many other VPNs , an association of VPN pathways with specific behaviors is not unique to LC cell types . The notion that individual neuronal pathways are tuned for specific behavioral requirements is a prominent theme in invertebrate neuroethology , with these neurons described as ‘matched filters’ for behaviorally relevant features of the external world ( Warrant , 2016; Wehner , 1987 ) . A number of previously studied VPN pathways , outside of the LC subgroup , have been described as encoding specific behaviorally related visual stimuli . In particular , very similar to our results for LC6 and LC16 , a group of tangential cells of the lobula and lobula plate ( Foma-1 neurons ) were found to respond to looming visual stimuli and , upon optogenetic activation , trigger escape responses ( de Vries and Clandinin , 2012 ) . And perhaps most famously , the long-studied LPTCs , such as the HS and VS cells , integrate local motion signals so as to preferentially respond to global optic flow patterns that are remarkably similar to visual motion encountered during specific behavioral movements ( Hausen , 1976 , 1982a; Krapp et al . , 1998 ) . Both our results and these findings are consistent with the idea that , at the outputs of the fly visual system , we find VPN pathways whose encoding properties are already well matched to particular fly behaviors or groups of behaviors . Matching the response properties of these deep sensory circuits to behavioral needs may be a general evolutionary solution to the challenge of dealing with the complexity of the visual world with limited resources . LC neurons have long been recognized as a potential entry point for the circuit-level study of visual responses outside of the canonical motion detection pathways . We provide a comprehensive anatomical description of LC cell types and genetic reagents to facilitate such further investigations . We also show that activation of several LC cell types results in avoidance behaviors and that some of these same LC types respond to stimuli that can elicit such behaviors . Other LC neurons appear to mediate attractive behavioral responses . Our work provides a starting point for exploring the circuit mechanisms both upstream and downstream of LC neurons . Split-GAL4 transgenes were selected based on GAL4-line expression patterns ( Jenett et al . , 2012; Kvon et al . , 2014 ) ( Barry J Dickson , personal communication ) and constructed as previously described ( Pfeiffer et al . , 2010 ) . Tables with details of genotypes are included as Supplementary file 1B , 1C and 1D . Supplementary file 1B summarizes genotypes and results of behavioral experiments . Supplementary file 1C lists all LC cell types described here and the driver lines for each type; Supplementary file 1D provides details of the genetic reagents used for each panel of the anatomy Figures and for Videos 1 and 2 . With a few exceptions indicated in Supplementary file 1B and 1C , all split-GAL4 AD transgenes are inserted in attP40 and all DBD transgenes in attP2 . Drosophila melanogaster flies were reared on standard cornmeal/molasses food at 25°C and 50% humidity unless otherwise indicated . For optogenetic activation experiments in the arena and stochastic activation experiment , flies were reared on standard food supplemented with retinal ( 0 . 2 mM all-trans-retinal prior to eclosion and then 0 . 4 mM post eclosion ) at 22°C and 60% humidity in darkness . For optogenetic activation experiments in the single-fly assay , flies were reared on standard food supplemented with retinal ( 0 . 4 mM all-trans-retinal throughout ) at 22°C and 60% humidity in darkness . Flies of both sexes were used for behavioral experiments unless otherwise indicated . All anatomical analyses were done with female flies . Vitamin A-deficient food was described previously ( Nichols and Pak , 1985 ) . Briefly , per 500 ml of food: 270 ml H2O , 230 ml grape juice ( Welch's , Welch Foods Inc . , Concord , MA ) , 11 g Bacto-agar ( Difco , Franklin Lakes , NJ ) , 30 g glucose , 10 g sucrose , 5 g fructose , 10 g yeast ( Fleischmann's dry ) , 10 ml 1 M NaOH , 2 ml proprionic acid and 0 . 2 ml phosphoric acid . In addition to the split-GAL4 lines ( see Supplementary file 1B , 1C and 1D ) , the following fly strains were used: ( 1 ) 20XUAS-CsChrimson-mVenus in attP18 ( Klapoetke et al . , 2014 ) ; ( 2 ) pBDPGAL4U in attP2 , an enhancerless GAL4 driver ( Pfeiffer et al . , 2010 ) ; used as a control driver in behavioral assays; ( 3 ) pBPhsFLP::2PEST in attP3 ( Nern et al . , 2015 ) ; ( 4 ) pJFRC300-20XUAS-FRT>-dSTOP-FRT>-CsChrimson-mVenus in attP18; created by transferring the stop cassette from pJFRC177-10XUAS-FRT>-dSTOP-FRT>-myr::GFP ( Nern et al . , 2011 ) into UAS-CsChrimson-mVenus; ( 5 ) MCFO-1 ( Nern et al . , 2015 ) ; ( 6 ) MCFO-7 ( Nern et al . , 2015 ) ; ( 7 ) pJFRC200-10XUAS-IVS-myr::smGFP-HA in attP18 , pJFRC216-13XLexAop2-IVS-myr::smGFP-V5 in su ( Hw ) attP8 ( HA_V5 ) ( Nern et al . , 2015 ) ; ( 8 ) pJFRC7-20XUAS-IVS-GCaMP6m in VK00005 ( Chen et al . , 2013 ) in DL background; ( 9 ) w;; pJFRC51-3XUAS-IVS-syt::smHA in su ( Hw ) attP1 , pJFRC225-5XUAS-IVS-myr::smFLAG in VK00005 ( Nern et al . , 2015 ) ; ( 10 ) norpA[36] ( de Vries and Clandinin , 2012 ) ; ( 11 ) w+ DL; DL; pJFRC49-10XUAS-IVS-eGFPKir2 . 1 in attP2 ( DL ) ( DL denotes chromosomes derived from the DL fly strain ) ( von Reyn et al . , 2014 ) ; ( 12 ) DL ( wild-type strain from M . H . Dickinson , University of Washington ) . For screening of split-GAL4 combinations , we adopted a protocol previously applied to large-scale characterization of GAL4 expression patterns of larval fly brains ( Li et al . , 2014 ) . Brains of adult flies carrying the split-GAL4 hemidriver combination of interest and a UAS reporter ( pJFRC200-10XUAS-IVS-myr::smGFP-HA in attP18 ) were dissected in insect cell culture medium , incubated ~12–24 hr at 4°C in 10 µl 1% ( w/v ) paraformaldehyde in the same medium in 60-well or 72-well Terasaki plates , washed with PBS and subsequently attached to Poly-L-Lysine coated coverslips while immersed in PBS . Further processing was in Copeland jars in a total volume of ~10 ml . This included the following steps: 2 × 5 min in PBT ( PBS + 0 . 5% Triton X-100 ) , 1 hr in PBT with 0 . 5% Normal Goat Serum ( PBT-NGS ) , overnight at 4°C in PBT-NGS with primary antibodies ( rabbit anti-GFP 1:1000 , mouse anti-Brp or mAb Nc82 ( Wagh et al . , 2006 ) 1:50 ) , 2 × 5 min in PBT , 1 hr in PBT-NGS , overnight at 4°C in PBT-NGS with secondary antibodies , 2 × 5 min in PBT and 1 × 5 min in PBS . Brains were then post-fixed for 4 hr with 4% ( w/v ) PFA in PBS , further rinsed with PBS and subsequently dehydrated and mounted in DPX as described ( Nern et al . , 2015 ) . A detailed updated version of this screening protocol is available online ( https://www . janelia . org/project-team/flylight/protocols under ‘IHC - Adult Split Screen’ ) . We used two sets of markers to visualize split-GAL4 expression patterns . 20XUAS-CsChrimson-mVenus in attP18 ( Klapoetke et al . , 2014 ) was used to reveal the overall expression patterns of split-GAL4 lines used in behavioral experiments with this effector . For most other anatomical analyses ( except stochastic labeling ) , a combination of pJFRC51-3XUAS-IVS-Syt::smHA in su ( Hw ) attP1 and pJFRC225-5XUAS-IVS-myr::smFLAG in VK00005 ( Nern et al . , 2015 ) was used . Immunolabeling of fly brains to detect these markers together with anti-Brp as a neuropil label was performed as described ( Aso et al . , 2014a ) . Brains were mounted in DPX . Detailed protocols can be found online ( https://www . janelia . org/project-team/flylight/protocols under ‘IHC - Anti-GFP’ , ‘IHC - Polarity Sequential’ and ‘DPX mounting’ ) . Stochastic labeling of LC neurons in multiple colors was performed using Multicolor FlpOut ( MCFO ) ( Nern et al . , 2015 ) . MCFO fly stocks used for specific experiments are listed in Supplementary file 1D . MCFO samples were processed for immunolabeling of three epitope-tagged marker proteins ( smGFP-HA , smGFP-V5 and smGFP-FLAG [Viswanathan et al . , 2015] ) together with the anti-Brp reference pattern and mounted in DPX as described ( Nern et al . , 2015 ) . Detailed protocols can be found online ( https://www . janelia . org/project-team/flylight/protocols under ‘IHC - MCFO’ ) . Images were acquired on Zeiss LSM 710 or 780 confocal microscopes with 20 × 0 . 8 NA or 63 × 1 . 4 NA objectives at 0 . 62 µm x 0 . 62 µm x 1 µm ( 20x ) or 0 . 19 µm x 0 . 19 µm ( in a few cases 0 . 38 µm x 0 . 38 µm ) x 0 . 38 µm ( 63x ) voxel size . In some images ( e . g . panel A of Figure 3—figure supplement 2 ) , signal from AlexaFluor 647 or DyLight 649 dyes was also detected in the reference pattern ( Alexa488 ) channel . This crosstalk appears to be mainly due to altered spectral properties of these dyes in the DPX mounting medium rather than microscope set-up or antibody cross-reaction . Four channel MCFO images ( HA , V5 , FLAG plus anti-Brp ) were acquired as two separate stacks which were combined post-imaging . For 63x images , brain regions larger than a single field of view were imaged as up to five overlapping tiles; multiple tiles were combined ( ‘stitched’ ) . Brain alignment to a template brain was achieved as described ( Aso et al . , 2014a ) ; to facilitate alignment of 63x tiles , most samples imaged as 63x were also imaged as whole brains at 20x . Initial image processing steps applied to most or all images ( such as stitching and alignment ) were as previously described ( Aso et al . , 2014a ) . Alignment quality showed some variation between specimens and within subregions of the same specimen; samples with acceptable alignment quality in the relevant brain regions were identified by visual inspection of an overlay of the aligned anti-Brp patterns of the sample and the template brains . Some processed ( e . g . aligned or stitched ) images used for anatomical analyses were stored using a ‘visually lossless’ compression ( h5j format ) . This compression did not appear to have a detectable effect on the neuroanatomical features that were characterized using these images . Fiji ( http://fiji . sc ) and Vaa3D ( Peng et al . , 2010 ) were used for most analyses and processing of individual images . To generate specific views from three dimensional image stacks , appropriately oriented substack views were generated using the Neuronannotator mode of Vaa3D and exported as TIFF format screenshots . The scale of these images was calibrated using the known dimensions and pixel resolution of the starting image and the pixel resolution and zoom of the exported image . In some cases , multiple figure panels ( in either the same figure or different figures ) show different views of the same cells that were generated from the same image stack ( using Vaa3D ) to illustrate distinct anatomical features . To display images in similar orientations within a figure , some images were rotated or mirrored . To fill in empty space outside the original field of view in some panels with rotated images , canvas size was increased and space outside the original image filled in with zero pixels using Fiji . Some images ( for example in the overlays in Figure 3 ) were manually segmented to remove background or labeled cells or structures other than those of interest; instances of such processing are specifically noted in the respective figure legends . Overlays of aligned images were assembled in Fiji with the exception of Figure 11A for which aligned images of LC6 and LC16 were segmented and overlaid using FluoRender ( Wan et al . , 2012 ) , as previously described ( Aso et al . , 2014a ) . The figures were assembled using Adobe Indesign with some schematics generated in Adobe Illustrator . All reported anatomical features were confirmed with multiple specimens . For the analyses of glomerulus shape and location and overall lobula layer patterns , at least three brains per cell type were imaged at high resolution ( 63x ) with pJFRC51-3XUAS-IVS-Syt::smHA in su ( Hw ) attP1 and pJFRC225-5XUAS-IVS-myr::smFLAG in VK00005 as reporters ( see above ) . For the illustrations in Figures 3 and 5 , individual aligned brains were selected based on alignment quality in the regions of interest; however the features shown were also examined in unaligned samples and for at least two additional brains . The number of MCFO labeled single cells that were imaged and examined at high resolution varied: the lowest numbers was for LC26 ( 7 cells ) ; the highest numbers were for LC10 subtypes ( see Figure 10—figure supplement 2B ) . Estimates of cell numbers in Supplementary file 1A are based on manual counts using high resolution ( 63x ) confocal stacks of the indicated split-GAL4 driver lines with pJFRC225-5XUAS-IVS-myr::smFLAG in VK00005 as reporter ( see above ) . Numbers are averages of counts of cells from three or more optic lobes . Lateral arbor spreads of LC neurons within lobula layers were estimated using substack projection images similar to those shown in Figure 7—figure supplement 1 . A segmented line was drawn along the part of a lobula layer covered by a cell’s arbors and the length of this line ( measured using Fiji ) divided by the maximum length of the entire layer determined in the same way . Arbor spreads were measured along both the AP and DV axes of the lobula . To estimate visual column coverage from these numbers , we assumed a circular eye of ~750 ommatidia , a uniform distribution of the corresponding visual columns across the lobula ( i . e . ~31 columns along each lobula axis ) and treated LC neuron arbors within each layer as planar and ellipse-shaped . All of these simplifications , which are approximations of described features of the visual system ( Wolff and Ready , 1993; Meinertzhagen and Hanson , 1993 ) , limit the precision of these estimates , with larger overestimates expected for large arbors . Fly tracking videos were analyzed using either manual human annotation or software ( https://github . com/wryanw/bowl_assay_tracking_and_analysis ) written in MATLAB . The automated tracking identifies the center of mass of each fly and determines the fly’s azimuthal orientation for every frame using a template-matching algorithm . The center of mass is determined in two dimensions in the arena assay and three dimensions in the single-fly assay , with 95% of the data having an error of less than five pixels . Position and orientation information for each fly was converted to velocity ( to measure forward and backward walking ) using a smoothing filter or angular velocity ( to measure turning ) using a sliding Savitzky-Golay filter ( Orfanidis , 1996 ) . Forward and backward walking velocity was determined relative to the orientation of the fly , along its path of movement . For the arena assay , mean velocities were calculated for the full 1 s of optogenetic activation . For the single-fly assay , mean velocities for motion and turning were calculated for the first 350 ms after the onset of light stimulation during which most flies remained in the field of view on the assay’s small platform . Reaching and jumping were elicited later than 350 ms in some cases , so a 500 ms cutoff was used for these behaviors . These average motion and turning values for individual flies were used to generate the distributions seen in Figure 8—figure supplement 1 and Figure 9—figure supplement 2 . For assessing behavioral phenotype penetrance , we used thresholds that required turning rates for experimental data to be beyond the 97 . 7th percentile of control data and walking velocity to be outside of the 2 . 3rd percentile ( negative values , backward walking ) or the 97 . 7th percentile ( positive values , forward walking ) to be scored as positive . These percentiles roughly correspond to two standard deviations away from the mean in a normal distribution; we use percentiles because the distributions are non-normal . The thresholds were determined independently for the arena and the single-fly assays and are indicated by horizontal red lines in Figure 8—figure supplement 1 and Figure 9—figure supplement 2 . The thresholds were used to convert mean velocities into true/false values for forward walking , backward walking and turning . Jumping and reaching were scored manually . Jumping in the arena assay was defined as events in which a fly’s legs all left the ground , usually accompanied by the initiation of wing flapping and change in fly shape . Jumping in the single-fly assay was defined as a fly leaving the platform due to a sudden leg extension . Reaching in the arena assay was defined as when the fly pitches its body back and reaches with a forward and upward extension of the forelegs , while staying in the same location . Reaching in the single-fly assay was defined as a head-up elevation of the long body axis , with both front legs leaving the platform and at least one leg elevating above a horizontal plane at the level of the dorsal tip of the head . For box plots , the dividing line in the box indicates the median , the boxes contain the inner quartile range , the notches give the 95% confidence interval , the lines extending beyond the box include 95% of the data , and the dots beyond those lines are outliers . All flies used for calcium imaging experiments were reared under standard conditions ( 25°C , 60% humidity , 12 hr light/12 hr dark , standard cornmeal/molasses food ) , and all imaging experiments were performed on females 3–6 d post-eclosion . To image from individual lobula columnar cell-types , split-GAL4 driver lines ( LC6: OL0070B , LC16: 0L0046B , LC11: OL0015B ) were crossed to pJFRC7-20XUAS-IVS-GCaMP6m in VK00005 ( DL background ) effector line . The imaging preparation was almost identical to that described in ( Strother et al . , 2014 ) . Briefly , flies were cold anesthetized and tethered to a fine wire at the thorax using UV-curing adhesive . The two most anterior legs ( T1 ) were severed and glued down along with the proboscis to prevent grooming of the eyes and to immobilize the head . Tethered flies were glued by the head capsule into the fly holder and after addition of saline to the bath , the cuticle at the back of the head was dissected away to expose the brain . Muscles 1 and 16 were severed to reduce the motion of the brain within the head capsule , and the post-ocular air sac on the imaged side was removed to expose the optic glomeruli . The optic glomeruli were imaged using a two-photon microscope ( Prairie Ultima IV , Bruker Optics Inc . , Billerica , MA ) with near-infrared excitation ( 930 nm , Coherent Chameleon Ultra II , Coherent Inc . , Santa Clara , CA ) and a 60x objective ( Nikon CFI APO 60XW ) . The excitation power was never greater than 20 mW at the sample . Imaging parameters varied slightly between experiments but were within a small range of our typical acquisition parameters: 128 × 90 pixel resolution , and 10 Hz frame rate ( 10 . 0–10 . 5 Hz ) . LC cell axon calcium data were collected from single planes selected to capture a consistently large slice of each glomerulus . Flies were placed in the center of a modular LED display ( Reiser and Dickinson , 2008 ) on which visual stimuli were presented . The display was configured to cover 60% of a cylinder , with LEDs subtending 72° in elevation and 216° in azimuth ( maximum pixel size of 2 . 25° ) as seen by the fly in the center of the cylinder . The display consists of 574 nm peak output LEDs ( Betlux ultra-green 8 × 8 LED matrices , #BL-M12A881UG-XX , Betlux , Ningbo , China ) covered with a gel filter ( LEE #135 Deep Golden Amber ) to greatly reduce stimulus emission at wavelengths that overlap with those of GCaMP emission . The stimuli were generated using custom MATLAB scripts ( https://github . com/mmmorimoto/visual-stimuli ) . The dark loom stimulus consisted of a series of 35 disk sizes , with the edge pixel intensity interpolated to approximate a circle on the LED screen . The luminance-matched stimulus was created using the dark looming disk stimulus , spatially scrambling the location of dark pixels of each frame only within the area of the final size of the disk . The time series of looming stimuli sizes were presented based on the classic parameterization for looming stimuli assuming a constant velocity of approach . The speed of the loom is represented by a single parameter ( r/v ) that describes the ratio of the stimulus radius to its approach speed ( Gabbiani et al . , 1999 ) . All looming stimuli appear as 4 . 5° spots and increase to a maximum diameter of 54° . The experimental protocol consisted of 3 repetitions of each stimulus type presented using a randomized block trial structure . Data analysis was performed with software written in MATLAB . Motion compensation was performed by cross-correlating each frame to a reference image , using software written by James Strother ( https://bitbucket . org/jastrother/neuron_image_analysis ) . The fluorescence signal is determined within hand-drawn regions of interest selected to tightly enclose the entire slice of each glomerulus captured within the imaging plane . ΔF/F is calculated as the ratio of ( F −- F0 ) / F0 , where F is the instantaneous fluorescence signal and F0 is calculated as the 10th percentile of the fluorescence signal within a sliding 300 frame window . For combining the responses of individual flies across animals , we normalized the ΔF/F responses from each individual fly to the 98th percentile of the ΔF/F across all visual stimuli within one experiment . All responses are the mean of the mean response ( across repeated stimulus presentations ) of each of five flies . Error bars indicate mean ± SEM . All significance results presented for Ca2+ imaging were determined with the Mann-Whitney test .	Many animals rely heavily on what they can see to interact with the world around them . But how does the brain use such visual information to guide behavior ? Light-sensitive neurons in the eye cannot distinguish between the visual signals associated with , say , an approaching predator or a source of food . Yet the brain can make this distinction . Networks of neurons in the brain perform computations to extract information from a visual scene that indicates the need for a particular behavior , such as an escape response . These networks are found in regions of the brain that communicate closely with the eyes . Cells known as visual projection neurons then relay the output of these networks to more central parts of the brain . By studying visual projection neurons , it is possible to work out what the eye tells the brain , and how the brain uses this information to control behavior . The fruit fly Drosophila is a suitable model organism in which to study these phenomena . This insect shows a range of behavioral responses to visual stimuli , and can be studied using sophisticated genetic tools . Wu , Nern et al . set out to explore how a group of visual projection neurons known as lobula columnar cells help fruit flies respond appropriately to visual stimuli . Experiments revealed that individual subtypes of lobula columnar cells convey information about the presence and general location of specific visual features . Wu , Nern et al . identified a number of lobular columnar subtypes involved in triggering escape responses to specific stimuli – such as walking backwards or taking off in flight – as well as others that can trigger the flies to approach a target . A next step is to map the circuits of neurons that act upstream and downstream of lobula columnar cells . This can help to reveal how these neurons detect specific visual features and how the fly then chooses and executes an appropriate behavior in response . Such studies in flies can provide insights into general principles of how brains use sensory information to guide behavior .	[ "Abstract", "Introduction", "Results", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	Visual projection neurons in the Drosophila lobula link feature detection to distinct behavioral programs
Myelination of axons facilitates rapid impulse propagation in the nervous system . The axon/myelin-unit becomes impaired in myelin-related disorders and upon normal aging . However , the molecular cause of many pathological features , including the frequently observed myelin outfoldings , remained unknown . Using label-free quantitative proteomics , we find that the presence of myelin outfoldings correlates with a loss of cytoskeletal septins in myelin . Regulated by phosphatidylinositol- ( 4 , 5 ) -bisphosphate ( PI ( 4 , 5 ) P2 ) -levels , myelin septins ( SEPT2/SEPT4/SEPT7/SEPT8 ) and the PI ( 4 , 5 ) P2-adaptor anillin form previously unrecognized filaments that extend longitudinally along myelinated axons . By confocal microscopy and immunogold-electron microscopy , these filaments are localized to the non-compacted adaxonal myelin compartment . Genetic disruption of these filaments in Sept8-mutant mice causes myelin outfoldings as a very specific neuropathology . Septin filaments thus serve an important function in scaffolding the axon/myelin-unit , evidently a late stage of myelin maturation . We propose that pathological or aging-associated diminishment of the septin/anillin-scaffold causes myelin outfoldings that impair the normal nerve conduction velocity . Fast nerve conduction is crucial for normal motor , sensory , and cognitive abilities . In vertebrates , rapid ( ‘saltatory‘ ) signal propagation is achieved by the myelination of axons . Myelination is largely completed before adulthood , notwithstanding that adult-born oligodendrocytes can assemble new myelin sheaths throughout life ( Nave and Werner , 2014 ) . Yet , myelin is one of the most long-lived structures in the CNS ( Toyama et al . , 2013 ) . With age however , the abundance and dynamics of CNS myelin decreases ( Lasiene et al . , 2009 ) while there is an increase in the frequency of pathological redundant myelin sheaths ( Peters , 2002; Sturrock , 1976 ) , i . e . , local outfoldings of compact myelin with normal-appearing axo-myelinic interface . These myelin outfoldings emerge from the innermost , adaxonal myelin layer , a part of the non-compacted , cytoplasmic channel system through myelin . This compartment of myelin has previously been implicated in glia-axonal metabolic coupling ( Nave and Werner , 2014 ) . Analogous focal outfoldings of peripheral myelin are the pathological hallmark of tomaculous neuropathies affecting the PNS . The normal structure of myelin thus requires stabilization , which can fail upon normal aging and in myelin-related disorders . However , the molecular mechanisms that might prevent myelin outfoldings have remained unknown . A striking variety of myelin-related genes causes – when mutated in human disorders and in animal models - common pathological features including axonopathy , neuroinflammation , hypomyelination , and structural impairments affecting myelin . For example , abundant constituents of CNS myelin such as proteolipid protein ( PLP ) , cyclic nucleotide phosphodiesterase ( CNP ) , and myelin associated glycoprotein ( MAG ) , are not essential for the biogenesis of myelin per se but their deficiency in mice causes complex CNS pathology ( Edgar et al . , 2009; Griffiths et al . , 1998; Li et al . , 1994; Montag et al . , 1994 ) . The neuropathological profiles observed in these and other myelin mutants are highly overlapping , which has made it difficult to explain distinct aspects of neuropathology by the loss of individual structural myelin proteins . Here , we followed the hypothesis that mutations of single genes can have secondary consequences for the entire protein composition of myelin , which allow elucidating the molecular cause of distinct neuropathological features . We report the discovery of a previously unrecognized filamentous scaffold in the innermost layer of CNS myelin that extends longitudinally along myelinated axons . This filament is composed of distinct septin monomers ( SEPT2/SEPT4/SEPT7/SEPT8 ) and associated with the adaptor protein anillin ( ANLN ) . The formation of myelin septin filaments is a late stage of myelin maturation , thereby avoiding that the property of septins to rigidify the membranes they are associated with ( Gilden and Krummel , 2010; Tooley et al . , 2009 ) could hinder the normal developmental ensheathment of axons . Importantly , this sub-membranous septin/anillin-scaffold ( SAS ) is required for the normal structure of the axon/myelin unit in vivo as its deficiency causes a specific structural impairment of the myelin sheath , myelin outfoldings , and reduced nerve conduction velocity . These findings were possible by systematic label-free myelin proteome analysis in several models of complex neuropathology . By quantitative evaluation of electron micrographs ( Figure 1A; Figure 1—figure supplement 1A–E ) , hypomyelination was a significant feature in Plp1null and Magnull mice , swellings of the inner tongue of myelin in Magnull and Cnpnull mice , axonal spheroids in Plp1null mice , and degenerated axons in Cnpnull mice . Split myelin lamellae ( not quantified ) were observed in Plp1null mice . Together , these mouse mutants display distinct but overlapping profiles of neuropathology . Notably , myelin outfoldings ( Figure 1B; Figure 1—figure supplement 1A ) were common to all three models , suggesting a common molecular mechanism . 10 . 7554/eLife . 17119 . 003Figure 1 . Mouse mutants with complex CNS pathology exhibit distinct but overlapping changes of myelin composition . ( A ) Venn diagram summarizing CNS pathology in mice lacking the myelin proteins CNP , MAG , or PLP , according to quantitative evaluation of electron micrographs of optic nerves at P75 . Note that myelin outfoldings are common to all analyzed mutants . See Figure 1—figure supplement 1A–E for quantification of these experiments . ( B ) Electron micrographs of P75 optic nerve cross-sections showing several myelinated axons . Myelin outfoldings and associated axons are labelled with stippled lines and asterisks , respectively . Images are representative of 4 animals per genotype , as quantified in Figure 1—figure supplement 1A . ( C ) Venn diagram summarizing myelin proteome alterations in Cnpnull , Magnull , and Plp1null-mice determined by quantitative mass spectrometry . Given are the numbers of proteins exhibiting significantly changed abundance in myelin purified from the brains of respective mutants at P75 . Note that several septins are diminished in all analyzed mutants . n=3 animals per genotype . See Figure 1—source data 1 for entire dataset and Figure 1—figure supplement 1G–I for validation by immunoblot . ( D ) Immunolabelling validates diminishment of SEPT8 ( green ) in myelinated fibre tracts of Cnpnull , Magnull , and Plp1null-mice . Longitudinally sectioned optic nerves of P75 mice are shown . The axonal marker TUJ1 ( red ) was co-labelled as a control . Images are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 00310 . 7554/eLife . 17119 . 004Figure 1—source data 1 . Dataset file ( differential myelin proteome analyses ) related to Figure 1C , Figure 1—figure supplement 1F , and Figure 6E . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 00410 . 7554/eLife . 17119 . 005Figure 1—figure supplement 1 . Neuropathological and molecular analysis in mouse models of complex CNS pathology ( Cnpnull , Magnull , and Plp1null mice ) . ( A–E ) Quantification of CNS pathology as summarized in main Figure 1A . Electron micrographs of optic nerves were evaluated for the indicated pathologies . Note that myelin outfoldings are common to all three mutants . n=4 mice per genotype; age P75; one-way ANOVA with Bonferroni’s multiple comparisons test comparing to WT . p-values are as follows: ( A ) WT vs . Cnpnull p=0 . 002 , WT vs . Magnull p=0 . 02 , WT vs . Plp1null p=0 . 002; ( B ) WT vs . Cnpnull p=0 . 11 , WT vs . Magnull p=0 . 02 , WT vs . Plp1null p<0 . 001; ( C ) WT vs . Cnpnullp<0 . 001 , WT vs . Magnull p<0 . 001 , WT vs . Plp1null p=0 . 56; ( D ) WT vs . Cnpnull p<0 . 001 , WT vs . Magnull p>0 . 999 , WT vs . Plp1null p=0 . 37; ( E ) WT vs . Cnpnull p=0 . 22 , WT vs . Magnull p>0 . 999 , WT vs . Plp1null p=0 . 0012 . ( F ) Diminishment of myelin septins in models of myelin-related pathology according to label-free proteomics . Myelin was purified from the brains of the indicated mutants and littermate controls , and analyzed by quantitative mass spectrometry . Given is the abundance in myelin of SEPT2 , SEPT4 , SEPT7 , SEPT8 relative to their respective littermate controls . Note that the abundance in myelin of SEPT4 , SEPT7 , and SEPT8 , was significantly reduced in all analyzed mutants . The diminishment in myelin of SEPT2 was significant in Cnpnulland Plp1nullmice and at the border of significance in Magnullmice . See Figure 1—source data 1 for the entire dataset . n=3 mice per genotype; age P75; unpaired two-tailed t-test; p-values are as follows: SEPT2: p=0 . 012 ( Cnpnull ) , p=0 . 0614 ( Magnull ) , p=0 . 013 ( Plp1null ) ; SEPT4: p<0 . 001 ( Cnpnull ) , p=0 . 002 ( Magnull ) , p<0 . 001 ( Plp1null ) ; SEPT7: p<0 . 001 ( Cnpnull ) , p=0 . 0114 ( Magnull ) , p=0 . 009 ( Plp1null ) ; SEPT8: p=0 . 007 ( Cnpnull ) p=0 . 011 ( Magnull ) p=0 . 011 ( Plp1null ) . ( G–I ) Immunoblotting validates the diminished abundance of septins ( SEPT2 , SEPT4 , SEPT7 , SEPT8 ) in CNS myelin of all three models . Genotypes and equal loading were controlled by immunoblotting . Blots are representative of 3 animals per genotype . ( J–M ) Abundance of septin-mRNAs in models of myelin-related pathology . The abundance of the indicated mRNAs in a myelin-rich white matter tract ( corpus callosum ) of the indicated models was determined by qRT-PCR . The abundance of Sept2 , Sept7 , and Sept8 mRNA was unaltered in all models ( J , L , M ) . The abundance of Sept4 was moderately reduced in Cnpnull and Magnull corpus callosi ( K ) . Corpus callosi of n=4 animals per genotype were analyzed . One-way ANOVA with Bonferroni’s multiple comparison test comparing to WT . ( J ) WT vs . Cnpnull p=0 . 0767 , WT vs . Magnull p=0 . 3661 , WT vs . Plp1null p> 0 . 9999; ( K ) WT vs . Cnpnull p=0 . 006 , WT vs . Magnull p=0 . 0014 , WT vs . Plp1null p=0 . 175; ( L ) WT vs . Cnpnull p=0 . 45 , WT vs . Magnull , p>0 . 999 , WT vs . Plp1null p=0 . 085; ( M ) WT vs . Cnpnull p>0 . 999 , WT vs . Magnull p>0 . 999 , WT vs . Plp1null p=0 . 14 . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 005 By subjecting myelin purified from the brains of these models to label-free quantitative mass spectrometry ( Distler et al . , 2014 ) , we found distinct but overlapping alterations of the myelin proteome ( Figure 1C; Figure 1—source data 1 ) . Notably , the abundance of several septins was reduced in all analyzed mutants ( Figure 1C ) . Septins have important functions in physiology and cell division ( Dolat et al . , 2014; Fung et al . , 2014 ) , but their role in myelinating cells is unknown ( Buser et al . , 2009; Patzig et al . , 2014 ) . By mass spectrometry ( Figure 1—source data 1 ) , the most abundant septins in wild-type CNS myelin are SEPT2 , SEPT4 , SEPT7 , and SEPT8 . The abundance of all four septins was reduced in myelin of all three mutants ( Figure 1—figure supplement 1F ) , as validated by immunoblotting ( Figure 1—figure supplement 1G–I ) . Immunohistochemical analysis of optic nerves confirmed the diminishment of SEPT8 in all three mutants ( Figure 1D ) . By qRT-PCR , we could amplify cDNA fragments for Sept2 , Sept4 , Sept7 , and Sept8 from mutant and control corpus callosi with approximately similar efficiency , suggesting that transcriptional regulation is unlikely to cause the loss of myelin septins ( Figure 1—figure supplement 1J–M ) . Together , the presence of myelin outfoldings correlates with loss of myelin septins in three models of complex CNS pathology . Forming membrane-associated hetero-oligomers and higher-order structures ( Bridges et al . , 2014; Sirajuddin et al . , 2007 ) , septins can rigidify plasma membranes ( Gilden and Krummel , 2010 ) . By mass spectrometry ( Figure 1—source data 1 ) , SEPT2 , SEPT4 , SEPT7 , and SEPT8 have a molar stoichiometry of about 1:1:2:2 in normal myelin . SEPT9 was also detected , but at much lower abundance . Thus , myelin comprises septin subunits from all four homology groups , a likely prerequisite for their assembly ( Dolat et al . , 2014; Fung et al . , 2014; Sirajuddin et al . , 2007 ) . To determine the localization and higher-order structure of myelin septins , we performed immunohistochemistry and confocal microscopy of longitudinal sections of optic nerves and spinal cords . SEPT7 and SEPT8-labelling was found to parallel ( but not overlap with ) axonal neurofilament labelling ( Figure 2A–C , Video 1 ) , suggesting the presence of longitudinal septin filaments in myelin . We therefore colabelled SEPT8 and a marker for adaxonal myelin ( MAG ) . In cross sections , SEPT8-immunopositive puncta appeared contained within the ring-shaped compartment defined by MAG-immunopositivity ( Figure 2D ) . Any ring-shaped axon/myelin-unit identified by MAG-labelling exhibited between 0–3 puncta of SEPT8-labeling ( Figure 2E ) independent of the axonal diameter ( Figure 2F ) . Aiming to reveal the exact localization of SEPT8 in the adaxonal cytoplasmic ( i . e . non-compacted ) compartment of myelin at higher resolution , we used cryo-immuno electron microscopy of optic nerves . Immunogold labelling of SEPT7 and SEPT8 supported the localization in adaxonal myelin ( Figure 2G–H , Figure 2—figure supplement 1A–B ) . Interestingly , within this compartment SEPT8 immunogold was mostly associated with the innermost membrane of compact myelin ( Figure 2H‘–I ) , in difference to MAG , a transmembrane protein localized to the adaxonal membrane . 10 . 7554/eLife . 17119 . 006Figure 2 . Septins form longitudinal filaments in the adaxonal compartment of mature CNS myelin . ( A–C ) Immunofluorescent signal of SEPT8 and SEPT7 ( green ) extends longitudinally alongside axons identified by neurofilament-labelling ( red ) . All panels show maximal projections of confocal stacks , and 3-dimensional reconstructions of longitudinally sectioned WT optic nerve ( A ) or spinal cord ( B , C ) at age P75 . Images are representative of three animals . ( D–F ) SEPT8 ( green ) immunolabelling indicates that septin filaments localize to the adaxonal non-compact myelin compartment marked by MAG-immunolabelling ( red ) ( confocal micrograph , D ) . The number of filaments represented by SEPT8-puncta is plotted in relation to their frequency per axon/myelin-unit ( E ) and the axonal diameter ( F ) . Data are represented as mean ± SEM . Analysis of 4 animals; repeated-measures-ANOVA; p=0 . 0014 ( E ) , n . s . , p=0 . 26 ( F ) . ( G , H ) Immunogold-labelling of cryosections identifies the localization of SEPT8 in the adaxonal myelin compartment in longitudinally ( G ) and cross-sectioned ( H ) optic nerves . Black arrowheads point at immunogold; white arrowhead points at the inner mesaxon . Images are representative of three animals . ( H‘ ) Enlargement of the boxed area in H shows immunogold labelling of SEPT8 associated with the innermost membrane layer of compact myelin ( green false colour in the overlay ) , not with the adaxonal myelin membrane ( red false colour in the overlay ) . ( I ) Quantification of immunogold labelling of SEPT8 and MAG relative to the innermost membrane layer of compact myelin ( type 1 ) and the adaxonal myelin membrane ( type 2 ) . Note that SEPT8 immunogold labelling is associated with the innermost membrane layer of compact myelin while MAG labelling is associated with the adaxonal myelin membrane . Mean ± SEM . Analysis of 3 animals; two-tailed paired t-test; SEPT8 p=0 . 002 , MAG , p=0 . 02 . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 00610 . 7554/eLife . 17119 . 007Figure 2—figure supplement 1 . Localization of myelin septins . ( A , B ) Localization of SEPT7 and SEPT8 by cryo-immuno electron microscopy . ( a ) Immunodetection of SEPT8 ( 15 nm gold particles , white arrowheads ) together with SEPT7 ( 10 nm gold particles , black arrowheads ) on optic nerve sections . Note that SEPT8 labeling was frequently found in close proximity to SEPT7 labeling and mostly confined to the adaxonal myelin compartment . Occasional SEPT7-labeling was also observed in axons ( star ) . ( B ) As a control , SEPT7 ( 10 nm gold particles , black arrowheads ) was detected individually , also displaying labeling in adaxonal myelin . ( C , D ) Myelin septins extend from the internodal segment into the juxtaparanodal but not the paranodal segment ( C–D ) Immunohistochemistry on longitudinal spinal cord sections at P75 detecting SEPT8 together with a marker for the juxtparanodal segment ( Kv1 . 2 ) ( C ) or the paranodal segment ( CASPR ) ( D ) . Note that SEPT8-labeling occasionally co-distributed with Kv1 . 2-labeling but not with CASPR-labeling . JX , juxtaparanode; PN , paranode . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 00710 . 7554/eLife . 17119 . 008Video 1 . 3-dimensional reconstruction of a myelin septin filament ( SEPT8 immunolabelling , green ) alongside an axon ( neurofilament immunolabelling , red ) Related to Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 008 Taken together , we have identified a previously unrecognized network of longitudinal septin filaments in the adaxonal non-compact layer of CNS myelin ( Figure 3A ) . These filaments extend straight or slightly undulating for various lengths up to 20 µm . Septin filaments extend beyond the internodal segments of myelinated axons into the juxtaparanodal regions , as evidenced by occasional proximate-labelling of SEPT8 with Kv1 . 2 ( Figure 2—figure supplement 1C ) . Proximate-labelling with the paranodal marker CASPR was not found ( Figure 2—figure supplement 1D ) . 10 . 7554/eLife . 17119 . 009Figure 3 . Septins in myelin development and aging . ( A ) Scheme showing the localization of septin filaments ( red ) in healthy adaxonal myelin . Note that all experimental data support a model that myelin septins ( SEPT2 , SEPT4 , SEPT7 , SEPT8 ) assemble as longitudinal filaments in the non-compact adaxonal compartment of myelin , in which they underlie the innermost membrane of compact myelin . ( B ) Immunoblotting indicates that the abundance of several septins ( SEPT4 , SEPT7 , SEPT8 ) increases with age in myelin purified from wild-type brains at P10 , P15 , P21 , and P28 , reflecting the maturation of myelin . Note that the abundance of the classical myelin markers MBP and PLP is unaltered . ( C ) Immunolabelling of WT optic nerves detects SEPT8 ( green ) at P28 but not at P15 or P21 . Note that the myelin marker MAG ( red ) is detectable at all time points , reflecting that the optic nerve is largely myelinated by P14 . Images are representative of three experiments . ( D ) Quantitative evaluation of electron micrographs of WT optic nerves indicates that the frequency of axon/myelin-units with myelin outfoldings declines between P15 and P75 ( i . e . with myelin maturation ) and is strongly increased at 2 years of age ( i . e with normal aging ) . Mean ± SEM . n=4 animals were analyzed . One-way ANOVA with Tukey's multiple comparison test; P15 vs . P75 , p=0 . 044; P15 vs . 2 yrs , *p<0 . 001 , P75 vs . 2 yrs , p<0 . 001 . ( E ) Immunoblotting indicates that the abundance of myelin septins is decreased in myelin purified from mice at the age of two years compared to P75 , reflecting normal aging . Blot is representative of 3 animals per age . ATPase-alpha1 ( ATP1A1 ) served as a control . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 009 During oligodendroglial maturation , the abundance of Sept8-mRNA increases , but later than most myelin markers ( de Monasterio-Schrader et al . , 2012 ) . When we immunoblotted myelin ( purified from wild-type mouse brains at P10 , P15 , P21 , P28 , representing ongoing myelin maturation ) , the abundance of SEPT4 , SEPT7 , and SEPT8 increased with age whereas the abundance of the classical myelin markers PLP and myelin basic protein ( MBP ) did not change ( Figure 3B ) . When co-immunolabelling SEPT8 and MAG in optic nerves ( ages P15 , P21 , P28 ) , MAG was readily detectable at all examined ages ( Figure 3C ) . Conversely , SEPT8 was detectable at P28 but not at P15 or P21 ( Figure 3C ) . Thus , the assembly of myelin septins represents a late stage of myelin maturation coinciding with a developmentally decreased frequency of myelin outfoldings ( Figure 3D ) . Considering the increase of myelin outfoldings in the aging brain ( [Peters , 2002; Sturrock , 1976] and Figure 3D ) , we also assessed the abundance of septins ( SEPT2 , SEPT4 , SEPT7 , SEPT8 ) in myelin purified from wild-type mouse brains at two years of age . Interestingly , their abundance was decreased compared to myelin from young adult brains ( Figure 3E ) , though not as strongly as in myelin of Magnull , Cnpnull and Plp1null mice . Together , the developmental assembly of myelin septins represents a late stage of myelin maturation correlating with a reduced frequency of myelin outfoldings . On the other hand , the increase of myelin outfoldings with normal aging coincides with a reduced abundance of myelin septins . The mRNA abundance profile of Sept8 during oligodendroglial maturation coincides with that of Anln ( de Monasterio-Schrader et al . , 2012 ) , which encodes the pleckstrin homology ( PH ) domain-containing protein anillin ( ANLN ) ( Menon et al . , 2014; Oegema et al . , 2000 ) that can serve as an adaptor to recruit septins onto membranes , at least in yeast ( Liu et al . , 2012 ) . We have thus tested whether anillin is also associated with septins in myelin . When co-immunolabelling SEPT8 and ANLN in the optic nerves of mature wild-type mice , ANLN was readily detectable and co-localized with SEPT8 ( Figure 4A ) . Importantly , ANLN immunolabelling was strongly reduced in Magnull , Cnpnull and Plp1null mice , similar to that of SEPT8 ( Figure 4B ) . The diminishment of ANLN was confirmed by immunoblotting of myelin purified from the brains of these mutants ( Figure 4C ) . Together , these data imply that myelin septin filaments are associated with anillin . 10 . 7554/eLife . 17119 . 010Figure 4 . Anillin co-distributes with myelin septins . ( A ) Confocal microscopy reveals proximity-labelling of SEPT8 ( green ) and anillin ( red ) . The merge additionally shows the axonal marker Tuj1 ( white ) . Longitudinally sectioned optic nerves of P75 WT mice are shown . Example is representative of three independent experiments . ( B ) Immunolabelling indicates diminishment of anillin ( red ) in myelinated fibre tracts of Cnpnull , Magnull , and Plp1null-mice similar to SEPT8 ( green ) . The axonal marker TUJ1 ( white ) was co-labelled as a control . Longitudinally sectioned optic nerves of P75 mice are shown . Examples are representative of three independent experiments . ( C ) Immunoblotting of myelin purified from brains at P75 shows diminished abundance in Cnpnull , Magnull , and Plp1null-mice compared to WT controls . ATPase-alpha1 ( ATP1A1 ) served as a control . Blot is representative of n=3 animals . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 010 In yeast , the cytoplasmic headgroups of the membrane lipid phosphatidylinositol ( 4 , 5 ) -bisphosphate ( PI ( 4 , 5 ) P2 ) recruit ANLN to the cleavage furrow ( Liu et al . , 2012 ) , thereby mediating the submembraneous polymerization of septins ( Bertin et al . , 2010 ) . PI ( 4 , 5 ) P2 can also recruit mammalian SEPT4 onto the plasma membrane ( Zhang et al . , 1999 ) , at least in vitro . We thus tested whether PI ( 4 , 5 ) P2 also affects myelin septin assembly in vivo . To this aim , we re-investigated Ptenflox/flox;CnpCre/WT mice ( Goebbels et al . , 2010 ) , in which reduced PI ( 4 , 5 ) P2–levels in myelin ( Figure 5A–B ) cause myelin outfoldings ( Figure 5C ) via a yet unknown mechanism . Indeed , the abundance of septins ( SEPT2 , SEPT4 , SEPT7 , SEPT8 ) and ANLN was strongly reduced in myelin purified from the brains of Ptenflox/flox;CnpCre/WT mice ( Figure 5D ) . This implies that the PI ( 4 , 5 ) P2–dependent membrane-recruitment of ANLN and septins is principally conserved between the yeast cleavage furrow and murine myelin . 10 . 7554/eLife . 17119 . 011Figure 5 . Membrane phosphoinositides mediate septin/anillin assembly in myelin . ( A ) Scheme of the reaction catalyzed by the phosphatase PTEN , which converts PI ( 3 , 4 , 5 ) P3 into PI ( 4 , 5 ) P2 . The reverse reaction is mediated by phosphatidylinositol-3-kinase ( PI3K ) . Previous reports indicated that SEPT4 and ANLN can be recruited onto membranes via PI ( 4 , 5 ) P2 , at least in vitro and in yeast , respectively; however an association has not yet been demonstrated in mice in vivo . ( B ) Ptenflox/flox;CnpCremice ( Pten cKO ) provide an established model in which the Cre-mediated deletion of Pten in myelinating cells causes a reduced abundance of PI ( 4 , 5 ) P2 ( Goebbels et al . , 2010 ) . ( C ) Ptenflox/flox;CnpCremice display myelin outfoldings ( Goebbels et al . , 2010 ) . The electron micrograph of a P75 optic nerve cross-section showing several myelinated axons . A myelin outfolding and the associated axon are labelled with stippled line and asterisk , respectively . Image is representative of 4 animals per genotype . ( D ) Immunoblotting of myelin purified from mouse brains at P75 indicates that the abundance of myelin septins and anillin is reduced when oligodendroglial PTEN is lacking in Ptenflox/flox;CnpCremice ( Pten cKO ) . The blot is representative of 3 animals per genotype . ATP1A1 served as a control . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 011 To examine whether a causal relationship exists between myelin septins and myelin outfoldings , we generated mouse mutants by gene targeting ( Figure 6—figure supplement 1A ) . Promotor activity of the Sept8-gene in brains was monitored by lacZ histochemistry in Sept8lacZ/WT mice ( Figure 6—figure supplement 1C ) . LacZ-staining was observed in grey and white matter areas ( Figure 6—figure supplement 1C ) and presumably not restricted to oligodendrocytes . This prompted us to generate mice lacking SEPT8 selectively from myelinating cells ( Sept8flox/flox;CnpCre/WT , also termed conditional mutants , COND ) in addition to mice lacking SEPT8 from all cells ( Sept8null/null , also termed constitutive knockout , KO ) ( Figure 6—figure supplement 1C ) . Conditional and constitutive Sept8-mutants were born at expected frequencies , and their major myelinated fibre tracts formed normally as judged by light microscopic examination of histochemically stained myelin ( Figure 6A ) . Electron microscopy did not reveal abnormalities of developmental myelination ( Figure 6B ) , the percentage of myelinated axons ( Figure 6C ) , and myelin thickness ( Figure 6D ) . 10 . 7554/eLife . 17119 . 012Figure 6 . Myelination in mice lacking myelin-associated septins . ( A ) Silver impregnation visualizes myelinated fibre tracts in mice lacking SEPT8 from myelinating cells ( Sept8flox/flox;CnpCre/WT-mice; COND ) or from all cells ( Sept8null/null-mice; KO ) . Images are representative of three animals per genotype . See Figure 6—figure supplement 1 for generation and validation of Sept8 mutant mice . ( B ) Electron micrographs of cross-sectioned WT and Sept8flox/flox;CnpCre/WT-mice ( COND ) optic nerves fixed by high-pressure freezing indicate indistinguishable myelin ultrastructure at P15 . Images are representative of three animals per genotype . ( C ) Quantitative evaluation of electron micrographs of optic nerves at P15 and P75 reveals a normal frequency of myelinated axons in Sept8-mutant mice ( COND , KO ) . Mean +/ SEM . n=5 animals per condition; not significant ( n . s . ) according to one-way ANOVA with Tukey’s post test; see Materials and methods section for p-values . ( D ) g-ratio analysis of electron micrographs of optic nerves at P75 indicates normal myelin sheath thickness in Sept8flox/flox;CnpCre/WT-mice ( COND ) . Mean +/ SEM . Not significant ( n . s . ) according to two-way ANOVA ( p=0 . 7823 ) . ( E ) Differential myelin proteome analysis reveals that septins ( SEPT2 , SEPT4 , SEPT7 ) and anillin ( ANLN ) are diminished in myelin purified from Sept8null/null-mice ( KO ) at P75 , whereas classical myelin proteins are not affected . Mean +/ SEM . n=3 animals per genotype; two-tailed unpaired t-test p<0 . 001 . See Figure 1—source data 1 for the entire dataset . ( F ) Immunoblotting validates the lack of myelin septins and anillin ( ANLN ) in myelin purified from the brains of Sept8-mutant mice ( COND , KO ) . Tubulin-beta4 ( TUBB4 ) was detected as a control . The blot is representative of three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 01210 . 7554/eLife . 17119 . 013Figure 6—figure supplement 1 . Generation of mice lacking expression of SEPT8 . ( A ) Scheme of the engineered Sept8-allele before and after recombination ( see Materials and methods section for details ) . ( B ) Genotyping PCR identifying the indicated alleles . ( C ) Histochemistry to visualize Sept8 gene activity in Sept8lacZ/WT mice . Shown is a sagitally sectioned brain ( to the right ) and enlarged the white matter ( inset ) . Note strong labeling in cortex , hippocampus , and cerebellum , suggesting gene activity in neurons . Also note labeling in the white matter , in agreement with gene activity in oligodendrocytes . Image representative of 3 animals . ( D ) qRT-PCR to determine the abundance of myelin septin mRNAs in the white matter ( corpus callosum ) of wild-type ( WT ) and Sept8flox/flox;CnpCre/WT ( COND ) mice . Note that Sept2 , Sept4 , and Sept7 mRNAs were unaltered in abundance in COND compared to WT mice . Sept8-mRNA in COND probably represents gene activity in non-oligodendroglial cells . 4 mice per genotype; unpaired two-tailed t-test; Sept8 p<0 . 001 , Sept2 p=0 . 36 , Sept4 p=0 . 59 , Sept7 p=0 . 63 , Anln p=0 . 46 . ( E ) Immunoblot analysis of brain lysates and myelin purified from wild-type ( WT ) , Sept8flox/flox;CnpCre/WT ( COND ) , and Sept8null/null ( KO ) mice . Note that none of the three isoforms of SEPT8 ( arising from alternative splicing ) is detectable in KO mice , whereas in COND mice only the single myelin-enriched isoform is undetectable . SEPT8-signal in COND mice represents non-oligodendroglial expression . Myelin markers ( CNP , PLP ) and axonal beta3-tubulin ( TUBB3 ) were detected as controls . ( F , G ) Immunohistochemical detection of SEPT8 ( E ) or SEPT7 ( F ) ( green ) together with the axonal marker neurofilament ( NF , red ) on cross sectioned spinal cords of wild-type ( WT ) and Sept8flox/flox;CnpCre/WT ( COND ) mice at P75 . Note that punctate septin labeling was readily detectable adjacent to NF labeling in WT in agreement with localization in the adaxonal myelin compartment but not detected in COND . Representative images of three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 013 Next , we examined if SEPT8-deficiency affects the protein composition of myelin . By label-free quantitative mass spectrometry , septins and ANLN were almost undetectable in myelin purified from Sept8null/null mice whereas the abundance of all conventional myelin markers was unaltered compared to littermate controls ( Figure 6E; Figure 1—source data 1 ) . Loss of myelin septins in Sept8-mutant mice was validated by immunoblotting ( Figure 6F; Figure 6—figure supplement 1E ) and immunohistochemistry ( Figure 6—figure supplement 1F , G ) . By immunoblotting , abundance and phosphorylation of Akt and Erk were similar in Sept8null/null and control corpus callosi ( data not shown ) . By qRT-PCR , cDNA-fragments for Sept2 , Sept4 , and Sept7 , were amplified with about equal efficiency from mutant and control corpus callosi ( Figure 6—figure supplement 1D ) , implying that the loss of myelin septins ( secondary to SEPT8-deficiency ) is a posttranscriptional event . Together , these results strongly suggest that SEPT2 , SEPT4 , SEPT7 , and ANLN are degraded if not stabilized by their incorporation into a myelin septin filament , which requires SEPT8 in its core oligomer . When analyzing mature Sept8-mutants by electron microscopy , myelin outfoldings were obvious ( Figure 7A ) . However , myelin outfoldings were a significant feature only in adult Sept8-mutants , but not at the developmental age of P15 ( Figure 7B ) . We did not observe hypomyelination ( Figure 6C , D ) , axonal spheroids ( not shown ) , degenerated axons ( Figure 7C ) , inner tongue swellings ( Figure 7E ) , astrogliosis ( Figure 7—figure supplement 1A ) , or microgliosis ( Figure 7—figure supplement 1B ) . Thus , the deficiency of myelin septin filaments very specifically caused myelin outfoldings but not complex neuropathology , in marked difference to the lack of CNP , MAG , PLP , or oligodendroglial PTEN . It is noteworthy that myelin outfoldings per se do not cause neuroinflammation . 10 . 7554/eLife . 17119 . 014Figure 7 . Lack of myelin septins causes myelin outfoldings and decelerated nerve conduction . ( A ) Electron micrographs of optic nerves exemplify myelin outfoldings at P75 . Stippled lines highlight myelin outfoldings; associated axons are marked with asterisks . ( B ) Quantitative evaluation of optic nerve electron micrographs reveals progressive myelin outfoldings in adult Sept8-mutant mice ( COND , KO ) . Mean +/ SEM . n=3–6 animals per condition; significant according to two-way ANOVA for the effects of genotype ( p<0 . 0001 ) and age ( p<0 . 0001 ) ; see Materials and methods section for p-values of Tukey’s post-test . ( C–D ) Quantitative evaluation of electron micrographs of optic nerves at age P75 indicates the absence in Sept8-mutant mice ( COND , KO ) of other myelin-related pathology such as degenerated axons ( C ) or inner tongue swellings ( D ) . Mean +/ SEM . n=3 animals; not significant ( n . s . ) according to one-way ANOVA; degenerated axons p=0 . 25 ( C ) , inner tongue swellings p=0 . 92 ( D ) . ( E ) Electrophysiological measurement reveals decelerated nerve conduction in spinal cords of Sept8-mutant mice ( COND , KO ) compared to controls at 6 and 18 months of age . Mean +/ SEM . n=4–9 animals per genotype and age; two-way ANOVA shows the significance for genotype-dependency of the effect ( p<0 . 0001 ) ; for p-values of Tukey’s post-test see Materials and methods section . ( F ) Hypothetical model of pathological outfoldings of compact myelin ( green ) upon loss of myelin septins . Shown are longitudinal and cross-sectional views . Arrows indicate hydrostatic outward pressure and cytoplasm flow in adaxonal myelin ( orange ) unless stabilized by septin filaments ( red ) . Inset , the septin/anillin scaffold is assembled from core-hexamers in dependence of PI ( 4 , 5 ) P2-levels . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 01410 . 7554/eLife . 17119 . 015Figure 7—figure supplement 1 . SEPT8 deficiency does not induce astrogliosis , microgliosis or altered structure or density of the nodes of Ranvier . ( A–B ) Immunohistochemical analysis of GFAP-immunopositivity ( B ) and MAC3-immunopositivity ( B ) in the white matter ( hippocampal fimbria ) of wild-type ( WT ) , Sept8null/null ( KO ) , and Sept8flox/flox;CnpCre/WT ( COND ) mice at one year of age . Note the similarity of immunolabeling in KO compared to WT mice . Mild astrogliosis and microgliosis indicated by enhanced GFAP-immunopositivity and MAC3-immunopositivity was found in COND mice , which is independent of the Sept8-allele . Instead this is attributable to the heterozygosity of the Cnp gene introduced by the utilized Cre-driver line . This is indicated by similar GFAP-immunopositivity and MAC3-immunopositivity in COND mice compared to CnpCre/WT and Sept8null/null;CnpCre/WT ( KO;CnpCre/WT ) mice . See Materials and methods section for the procedure of quantification of the relative area of immunopositivity . n=5–7 animals per genotype; one-way ANOVA with Tukey’s multiple comparison test; p-values: ( A ) ANOVA p<0 . 001 post tests: WT vs . KO p=0 . 999 , WT vs . COND p=0 . 001 , WT vs . CnpCre/WT p<0 . 001 , WT vs . KO;CnpCre/WT p=0 . 0352 , KO vs . COND p<0 . 001 , KO vs . CnpCre/WT p<0 . 001 , KO vs . KO; CnpCre/WT p=0 . 025 , COND vs . CnpCre/WT p=0 . 999 , COND vs . KO;CnpCre/WT p=0 . 75 , CnpCre/WT vs . KO; CnpCre/WT p=0 . 622 , ( B ) ANOVA p<0 . 001 post tests: WT vs . KO p=0 . 9917 , WT vs . COND p=0 . 0323 , WT vs . CnpCre/WT p=0 . 4072 , WT vs . KO; CnpCre/WT p=0 . 0091 , KO vs . COND p=0 . 0079 , KO vs . CnpCre/WT p=0 . 18 , KO vs . KO; CnpCre/WT p=0 . 002 , COND vs . CnpCre/WT p=0 . 70 , COND vs . KO;CnpCre/WT p=0 . 98 , CnpCre/WT vs . KO;CnpCre/WT p=0 . 37 . ( C–D ) Immunohistochemical analysis of a marker of the nodes of Ranvier ( sodium channels; NaCh , green ) and a paranodal marker ( CASPR ) on longitudinal spinal cord sections of wild-type ( WT ) and Sept8null/null ( KO ) mice . Note that the node organization is well preserved and that the density of nodes is unaltered according to unpaired two-tailed t-test; n=3 animals per genotype; p=0 . 16 . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 01510 . 7554/eLife . 17119 . 016Figure 7—figure supplement 2 . Dynamics of septin filament degradation in myelin and emergence of myelin outfoldings in adult mice . ( A ) Experimental scheme of Tamoxifen injection into adult Sept8flox/flox;PlpCre-ERT2 and Sept8flox/flox control mice , immunoblotting , and transmission electron microscopic analysis ( TEM ) . wks pti , weeks post Tamoxifen injection . ( B ) Immunoblotting indicates that the abundance of myelin septins ( SEPT2 , SEPT4 , SEPT7 , SEPT8 ) is about halved in myelin purified from the brains of Sept8flox/flox;PlpCre-ERT2 ( Sept8 icKO ) mice 4 weeks pti and greatly reduced 8 weeks pti . ( C ) Electron micrograph showing several myelinated axons in the cross-sectioned optic nerve of a Sept8flox/flox;PlpCre-ERT2 mouse 18 weeks pti . A myelin outfolding is highlighted ( stippled line ) . Image is representative of 4 animals as quantified in Figure 7—figure supplement 1D . ( D ) Quantitative evaluation of optic nerve electron micrographs reveals myelin outfoldings in Sept8flox/flox;PlpCre-ERT2 ( Sept8 icKO ) mice 18 weeks pti . Mean ± SEM; n=3–4 animals per condition; significant for the effect of genotype 18 weeks pti ( p=0 . 0068 ) and not significant ( n . s . ) 4 weeks pti ( p=0 . 90 ) according to two-way ANOVA . p-values according to Tukey’s post-test . DOI: http://dx . doi . org/10 . 7554/eLife . 17119 . 016 To elucidate the dynamics of the degradation of the septin/anillin scaffold in mature myelin , we have induced recombination of the Sept8 gene by injecting Tamoxifen into adult Sept8flox/flox;PlpCre-ERT2 mice ( Figure 7—figure supplement 2A ) , thereby preventing the assembly of new myelin septin core oligomers in myelinating cells . Sept8flox/flox mice subjected to Tamoxifen injections served as controls . By immunoblot , the abundance of myelin septins ( SEPT2 , SEPT4 , SEPT7 , SEPT8 ) was about halved in myelin purified from the brains of Sept8flox/flox;PlpCre-ERT2 mice 4 weeks post Tamoxifen injection ( pti ) and greatly reduced 8 weeks pti ( Figure 7—figure supplement 2B ) , reflecting the degradation of myelin septin filaments . Using the same model we also tested whether depleting the septin/anillin-scaffold causes myelin outfoldings when induced after developmental myelination has been completed . Indeed , myelin outfoldings were evident 18 weeks pti according to electron microscopic analysis ( Figure 7—figure supplement 2C , D ) . Thus , continuous replenishment of the myelin septin scaffold is required to prevent the formation of myelin outfoldings . To test whether the septin/anillin-scaffold of myelin is relevant for CNS function in vivo , we measured nerve conduction in the spinal cord of Sept8null/null and Sept8flox/flox;CnpCre/WT mice at various ages . Indeed , conduction velocity was reduced by about 20% in Sept8-mutants ( Figure 7E ) . Considering that the density of nodes of Ranvier is unaltered in Sept8null/null mice ( Figure 7—figure supplement 1C , D ) , this finding suggests that myelin outfoldings affect saltatory conduction . While it is plausible that myelin outfoldings impede current flow along myelinated fibre tracts , we cannot formally rule out that unidentified secondary effects also contribute to decelerated conduction . Myelin outfoldings are frequently observed during developmental myelination , and these membranes have been interpreted as a reservoir for the elongation of axon/myelin-units during body growth ( Cullen and Webster , 1979; Rosenbluth , 1966; Snaidero et al . , 2014 ) . Our data reveal that the assembly of septin filaments is a late stage of myelin maturation . By forming a scaffold underlying the innermost membrane layer of compact myelin , septins prevent outfoldings of the entire stack of compact myelin layers in the adult CNS . Septin filaments assemble by associating with membranes ( Bridges et al . , 2014 ) involving the interactions of particular septin monomers and the adaptor protein anillin with PI ( 4 , 5 ) P2 ( Bertin et al . , 2010; Liu et al . , 2012; Zhang et al . , 1999 ) , a recruitment mechanism conserved between the yeast cleavage furrow and mammalian myelin . As submembraneous filaments , septins can provide structure and rigidity to the membranes they are associated with ( Gilden and Krummel , 2010; Spiliotis and Gladfelter , 2012; Tooley et al . , 2009 ) . The adaxonal myelin compartment is regularly spaced , and the cytoplasmic surfaces of the flanking membranes are in close proximity but not compacted . Adaxonal myelin differs from compact myelin by lacking MBP and instead containing oligodendroglial cytoplasm . In our model ( Figure 7F ) , the septin/anillin scaffold stabilizes compact myelin by associating with its innermost membrane surface . This architecture counteracts hydrostatic pressure and lateral membrane flow ( Gilden and Krummel , 2010 ) , which in the absence of adhesive forces would push compact myelin outward , e . g . as a consequence of the net growth of individual myelin membranes or the slow but lifelong intercalation of new myelin segments between existing sheaths ( Young et al . , 2013 ) . High PI ( 3 , 4 , 5 ) P3-levels trigger the active net growth of myelin sheaths and thus myelin thickness via the Akt/mTOR pathway ( Chan et al . , 2014; Goebbels et al . , 2010; Macklin , 2010 ) whereas high PI ( 4 , 5 ) P2-levels are required to recruit the septin/anillin-scaffold of myelin . Loss of septins in the myelin sheath is likely owing to dysregulated filament assembly or stability , which in addition to membrane-phosphoinositides and ANLN may also involve other regulatory factors ( Joberty et al . , 2001 ) . Several mechanisms are thus feasible how mutations affecting structural myelin proteins may cause the post-transcriptional decrease of septins . For example , PLP is required for a normal myelin lipid composition ( Werner et al . , 2013 ) and CNP prevents premature closure of the cytoplasmic channels through myelin ( Snaidero et al . , 2014 ) , which probably provide transport routes for molecules and vesicles into the adaxonal myelin layer ( Nave and Werner , 2014 ) . It is plausible that both phenomena may directly or indirectly affect the septin/anillin-scaffold that prevents the formation of myelin outfoldings . The axon/myelin-unit comprises various stabilizing molecules . Recently , a cytoskeletal actin/spectrin-lattice has been discovered that underlies the axonal membrane along its entire length ( Xu et al . , 2013 ) . Moreover , neuronal and glial Ig-CAMs interact along the axon-myelin interface ( Kinter et al . , 2012 ) . We note that at the sites of pathological myelin outfoldings , the axo-myelinic membrane apposition is not disrupted . However , the lack of myelin septins causes the entire stack of compacted myelin to focally detach from the adaxonal oligodendroglial membrane and to ‘fold out‘ . This reflects strong adhesive forces between adjacent layers of compact myelin mediated by myelin proteins such as MBP . The non-compact compartments of myelin are considered as cytoplasmic routes enabling the axon-supportive function of oligodendrocytes ( Nave and Werner , 2014 ) . We conclude that the septin/anillin-scaffold reported here stabilizes mature CNS myelin while allowing diffusion and transport processes in the non-compact adaxonal myelin layer to take place . Plp1null , Cnpnull , Magnull , and Ptenflox/flox;CnpCre/WT mice were described previously ( Goebbels et al . , 2010; Klugmann et al . , 1997; Lappe-Siefke et al . , 2003; Montag et al . , 1994 ) . To homogenize the genetic background , the lines were backcrossed to the C57BL/6N strain for >10 generations before breeding the experimental animals for the present study . Genotyping of the Plp1 allele was by genomic PCR with primers 1864 ( 5‘-TTGGCGGCGA ATGGGCTGAC ) , 2729 ( 5‘-GGAGAGGAGG AGGGAAACGAG ) , and 2731 ( 5‘-TCTGTTTTGC GGCTGACTTTG ) . PCR genotyping of the Cnp allele was with primers 2016 ( 5´-GCCTTCAAAC TGTCCATCTC ) , 7315 ( 5´-CCCAGCCCTT TTATTACCAC ) , 4193 ( 5´-CCTGGAAAAT GCTTCTGTCCG ) , and 4192 ( 5´-CAGGGTGTTA TAAGCAATCCC ) . PCR genotyping of the Mag allele was with primers 1864 ( 5’-TTGGCGGCGA ATGGGCTGAC ) , 7650 ( 5’-ACGGCAGGGA ATGGAGACAC ) , and 7649 ( 5’-ACCCTGCCGC TGTTTTGGAT ) . PCR genotyping of the Pten allele was with primers 5495 ( 5´-ACTCAAGGCA GGGATGAGC ) and 20515 ( 5´-CAGAGTTAAG TTTTTGAAGGCAAG ) . Embryonic stem cells ( ES ) harbouring an engineered allele of the Sept8 gene were acquired from the European Conditional Mouse Mutagenesis Program ( Eucomm ) . ES were microinjected into blastocysts derived from FVB mice , and embryos were transferred to pseudo-pregnant foster mothers , yielding 6 chimeric males . For ES clone EPD0060_3_G04 , germline transmission was achieved upon breeding with C57BL/6N-females , yielding mice harbouring the Sept8lacZ-neo allele . The lacZ/neo cassette was excised in vivo upon interbreeding with mice expressing FLIP recombinase ( 129S4/SvJaeSor-Gt ( ROSA ) 26Sortm1 ( FLP1 ) Dym/J; backcrossed into C57BL/6N ) , yielding mice carrying a Sept8flox allele . To inactivate expression of SEPT8 whole-body-wide , exon 2 of the Sept8 gene was excised in vivo upon interbreeding with mice expressing Cre recombinase under control of the adenoviral EIIA promoter ( Holzenberger et al . , 2000 ) ( backcrossed into C57BL/6N ) , yielding mice carrying a Sept8null allele . When heterozygous Sept8null mice were interbred to obtain homozygous mutants , those were born at Mendelian frequency . For simplicity , Sept8null/null mice are also termed knockout ( ‘KO‘ ) . To inactivate expression of SEPT8 in myelinating cells , exon 2 was excised in vivo upon appropriate interbreedings of Sept8flox mice with mice expressing Cre recombinase under control of the Cnp promoter ( Lappe-Siefke et al . , 2003 ) . For simplicity , Sept8flox/flox;CnpCre/WT mice are also termed conditional mutant ( ‘COND‘ ) . Corresponding control mice ( genotypes Sept8WT/WT;CnpWT/WT , Sept8flox/WT;CnpWT/WT , and Sept8flox/flox;CnpWT/WT ) are labelled ‘WT ‘throughout . For Tamoxifen-induced inactivation of SEPT8 expression in myelinating cells , Sept8flox mice were interbred with mice expressing Cre-ERT2 in myelinating cells under control of the Plp-promoter ( Leone et al . , 2003 ) . Sept8flox/flox;PlpCre-ERT2 ( inducible conditional mutant; Sept8 icKO ) and control Sept8flox/flox mice were injected with 1 mg Tamoxifen per day beginning at 8 weeks of age for five consecutive days , then after a pause of two days again for five consecutive days . Routine genotyping of the Sept8 allele was by PCR with sense primer P1 ( 5’-GAAGCAGCCA TAGAGGAGATCC; binding 5 ‘of the first loxP-site ) in combination with antisense primers P2 ( 5’-GGTGGCTTTG AACTTGCTATCC; binding the segment flanked by loxP-sites ) , and P3 ( 5’-CAGGCAGATG TATATGCAGCAG; binding to the lacZ cassette ) or P4 ( 5’-CAGGCAGATG TATATGCAGCAG; binding 3 ‘of the third loxP site ) . The PCR shown in Figure 6—figure supplement 1B was performed with P1 , P2 , and P3 . All experimental animals were progeny of heterozygous parents to facilitate that mutant animals were analyzed together with littermate controls as far as possible . Mice were kept in the mouse facility of the Max-Planck-Institute of Experimental Medicine with a 12 hr light/dark cycle and 2–6 animals per cage . All experiments were approved by the local Animal Care and Use Committee in agreement with the German Animal Protection Law . All quantifications were performed blinded to the genotypes . All graphs display mean values and standard error of the mean ( SEM ) as error bars . All statistical tests were performed using GraphPad Prism 6 . 0 . Tests were chosen depending on experimental groups and as suggested by the software . To test for variance , F-test was performed using GraphPad Prism 6 . 0 . GraphPad online test was used to detect outliers ( http://graphpad . com/quickcalcs/Grubbs1 . cfm ) ; however no outliers were identified . The levels of significance were set at p<0 . 05 ( * ) , p<0 . 01 ( ) , and p<0 . 001 ( * ) . For quantifications displayed in Figure 6C , p-values are as follows: P15: WT vs . COND p=0 . 19 , WT vs . KO p=0 . 19; P75: WT vs . COND p=0 . 64 , WT vs . KO p=0 . 55 . For quantifications displayed in Figure 7B , p-values for genotype-dependent comparisons are as follows: P15: WT vs . COND p=0 . 29 , WT vs . KO p=0 . 35 , COND vs . KO p=0 . 99; P75: WT vs . COND p<0 . 0001 , WT vs . KO p<0 . 0001 , COND vs . KO p=0 . 96; 6 mo: WT vs . COND p<0 . 0001 , WT vs . KO p<0 . 0001 , COND vs . KO p=0 . 14; 12 mo: WT vs . COND p<0 . 0001 , WT vs . KO p<0 . 0001 , COND vs . KO p<0 . 001 . p-values for age-dependent comparisons in Figure 7B are as follows: WT: P75 vs . P15 p=0 . 04 , 6mo vs . P15 p=0 . 5074 , 12mo vs . P15 p=0 . 98 , 6mo vs . P75 p=0 . 43 , 12mo vs . P75 p=0 . 04 , 12mo vs . 6mo p=0 . 39; COND: P75 vs . P15 p<0 . 001 , 6mo vs . P15 p<0 . 0001 , 12mo vs . P15 p<0 . 0001 , 6mo vs . P75 p=0 . 12 , 12mo vs . P75 p<0 . 0001 , 12mo vs . 6mo p<0 . 0001; KO: P75 vs . P15 p<0 . 001 , 6mo vs . P15 p<0 . 0001 , 12mo vs . P15 p<0 . 0001 , 6mo vs . P75 p=0 . 003 , 12mo vs . P75 p<0 . 0001 , 12mo vs . 6mo p<0 . 0001 . For quantifications displayed in Figure 7E , two-way ANOVA shows the significance for genotype-dependency of the effect ( p<0 . 0001 ) , but not for age-dependency ( p=0 . 113 ) . p-values of Tukey's multiple comparison tests are as follows: 6mo: WT vs . COND p=0 . 03 , WT vs . KO p=0 . 04 , COND vs . KO p=0 . 99; 18mo: WT vs . COND p=0 . 006 , WT vs . KO p=0 . 002 , COND vs . KO p=0 . 998 . For other p-values , see the respective figure legends . Immungold labelling of cryosections was performed as described ( Werner et al . , 2007 ) . Optic nerves dissected from male WT ( C57Bl/6N ) mice at P75 were used . Antibodies were specific for SEPT8 ( ProteinTech Group 11769-1-AP , 1:50 ) ( yielding similar results as a previously published antibody ( Ihara et al . , 2003 ) kindly provided by M . Kinoshita , 1:50 ) , SEPT7 ( IBL18991 , 1:50 ) , or MAG ( Chemicon MAB1567 , 1:50 ) . To quantitatively assess the exact protein localization within the adaxonal non-compact myelin layer , all gold particles located to this compartment on the micrographs ( a minimum of 200 gold particles per animal ) were assigned to one of two categories as schematically depicted in Figure 2H': associated with the innermost membrane layer of compact myelin ( type 1 ) or associated with the adaxonal myelin membrane ( type 2 ) . Preparation of samples for transmission electron microscopy by high pressure freezing and freeze substitution was performed essentially as described ( Möbius et al . , 2010 ) . Briefly , optic nerves were dissected and placed into aluminum specimen carriers with an indentation of 0 . 2 mm . The remaining space was covered with 20% PVP ( Sigma ) in PBS , and the sample was cryofixed using a HPM100 high-pressure freezer ( Leica ) . Freeze substitution was carried out in a Leica AFS ( Leica , Vienna , Austria ) as follows: samples were initially kept in tannic acid ( 0 . 1% in acetone ) at −90°C for 100 hr , washed with acetone ( 4 × 30 min , −90°C ) and then transferred into OsO4 ( EMS; 2% ) and uranyl acetate ( SPI Chem , 0 . 1% ) in acetone at −90°C . The temperature was raised from −90 to −20°C in increments of 5°C/hr , then kept unaltered at −20°C for 16 hr , and then raised to +4°C in increments of 10°C/hr . The samples were then washed with acetone ( 3 × 30 min at 4°C ) , allowed to adjust to room temperature , and finally transferred into Epon ( Serva ) ( 25% , 50% , and 75% Epon in acetone for 1–2 hr each , 90% Epon in acetone for 18 hr , 100% Epon for 4 hr ) . The samples were placed in an embedding mold and polymerized ( 60°C , 24 hr ) . Preparation of samples for transmission electron microscopy by conventional aldehyde fixation was performed as described ( Möbius et al . , 2010 ) . Briefly , mice were perfused with 2 , 5% glutaraldehyde and 4% formaldehyde in phosphate buffer containing 0 . 5% NaCl . Spinal cord samples , optic nerves , and brain samples were postfixed in 1% OsO4 in 0 . 1 M phosphate buffer and embedded in epoxy resin ( Serva ) . Ultrathin sections ( 50 nm ) were cut using a Leica Ultracut S ultramicrotome ( Leica ) and contrasted with an aqueous solution of 4% uranyl acetate ( SPI Chem ) followed by lead citrate . The samples were examined in a LEO 912AB Omega transmission electron microscope ( Zeiss ) . Pictures were taken with an on-axis 2048 × 2048-CCD-camera ( TRS ) . For assessment of pathology , mice were analyzed at postnatal day 75 ( P75 ) unless indicated otherwise; 3–6 male mice were used per genotype and age . 10–15 randomly selected , non-overlapping images were taken per optic nerve at 7000 × magnification ( 1 field=220 µm ) . Electron micrographs were analyzed using ImageJ ( Fiji ) . To quantify axonal pathology and the proportion of nonmyelinated axons , all axons on the micrographs ( a minimum of 700 axons per animal ) were assigned to one of five categories: healthy-appearing myelinated axons , healthy-appearing nonmyelinated axons , axons with myelin comprising a swollen adaxonal compartment ( inner tongue ) , axonal spheroids , and degenerated axons . Axons were counted as myelinated if ensheathed by at least one complete layer of compacted myelin . Axonal changes were identified by mild signs of pathology including invaginations of adaxonal myelin membrane into the axon . Degenerated axons were identified by tubovesicular structures and amorphous cytoplasm . The area occupied by myelin outfoldings was quantified by a point counting method ( Edgar et al . , 2009 ) . Briefly , a regular grid of 0 . 25 µm2 was placed on the images . The number of intercepts coinciding with myelin outfoldings was related to the evaluated area . When quantifying myelinated axons , axonal pathology , axonal diameters , and myelin outfoldings , to compare Cnpnull , Magnull , and Plp1null mice with WT mice significance was determined using GraphPad Prism 6 . 0 . To compare SEPT8 mutant mice with WT littermates , significance was determined using GraphPad Prism 6 . 0 . The g-ratio was calculated as the ratio between the axonal Feret diameter and the Feret diameter of the corresponding myelin sheath . To determine g-ratios , a regular grid was placed on the images for randomization . All axons crossed by the grating were assessed , yielding a minimum of 100 myelinated axons per animal . The g-ratio was assessed using GraphPad Prism 6 . 0 . All quantifications were performed blinded to the genotype . Immunohistochemistry to determine neuropathology on sections of paraffin-embedded brains and spinal cords was essentially as described ( de Monasterio-Schrader et al . , 2013 ) . Antibodies were specific for MAC3 ( 1:400; Pharmingen ) or glial fibrillary acidic protein ( GFAP; 1:200; NovoCastra ) . Images were captured at 20x magnification using a bright-field light microscope ( Zeiss AxioImager Z1 ) coupled to a Zeiss AxioCam MRc camera controlled by Zeiss ZEN 1 . 0 software and processed using ImageJ 1 . 46 and Adobe Photoshop . Mice were 12 months old; sections from 4–6 male mice per genotype including the mean of both fimbriae were quantified . To quantify white matter area immunopositive for MAC3 or GFAP , the hippocampal fimbria was selected and analyzed using an ImageJ plugin for semiautomated analysis . Data were normalized to wild-type levels . All quantifications were performed blinded to the genotype . Statistical analysis was performed using GraphPad Prism 6 . 0 . Silver impregnation of myelin on histological sections was as described ( Gallyas , 1979 ) . Images were captured at 10x magnification ( Zeiss AxioImager Z1 ) and stitched using Zeiss Zen2011 . Figure 6A shows sections from mice of the indicated genotypes at one year of age . To visualize cells with Sept8 gene activity , we performed lacZ immunohistochemistry on the brains of heterozygous Sept8lacZmice . Mice were perfused with 4% PFA . Vibratome sections of 100 µm thickness ( Leica VT 1000S ) were incubated at 37°C with X-gal solution ( 1 . 2 mg X-gal per ml , 5 mM potassium ferricyanide , 5 mM potassium ferrocyanide , and 2 mM MgCl2 in PBS ) . After the sections had incubated for about 2 hr in the dark , they were rinsed with PBS , dried , and mounted using Aqua-Poly/Mount ( Polysciences ) . Images of the whole brain were captured at 10x magnification ( Zeiss AxioImager Z1 ) and stitched ( Zeiss Zen2011 ) ; the cerebellum was imaged at 40x magnification . To determine the expression and localization of myelin septins , immunohistochemistry was performed on cryosectioned optic nerves and spinal cords . Blocking solution contained 10% horse serum , 0 . 5% triton X-100 , and 1% BSA in PBS . Antibodies were specific for SEPT7 ( IBL18991; 1:1000 ) , SEPT8 ( ProteinTech Group 11769-1-AP; 1:500 ) , ANLN ( Acris AP16165PU-N; 1:200 ) , TUJ1 ( Covance; 1:1000 ) , neurofilament ( SMI31; Covance; 1:1500 ) , myelin-associated glycoprotein ( MAG 513; Chemicon; 1:50 ) , voltage-gated sodium channel Nav1 , 6 ( alomonelabs; 1:500 ) , or contactin-associated protein ( CASPR; Neuromabs; 1:500 ) . Secondary antibodies were donkey α-rabbit-Alexa488 ( dianova ) , donkey α-goat-Cy3 ( dianova ) , and donkey α-mouse Dyelight633 ( Yo-Pro ) . Images were obtained by confocal microscopy ( Leica SP5 ) . The signal was collected sequentially with the objective HCX PL APO CS 63 . 0 × 1 . 30 GLYC 21°C UV . An argon laser with the excitation of 488 nm was used to excite the Alexa488 fluorophore , and the emission was set to 500–573 nm . The laser DPSS 561 was used to excite the Cy3 fluorophore , and the emission was set to 573–630 nm . The HeNe laser 633 was used to excite Dyelight633 , and emission was detected between 645–738 nm . DAPI was excited with 405 nm and collected between 417–480 nm . The LAS AF lite and Fiji were used to export the images as tif-files . Imaris was used for 3D-reconstructions . The number of SEPT8-puncta ( in Figure 2D–F ) was determined using Fiji . Axonal diameter was determined using a threshold for the neurofilament immunolabelling signal and the particle analyzer plugin . SEPT8-punctae per myelinated axon as identified by the MAG-immunopositive ring were counted in WT at P75 ( n=4 animals , 1 section each ) . Per animal , 6 random confocal micrographs of spinal cord white matter with a size of 2500 µm2 per micrograph were quantified , yielding about 100 axons per animal . Statistical analysis was performed using GraphPad Prism 6 . 0 . For the quantifications of nodal density , the frequency of occurrence of two CASPR-immunopositive paranodes was analyzed using Fiji . CASPR-immunopositivity was converted using a threshold and counted using ITNC plugin ( n=3 animals per genotype , 1 section each , 8 random confocal micrographs of spinal cord white matter with a size of 2500 µm2 per micrograph ) . Statistical analysis was performed using GraphPad Prism 6 . 0 . A light-weight membrane fraction enriched for myelin was purified from mouse brains by sucrose density centrifugation and osmotic shocks as described ( Jahn et al . , 2013 ) . For immunoblot analyses of myelin during development and aging , male wild-type ( C57Bl/6N ) mice of the indicated ages were used . For proteome or immunoblot analyses of mutant mice and their respective wild-type littermates , male mice at the age of 75 days were used . Protein concentrations were determined using the DC protein assay ( BioRad ) . Differential quantitative myelin proteome analyses were based on previously described label-free , gel-free procedures ( Distler et al . , 2014; Jahn et al . , 2013; Patzig et al . , 2011 ) . Briefly , brains were dissected from three male Cnpnull , Magnull , Plp1null , or Sept8null/null mice at the age of 75 days and three respective control littermates , and myelin was biochemically purified ( see above ) . Per sample , 25 µg of total myelin protein was precipitated using the ProteoExtract Protein Precipitation Kit ( Calbiochem ) , solubilized in 50 mM NH4HCO3 with 0 . 1% Rapigest ( Waters ) ( 10 min , 85°C ) , reduced with 5 mM DTT ( 45 min , 56°C ) , and alkylated with 15 mM iodoacetamide ( 45 min , room temperature ) in the dark . Solubilized proteins were digested with 0 . 5 µg of sequencing grade trypsin ( Promega ) for 16 hr ( 37°C ) . After digestion , Rapigest was hydrolyzed by adding 10 mM HCl , the resulting precipitate was removed by centrifugation ( 13 , 000 g , 15 min , 4°C ) , and the supernatant was transferred into an autosampler vial . Data for the analysis of myelin purified from Cnpnull , Plp1null , and Sept8nul/nulll mice and the respective littermate controls were obtained using separation of tryptic peptides by nanoscale ultraperformance liquid chromatography ( nanoUPLC ) coupled to a quadrupole time of flight ( QTOF ) Premier mass spectrometer ( Waters ) with an alternating low and elevated ( E ) energy mode of acquisition ( MSE ) ( UPLC-MSE ) . Data for the analysis of myelin purified from Magnull mice and littermate controls were obtained using a Synapt G2S mass spectrometer ( Waters ) . For each biological replicate , 2–4 technical replicates were measured . The continuum LC-MSE data were processed and searched using the IDENTITYE-algorithm of ProteinLynx Global Server ( PLGS ) version 2 . 3 ( Waters ) . Protein identifications were assigned by searching the Uni-ProtKB/Swiss-Prot Protein Knowledgebase Release 52 . 3 for mouse proteins ( 12 , 920 entries ) supplemented with known possible contaminants ( porcine trypsin , human keratins ) using the precursor and fragmentation data afforded by LC-MS acquisition as described previously ( Patzig et al . , 2011 ) . Mass tolerances for peptide and fragment ions were set at 15 and 30 ppm , respectively . Peptide identifications were restricted to tryptic peptides with no more than one missed cleavage . Carboxyamidomethylation of Cys was set as fixed modification , and oxidation of Met , acetylation of protein N termini , and deamidation of Asn and Gln were searched as variable modifications . For a valid protein identification , the following criteria had to be met: at least two peptides detected together with at least seven fragments . The false-positive rate for protein identification was set to 1% based on search of a 5x randomized database , which was generated automatically using PLGS 2 . 3 by reversing the sequence of each entry . By using the replication rate as a filter , the false-positive rate was further minimized , as false-positive identifications do not tend to replicate across injections due to their random nature . By requiring a protein identification to be made in at least three technical replicates , the effective false-positive rate was lowered to <0 . 2% . For label-free absolute quantification of protein abundance by mass spectrometry based on the TOP3-method ( Silva et al . , 2006 ) , data were post-processed using ISOQuant software ( Distler et al . , 2014 ) . The relative abundance of a protein in myelin was accepted as altered if both , significant according to unpaired two-tailed t-test and exceeding a threshold of 25% . Immunoblotting was performed as described ( Werner et al . , 2007 ) . Antibodies were specific for SEPT2 [ProteinTech Group; 1:500; yielding similar results as a previously described antibody ( Buser et al . , 2009] ) , SEPT4 ( IBL; 1:500 ) , SEPT7 ( IBL; 1:5000 ) , SEPT8 ( ProteinTech Group; 1:2000 ) , ANLN ( Acris AP16165PU-N; 1:1000 ) , MAG ( Covance; 1:500 ) , myelin basic protein ( MBP; DAKO; 1:500 ) , PLP/DM20 ( A431; 1:5000 ) , cyclic nucleotide phosphodiesterase ( CNP; Sigma; 1:1000 ) , myelin-oligodendrocyte glycoprotein ( MOG; 1:5000; kindly provided by C . Linington , Glasgow ) , ATPaseα1 ( abcam; 1:2500 ) , actin ( Milipore; 1:5000 ) , beta3-Tubulin ( TUBB3/Tuj1; Sigma; 1:5000 ) , or beta4-Tubulin ( TUBB4; Sigma; 1:500 ) . Antibodies specific for phosphorylated or total Akt or Erk1/2 were obtained as a kit and applied as suggested by the manufacturer ( CellSignaling , Leiden , The Netherlands ) . Secondary HRP-coupled anti-mouse , -rabbit , or -goat antibodies were from dianova . Immunoblots were scanned using the Intas ChemoCam system . qRT-PCR was essentially as described ( de Monasterio-Schrader et al . , 2013 ) . Briefly , corpus callosi of 10-week-old male mice of the indicated genotypes were homogenized in TRIzol ( Invitrogen ) using Polytron PT 3100 ( Kinematica ) . RNA was extracted and purified using RNeasy Miniprep kit ( Qiagen ) . The integrity of purified RNA was confirmed using the Agilent RNA 6000 Nano kit and the Agilent 2100 Bioanalyser ( Agilent Technologies ) . cDNA was synthesized using random nonamer primers and the SuperScript III RNA H Reverse Transcriptase ( Invitrogen ) . Quantitative RT-PCR was performed using the Power SYBR Green PCR Master Mix ( Applied Biosystems ) and 7500 Fast Real-Time PCR system ( Applied Biosystems ) . mRNA abundance was analyzed in relation to the mean of the standards Ppia and Rps13 , which both did not differ between genotypes . Statistical analysis was performed in GraphPad Prism 6 . 0 . Primers were specific for Sept2 ( forward 5‘-TCCTGACTGA TCTCTACCCAGAA , reverse 5‘-AAGCCTCTAT CTGGACAGTTCTTT ) , Sept4 ( forward 5‘-ACTGACTTGT ACCGGGATCG , reverse 5‘-TCTCCACGGT TTGCATGAT ) , Sept7 ( forward 5‘-AGAGGAAGGC AGTATCCTTGG , reverse 5‘-TTTCAAGTCC TGCATATGTGTTC ) , Sept8 ( forward 5‘-CTGAGCCCCG GAGCCTGT , reverse 5‘-CAATCCCAGT TTCGCCCACA ) , Anln ( forward 5’-ACAATCCAAG GACAAACTTGC , reverse 5’- GCGTTCCAGG AAAGGCTTA , Ppia ( forward 5‘-CACAAACGGT TCCCAGTTTT , reverse 5‘-TTCCCAAAGA CCACATGCTT ) , and Rps13 ( forward 5‘-CGAAAGCACC TTGAGAG GAA , reverse 5‘-TTCCAATTAG GTGGGAGCAC ) . Nerve conduction velocity in the CNS was measured in vivo on 39 adult male mice ( 23 mice at the age of 6 months: 9 WT , 7 KO , and 7 COND; 16 mice at the age of 18 months: 6 WT , 6 KO , 4 COND ) . Electrophysiology was essentially as described ( Dibaj et al . , 2012; Steffens et al . , 2012 ) , with moderately modified surgery procedure . Briefly , anaesthesia was initiated by 80 mg/kg pentobarbital injected i . p . The rectal body temperature was measured and kept at 37°C by a heated plate . After cannulation of the jugular vein , anesthesia was continued with 40–60 mg methohexital per kg and hour . Tracheotomy was performed , and a tube for artificial ventilation was inserted . Active respiratory movements were abolished by paralysis with pancuronium ( 800 µg per kg supplemented i . p . every hour ) and artificial ventilation with a gas mixture of CO2 ( 2 . 5% ) , O2 ( 47 . 5% ) , and N2 ( 50% ) at 120 strokes/min ( 100–160 µl/stroke depending on the body weight ) . The vertebral column was rigidly fixed with two custom-made clamps . Electrocardiograms were monitored throughout . Changes of heart rate and temperature were used to control the anaesthetic state . Blood O2 saturation was monitored by a sensor in the inguinal region ( MouseOx system , Starr Live Sciences Corp . ) . Laminectomy was performed from vertebrae TH13 to L5 to expose spinal cord segments L1-L4 and dorsal roots L3-L5 . For electrophysiology , the spinal cord was covered with mineral oil . Stimulation and recording were performed with bipolar platinum electrodes . Rectangular constant voltage pulses were used with a duration of 0 . 1 ms to stimulate dorsal root L4 . Supramaximal stimulation strength of 5T was used , which is 5 times the electrical threshold for the lowest threshold fibres ( T=threshold; thresholds were frequently tested and adjusted during the experiment ) . Recording was performed with a sampling rate of 50 kHz on the ipsilateral fasciculus gracilis at spinal cord level L1 . The signal was appropriately pre-amplified and filtered . At the end of the experiment , the distance between the stimulation electrode ( cathode ) and the recording electrode was measured in situ by using a thin cotton thread .	Normal communication within the brain or between the brain and other parts of the body requires information to flow quickly around the nervous system . This information travels along nerve cells in the form of electrical signals . To speed up the signals , a part of the nerve cell called the axon is frequently wrapped in an electrically insulating sheath made up of a membrane structure called myelin . The myelin sheath becomes impaired as a result of disease or ageing . In order to understand what might produce these changes , Patzig et al . have used biochemical and microscopy techniques to study mice that had similar defects in their myelin sheaths . The study reveals that forming a normal myelin sheath around an axon requires a newly identified ‘scaffold’ made of a group of proteins called the septins . Combining with another protein called anillin , septins assemble into filaments in the myelin sheath . These filaments then knit together into a scaffold that grows lengthways along the myelin-wrapped axon . Without this scaffold , the myelin sheath grew defects known as outfoldings . Axons transmitted electrical signals much more slowly than normal when the septin scaffold was missing from the myelin sheath . Future studies are needed to understand the factors that control how the septin scaffold forms . This could help to reveal ways of reversing the changes that alter the myelin sheath during ageing and disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	Septin/anillin filaments scaffold central nervous system myelin to accelerate nerve conduction
Wiscott Aldrich Syndrome protein ( WASP ) deficiency results in defects in calcium ion signaling , cytoskeletal regulation , gene transcription and overall T cell activation . The activation of WASP constitutes a key pathway for actin filament nucleation . Yet , when WASP function is eliminated there is negligible effect on actin polymerization at the immunological synapse , leading to gaps in our understanding of the events connecting WASP and calcium ion signaling . Here , we identify a fraction of total synaptic F-actin selectively generated by WASP in the form of distinct F-actin ‘foci’ . These foci are polymerized de novo as a result of the T cell receptor ( TCR ) proximal tyrosine kinase cascade , and facilitate distal signaling events including PLCγ1 activation and subsequent cytoplasmic calcium ion elevation . We conclude that WASP generates a dynamic F-actin architecture in the context of the immunological synapse , which then amplifies the downstream signals required for an optimal immune response . Upon encountering antigen-presenting cells ( APCs ) displaying T cell receptor ( TCR ) ligands and adhesion molecules , T cells undergo a series of actin dependent shape changes and signaling steps . These ligands include MHC complexed with antigentic peptide ( MHCp ) for mouse monoclonal T cells , agonist anti-CD3 antibody for polyclonal mouse ( 2C11 ) or human ( OKT3 ) T cells . All of these ligands when presented with ICAM1 in supported lipid bilayers ( SLB ) trigger a series of changes in T cell synaptic cytoskeleton and morphology ( Campi et al . , 2005; Kaizuka et al . , 2007; Burkhardt et al . , 2008; Ilani et al . , 2009; Beemiller et al . , 2012; Kumari et al . , 2012 ) . First , TCR signaling is initiated at F-actin-based protrusions of polarized T cells ( Valitutti et al . , 1995; Negulescu et al . , 1996 ) . Next , these early activation events lead to F-actin dependent formation of a broad immunological synapse ( IS ) ( Wülfing and Davis , 1998; Grakoui et al . , 1999 ) containing T cell antigen receptor ( TCR ) microclusters ( MCs ) ( Bunnell et al . , 2002; Campi et al . , 2005 ) . Synapse spreading is followed by myosin II dependent contraction to form supramolecular activation clusters ( SMACs ) ( Monks et al . , 1998; Griffiths et al . , 2001; Ilani et al . , 2009 ) . TCR MCs form in the outer lamellipodia-like ‘distal’ ( d ) SMAC , after which signaling MCs move centripetally through the lamella like ‘peripheral’ ( p ) SMAC , driven by centripetal flow of F-actin ( Ponti et al . , 2004; Varma et al . , 2006; DeMond et al . , 2008; Vardhana et al . , 2010; Kumari et al . , 2012; Yi et al . , 2012 ) to eventually reach the ‘central’ ( c ) SMAC . The cSMAC is an F-actin depleted zone ( Stinchcombe et al . , 2006 ) , in which TCR is destined for down-regulation via extracellular vesicle formation ( Vardhana et al . , 2010; Choudhuri et al . , 2014 ) . Actin polymerization and remodeling continues throughout the lifetime of the immunological synapse , and studies using a variety of T cell activation systems have identified a critical requirement for F-actin and its molecular effectors in optimal TCR signaling and T cell function ( Valitutti et al . , 1995; Holsinger et al . , 1998; Snapper et al . , 1998; Campi et al . , 2005; Burkhardt et al . , 2008 ) . Pharmacological disruption of global F-actin or genetic manipulations that reduce synaptic F-actin polymerization result in defects in early signaling events such as cell adhesion , TCR MC formation , early TCR signaling , TCR MC transport , as well as the TCR-distal events including intracellular calcium rise , and store operated calcium ion entry ( DeBell et al . , 1992; Valitutti et al . , 1995; Campi et al . , 2005; Nolz et al . , 2006; Varma et al . , 2006 ) . Since global actin perturbation impairs all the above steps , it has been a challenge to dissect the mechanistic role of F-actin in TCR signaling and to identify whether there exist functionally distinct F-actin networks within SMACs that may play distinct roles at various stages of the pathway . One way to investigate functional diversity within the subsynaptic F-actin network , during TCR signaling , is to utilize actin perturbations that dissociate early TCR signaling from late TCR signaling . One such context is provided by the loss of an actin effector protein—Wiscott Aldrich Syndrome Protein ( WASP ) . F-actin polymerization relies on regulatory factors such as nucleation promoting factors ( NPFs ) and downstream nucleation factors such as Arp2/3 complex ( Blanchoin et al . , 2000 ) . WASP is a hematopoietic-cell-specific NPF activated downstream of TCR , originally identified as a target of loss-of-function mutations in the X-linked immunodeficiency Wiskott-Aldrich syndrome ( WAS ) ( Derry et al . , 1994; Machesky and Insall , 1998; Takenawa and Miki , 2001 ) . WAS T cells exhibit a variety of T cell activation defects such as aberrant synapse morphology , impaired proliferation in response to TCR-activation stimuli , and impaired calcium ion signaling ( Molina et al . , 1993; Cianferoni et al . , 2005; Calvez et al . , 2011 ) . Similar to WAS patients , T cells from Was−/− mice also display profound defects in antigen receptor-induced proliferation , IS stability , nuclear NFAT translocation and IL-2 production ( Snapper et al . , 1998; Zhang et al . , 1999 , 2002; Cannon and Burkhardt , 2004 ) . T cells from Was−/− mice ( Zhang et al . , 1999; Krawczyk et al . , 2002; Cannon and Burkhardt , 2004; Sims et al . , 2007 ) and human WAS T cells ( Molina et al . , 1993; Dupre et al . , 2002; Calvez et al . , 2011 ) have apparently normal total F-actin levels as well as SMAC organization within the immunological synapse , while initial TCR–associated kinase signaling in response to MHC-peptide complexes in the context of adhesion ligands is also intact ( Rengan et al . , 2000; Sato et al . , 2001; Krawczyk et al . , 2002; Cannon and Burkhardt , 2004; Sims et al . , 2007 ) . Despite many years of study , the F-actin network to which WASP contributes , and the specific TCR-signaling steps in which it participates to regulate calcium signaling , remain unknown . How might WASP regulate T cell calcium ion responses without affecting total synaptic F-actin ? As an NPF , WASP binds to Arp2/3 and G-actin , increasing the ability of Arp2/3 to nucleate actin branches from existing filaments . Moreover , WASP binds hematopoietic lineage cell-specific protein 1 ( HS1 ) through its SH3 domain ( Dehring et al . , 2011 ) . HS1 is also activated in response to TCR stimulation ( Taniuchi et al . , 1995; Gomez et al . , 2006 ) and can weakly activate Arp2/3 complex , as well as stabilize branched F-actin filaments ( Weaver et al . , 2001 ) . HS1 deficient T cells show defects similar to WASP−/− T cells in TCR activation dependent calcium elevation , proliferation , IL-2 secretion and NFAT activation ( Taniuchi et al . , 1995; Hutchcroft et al . , 1998; Gomez et al . , 2006 ) . It is therefore possible that a previously uncharacterized subclass of the synaptic F-actin network at the TCR MC that represent a small fraction of total synaptic F-actin , is generated by WASP and stabilized by HS1 , supports calcium signaling . Alternatively , it has also been proposed that WASP is a modular scaffolding protein capable of interacting with other proteins of the TCR signalosome , independent of its role as an NPF ( Huang et al . , 2005 ) . Although these two hypotheses are not mutually exclusive , an F-actin dependent role could be addressed by identifying the F-actin network in the immunological synapse to which WASP contributes , and independently targeting this network to investigate the role of the WASP-generated F-actin subpopulation in calcium signaling at the synapse . Thus , WASP can be utilized as a tool to probe for functionally distinct organizational categories of F-actin within the synapse . The signaling cascade leading up to calcium ion elevation in response to TCR engagement has been studied in much detail ( Braiman et al . , 2006; Mingueneau et al . , 2009; Sherman et al . , 2011 ) . TCR ligation triggers a molecular program that results in activation of phospholipase C-γ1 ( PLCγ1 ) , through phosphorylation on Y-783 by Itk ( Park et al . , 1991 ) . Once it has been activated , phospho-PLCγ1 catalyzes the conversion of phosphatidylinositol-4 , 5 bisphosphate ( PIP2 ) to inositol trisphosphate ( IP3 ) and diacylglycerol . IP3 then acts as a second messenger and facilitates release of calcium ions from intracellular stores . Following TCR activation , PLCγ1 recruitment at the synapse is primarily mediated via binding to linker of activated T cells ( LAT ) ( Braiman et al . , 2006 ) . Additionally , recent studies using Jurkat T cells and thymocytes have reported a role for the cortical cytoskeleton in both promoting and inhibiting PLCγ1 activation ( Babich et al . , 2012; Tan et al . , 2014 ) . Although PLCγ1 binds F-actin in biochemical assays , and loss of F-actin dynamics led to reduced PLCγ1 phosphorylation in Jurkat T cells ( DeBell et al . , 1992; Carrizosa et al . , 2009; Patsoukis et al . , 2009; Babich et al . , 2012 ) , the dependence of PLCγ1 activation on WASP activity has not been tested in primary T cells . We hypothesize that WASP and HS1 generate an F-actin network that maintains phosphorylation of PLCγ1 at the synapse , accounting for their role in calcium ion elevation ( Carrizosa et al . , 2009 ) . In this study , we tested these hypotheses by characterizing the F-actin microarchitecture at the immunological synapse that is selectively regulated by WASP , and evaluating its role in early signaling , HS1 and PLCγ1 dynamics , and calcium signaling at the immunological synapse . The results presented here identify and functionally characterize a WASP-dependent actin network at the immunological synapse that regulates phospho-PLCγ1 levels at TCR MC and calcium ion elevation in T cells . This network is visualized as F-actin foci that result from new F-actin actin polymerization at TCR MC . Disruption of these actin foci does not impair initial synapse formation or early TCR signaling . Importantly , disruption of the F-actin foci using a selective Arp2/3 complex inhibitor also spares early TCR signaling , but results in the same impairment in HS-1 and PLC-γ1 activation that has been observed in WASP deficient T cells . Defining a new F-actin network in the immunological synapse , and its molecular regulation , furthers our mechanistic understanding of the cytoskeletal regulation of T cell activation and its dysfunction in WAS pathology . Utilizing a planar surface functionalized with anti-mouse CD3 antibody and ICAM1 as an antigen-presenting surface , we first compared the F-actin cytoskeleton in polyclonal primary CD4 T cells from either wild type ( WT ) or WASP deficient C57BL/6J mice at the early spreading stage of immunological synapse formation . To resolve dynamic F-actin networks , we employed a labeling strategy designed to highlight growing filament ends ( Furman et al . , 2007 ) . A 1 min pulse of Rhodamine-ATP-G-actin in live-permeabilized cells identified freshly incorporated subunits ( Fresh F-actin ) , while subsequent fixation and Alexa488-phalloidin staining identified total F-actin . T cells synapses were visualized using total internal reflection fluorescence microscopy ( TIRF ) to selectively illuminate structures within 200 nm of the interface . WT T cells exhibited discrete F-actin rich features ( Figure 1A , arrowheads ) . We refer to these structures as F-actin ‘foci’ . To selectively extract and quantify F-actin foci from the lamellar actin background , we utilized a spatial frequency-filter based local background correction method ( Figure 1A graph , Figure1—figure supplement 1A , 2 , ‘Materials and methods’ ) . This foci extraction method is capable of reliably identifying foci on a variable background of lamellar F-actin ( Figure 1—figure supplement 1B ) . Assessment of the amount of newly polymerized actin in foci vs surrounding lamellar areas using the above-mentioned labeling and quantification method revealed that the F-actin foci incorporate fresh actin subunits ( Figure 1B ) . This indicates that the foci are dynamically polymerizing structures , rather than sites where F-actin accumulates due to rearrangement of preexisting filaments . Furthermore , the polymerization rate of individual filaments ( Fresh/Total ratio ) is similar between foci and surrounding areas in WT T cells ( Figure 1B ) , indicating that the foci result from preferential local nucleation , rather than enhanced polymerization rate of existing filaments . When compared at the whole synapse scale , overall synaptic F-actin levels , as well as the rate of actin polymerization were similar in WT and WASP−/− T cells , as expected ( Cannon and Burkhardt , 2004 ) . However , there was a significant reduction in F-actin foci in WASP−/− cell synapse ( Figure 1C ) , suggesting that dynamic F-actin foci are formed in a WASP dependent manner . The loss of foci in WASP−/− T cells is not due to non-specific developmental defects , since acute siRNA mediated knockdown of WASP in activated mouse CD4 T cells resulted in a similar phenotype to the WASP−/− T cells . No reduction in total F-actin and significant reductions in foci were observed in WASP siRNA transfected cells imaged on SLB with anti-CD3 and ICAM1 ( Figure 1—figure supplement 3 ) . This lack of effect on total F-actin in WASP deficient or depleted cells may be attributed to the fact that these foci contribute to <10% of the total synaptic F-actin intensity ( Figure 1—figure supplement 1B top graph ) , or perhaps to a compensatory increase in the lamellar and lamellipodial actin when more ATP-G-actin is available for polymerization regulated by other NPFs such as WAVE2 . These results indicate that WASP dependent F-actin foci form in the T cell synapse . 10 . 7554/eLife . 04953 . 003Figure 1 . WASP dependent dynamic F-actin foci at the T cell synapse . ( A ) Barbed-end decoration of freshly polymerized actin filaments reveals the dynamic behavior in WT ( top panel ) and WASP−/− ( bottom panel ) T cell synapses . CD4 T cells isolated from WT and Was−/− C57BL/6J mice were processed for barbed end labeling ( fresh actin ) to identify the actin incorporation sites within 1 min of polymerization , as well as total F-actin labeling , as described in ‘Materials and methods’ . Arrowheads indicate the sites of F-actin foci in both ‘Total’ as well as ‘Fresh’ F-actin images . ( B ) The graph shows average incorporation of Rhodamine actin per pixel within the foci or the surrounding lamellar pixels ( Fresh F-actin ) ; and a ratio of Rhodamine and Alexa488 actin intensities in foci or surrounding pixels , in the WT cells . The foci areas in the total F-actin image were identified and outlined by intensity rank-based filtering as described in ‘Materials and methods’ . The synaptic area outside the foci was defined as lamellar surround . These outlined areas were then analyzed in both ‘fresh F-actin’ , and ‘total F-actin’ raw images and per pixel intensity is plotted in the graph . Note that while the rate of polymerization ( Fresh/Total ) ratio is not altered , there is higher incorporation of actin at foci sites ( Fresh ) . n = 20 cells , p1 < 0 . 0001 , p2 = 0 . 181 ( Wilcoxon nonparametric pairwise comparison ) . ( C ) The graph shows total F-actin intensity , fresh F-actin intensity , or the ratio of the two at the synapse , each point represents value obtained from single cell , n in WT = 33 , n in WASP−/− = 35; p1 = 0 . 06 , p2 = 0 . 059 , p3 = 0 . 0008 . Scale bars , 5 μm . ( D ) WASP is critical for F-actin foci generation . Freshly purified CD4 T cells from WT 129 ( top left ) , Was−/− 129 ( bottom left ) , Wasl−/− ( top right ) , or Was/− Wasl−/− 129 ( bottom right ) mice were activated on SLB containing ICAM1 and anti-CD3 antibody , fixed and stained with Alexa488-phalloidin . Note that only the lack of WASP , and not N-WASP , causes loss of actin foci . n1 = 54 , n2 = 45 , n3 = 53 , n4 = 28 . p1 = 0 . 30 , p2 < 0 . 0001 , p3 < 0 . 0001 . ( E ) F-actin foci in CD4 T cells freshly purified from wild type ( WT ) mouse ( left ) or from Hcls1−/− mouse ( center ) or Hcls1−/− Was−/− ( right ) C57BL/6J mice were incubated with bilayer containing anti-CD3 antibody and ICAM1 for 2 min . Cells were then processed for F-actin staining ( Alexa488-phalloidin ) and visualized . Note that , while there is a minor alteration in F-actin foci in HS1−/− T cells , there is a gross deficit in HS1−/− WASP−/− double knockout T cells . The graph shows the quantification of total intensity of F-actin foci at the synapse in individual T cells derived from the indicated backgrounds . n1 = 48 , n2 = 70 , n3 = 64 , p1 = 0 . 031 , p2 = 0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 00310 . 7554/eLife . 04953 . 004Figure 1—figure supplement 1 . Method of analyzing F-actin foci from the raw F-actin TIRF image . ( A ) The raw image ( left ) was processed for local background subtraction . Briefly , we first used a rank filter selecting for the 25th , 50th ( median ) or 75th percentile intensity value in a 1 . 6 × 1 . 6 µm moving window , to extract the ‘background image’ . This ‘background image’ was then subtracted from the raw image to generate a high-pass filtered image where we could clearly visualize F-actin foci . The processed image was manually checked for co-localization of F-actin foci with the F-actin foci in the raw image . As shown in the ‘merge’ panel , the 50% rank filter was found to be most optimal for F-actin foci extraction , while the 25% filter overestimated and 75% filter underestimated the number of F-actin foci . The 50% rank filter corrected image was therefore used to quantify F-actin foci per cell , as indicated in the corresponding figure legends . In some experiments , this processing was also carried out for TCR and ICAM1 image , to estimate co-localization between F-actin foci and TCR MC . ( B ) Characterization of range of F-actin foci intensities to assess the efficacy of foci extraction method . Intensities of pixels corresponding to foci were quantified in raw images ( upper graph ) or in images generated using 50% intensity rank filtering ( ‘Background images’ , lower graph ) . CD4 T cells were incubated with bilayer containing MHCp and ICAM1 for 2 min , fixed , and stained with Alexa488-phalloidin ( F-actin ) . TIRF images of the T cell synapse were processed to extract F-actin foci as described above in ( A ) . These foci were used to create a binary mask to outline the foci or total synaptic cell regions in either original raw images ( upper graph ) or background images ( lower graph ) . Each graph shows the intensity distribution of individual pixels corresponding to foci or total synapse area , and the X-axis represents the binned pixel intensity , normalized to the mean synaptic F-actin intensity from each cell's raw image . The Y-axis represents the number of pixels in the foci or synapse area , normalized to either the total number of pixels in the synapse ( ‘/total number of pixels’ traces ) , or to only the total number of pixels in the foci area ( ‘/foci pixels’ trace ) . The data is represented as the mean number of pixels from 13 cells ±SEM . It is notable that the number of pixels encompassing foci regions is <10% of total number of pixels ( top left graph ) . Also note that in both raw and background images , the pixels in foci do not exclude the high-intensity range , suggesting that our foci identification method does not preclude foci in high intensity neighborhoods , and identifies higher intensity patches even in high background neighborhood . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 00410 . 7554/eLife . 04953 . 005Figure 1—figure supplement 2 . Test of rank-filter based processing method for foci detection . T cells purified from WT or Was−/− mice were incubated with anti-CD3/ICAM1 reconstituted lipid bilayers for 2 min and then stained with Alexa488-phalloidin . WT and WASP−/− cells have comparable levels of total F-actin at synapse ( raw images in left panels ) , as shown in Figure 1 . However , WASP−/− cells visibly lack F-actin foci . When processed using local background correction method , the processed images indeed show quantifiably different numbers of foci . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 00510 . 7554/eLife . 04953 . 006Figure 1—figure supplement 3 . WASP silencing causes a reduction in F-actin foci . Mouse CD4 T cell blasts were electroporated with WASP siRNAs , or with non-targeting sequences ( control ) . The extent of depletion of protein levels was estimated using western blots . ∼40% of protein depletion was achieved using WASP siRNAs ( relative densitometry value mentioned on top of gel lanes ) . Cells electroporated with WASP siRNA exhibited significantly fewer F-actin foci . AND T cell blasts electroporated with siRNA sequences , as described above , were incubated with bilayer containing anti-CD3 and ICAM1 for 2 min , followed with fixation and staining with Alexa488-phalloidin ( left panels ) , and imaged using TIRF microscopy . The graphs show quantitation of total F-actin ( right graph ) or F-actin foci ( left graph ) per cell , in control or WASP siRNA-treated cells . n1 = 35 , n2 = 60 , p = 0 . 238 ( right ) , p = 0 . 0025 ( left graph ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 00610 . 7554/eLife . 04953 . 007Figure 1—figure supplement 4 . WASP family members NWASP and WAVE2 are not associated with F-actin foci . ( A ) TCR-induced recruitment of NWASP and WAVE2 to IS . Mouse primary WT CD4 T cells were incubated with bilayer containing ICAM1 alone ( − ) or both ICAM1 and anti-CD3 ( + ) for 2 min at 37°C , fixed and immunostained for endogenous proteins . Stained cells were visualized using TIRF microscopy . The graph shows quantitation of antibody fluorescence at IS , where each point represents the value obtained from a single cell . n1 = 16 , n2 = 54 ( for WAVE2 ) , n3 = 16 , n4 = 78 ( for NWASP ) ; p1 , p2 < 0 . 0001 . Each point represents average levels of indicated protein at synapse in a single cell . ( B ) The images shown are TIRF plane distributions of the indicated proteins . As elaborated in the magnified areas marked with white boundary in original ‘merge’ image , there is a lack of co-localization between either of these proteins and TCR MCs . Scale bar , 5 μm . Insets in ( B ) have been intensity scaled differently from original ‘Merge’ panel to highlight protein distribution with more clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 00710 . 7554/eLife . 04953 . 008Figure 1—figure supplement 5 . WASP dependent HS1 recruitment and F-actin foci . ( A ) HS1 silencing and F-actin foci . AND CD4 T cell blasts were electroporated with non-specific ( control siRNA oligos ) or oligos targeting HS1 . The depletion of protein was assessed using western blotting ( top left , gel ) . Cells electroporated with control or HS1 siRNA oligos were incubated with bilayers for 2 min at 37°C , fixed , and stained with Alexa488-phalloidin , and then imaged using TIRF illumination . The images ( right ) show F-actin distribution in their respective backgrounds . The graph shows quantification of F-actin foci in control or HS1 depleted cells . n1 = 70 , n2 = 100 , p = 0 . 0268 . ( B ) WASP determines phospho-HS1 levels at the synapse . CD4 T cells purified from WT-129 mouse , Was−/− , or Wasl−/− 129 mice were incubated with bilayer containing ICAM1 and anti-CD3 for 2 min , and were fixed and stained with anti-phospho-HS1 antibody . The TIRF images were quantified for phospho-HS1 levels . Each dot represents average phospho-HS1 intensity per cell . n1 = 42 , n2 = 43 , n3 = 46 , n4 = 27 . p1 = 0 . 637 , p2 , p3 < 0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 008 We next assessed the role of other WASP-family proteins—NWASP and WAVE2—in generating F-actin foci in WT cells . NWASP ( WASL ) is a close homolog of WASP and is capable of compensating for WASP's function in T cell development ( Cotta-de-Almeida et al . , 2007 ) . However , primary CD4 T cells from Wasl−/− mice ( Figure 1D ) formed immunological synapses with F-actin foci similar to T cells from WT mice . WAVE2 is required for immunological synapse formation itself ( Nolz et al . , 2006 ) , thus we could not examine synaptic foci in WAVE2 deficient T cells . Instead , we determined the localization of NWASP and WAVE2 in the immunological synapse in order to gain further insight into their possible role in foci formation . Activation of T cells with anti-CD3 and ICAM1 recruited NWASP and WAVE2 to the IS ( Figure 1—figure supplement 4A ) ( Nolz et al . , 2006 ) , but the recruited proteins failed to co-localize with F-actin foci ( Figure 1—figure supplement 4B , NWASP coloc . = 9 . 48% ± 0 . 76 n = 78; WAVE2 coloc . = 14 . 36% ± 1 . 25 , n = 54 ) . We were unable to identify suitable reagents for staining of endogenous WASP in the immunological synapse . These results demonstrate that NWASP does not contribute to F-actin foci and suggest that WAVE2 is unlikely to contribute directly to formation of F-actin foci . In addition to WASP , deficiency of HS1 , a cortactin-related NPF , also leads to reduced calcium ion signaling in T cells ( Gomez et al . , 2006 ) . Therefore , we examined the role of HS1 in generation of F-actin foci using T cells from WT and HS1−/− mice . F-actin foci formation was only marginally impaired in HS1−/− T cells , this reduction was modest compared to T cells from mice lacking both HS1 and WASP ( Figure 1E ) . A similar marginal reduction in F-actin foci intensity was obtained by siRNA-mediated knockdown of HS1 in activated mouse CD4 T cells ( Figure 1—figure supplement 5A ) . HS1 is phosphorylated on Y397 ( phospho-HS1 ) during T cell activation ( Hutchcroft et al . , 1998; Gomez et al . , 2006; Carrizosa et al . , 2009 ) . We next determined if this phosphorylation is WASP dependent . WT mouse CD4 T cells activated with anti-CD3 and ICAM1 displayed foci of phospho-HS1 in the synapse ( Figure 1—figure supplement 5B ) . These phospho-HS1 foci were significantly reduced in WASP−/− or WASP−/− NWASP−/− CD4 T cells , but not in CD4 T cells from NWASP deficient mice ( Figure 1—figure supplement 5B ) . Thus , HS1 is recruited to the T cell synapse in a WASP dependent manner . To understand the mechanism underlying formation of the F-actin foci , we determined if they were associated with TCR MCs . In murine polyclonal CD4 T cells activated on anti-CD3 and ICAM1 containing bilayers , as in Figure 1 , TCR MC localized with F-actin foci ( Figure 2A , upper panels , arrows ) . Similar co-localization of F-actin foci with TCR MC was observed with monoclonal AND TCR transgenic T cells stimulated with MHCp and ICAM1 ( Figure 2A , lower panels ) . The tracking of TCR with H57 Fab allowed accurate quantification of co-localization between TCR and F-actin . After 1 min of cell attachment , 40 ± 1 . 7% of TCR MCs co-localized with F-actin foci in these cells ( Figure 2—figure supplement 1A , ‘Materials and methods’ , Figure 2B ) , significantly above chance level co-localization ( Figure 2—figure supplement 1B ) , whereas only 15 ± 0 . 07% of total ICAM1 exhibited association with actin foci ( Figure 2B ) , which was not different from chance . When examined in Human primary CD4 T cells as well , F-actin foci formed in WASP-dependent manner ( Figure 2—figure supplement 2A , B ) , and TCR MCs , but not ICAM1 MCs , were associated with F-actin foci ( TCR co-localization with F-actin foci was 34 . 5% ± 2 . 0 , n = 76 , co-localization of ICAM1 with F-actin foci was 13 . 1% ± 0 . 8 , n = 66 ) ( Figure 2C , D ) . Furthermore , TCR-activation led to HS1 phosphorylation at the T cell synapse , as shown previously ( Figure 2—figure supplement 3A ) ( Hutchcroft et al . , 1998; Gomez et al . , 2006 ) , and the phospho-HS1 co-localized with F-actin foci in both TCR-activation stimuli ( Figure 2—figure supplement 3B ) . This indicated that TCR-MCs associated foci exist in a variety of primary CD4 T cells including mouse and human primary T cells , stimulated with diverse TCR-triggering contexts , namely anti-CD3 and peptide-MHC complexes . 10 . 7554/eLife . 04953 . 010Figure 2 . F-actin foci co-localize with TCR MC and not ICAM1 . ( A ) Freshly isolated mouse AND CD4 T cells were incubated with lipid bilayer reconstituted with Alexa568 tagged anti-CD3 ( TCR , red ) and Alexa647-ICAM1 ( blue ) , for 2 min at 37°C . Post incubation , cells were fixed and stained for F-actin using Alexa488-phalloidin ( green ) , and imaged using TIRF microscopy . The region marked 1 in the ‘merge’ panel is magnified to clearly show the co-localization of actin foci with TCR-containing MCs . Lower panels: AND mouse CD4 T cell blasts exhibit F-actin enrichment at TCR MCs sites . AND mouse CD4 T cell blasts were labeled with Alexa568-H57 Fab ( TCR ) , and incubated with bilayer reconstituted with MHCp and Alexa405-ICAM1 for 2 min at 37°C , fixed and stained for F-actin . Region marked 2 in the ‘merge’ image is further magnified to show the overlap between TCR and F-actin ( arrows ) . The insets 1 and 2 in both MHC-activated and anti-CD3-activated mouse T cells are contrasted differently from the original ‘merge’ image to highlight the TCR and actin distribution . ( B ) Quantitation of the fraction of TCR or ICAM1 localized with F-actin foci . AND CD4 T blasts were incubated with antigen containing bilayer for 2 min , as described above and the images acquired were processed for colocalization assessment as described in ‘Materials and methods’ section . Each point represents fraction of total synaptic TCR or ICAM1 associated with F-actin foci in a single cell . n1 = 99 , n2 = 100 . p < 0 . 0001 . ( C , D ) Freshly isolated human peripheral blood CD4 T cells were incubated with bilayer containing Alexa568 tagged anti-CD3 ( TCR ) , Cy5-ICAM1 at 37°C , fixed and stained with Alexa-488 phalloidin and subsequently imaged using TIRF microscopy . The image shows a freshly attached cell to the bilayer . The line marked by the arrow in the ‘merge’ panel shows the line-scan profile plotted in ( C ) , where relative intensities of ICAM1 and actin with single TCR MC across the pixels marked in the ‘merge’ image are shown . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 01010 . 7554/eLife . 04953 . 011Figure 2—figure supplement 1 . Test for co-localization of foci and MCs . ( A ) An example of image analysis carried out to assess co-localization between F-actin foci and TCR/ICAM1 , using 50% rank-filter based processing , as described in Figure 1–figure supplement 1A ( B ) TCR MC and F-actin foci co-localization is significantly higher than chance overlap . To measure chance level co-localization , the actin images from Figure 1 were shifted by 5 pixels in x and y dimensions ( schematic on top ) , and analyzed for co-localization with respect to the unshifted TCR ( left graph ) or ICAM1 ( right graph ) images . Note that while pixel shift results in significant reduction TCR-F-actin foci co-localization ( left graph , p < 0 . 0001 ) , levels of ICAM1/F-actin foci association are unaffected ( right graph , p = 0 . 53 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 01110 . 7554/eLife . 04953 . 012Figure 2—figure supplement 2 . Loss of WASP in Human CD4 T cells , and its impact on foci induction . ( A ) Human CD4 T cells were incubated with culture media containing lentiviral particles carrying WASP shRNA or non-specific ( control ) shRNA for 48 hr , and were lysed and assess for WASP depletion using western blotting ( ‘Materials and methods’ ) . ( B ) Human CD4 T cells were incubated with lentiviral particles carrying WASP shRNA or mock shRNA ( Control ) for 60 hr , and were further activated using anti-CD3 and ICAM1 containing surface . Cells were then fixed , stained with Alexa488-phalloidin , imaged using TIRFM and analyzed for foci induction . Note that transduction with WASP shRNA leads to significant reduction in foci at synapse . The graph represents foci intensity per cell , normalized to mean intensity in Control T cells , n1 = 29 , n2 = 56 , p < 0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 01210 . 7554/eLife . 04953 . 013Figure 2—figure supplement 3 . ( A ) Increased phosphorylation of HS1 in response to TCR activation . AND CD4 T cell blasts were incubated with bilayers reconstituted with ICAM1 alone ( left image ) or both ICAM1 and MHCp ( right image ) for 2 min at 37°C , and then fixed and immunostained for phosphorylated HS1 . The images show phospho-HS1 ( p-HS1 ) signal in the cells acquired in the TIRF field . The graph ( far right ) shows quantitation of total p-HS1 intensities , where each point represents integrated intensity per cell in the TIRF field . n1 = 33 , n2 = 33 , p < 0 . 0001 . ( B ) Phospho-HS1 localizes to F-actin foci and TCR MCs . Mouse AND CD4 T cell blasts ( left panels ) or primary CD4 T cells ( panels on the right ) were incubated with bilayer containing ICAM1 and MHCp ( for blasts ) or ICAM1 and anti-CD3 ( right panels ) for 2 min , fixed and stained with phospho-Y397 HS1 antibody ( red ) and Alexa488-phalloidin ( green ) . The area marked in ‘Merge’ image has been further enlarged in insets; the arrows indicate F-actin foci sites that localize with phospho-HS1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 01310 . 7554/eLife . 04953 . 014Figure 2—figure supplement 4 . Lack of F-actin foci in the cSMAC of primary T cells and the Jurkat T cell line . AND CD4 T cell blasts were either incubated with bilayers containing Alexa568 tagged anti-CD3 and ICAM1 ( far left panels ) or were labeled with Alexa568-H57 Fab and incubated with MHCp/ICAM1 bilayers ( center left panels ) , for 2 min . Cells were fixed and stained for F-actin using Alexa488-phalloidin ( middle row , green ) and visualized using spinning disc confocal microscopy . Each image is a maximum intensity projection of the bottom three planes that show high intensity signals from the bilayer . Note that TCR MCs ( bottom row , red ) in the F-actin depleted central zone of the cell ( asterisk ) exhibit no significant co-localization with actin foci . This phenomenon where central microclusters lack F-actin foci was observed in >90% cells exhibiting well-defined cSMAC ( number of experiments >3 ) . F-actin foci were not detected in Jurkat T cells ( center right panel , and far right insets ) . Jurkat T cells were activated on bilayer containing Aexa568 tagged anti-CD3 ( TCR , red , bottom row ) and ICAM1 for 2 min and were fixed and stained with Alexa488-phalloidin ( Actin , green , middle row ) . As highlighted in insets , visible F-actin enrichment is missing from the TCR MCs in these cells . Scale bars , 5 µm ( AND T cell ) and 10 µm ( Jurkat T cell ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 01410 . 7554/eLife . 04953 . 015Figure 2—figure supplement 5 . Association of TCR MC and F-actin foci in live T cell . Live cell imaging of primary human CD4 T cell electroporated with Lifeact-GFP DNA expression construct ( Actin , green , top panels ) on lipid bilayer reconstituted with Alexa568 tagged anti-CD3 ( TCR , pseudocolored in red , middle panels ) and ICAM1 . The images show snapshots of TIRF images for the indicated time points from the start of observation . The circle on the top left of the t = 235 s image tracks an example TCR MC that exhibits movement along with F-actin foci in subsequent frames , until it finally merges with cSMAC , concomitant with disappearance of actin foci . The images on the right show kymographs obtained from the central section of the cell ( marked with white bar in 260 s ‘merge’ panel ) . The arrows point to the MCs that colocalize with LifeAct and co-migrate until the MC reaches the cSMAC . This association was readily observed in >60% cells monitored ( n = 7 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 015 Since a fraction of TCR MC did not localize with foci , we sought to examine this population . High magnification microscopy of T cell synapses formed by mouse CD4 T cells on anti-CD3 and ICAM1 SLB provided insight into both synaptic and non-synaptic F-actin organization ( Figure 2—figure supplement 4 , Videos 1–2 ) . As expected ( Stinchcombe et al . , 2006 ) , the F-actin signal progressively decreases toward the synapse center where central TCR MC are devoid of F-actin foci ( Figure 2—figure supplement 4 , asterisks in the ‘Merge’ panel ) . To eliminate the possibility that the lack of foci on TCR MC in the nascent cSMAC is a consequence of cell fixation methodology , we examined foci in live T cells by transfecting them with LifeAct-GFP , a peptide construct that selectively labels F-actin ( Riedl et al . , 2008 ) . Since primary mouse T cells showed poor survival after transfection , we used human primary CD4 T cells that exhibit higher viability ( Chicaybam et al . , 2013 ) . In live cell synapses , LifeAct-GFP was observed to form foci on TCR MCs , which continued to co-migrate until the delivery of MCs to the cSMAC ( Figure 2—figure supplement 5 , Video 3 ) . Analysis of kymographs revealed that the average TCR MC speed was 5 . 66 ± 2 . 2 μm/min , average Foci speed was 6 . 63 ± 3 . 1 μm/min ( p = 0 . 21 , ns ) . The TCR signals persisted in the cSMAC , whereas the GFP foci were extinguished , consistent with continual nucleation of F-actin at the MC site ( Figure 1A ) until it reaches the cSMAC , where TCR signals are known to be terminated ( Vardhana et al . , 2010; Choudhuri et al . , 2014 ) . 10 . 7554/eLife . 04953 . 016Video 1 . AND CD4 T cells were incubated with bilayer containing ICAM1 and Alexa568 tagged anti-CD3 ( red ) for 2 min , fixed and stained with Alexa488-phalloidin ( green ) . Cells were then visualized using spinning disc confocal microscopy . Video shows 3D reconstruction of images acquired . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 01610 . 7554/eLife . 04953 . 017Video 2 . AND CD4 T cells were labeled with Alexa568-H57 Fab ( red ) , were activated on bilayer containing MHCp and ICAM1 for 2 min , then fixed and stained with Alexa488-phalloidin ( green ) , and eventually imaged using spinning disc confocal microscopy . Video shows 3D reconstruction of images acquired . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 01710 . 7554/eLife . 04953 . 018Video 3 . Human CD4 T cells were transfected with Lifeact-GFP ( pseudocolored green ) and were incubated with bilayer containing ICAM1 and Alexa568 tagged anti-CD3 ( pseudocolored red ) . The video was acquired with 5 s intervals between frames . The play rate is 80 times faster than the acquisition rate . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 018 We also note that the Jurkat T cell line stimulated using a similar anti-CD3 and ICAM1 activation system displayed only random co-localization of TCR MC and F-actin ( 9 . 0% ± 0 . 09 TCR colocalized with F-actin , n = 47 , Figure 2—figure supplement 4 , right panels , Video 4 ) . This suggests that a distinct F-actin organization may exist in this commonly used leukemia derived cells line . 10 . 7554/eLife . 04953 . 019Video 4 . Jurkat T cells were transfected with Lifeact-GFP ( pseudocolored green ) and incubated with bilayer containing ICAM1 and Alexa568 tagged anti-CD3 ( pseudocolored red ) . The video was acquired with 15 s interval between frames . The play rate is 33 times faster than the acquisition rate . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 019 Since F-actin foci are associated with TCR MC sites in pSMAC and dSMAC zones , we sought to investigate the requirement of TCR signals for their formation . AND T cells activated using ICAM1 alone displayed some interface F-actin ( Figure 3A , top images ) ; however , a significant induction of F-actin foci only occurred when AND TCR transgenic T cells were incubated with both ICAM1 and agonist peptide-MHC ( Figure 3A , top graph ) . A similar requirement of TCR triggering for F-actin foci formation was observed in primary human CD4 T cells ( Figure 3A , bottom graph ) , although it did not scale with the dose of TCR agonist; in T cells incubated with low or high concentrations of anti-CD3 antibody , we did not observe a corresponding increase in F-actin foci ( Figure 3—figure supplement 1 ) . These results show that TCR activation triggers robust foci formation above a threshold level of TCR engagement , in the presence of a constant level of ICAM1 mediated adhesion . The residual foci intensity observed in T cells in the ICAM1 alone case ( Figure 3A graph ) is primarily due to fluctuations in lamellipodial F-actin detected as foci by our analysis method ( Figure 1—figure supplement 1 , 2 ) . 10 . 7554/eLife . 04953 . 020Figure 3 . Relationship of F-actin foci to TCR engagement and TCR proximal signaling . ( A ) AND mouse CD4 T cell blasts were incubated with lipid bilayer containing Alexa405-ICAM1 and MHCp , or bilayer containing Alexa405-ICAM1 alone ( ICAM1 ) at 37°C for 2 min . The cells were fixed and stained with Alexa488-phalloidin ( Actin ) and imaged using TIRF microscopy . The images in the right panel are reflection ( interference reflection microscopy – IRM ) images in the two conditions , showing cell-bilayer contact area . Actin images were further high-pass filtered using a rank-filter based subtraction method ( see ‘Materials and methods’ ) to reveal spatially localized actin features within each cell , under both incubation conditions . The graph ( top right ) shows average intensity of actin features ( F-actin foci ) per cell . n1 = 31 , n2 = 53 , p < 0 . 0001 . The bottom graph shows foci induction in human CD4 T cells . Primary human CD4 T cells were incubated with bilayer containing Alexa405-ICAM1 alone , or both Alexa405-ICAM1 and Alexa568 tagged anti-CD3 at 37°C for 2 min , and were then fixed and stained for F-actin using Alexa488-phalloidin , for imaging using TIRF microscopy . The graph shows integrated intensity of actin spots per cell , each point on the graph represents a single cell . n1 = 117 , n = 79 , p < 0 . 0001 ( B ) Formation of TCR MC associated F-actin foci requires SFK signaling . AND T cell blasts were treated with PP2 for 10 min at 37°C and were labeled with Alexa568-H57 Fab . This was followed by further incubation with ICAM-1/MHCp containing bilayer at 37°C for 2 min , and fixation and staining for F-actin . The TIRF images show Actin ( Alexa488-phaloidin , green ) and TCR ( red ) distribution in DMSO treated ( control , left ) and PP2 treated ( right ) cells . Graph ( far right ) shows mean intensity of F-actin features in control and PP2 treated cells . n1 = 32 , n2 = 35 , p < 0 . 0001 . ( C ) Phospho-Zap70 levels are normal in WASP deficient cells . Freshly isolated C57BL/6J WT or WASP−/− CD4 T cells were activated on bilayer containing anti-CD3 and ICAM1 for 2 min , fixed and stained with anti-phospho-Zap70 antibody ( left graph ) . The top images show phospho-Zap70 intensity at the synapse , and the left graph shows integrated intensity of phospho-Zap70 per cell . n1 = 30 , n2 = 49 , p = 0 . 17 . The WASP−/− cells display normal whole cell levels of TCR-induced F-actin , as marked by Alexa488-phalloidin staining ( right graph ) . The cells were imaged under wide-field illumination settings – each point represents total F-actin intensity per cell . n1 = 53 , n2 = 212 , p = 0 . 583 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 02010 . 7554/eLife . 04953 . 021Figure 3—figure supplement 1 . Effect of TCR-ligand concentration on F-actin foci formation . CD4 human T cells were incubated with bilayer containing indicated concentration of Alexa568 tagged anti-CD3 and ICAM1 for 5 min , fixed and stained with Alexa488-phalloidin , and then imaged using TIRF microscopy . The graph on the left shows total intensities of anti-CD3 gathered by cells at synapse , at indicated anti-CD3 reconstitution densities . n1 = 84 n2 = 112 , p < 0 . 0001 . The graph on the right shows total intensity of actin features in the synapse , obtained using the same images used for analysis in the left graph . p = 0 . 06 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 02110 . 7554/eLife . 04953 . 022Figure 3—figure supplement 2 . Localization of phosphorylated form of Src family kinases ( SFK ) at TCR MC/F-actin foci . AND mouse CD4 T cell blasts were labeled with Alexa568-H57 Fab , incubated with MHCp/ICAM1 containing bilayer for 2 min , fixed and then immunostained for phospho-SFK ( p-SFK ) . An example line scan profile for the line ( white ) indicated on the left image is displayed ( right graph ) to show the relative distribution of TCR , actin and phospho-SFK across the row of pixels . Scale bar , 6 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 022 To further investigate the role of TCR signaling in foci formation , we examined the role of Src family kinases ( SFK ) signaling . TCR signaling proceeds by accumulation of active Lck , as indicated by phosphorylation of the activation loop , in MCs ( Campi et al . , 2005 ) . Consistent with this , phospho-SFK localized to F-actin foci associated MC ( Figure 3—figure supplement 2 ) . Furthermore , inhibition of SFK phosphorylation using PP2 ( Varma et al . , 2006 ) significantly reduced the number of F-actin foci ( Figure 3B ) , indicating that TCR-driven SFK signaling is required for their formation . We conclude that F-actin foci correspond to sites of early TCR signaling and require active SFKs . The next step in TCR proximal signaling is the SFK mediated recruitment of Zap70 to the TCR cytoplasmic domains and the phosphorylation of Zap70 by SFK on sites including Y319 in the interdomain linker . TCR-induced Zap70 Y319 phosphorylation in MC and overall F-actin levels at the synapse were comparable between primary WT and WASP−/− cells ( Figure 3C ) . This result demonstrates that WASP is dispensable for TCR-proximal signaling including Zap-70 recruitment and activation . Having found that TCR MC formation and proximal signaling is independent of WASP , we investigated further TCR-distal signaling events that would lead to WASP-dependent calcium ion flux , as reported previously in human and mouse T cells ( Zhang et al . , 2002; Calvez et al . , 2011 ) . PLCγ1 is one of the key effector molecules regulating calcium ion flux downstream of TCR-activation , which mediates inositol-1 , 3 , 4-trisphosphate synthesis in the plasmamembrane , thereby facilitating calcium ion release from endoplasmic reticulum stores ( DeBell et al . , 1992; Babich et al . , 2012 ) . In addition , there is evidence that PLCγ1 can bind to F-actin ( Carrizosa et al . , 2009 ) . We therefore reasoned that F-actin foci may locally enrich PLCγ1 at the TCR MC , supporting PLCγ1 interaction with TCR signalosome effectors such as Itk and LAT , thereby promoting its phosphorylation and activation ( Braiman et al . , 2006 ) . We therefore assessed the role of WASP in PLCγ1 activation in WT and WASP−/− T cells . WASP−/− T cells exhibited significantly reduced phospho-PLCγ1 levels at the synapse ( Figure 4A ) . The reduction in phospho-PLCγ1 levels was not due to a general decrease in total PLCγ1 levels in the WASP−/− T cells , as the levels of total cellular PLCγ1 were comparable in WT and WASP−/− whole T cell lysates ( Figure 4B ) . Thus , a loss of WASP-driven F-actin foci is correlated with impairment of PLCγ1 activation at the synapse . 10 . 7554/eLife . 04953 . 009Figure 4 . WASP regulates calcium ion signaling via generation of F-actin foci . ( A ) WASP deficient T cells exhibit impaired TCR-induced PLCγ1-Y783 phosphorylation . CD4 T cells freshly isolated from C57BL/6J WT or Was−/− mice were activated with ICAM1 alone , or both ICAM1 and anti-CD3 for 2 min . Cells were fixed and immunostained for phospho-PLCγ1 ( image not shown ) , and visualized using confocal microscopy . The images show phospho-PLCγ1 staining in the bottom section ( synapse plane ) of WT ( left ) or WASP−/− ( right ) cells . The graph on the left shows phospho-PLCγ1 levels in the synapse planes in the cells in WASP deficiency background . n1 = 104 , n2 = 60 , n3 = 94 , p < 0 . 0001 . ( B ) Assessment of total cellular PLCγ1 phosphorylation in WASP−/− T cells using western blot . Freshly isolated WT or WASP−/− CD4 T cells were incubated with anti-CD3/CD28 beads for 5 min , lysed , and the lysates were analyzed using western blotting . Note that , in WASP−/− T cells TCR-induced PLCγ1 phosphorylation is defective , while total PLCγ1 is comparable to the WT cells . HS1 phosphorylation was included as a control that exhibits diminished phosphorylation in WASP−/− T cells . These experiments were repeated twice with similar results . ( C ) Arp2/3 activation by WASP is essential for F-actin foci generation and optimal phospho-PLCγ1 at the synapse . Human CD4 T cells were transfected with GFP-WASP ( WT ) , GFP-WASPΔC , or GFP-WASP291F ( shown in the schematic on the left ) for 16 hr and were then incubated with anti-CD3/ICAM1-reconstituted bilayers for 2 min , fixed and stained with Alexa568-phalloidin and anti-phospho-PLCγ1 antibody . The images show the GFP ( green ) , F-actin ( blue ) and phospho-PLCγ1 ( red ) distribution at the synapse for WT ( left ) and WASPΔC ( right ) T cells . ( D ) The graph shows levels of GFP-tagged constructs at the synapse , analyzed and obtained via 50% rank filtering of images shown in ( D ) , as described in ‘Materials and methods’ , as well as the quantitation of total synaptic F-actin , foci and phospho-PLCγ1 in the same cells , normalized to mean values obtained for WT cells . n1 = 23 , n2 = 27 , n3 = 33 , n4 = 27 p values , p < 0 . 0001 between WT and WASPΔC cells for foci and phospho-PLCγ1 levels , and between untransfected and cells expressing GFP tagged constructs for ‘GFP puncta’ . For all other comparisons , p > 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 009 We next asked whether the well-established role of WASP in T cell calcium ion signaling pathway could be accounted for by F-actin foci , or by other pathways involving WASP-dependent protein–protein interactions . WASP is a scaffolding protein and interacts with a variety of proteins at MCs ( Thrasher and Burns , 2010 ) . To distinguish between two possible roles of WASP , activating Arp2/3 complex to generate F-actin foci , or acting as a direct molecular scaffold , in the regulation of PLCγ1 activation at the synapse , we overexpressed mutant forms of human WASP , including an Arp2/3-activation deficient form ( WASP Δ473–480 or WASPΔC ) and SFK scaffolding deficient form ( WASPY291F ) in human CD4 T cells . WASP binds with Arp2/3 via its C-terminal verprolin-homology , central , acidic ( VCA ) domain , while a stretch of hydrophobic residues in the central domain is reported to facilitate Arp2/3 activation without affecting the binding of Arp2/3 to the acidic domain ( Machesky and Insall , 1998; Panchal et al . , 2003 ) . WASPΔC lacks this Arp2/3 activation sequence within the C-region ( Kato et al . , 1999 ) ( Figure 4C , schematic ) . Indeed , T cells expressing GFP-WASPΔC showed normal levels of Arp2/3 complex recruitment ( data not shown ) , as well as total F-actin at the synapse , and synaptic recruitment of WASPΔC was comparable to WT WASP ( Figure 4C , D ) . However , WASPΔC expressing cells were severely defective in F-actin foci generation , as well as in PLCγ1 phosphorylation at the synapse ( Figure 4C , D ) . This result indicates that WASP-dependent Arp2/3 activation is essential for foci genesis and activation of calcium ion signaling effector PLCγ1 at the synapse . Overexpression of WASP Y291F ( Dovas et al . , 2009 ) , a mutant form defective in phosphorylation dependent interaction with SH2 domain of SFKs ( Padrick and Rosen , 2010 ) , had no effect on foci formation or PLCγ1 activation . This result suggests that WASP's scaffolding of SFKs via Y291 does not play a role in foci formation or PLCγ1 phosphorylation ( Figure 4D ) . Together these results demonstrate a critical requirement of WASP's Arp2/3 activation and foci nucleation ability for PLCγ1 activation . Thus , decoupling of WASP's interactions in MC from the ability to activate the Arp2/3 complex , but not SFKs , results in a strong disruption of F-actin foci . Next , we more directly addressed the role of Arp2/3 complex , which is known to nucleate actin filaments de novo downstream of WASP ( Dominguez , 2010 ) , in foci induction . A previous study utilizing shRNA-mediated silencing of the Arp2/3 complex subunits in Jurkat T cells had demonstrated that actin polymerization at the synapse can occur independent of the Arp2/3 complex ( Gomez et al . , 2007 ) . However , the high-resolution subsynaptic distribution of Arp2/3 complex at the primary T cell synapse remains unknown . We first assessed the association of Arp2/3 complex with TCR MCs . A significant fraction of endogenous Arp2/3 complex is recruited at the TCR MC sites ( Figure 5A , Figure 5—figure supplement 1A ) . In T cells treated with CK666 , a small molecule inhibitor affecting conformational change in Arp2/3 complex and thus actin nucleation ( Nolen et al . , 2009 ) , the association of both Arpc2 and 3 with TCR MCs was reduced ( Figure 5A , Figure 5—figure supplement 1A ) . The same treatment conditions also led to loss of F-actin foci ( Figure 5B ) . As a control , CK689 , an inactive isomer of CK666 ( Figure 5—figure supplement 1B ) , and the inhibitor for another class of actin nucleation factors , formins ( Rizvi et al . , 2009 ) , or inhibition of myosin II using blebbistatin ( Straight et al . , 2003 ) , did not inhibit F-actin foci ( Figure 5B , Figure 5—figure supplement 1C ) , indicating that these foci specifically require Arp2/3 complex for nucleation , and myosin-mediated actin rearrangement is not required for their formation . 10 . 7554/eLife . 04953 . 023Figure 5 . F-actin foci require activity of the Arp2/3 complex . ( A ) Arp2/3 complex is localized at TCR MCs . DMSO ( Control , top ) or CK666 treated- ( bottom ) AND mouse T cell blasts were labeled with Alexa568-H57 Fab and were incubated with bilayer containing ICAM1 and MHCp for 2 min at 37°C , and then fixed and immunostained with anti-Arp3 antibody . The TIRF images show relative distribution of Arp3 ( left , green ) and TCR ( middle , red ) in the TIRF plane . Note that while Arp3 distribution overlaps with MCs in control cells , MCs are significantly devoid of Arp3 in CK666-treated cells ( areas from ‘merge’ panels further magnified in insets ) . The graph ( far right ) shows the fraction co-localization of total synaptic TCR with Arp3 . n1 = 54 , n2 = 41 , p < 0 . 0001 . In this experiment , the mean intensity of Arp3 in synapse was 3 . 46 ( ±0 . 022 ) × 106 for control cells , and 2 . 38 ( ±0 . 015 ) × 106 for CK666 treated cells ( data not shown ) , and it was repeated twice with similar results . ( B ) TCR MCs associated F-actin enrichment is generated by Arp2/3 complex activation . AND T cell blasts were treated with DMSO alone ( control ) , or 100 μM CK666 , or 100 μM blebbistatin for 10 min at 37°C , were labeled with Alexa568-H57 Fab ( TCR , red ) , and incubated with Alexa405-ICAM1 ( ICAM1 , blue ) and MHCp-containing bilayer in the presence of the specific inhibitors at 37°C for 2 min . Cells were subsequently fixed , stained with Alexa488-phalloidin and imaged using TIRF microscopy . Note that there is a substantial reduction in number of actin foci in the cells treated with CK666 . The graph ( far right ) shows the quantitation of F-actin foci intensity in the cells in inhibitor treatment backgrounds . n1 = 52 , n2 = 28 , n3 = 56 . p1 < 0 . 0001 , p2 = 0 . 36 . ( C ) Immobilized anti-CD3 microdots induce localized actin polymerization that is dependent on Arp2/3 complex . Human CD4 T cells were transfected with LifeAct-GFP plasmid ( green ) , incubated with glass substrate coated with ICAM1 and printed with Alexa647 tagged anti-CD3 microdots ( red ) , and subsequently imaged live . The images represent snapshots taken from a time-lapse sequence , at the indicated time points , ‘0 s’ represents onset of the video sequence acquisition . Note that immediately upon contacting the microdot , cells display localized actin polymerization ( marked circle and * ) . CK666 treatment ( arrow after 95 s panel ) leads to immediate loss of F-actin from the foci . ( D ) Kymograph of F-actin events ( upper left image ) with anti-CD3 dot ( bottom left image ) during contact formation with the microdot prior to and during CK666 treatment . Arrowhead indicates the time of contact of the cell with anti-CD3 dot , and arrow indicates start of CK666 treatment . The line marked on the color combined image ( right ) shows the region of the cell used to create kymograph for the indicated time duration . ( E ) Enrichment of LifeAct ( Actin , black ) at the anti-CD3 microdot ( TCR , red ) in control and CK666-treated fixed cells . The graph shows the mean intensity per pixel obtained using the line-scan profile of LifeAct ( actin ) and anti-CD3 ( TCR ) across 31 microdots from 15 cells . Intensity profiles across identical lines for each microdot were obtained for both anti-CD3 and LifeAct; these values were normalized to the lowest pixel intensity for each line-profile , and their mean ± SEM values are plotted in the graph . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 02310 . 7554/eLife . 04953 . 024Figure 5—figure supplement 1 . Arp2/3 and formin’s role in F-actin foci . ( A ) Arp2/3 complex activation at the TCR MC . AND T cell blasts were labeled with Alexa568-H57 Fab fragments , and were then incubated with bilayer containing MHCp and ICAM1 for 2 min . Cells were then fixed and processed for p34-Arc ( Arpc2 ) immunofluorescence . The images show relative distribution of TCR MCs ( red , left ) and Arpc2 ( green , center ) , at the T cell interface . The graph ( far right ) shows the fraction of TCR localized with Arpc2/cell , in control and CK666-treated cells . n1 = 50 , n2 = 25 , p < 0 . 0001 . This experiment was repeated twice with matching results . ( B ) Effect of CK666 and its inactive control isomer , CK689 , on F-actin foci . AND T cell blasts were treated with DMSO alone ( control , left image ) or with 100 μM CK689 ( center image ) or CK666 ( right image ) for 10 min at 37°C , followed by incubation with ICAM1/MHCp-containing bilayer for 2 min . Cells were fixed and stained for F-actin with Alexa488-phalloidin . ( C ) Formin inhibition does not eliminate F-actin foci . C57BL/6J CD4 T cells were treated with DMSO ( control , left image ) or with 10 μM SMIFH2 ( right image ) for 30 min at 37°C , and were incubated with ICAM1/anti-CD3 bilayer in the presence of the inhibitor for 2 min . Cells were then fixed and stained with Alexa488-phalloidin . F-actin features were quantified in the images ( graph on the right ) . n1 = 63 , n2 = 44 , p = 0 . 18 . Each dot in the graph on the right represents total F-actin foci intensity per cell . Scale bar is 5 µm for all of the above images . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 02410 . 7554/eLife . 04953 . 025Figure 5—figure supplement 2 . F-actin foci on microdots are not enriched in phospholipid membranes . ( A ) F-actin enrichment at the microdot is not a consequence of cell morphology around the microdot . Human CD4 T cells were labeled with CM-DiI according to manufacturer's protocol and were then incubated with glass coverslip coated with patterned Alexa647 tagged anti-CD3 and ICAM1 for 5 min , fixed and stained with Alexa488-phalloidin . Cells were then imaged using TIRF microscopy . The graphs on the bottom panels show line-scan profiles of anti-CD3 ( TCR , red , left ) , actin ( Green , center ) and DiI ( blue , right ) acquired from 70 different microdots . The TCR , actin and DiI intensities were measured from identical pixel positions for a given microdot , were then normalized by the lowest pixel intensity per microdot for a given fluorescent channel , and plotted . Note that while actin shows enrichment at the microdot site , DiI is not enriched at the same position . ( B ) The same procedure as described above , except with the use of Cy3-anti- LFA1 Fab fragments instead of DiI , was carried out in T cells , and the normalized line-scan intensities were plotted as average value per pixel position ±SEM . The graph represents mean of normalized intensities across 25 pixels , acquired using 74 different microdots . In both these experiments ( A , B ) , n > 20 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 025 Since CK666 treatment can rapidly eliminate foci , we utilized CK666 to probe the kinetics of foci formation at individual TCR MCs . Given the small size and large lamellar actin background at the primary T cell synapse as well as highly mobile nature of MCs , this analysis would be extremely challenging to perform in the bilayer system . Therefore , we activated T cells on micron-scale anti-CD3 dots adsorbed on a glass surface where the remaining space was coated with ICAM1 in an attempt to restrict TCR signaling to a known , micron scale location in an artificial immunological synapse . Due to their aforementioned performance in time-lapse studies after transfection , human CD4 T cells were utilized for LifeAct-GFP transfection and live imaging on anti-CD3 microdots . Incubation with the patterned surface triggered rapid actin polymerization at the anti-CD3 microdot , reflected in LifeAct-GFP accumulation ( Figure 5C–E ) . This enrichment of LifeAct at the microdot was not due to the topography of the cell contact interface ( Owen et al . , 2013 ) , since a global membrane labeling dye , DiI , and integrin LFA1 were not enriched at these sites ( Figure 5—figure supplement 2A , B ) . Within 5 s of CK666 treatment , pre-existing LifeAct enrichment was lost from the microdot site ( Figure 5C–E Video 5 ) and fresh microdot contact events failed to elicit LifeAct-GFP enrichment in the presence of CK666 ( Video 5 ) . These data confirm that the TCR engagement sites dictate Arp2/3 complex-dependent F-actin foci formation at the synapse , de novo . Furthermore , ligand microdots can spatially pattern F-actin foci by predefining the sites of TCR engagement , and thus provide a suitable platform for spatially isolating TCR dependent mechanisms . 10 . 7554/eLife . 04953 . 026Video 5 . Human CD4 T cells were transfected with LifeAct-GFP plasmid ( green ) and incubated with ICAM1-coated glass substrate with 1 µm anti-CD3 printed dots ( red ) on stage at 37°C and imaged live . The play rate is 25 times faster than the acquisition rate . In all of the above images and videos , scale bar is 5 µm , unless otherwise noted . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 026 We have shown above that WASP's foci forming ability is required for PLCγ1 activation ( Figure 4B ) . In order to test whether the independent elimination of foci by perturbation of Arp2/3 complex would phenocopy this effect , we chose to deplete foci using CK666 and monitor the consequences of this treatment on early TCR signaling , PLCγ1 activation and subsequent calcium ion signaling . CK666 treatment , which resulted in severe inhibition of F-actin foci ( Figure 5B , Figure 6—figure supplement 1B ) , did not diminish TCR-proximal activation events . While fewer T cells formed synapses after CK666 treatment ( 37% reduction ) , upon adhesion , F-actin dependent processes including formation of TCR MC and exclusion of CD45 phosphatase from MCs ( Campi et al . , 2005; Varma et al . , 2006 ) ( Figure 6A ) , and the phosphorylation of Zap70 on Y319 ( Figure 6B ) proceeded normally in CK666-treated cells . In addition , other TCR-activation dependent signaling events leading to calcium flux , such as the phosphorylation of LAT on Y171 and SLP76 on Y145 were not altered in CK666-treated cells in comparison to control treated cells ( data not shown ) . However , a severe defect was observed in the levels of TCR-induced PLCγ1 phosphorylation on Y783 after CK666 treatment ( Figure 6C , D ) , where a drop was observed in both the number and intensity of phospho-PLCγ1 clusters at the synapse ( Figure 6C ) . Furthermore , while phospho-PLCγ1 localized with F-actin foci in control cells , treatment with CK666 led to loss of foci-associated phospho-PLCγ1 ( Figure 6E , Figure 6—figure supplement 1A ) . Upon attenuation of F-actin foci , total PLCγ1 levels were also reduced at the synapse , matching the decline in phospho-PLCγ1 levels ( Figure 6F ) and indicating that the impaired levels of phospho-PLCγ1 at the whole cell level may be accounted for by the reduction in PLCγ1 enrichment at the TCR signalosome . When assaying signaling events downstream of PLCγ1 activation , such as TCR-induced calcium ion elevation in the cytoplasm and nuclear mobilization of NFAT1 , CK666-treated cells showed significant reduction in these processes ( Figure 6G , H ) . Although a reduction in lamellipodial and lamellar actin was observed in CK666 treated cells ( mean total F-actin 42% reduced ) , this reduction was relatively lower than foci depletion ( mean 67% reduced ) or under global F-actin depletion ( CytoD , mean F-actin 90% reduced ) , and did not affect early TCR signaling ( Figure 6—figure supplement 1B ) . We thus conclude that WASP and Arp2/3 dependent foci play a non-redundant role in PLCγ1 activation and calcium ion signaling at the synapse . In line with this result , CK666-treated cells exhibited a loss of phospho-HS1 , as well as HS1 , at the synapse ( Figure 6—figure supplement 1C ) , similar to that observed in WASP−/− T cells ( Figure 1E ) . Similar to the TCR-distal signaling defects observed in CK666 treated mouse CD4 T cells described above , CK666-treated human CD4 T cells also displayed a reduction in phospho-PLCγ1 , but not in phospho-Zap70 ( Figure 6—figure supplement 2 ) , indicating that foci-dependent PLCγ1 activation cascade is conserved in T cells from diverse origins . 10 . 7554/eLife . 04953 . 027Figure 6 . Arp2/3 inhibition leads to defective TCR-distal signaling . ( A ) Formation of TCR MCs or CD45 exclusion does not require F-actin foci . AND CD4 T cell blasts labeled with Alexa568-H57 Fab ( to assess TCR clustering ) , or Alexa488-CD45 Fab ( for CD45 exclusion ) at 4⁰C , were then incubated with bilayers containing ICAM1 and MHCp for 2 min , fixed ( to assess TCR clustering ) or visualized live ( for CD45 exclusion ) using TIRF microscopy . The images were processed to assess TCR clustering , or CD45 and TCR colocalization using rank filter based filtering , as described in ‘Materials and methods’ section . For the bars showing TCR cluster intensities per cell , n1 = 45 , n2 = 28 , p = 0 . 6; for CD45 co-localization , n1 = 26 , n2 = 13 , p = 0 . 09 . ( B ) Phosphorylation of TCR-proximal molecule Zap70 is not reduced in the cells treated with CK666 . T cells were treated with DMSO or CK666 for 10 min , then incubated with surface containing ICAM1/anti-CD3 for 3 min , and processed for Y319-phospho-Zap70 and imaged using TIRF . The images were quantified to obtain the synaptic levels phospho-Zap70 , and plotted as normalized to mean value of the ‘control’ cells . In the graph shown here , n1 = 34 , n2 = 38 , p = 0 . 16 . ( C ) Synaptic phospho-PLCγ1 levels are reduced in cells lacking F-actin foci . DMSO ( control , top panel ) or CK666 ( bottom panel ) treated AND CD4 T cell blasts were incubated with bilayer containing anti-CD3 and ICAM1 for 2 min , fixed and stained with Alexa488-phalloidin ( green ) and phospho-Y783-PLCγ1 ( red ) , and visualized using TIRF microscopy . In these images , total synaptic levels of phospho-PLCγ1 were assessed ( top left , graph ) . These images were also analyzed using integrated morphometry and total number of phospho-PLCγ1 events per cell ( c , top right , graph ) or intensity distribution of the events across the population of T cells ( bottom right , histogram ) were measured . Note that CK666 treatment leads to a uniform reduction in phospho-PLCγ1 events across different intensity ranges ( c , bottom histogram ) . In all graphs in c , n1 = 50 , n2 = 67 , p < 0 . 0001 ( D ) Total cellular levels of phospho-PLCγ1 and PLCγ1 with or without treatment of cells with CK666 . CD4 T cell blasts were treated with 100 μM CK666 or DMSO ( control ) , and then incubated with anti-CD3/CD28 beads for 5 min in the presence of the inhibitor , lysed and analyzed using western blotting . ( E ) In CK666-treated cells , there is a substantial reduction in synaptic phospho-PLCγ1 co-localized with F-actin foci . Colocalization analysis was performed on images from ( C ) , to estimate phospho-PLCγ1 and F-actin colocalization , as described in ‘Materials and methods’ . ( F ) Pan- PLCγ1 levels at the synapse are reduced in CK666-treated cells . T cells were processed for PLCγ1 immunofluorescence and imaged as described above . In the graph , each dot represents PLCγ1 intensity per cell in the TIRF images . n1 = 29 , n2 = 34 , p < 0 . 0001 . ( G ) Defective calcium ion mobilization in CK666 treated cells . AND CD4 T cell blasts were loaded with Fluo4 , and treated with DMSO or 100 µM CK666 for 1 min , and then imaged live on anti-CD3 and ICAM1 , to monitor calcium ion flux ( ‘Materials and methods’ ) . Each point on the graph ( far right ) represents mean value of the baseline corrected fluorescence ±SEM ( n = 30 cells ) . ( H ) Nuclear translocation of NFAT1 in cells treated with CK666 . DMSO ( control , top panels ) or CK666-treated ( bottom panels ) AND CD4 T cell blasts were incubated with glass coverslips coated with ICAM1 alone ( left ) , or ICAM1 and anti-CD3 ( right ) for 10 min at 37°C , fixed and immunostained for NFAT1 ( red ) , and stained with Alexa488-phalloidin ( green ) . Cells were subsequently imaged using confocal microscopy . The middle sections from the z-stack of the images of the cells were used to quantitate nuclear levels of NFAT1 , by outlining phalloidin-free central area ( nucleus ) of the cell . The graph ( right ) shows nuclear NFAT intensity in the nuclear area , in a single section per cell . n1 = 98 , n2 = 40 , n3 = 156 , n4 = 31 . p-values , p1 = 0 . 92 , p2 , p3 < 0 . 0001 . ( I ) PLCγ1 dynamics at TCR microdots . Human CD4 T cells transiently expressing PLCγ1-YFP ( PLC , green ) were incubated with glass coverslips patterned with Alexa647 tagged anti-CD3 microdots ( TCR , red ) and coated with ICAM1 and imaged live using TIRF microscopy . The graphs on the bottom show linescan intensity profiles of PLCγ1 and anti-CD3 ( TCR ) across the pixels marked in the corresponding image in the top panels . The middle panel shows the intensity profiles during the addition of CK666 , and the right panel shows the profiles 30 s after the addition of CK666 . The graph in ( J ) shows anti-CD3 ( TCR ) and PLCγ1-YFP profiles obtained from at least 30 different microdots from >12 cells in each control and CK666 treatment background . PLCγ1-YFP expressing human T cells were incubated with patterned substrate for ( 1 + 4 ) min in the presence of DMSO ( Control ) or CK666 , fixed and imaged using TIRF microscope . The CK666 was added after 1 min of incubation of cells with the substrate . The graph shows mean value of pixel intensities calculated from the linescan profiles . The pixel intensities within a linescan were normalized to lowest pixel intensity within that linescan prior to calculating mean value across different linescans . For a given treatment background and linescan profile , TCR and PLCγ1 intensities were obtained from the identical pixel positions . Scale bar , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 02710 . 7554/eLife . 04953 . 028Figure 6—figure supplement 1 . CK666 treatment leads to loss of foci-associated PLCγ1 and phospho-HS1 from the synapse . ( A ) Images from Figure 4E were analyzed for phospho-PLCγ1 and F-actin foci chance level co-localization , after pixel shifts of phospho-PLCγ1 images by 5 pixels , as performed earlier , in Figure 3—figure supplement 1 . Note that both in DMSO ( control ) as well as CK666 treated cells , a shift of 5 pixels ( x and y ) in phospho-PLCγ1 image results in significant reduction in co-localization , indicating that loss of overlap in CK666 treatment is not just due to reduction in total phospho-PLCγ1 levels . ( B ) Characterization of the effect of CK666 vs Cytochalasin D ( CytoD ) on total synaptic F-actin , F-actin foci , phospho-Zap70 and phospho-PLCγ1 levels . Mouse CD4 T cells were treated with 100 µM CK666 or 5 µM CytoD for indicated time duration , were activated on anti-CD3 and ICAM1 containing surface for 2 min in the presence of inhibitors , fixed and processed for immunofluorescence of the indicated molecules . Note that , while both CK666 as well as CytoD treatments lead to loss of F-actin foci , CytoD treatment causes a greater reduction in total F-actin and phospho-Zap70 levels . For the left graph , for total actin and F-actin foci , n1 = 123 , n2 = 118 , n3 = 100 , n4 = 87; for phospho-Zap70 , n1 = 49 , n2 = 47 , n3 = 53 , n4 = 42; for phospho-PLCγ1 , n1 = 78 , n2 = 67 , n3 = 48 , n4 = 42 . For total phospho-Zap70 , p > 0 . 90 , and for the rest of the datasets , p < 0 . 0001 . In the CytoD graph , for total actin and F-actin foci , n1 = 128 , 67 , 88; for phospho-Zap70 , n1 = 80 , n2 = 27 , n3 = 57; for phospho-PLCγ1 n1 = 52 , n2 = 54 , n3 = 31 . For all of the data across treatment durations , p < 0 . 0001 . ( C ) CK666-treated or untreated AND T cells were incubated with anti-CD3/ICAM1 containing bilayer , and stained with phospho-HS1 antibody ( left graph ) or pan-HS1 antibody ( right graph ) . Cells were then imaged , and images were quantified for total synaptic phospho-HS1 ( left graph ) or total synaptic HS1 ( right graph ) intensities per cell . Each point on the graph represents integrated intensity of the indicated protein per cell . In the left graph , n1 = 47 , n2 = 18 , p < 0 . 0001; for the graph on the right , n1 = 42 , n2 = 59 , p < 0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 02810 . 7554/eLife . 04953 . 029Figure 6—figure supplement 2 . Effect of CK666 on TCR-induced Zap70 and PLCγ phosphorylation in human T cells . Freshly purified human CD4 T cells were treated with DMSO ( control ) or CK666 , and incubated with bilayer reconstituted with anti-CD3 and ICAM1 for 2 min , in presence of the inhibitor . Cells were then fixed and stained with Alexa488 phalloidin ( Actin , green ) and either of anti-phospho Zap70 ( upper panels ) or anti-phospho-PLCγ1 ( Lower panels ) antibodies , and subsequently imaged using TIRFM . The graph represents phospho-Zap70 ( upper graph ) or phospho-PLCγ1 levels at synapse , normalized to the mean value in control cells . For upper graph , n1 = 49 n2 = 47 , p = 0 . 976; for the lower graph , n1 = 78 , n2 = 67 , p = 0 . 004 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 02910 . 7554/eLife . 04953 . 030Figure 6—figure supplement 3 . Arp2/3 complex inhibition and synaptic dynamics of Zap70 and PLCγ1 in live cells . Human CD4 T cells were transfected with Zap70-GFP ( A ) or PLCγ1-GFP ( B ) for 16 hr , were then incubated with bilayers containing ICAM1 and Alexa568-OKT3 and imaged live at a rate of 1 frame per 5 s . The images represent maximum intensity projection of five frames ( 25 s of imaging ) immediately before ( -CK666 ) or after ( +CK666 ) CK666 treatment onstage . The projections have been scaled to an identical extent before ( upper panels ) and after ( lower panels ) CK666 treatment . An area in the ‘Merge’ panel has been further magnified in the insets ( rightmost panels ) . The insets have been scaled differently from the original ‘Merge’ panels ( but similarly between + CK666 and–CK666 cases ) , to visualize local protein distribution in better details . Note that while Zap70 localization to mobile TCR MC is maintained after CK666 treatment , PLCγ1 localization is severely reduced . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 030 To directly visualize the involvement of F-actin foci in PLCγ1 recruitment and stabilization at the TCR signalosome in live cells , we transiently overexpressed PLCγ1-YFP in human CD4 T cells , and examined its distribution at the synapse on SLB presenting anti-CD3 and ICAM1 in real time . PLCγ1 localized with TCR microclusters in live cell synapse , and this distribution was lost upon treatment of cells with CK666 ( Figure 6—figure supplement 3 ) . As a control , the association of human Zap70-GFP with MC was maintained after CK666 treatment ( Figure 6—figure supplement 3 ) . However , due the mobile nature of MC , it was challenging to follow individual cluster before and after CK666 treatment in this setting . We thus chose to utilize anti-CD3 microdots to activate human T cells . Human CD4 T cells expressing PLCγ1-YFP were incubated with micron-size immobile anti-CD3 dots , and visualized using live cell imaging immediately after cell attachment . Similar to F-actin foci assembly at the microdot , PLCγ1 enrichment was also visible at the microdot sites . CK666 treatment of cells led to a reduction of PLCγ1 enrichment at microdots ( Figure 6I , J ) . These data show that TCR MC associated F-actin foci assist in sustained accumulation of PLCγ1 at the TCR MC . While SLB and solid-phase adsorbed antigen presentation systems allowed for superior optical resolution of the synapse , it is important to further test the validity of predictions from these reductionist models in more physiological T cell-APC interactions . Thus we utilized an activated cultured endothelial cells ( EC ) based planar APC system , where a flat synaptic interface is formed that is ideal for en-face visualization ( Sage et al . , 2012 ) . Using this system , F-actin and HS1 rich ‘invadopod-like-protrusions’ ( ILPs ) are observed at TCR MC-like features in the periphery of the synapse ( Sage et al . , 2012 ) . We found that ILPs were associated with the Y397 phosphorylated form of HS1 , similar to F-actin foci ( Figure 7A ) . Furthermore , when pre-formed T cell-EC conjugates were treated with CK666 , there was concomitant loss of ILPs and phospho-HS1 , indicating that these structures also require Arp2/3-dependent continuous polymerization of F-actin , and support local phospho-HS1 levels ( Figure 7—figure supplement 1A–C ) . Since CK666-treatment also led to a reduction in total synaptic F-actin , which could be a consequence of inhibition of the EC cytoskeleton ( Figure 7—figure supplement 1A–C ) , we utilized specific shRNA-mediated reduction in WASP levels in T cells , and examined foci ( ILPs ) , total F-actin and phospho-HS1 levels at the T cell-EC conjugate interface ( Figure 7A , B ) . WASP silenced T cells display selective loss of both ILPs as well as phospho-HS1 , while maintaining total F-actin levels ( Figure 7A–C ) . Therefore , the F-actin foci that we have primarily been characterized using SLB in this study , are the counterparts of ILPs in the cell conjugate system , and are necessary for efficient T cell activation . Thus , F-actin foci may represent a cytoskeletal module that functions to optimally support TCR distal signaling ( Figure 7C ) in diverse antigen presenting contexts . 10 . 7554/eLife . 04953 . 031Figure 7 . ILP signaling in T-EC immunological synapse is WASP dependent . Human CD4 T cells were incubated with culture media containing lentiviral particles carrying WASP shRNA or non-specific ( control ) shRNA for 48 hr ( A ) T cells transduced with WASP shRNA or control shRNA carrying lentiviral particles were incubated with endothelial monolayer for 10 min , fixed and processed for Alexa594-phalloidin ( pseudo-colored green ) and phospho-HS1 ( pseudo-colored red ) immuno-staining . The conjugates were then imaged using an EMCCD-coupled spinning disc confocal microscope . Each image represents a single confocal plane of T cell synapse , where the planar endothelial interface is in focus . The area outlined in ‘merge’ panels was further scaled and magnified to show the details with more clarity ( bottom panels ) . The top panels show the image of the field of view in DIC ( left image ) or fluorescence settings . ( B ) A reduction in WASP levels results in defective phospho-HS1 accumulation at T cell-endothelial cell synapse . The upper graph shows quantitation of phalloidin intensity in the synaptic plane , while the lower graph shows phospho-HS1 levels in the same plane . For both the upper and lower graphs , n1 = 68 , n2 = 29 , p1 = 0 . 071 , p2 < 0 . 0001 . This experiment was repeated twice with similar results . ( C ) Model of temporal sequence of events leading to F-actin foci formation and PLCγ signaling at the immunological synapse . Multiple pathways can result in actin polymerization and remodeling at the synaptic interface , contributing to F-actin organization in different SMAC zones . One such pathway involves WAVE2 recruitment by activated LFA1 , followed by WAVE2 dependent Arp2/3 complex activation resulting in thick lamellipodial ( dSMAC ) and lamellar ( pSMAC ) F-actin meshworks . WAVE2-dependent F-actin pool is required for calcium-dependent calcium entry via the CRAC channel . Additional pathways including MyosinII-mediated actin remodeling is required for maintaining lamellar actin flow and directional persistence of microclusters ( MCs ) towards the cSMAC , and formin-mediated nucleation of F-actin promotes MTOC docking and stability of synapse . Another pool of F-actin or ‘F-actin foci’ is generated by the activity of WASP protein in the p- and dSMAC zones . Following TCR triggering , WASP is recruited at TCR signalosome via several possible mechanisms – such as via Vav , via NCK , via Zap70 and CrkL mediated WIP release and other effector mechanisms , and , through Fyn or PIP2 or PTP-PEST-binding at the plasma membrane ( PM ) . Once activated , WASP recruits Arp2/3 complex to the MC , which then leads to actin branch nucleation and polymerization at the MC , over and above the local background actin . This process continues even during MC movement in the lamellar region , with a high F-actin turnover at the foci until its delivery to the cSMAC . In the foci , HS1 is recruited via binding both the Arp2/3 complex as well as F-actin , and is subsequently phosphorylated . As a consequence of early TCR signaling , PLCγ1 is also recruited to the MC signalosome , where it is stabilized via interactions with both F-actin , and foci residing HS1 . F-actin foci dynamics in the proximity of the plasma membrane further support PLCγ1 phosphorylation , potentially by facilitating its interaction with PM-bound , upstream activators such as Itk . Phosphorylation of PLCγ1 by Itk then triggers phosphoinositide signaling , which in turn initiates calcium ion flux and NFAT1 activation . WASP deficiency or failure to activate Arp2/3 complex by WASPΔC mutant leads to selective loss of nucleation of foci at the MC . As a result , early signaling is not affected , however , both HS1 and PLCγ1 levels are severely reduced at the microcluster sites . The remaining PLCγ1 at synapse allows cell spreading and synapse formation , however , it is not sufficient to achieve calcium flux comparable to the control cells . Direct pharmacological inhibition of Arp2/3 complex using CK666 yields similar results; early TCR signaling is preserved while PLCγ1 phosphorylation and late signaling are severely perturbed . As actin polymerizing processes other than WASP also utilize Arp2/3 Complex , CK666-treated cells show a general reduction in lamellipodial and lamellar actin as well . However , the remaining F-actin levels are sufficient to support early TCR signaling . In contrast , total F-actin depolymerization at the synapse using CytochalasinD results in defects in early as well as late signaling , as has been reported in earlier studies . The image on the bottom shows a maximum intensity projection of synaptic contact interface of a human primary CD4 T cell , acquired using spinning disc confocal microscope . This cell was activated on a bilayer reconstituted with Alexa568 tagged anti-CD3 ( red ) and ICAM1 ( unlabeled ) , for 2 min , fixed and stained for F-actin ( green ) , and imaged . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 03110 . 7554/eLife . 04953 . 032Figure 7—figure supplement 1 . ILP F-actin and signaling in T-EC immunological synapse is dependent on Arp2/3 activity . ( A ) Visualization of loss of F-actin in ILPs following CK666 treatment . T cells were incubated with membrane-YFP ( green ) expressing endothelial cells in the presence of DMSO ( control , left ) or CK666 ( right ) for 5 min , then fixed , permeabilized and stained with Alexa594-phalloidin ( ‘Actin’ , pink ) . Note that F-actin rich protrusions are missing at the CK666-treated T cell interface . Scale bar , 5 µm . ( B ) Loss of ILPs correlated with reduced phospho-HS1 at the T cell-endothelial cell synaptic interface . T cells were incubated with endothelial cell layer , as described in ‘Materials and methods’ for 5 min and were then fixed and stained with Alexa594-phalloidin ( ‘Actin’ , green ) , phospho-HS1 ( red ) and anti-CD3 ( blue ) . Cells were then imaged using confocal microscopy . The insets show single T cells marked in the ‘Merge’ image . Scale bar , 10 µm . ( C ) The intensity of the indicated proteins was then analyzed in the synaptic plane without and with CK666 treatment . n1 = 58 , n2 = 71 , p1 = 0 . 031 , p2 , p3 < 0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 04953 . 032 Compartmentalization of signaling proteins is a well-established strategy for spatial regulation of signaling in mammalian cells . At the immunological synapse , the compartmentalization of TCR signalosome proteins is regulated for optimal function ( Fooksman et al . , 2010 ) . It is therefore surprising that cytoskeletal behavior is routinely measured as an all-or-none response of total polymerized F-actin , effectively ignoring different F-actin network subpopulations that could have distinct contributions to specific steps of the TCR signaling process . The WASP-dependent F-actin microarchitecture at the TCR MC that we report here is an important example . These F-actin ‘foci’ facilitate specific downstream TCR signaling steps , including PLCγ1 activation and calcium ion elevation , but are not required for TCR MC formation and other proximal signaling events such as Zap70 phosphorylation . Our study has shown that , when F-actin foci are depleted by WASP deficiency or CK666 treatment , the remaining lamellipodial and lamellar F-actin networks are not able to support optimal PLCγ1 activation , indicating that actin foci are indeed necessary for this later signaling step . Furthermore , we found that the molecular machinery associated with ‘flat’ TCR MC formed on planar substrates is the same as that in the distinctly three-dimensional ILP structures formed in T cell interfaces with antigen presenting endothelial cells . Others have noted the three dimensional nature of T-APC synapses that appear to have a classical immunological synapse/SMAC architecture ( Brossard et al . , 2005; Biggs et al . , 2011; Roybal et al . , 2013 ) . Our study thus defines a common F-actin rich module that bridges planar models and more complex , but more physiological , T-APC systems . F-actin ‘foci’ have been noted in prior studies ( Barda-Saad et al . , 2005; Beemiller et al . , 2012 ) , but they have not been studied in any detail . Ours is the first study , to our knowledge , to have performed a detailed characterization of foci as structures that are polymerized de novo via a WASP dependent mechanism . Previously , it was proposed that similar TCR associated F-actin foci were formed through a buildup of retrogradely flowing F-actin moving over artificially immobilized TCR on chrome micro-barriers in both Jurkat and primary T cells ( Smoligovets et al . , 2012; Yu et al . , 2010 ) . Further characterization of this buildup revealed that it is due to accumulation of pre-existing filaments ( Smoligovets et al . , 2013 ) . We avoided this pitfall here by utilizing a system in which TCR MC remain mobile or are pre-established in micron scale spots that are segregated and are less likely to be traversed by F-actin flows , preventing an accumulation of filaments . T cell ILP display molecular similarities to ‘podosomal’ actin structures that are a common feature of leukocyte contact with substrates ( Carman et al . , 2007; Dovas and Cox , 2011 ) . Composed of dynamic F-actin , such podosomes are rich in HS1 related cortactin and are formed in a WASP-dependent manner at the adhesion interface of leukocytes ( Dovas and Cox , 2011 ) . We speculate that the TCR utilizes similar WASP-HS1 based molecular machinery to generate F-actin foci/ILP that then serve to support calcium ion signaling through stabilization of PLCγ recruitment and phosphorylation . In the immunoreceptor driven synapses of non-T cells , this foci-like organization of actin microfilaments has not been observed thus far . F-actin architecture at the NK cell synapse appears to be uniform and the deep lamellipodial network dominates the synaptic interface ( Rak et al . , 2011 ) . Similarly , in B cells , a thick cortical actin meshwork that is inhibitory for early B cell receptor signaling has been reported ( Treanor et al . , 2011 ) , but evidence of any ‘foci-like’ organization is missing . In such cell types , WASP could serve other roles , and may not promote immunoreceptor signaling via foci generation at the synapse . In the case of both CD4 and CD8 primary T cells , foci are robustly triggered following TCR ligation . In activated mouse T cell blasts , antigen-dependent induction of foci was observed as well; however , a few low intensity actin features were visible in a fraction of T cell blasts activated on ICAM1 in SLB ( Figure 4A ) . We do not understand the origin of these low-intensity foci . It is possible that foci are generated at a basal level throughout the active signaling state of T cell blasts after initial priming , and antigen re-stimulation is a potent trigger for their formation . Alternatively , it is possible that integrins such as LFA-1 are capable of forming a few transient actin-rich features in activated lymphocytes in a specific chemokine environment , when its ligand ICAM1 is highly expressed . In contrast to the primary T cell systems , we have not detected foci in the Jurkat T cell line activated using the SLB system . A signaling network linking TCR to WASP has been outlined in the Jurkat cells ( Barda-Saad et al . , 2005; Thrasher and Burns , 2010 ) , so it is likely this signaling pathway is intact . However , any WASP-dependent foci that may exist in these cells are either smaller than the resolution limit , or are transient in nature and rapidly dissipate into the lamellar network before they can be observed . Interestingly , in Jurkat T cells , the actin polymerizing function of WASP is not required for its role in calcium ion signaling ( Silvin et al . , 2001 ) , and Arp2/3 silencing does not prevent PLCγ1 activation ( Gomez et al . , 2007 ) . Thus , it is also possible that these cells utilize an alternative WASP-independent cytoskeletal pathway to support TCR-induced calcium ion elevation ( Babich et al . , 2012 ) , and therefore do not need to utilize foci for this purpose . For example , the lack of negative regulators like PTEN may allow Jurkat cells to sustain calcium ion signals through alterations in lipid metabolism , without F-actin foci . A detailed comparison of the differential organization and function of F-actin modalities between Jurkat cells and the primary T cells will provide fundamental insights into the operation of compensatory cytoskeletal pathways that assist TCR distal signaling and may be important at different developmental stages in the T cell lineage or part of the failure of growth control in malignancy . Our study revealed at least two functional classes of F-actin at the synapse . The first component is critical for very early signaling events encompassing TCR triggering and the proximal signaling cascade . This fraction is represented by a slow turnover CK666-resistant cortical F-actin pool , which most likely relies on filament remodeling proteins other than actin branching for its dynamics . The second component , the WASP and Arp2/3-dependent F-actin foci network is responsible for assisting PLCγ1 activation and calcium ion signaling . Furthermore , in addition to these two networks , there is evidence for a third , CK666-sensitive network that we have not investigated in this study . This network is made up of a fraction of lamellar and lamellipodial F-actin that is rapidly eliminated under even short duration treatments of CK666 based on the reduction in ‘total F-actin’ ( Figure 6—figure supplement 1B ) . This observation indicates that this third sub-population is composed of dynamic F-actin . This network is not required for the early TCR signaling steps that we have tested , indicating that it may play other roles at the synapse that are known to be dependent on F-actin . For example , it may be equivalent to the WAVE2 dependent network that is not required for early TCR signaling or PLCγ1-dependent calcium ion efflux from endoplasmic reticulum , but is necessary for CRAC channel mediated calcium ion entry from extracellular medium ( Nolz et al . , 2006 ) . Considering that all the above-mentioned networks are made up of the same building blocks of G-actin , how are functional boundaries between them set up and regulated , and how is this related to the organizational differences between these networks ? Given the small size of T cells , the highly dynamic nature of their F-actin structures and the pleotropic action of pharmacological inhibitors on all F-actin organizations , it is challenging to dissect the specific roles of these sub-populations during TCR signaling . Thus , further interrogation of these networks will require a combination of genetic tools enabling conditional perturbation , as well as rescue , of Arp2/3 complex and NPFs ( Moulding et al . , 2007; Blundell et al . , 2009; Sato et al . , 2013 ) , in combination with super-resolution microscopy in primary T cells . The results from these experiments will reveal how these networks combine to generate an emergent multifunctional synaptic F-actin network during T cell activation . The primary recruitment of PLCγ1 from cytosol to the MC has been well characterized , and it involves its interaction with TCR signalosome proteins LAT and SLP76 , and subsequent phosphorylation at PLCγ1 residue Y783 by membrane-bound Itk ( Park et al . , 1991; Braiman et al . , 2006 ) . Loss of F-actin foci resulted in reduced phospho-Y783 PLCγ1 levels at the synapse ( Figure 6C–D ) . The microdot enrichment assays and the SLB experiments both indicated that this was largely due to defects in PLCγ1 accumulation at the MCs ( Figure 6 ) . Accumulation of PLCγ1 at actin foci could occur via its direct association with actin filaments , as reported earlier ( Carrizosa et al . , 2009; Patsoukis et al . , 2009 ) , as well as with foci-recruited HS1 . Both these mechanisms could simultaneously contribute to accumulation of PLCγ1 at TCR MCs . The precise manner by which accumulation at actin foci promotes PLCγ1 activation is not clear . There are several possible mechanisms that could underlie this process . First , it is probable that the continual nucleation of F-actin filaments at foci provides a sustained cytoskeletal microenvironment for the enrichment of PLCγ1 around the TCR signalosome . As the TCR MC traverses regions with highly variable levels of lamellar F-actin en route to the cSMAC , the foci would provide for the continuous presence of high levels of F-actin at the TCR microcluster , which may in turn ensure continuous local enrichment of PLCγ1 at the MC . Second , foci F-actin growing in juxtaposition to the plasma membrane may facilitate the interaction of F-actin bound PLCγ1 with its plasma-membrane anchored effectors such as Itk , thereby enhancing its phosphorylation . In this model , the growth of foci would exert force on and push the plasma membrane in its vicinity . Consistent with this mechanism , we find that foci zones mark areas of high membrane tension , observed as the foci-dependent localization and increased phosphorylation of mechanosensory protein CasL ( Kumari et al . , 2012;; Yu et al . , 2012 ) at the MC-foci sites ( Santos et al . , manuscript under review ) . This is also consistent with the manifestation as ILPs on the softer endothelial cell substrates . This observation also implies a possible role of F-actin foci in supporting synapse stability . Previously we have shown that WASP−/− T cell synapses form normally with normal pSMAC and dSMAC actin organizations , however they lost stability faster than those of WT cells ( Sims et al . , 2007 ) . It will therefore be interesting to investigate whether the synapse stabilizing role of WASP is mediated by actin foci , and if foci influence pSMAC oscillations , as reported earlier in WASP−/− cells ( Sims et al . , 2007 ) . Interestingly , in CK666-treated cells , there is a modest initial rise in intracellular calcium ion concentration , but final calcium ion concentrations are significantly lower than in DMSO-treated control cells and rapidly decline thereafter ( Figure 6G ) . The factors contributing to residual calcium ion elevation may be either residual phopspho-PLCγ1 in CK666-treated cells , or an additional mechanism of PLCγ1 activation that is insensitive to this inhibition . In accordance with this , we do not observe a complete loss of PLCγ1 at microdot sites in CK666 treated cells . In contrast to the above mentioned processes that augment calcium ion signaling in mature T cells , there are alternative actin cytoskeletal pathways in immature thymocytes that impinge upon PLCγ1 , and total F-actin is inhibitory for PLCγ1 activation and calcium ion flux ( Tan et al . , 2014 ) . Detailed examination of F-actin networks in T cells during different developmental stages will clarify how the actin cytoskeleton can generate different activation outcomes at the level of PLCγ1 ( Kumari et al . , 2014; Roybal et al . , 2013; Malissen et al . , 2014 ) . We find a specific requirement for WASP in the generation of F-actin foci . WASP−/− T cells show normal levels of total F-actin at the synapse , but no TCR MC associated F-actin foci . This lack of effect on total F-actin in WASP deficient or depleted cells may be attributed to the fact that the foci contribute to <10% of the total synaptic F-actin intensity ( Figure 1—figure supplement 1B ) , and perhaps to a compensatory increase in the background actin networks when more ATP-G-actin is available for polymerization regulated by other NPFs such as WAVE2 . We have shown that WASP's role in generating F-actin foci cannot be substituted for by NWASP . This selective requirement for WASP could be due to its specific interactions with upstream activators at the TCR signalosome . For example , a conventional mode of WASP and NWASP activation is through interaction with Cdc42 GTPase , However , WASP recruitment can occur independently of Cdc42 at the T cell synapse ( Cannon et al . , 2001 ) . In agreement with this , in preliminary experiments , Cdc42−/− T cells showed normal F-actin foci ( data not shown ) indicating that these foci form without the aid of Cdc42 . Additionally , Cdc42-dependent phosphorylation of WASP at Y291 by SFK is known to increase its affinity for Arp2/3 complex ( Torres and Rosen , 2003 ) . Overexpression of a phospho-deficient mutant ( WASP Y291F ) ( Blundell et al . , 2009; Dovas et al . , 2009 ) did not perturb F-actin foci formation or PLCγ1 phosphorylation , indicating that the WASP activation pathway utilized for foci formation may be distinct from the upstream regulatory pathways involving Cdc42 activation , or SFK binding at phospho-Y291 . While it was not possible to dominantly uncouple TCR MC from PLCγ1 activation by over-expressing WASP that is able to activate Arp2/3 , but does not interact with SFKs , it is possible that other regulatory phosphorylation sites can scaffold WASP's interaction partners , thereby playing an ancillary role in PLCγ1 activation ( Torres and Rosen , 2003; Blundell et al . , 2009; Macpherson et al . , 2012; Reicher et al . , 2012 ) . These results , and the failure of compensation by NWASP , suggest that WASP is activated and employed in a specialized manner in T cells to generate F-actin foci . The dimerization property of WASP and positive feedback with Arp2/3 complex ( Padrick and Rosen , 2010 ) , as well as its role in polarizing micron-scale ‘lipid rafts’ at the synapse ( Dupre et al . , 2002 ) may contribute towards a specific enrichment and involvement of WASP at foci . At this stage , the mechanistic basis for WASP vs NWASP utilization is not understood . To summarize , we have characterized a subsynaptic F-actin organization in T cells that is functionally distinct from larger scale lamellipodial and lamellar networks . These actin foci provide a signaling scaffold that is required for TCR-distal signaling and play a critical regulatory role in calcium ion signaling . There are significant questions that remain to be answered . For example , we do not yet know the MC-intrinsic properties that temporally regulate F-actin foci formation . We find that ∼40% of TCR MCs are associated with F-actin foci at steady-state . While a large fraction of remaining clusters are in the central cSMAC region of the cell , previously characterized to be an actin depleted zone containing signaling incompetent TCR ( Kaizuka et al . , 2007; Vardhana et al . , 2010 ) , a smaller fraction of foci-unassociated MC resides in the lamellar zone as well . At lamellar zone TCR MCs , it is unclear which MC-intrinsic attributes dictate foci induction , maintenance and subsequent termination of actin nucleation on entry into the cSMAC , and if integrin signaling assists in any of these steps . Further investigation will not only provide a deeper understanding of subsynaptic F-actin dynamics , but will also illuminate other specialized roles that F-actin subpopulations play during the lifetime of a signalosome . The results will also provide deeper insights into signaling compartmentalization of the T cell synapse that is regulated by the actin cytoskeleton and its alteration in WAS pathology . AND Mouse T cells used in this study were isolated from 2–3 months old AND B10 . Br mouse spleen and lymph nodes . For experiments involving activation with bilayer containing MHC-peptide , AND CD4 T cell blasts were used . Briefly , total splenocytes and lymph node cells were expanded in vitro , in the presence of 2 μM MCC peptide and 10 units/ml IL-2 for 6 days in Kanagawa's modified Dulbecco's modified Eagle Medium ( KDMEM ) supplemented with FBS , and were subsequently used for experiments . In other experiments freshly isolated CD4 T cells were used , as indicated in figure legends . For experiments with mouse CD4 T cells , spleens and lymph nodes were harvested , and CD4 T cells were purified using mouse CD4 isolation kit ( Miltenyi Biotec , Auburn , CA ) . These CD4 T cells were used immediately after isolation for the experiments . For WASP single knockout experiments , Was−/− C57BL/6J mice ( stock number 019458 ) were purchased from Jackson Laboratory ( Bar Harbor , Maine ) , and age-matched WT C57BL/6J mice were also purchased as controls ( Stock number 000664 ) . For WASP and HS1 single knockout , and Was−/− Hcls1−/− double knockout studies , C57BL/6J , Hcls1−/− , Was−/− and Hcls1−/− Was−/− mice ( Dehring et al . , 2011 ) were used , and CD4 T cells purified as described above . For experiments with WASP−/− , NWASP−/− and WASP−/− N-WASP−/− CD4 T cells , cells were purified from spleen and lymph nodes of WT 129 or knockout 129 mice ( obtained from Snapper laboratory [Cotta-de-Almeida et al . , 2007] ) , as described above . For experiments with human CD4 T cells , cells were purified from fresh peripheral blood samples ( New York Blood Center , NY ) using Rosette Sep kit ( Stemcell Technologies , Vancouver , BC , Canada ) according to the manufacturer's protocol . Human CD4 T cells were rested overnight in RPMI supplemented with FBS , prior to their use in experiments . N-WASP ( H-100 ) and WASP ( H-250 ) antibodies were purchased from Santa Cruz Biotechnology ( Dallas , TX ) . HS1 antibody ( D5A9 ) , Phospho-Y397 HS-1 antibody ( D12C1 ) , phospho-Y319 Zap70/Y352 Syk ( affinity purified antisera #2704 ) , PLCγ1 ( D9H10 ) , phospho-Y783 PLCγ1 ( #2821 ) , NFAT1 ( D43B1 ) , phospho-Y416 SFK ( #6943 ) , phospho-Y171 LAT ( #3581 ) and WAVE2 ( D2C8 ) antibodies were obtained from Cell Signaling Technology ( Beverly , MA ) . Anti-Arp3 monoclonal antibody ( FMS338 ) and anti-MyH9 ( Myosin IIA , #SAB2101542 ) antibody were purchased from Sigma Chemicals ( St . Louis , MO ) . Rabbit p34-Arc ( anti-Arpc2 ) antibody was purchased from EMD Millipore and phospho-Y145 SLP76 ( #NBP140511 ) was purchased from Novus Biologicals . Celltracker-CM-DiI was purchased from Life Technologies . Anti-rabbit secondary antibodies conjugated with Dylight647 and Dylight564 dyes or anti-mouse Fc-specific Alexa568 secondary antibodies were procured from Jackson Immunoresearch ( West grove , PA ) and tested for lack of cross-reactivity prior to their use in assays . For anti-TCR antibodies-mediated T cell activation on bilayers , 145-2C11 ( #16-0031 , for mouse cells ) and OKT3 ( #16-0037 for human cells ) clones were purchased ( eBioscience , San Diego , CA ) , mono-biotinylated and labeled with fluorophores , and tested for mono-biotinylation prior to use in experiments . For experiments involving assessment of PLCγ1 activation using western blots , T cells were activated with anti-CD3/CD28 coated Dynabeads ( Life Technology , Grand Island , NY , #11452D ) according to manufacturer's instructions . A plasmid encoding Lifeact-GFP was a kind gift of B Geiger ( Weizmann Institute , Rehovot , Israel ) , and GFP-tagged WASP constructs were a kind gift of Dianne Cox ( Albert Einstein School of Medicine , New York ) . Blebbistatin , Wiskostatin and PP2 were obtained from Sigma chemicals; CK666 and CK689 were obtained from Calbiochem ( EMD Millipore , Billerica , MA ) . Alexa488-phalloidin and Alexa568-phalloidin were purchased from Invitrogen ( Life Technologies ) . Other fluorescent probes were obtained from Invitrogen and were conjugated with desired proteins according to the manufacturer's protocols . I-Ek-6His and mouse ICAM-1-12His were produced in the S2 insect cell line . I-Ek-6His was either loaded with moth cytochrome C peptide 91–102 or in some cases was the same MCC peptide was covalently coupled to the β chain . The density I-Ek-6His and ICAM1-12His was 20–50 molecules/μm2 and 200 molecules/μm2 , as described previously ( Grakoui et al . , 1999; Choudhuri et al . , 2014 ) , unless otherwise noted . For siRNA experiments , siRNA were obtained from Dharmacon ( Thermo Fisher Life Sciences , Pittsburgh PA ) . Mouse WASP siRNA ( on-target plus , Cat no-LU-046528-01-0005 ) , Mouse HCSL-1 ( HS1 , on-target plus , Cat no-LU046134-01-0002 ) were purchased as a set of four . A mixture of these four siRNAs was used at a concentration of ∼600 nM for 5 million cells , for WASP , HS1 and non-specific ( control ) siRNA ( D-001810-10-20 ) cases . Cells were electroporated using Amaxa Nucleofector ( Lonza , Basel , Switzerland ) following the instructions from the manufacturer , and grown in K-DMEM for 72–80 hr before experiments . For lentiviral transduction experiments , GIPZ lentiviral expression system ( Thermo Scientific , Pittsburgh PA ) was used . Briefly , the HEK cells were transfected with GIPZ-WASP shRNA plasmids ( RHS4287-EG7454 , a mixture if three sequences-5′ TTCTTATCAGCTGGGCTAG 3′ ( V2LHS_92677 ) , 5′ TTGCTGATCTTCTTCTTCC 3′ ( V2LHS_92678 ) , and 5′ TCATCTTCATCGCCAGCCT 3′ ( V3LHS_352271 ) ) , or non-specific shRNA plasmid ( RHS4346 ) along with the trans-lentiviral packaging mix ( RHS4287-EG7454 ) , using the calcium phosphate method . The culture supernatant containing viral particles was isolated after 68 hr of transfection , and utilized to transduce CD4 T cells according to manufacturer's instructions . The transfected T cells were identified by GFP expression in both control and WASP shRNA transduction cases . WASP depletion was assessed using western blots . Approximately 2 million cells from the control and siRNA samples were harvested , lysed using RIPA buffer supplemented with protease inhibitor , and the resulting lysates were resolved in reducing conditions on a polyacrylamide gradient gel ( Biorad , Hercules , CA ) . After transfer to the nitrocellulose membrane , protein levels were assessed by staining with primary ( overnight at 4°C ) and IR-dye conjugated secondary antibodies , and subsequently imaged using an Odyssey infrared imager ( LI-COR Biosciences Lincoln , NB ) . CD4 T cells were incubated with bilayer reconstituted with MHCp ( I-Ek-MCC ) and ICAM1 , or 2C11 and ICAM1 ( AND mouse CD4 T cells and polyclonal C57BL/6J T cells ) , or with bilayer reconstituted with OKT3 and ICAM1 ( human polyclonal CD4 T cells ) for the indicated time duration in Flow chambers ( Bioptech Inc . , Butler , PA ) or Ibidi microscopy chambers ( Ibidi , Verona , WI ) . Bilayer deposition and ligand density determination was performed as described previously ( Grakoui et al . , 1999; Ilani et al . , 2009; Choudhuri et al . , 2014 ) . For immunostaining , cells were fixed using PHEMO buffer ( 10 mM EGTA , 2 mM MgCI2 , 60 mM Pipes , 25 mM HEPES , pH 6 . 9 ) supplemented with 3 . 7% paraformaldehyde at room temperature for 20 min and permeabilized using 0 . 1% Triton X-100 for 2 min . Cells were then blocked using 5% casein for 30 min , washed and incubated with primary antibody overnight at 4°C and appropriate secondary antibodies for 1 hr , followed by washing and visualization using the indicated microscopy method . For all of the colocalization experiments , mouse AND T cell blasts activated on MHCp bilayers were utilized . For WASP knockout T cell experiments , freshly isolated T cells were utilized . For all live imaging experiments , peripheral human CD4 T cells were utilized , unless otherwise noted . AND CD4 T cell blasts were loaded for 30 min with 2 µM of Fluo4 ( Life Technologies ) in HEPES buffered saline ( HBS ) which was supplemented with 1% HAS and pluronic acid . Cells were then treated with DMSO or 10 µM CK666 for 1 min , and then incubated in glass chambers ( Nunc Lab-Tek , Thermo Scientific , Rochester , NY ) containing 2C11 and ICAM1 . Cells were imaged for the indicated time , using a confocal microscope with fully open pinhole . The Fluo4 fluorescence obtained for each cell was normalized to baseline fluorescence , and mean values from individual cells were then plotted on the graph ±standard error of the mean ( SEM ) . Patterns of OKT3 antibody were created by micro-contact printing onto glass coverslips . First , topological masters were fabricated by electron beam using a 1 µm PMMA layer , and spun-coated onto silicon wafers . Then , a combination of a thin layer of hard PDMS and a thick layer of Sylgard184 was spun-coated onto these wafers , resulting in hydrophobic stamps for patterning . Stamps were incubated with 1:1 ratio of Alexa647-OKT3 and unlabeled isotype antibodies , maintaining a total concentration of 25 µg/ml . Following coating , stamps were rinsed with PBS , PBS + 0 . 05% Tween-20 , and deionized ( DI ) water , dried with N2 gas , and then placed in contact with plasma-cleaned coverslips for 2 min . Coverslips were washed and incubated with 1 µg/ml ICAM1 for 40 min , after which they were washed again and then used for the experiment . Barbed end labeling of F-actin was carried out as previously described ( Furman et al . , 2007 ) . Briefly , T cells were incubated with glass coverslip coated with ICAM1 and anti-CD3 for 5 min , and then incubated with 0 . 0125% saponin dissolved in Rinse buffer ( 20 mM HEPES , 138 mM KCl , 4 mM MgCl2 , 3 mM EGTA , pH 7 . 5 ) along with freshly dialyzed Rhodamine labeled G-actin ( 0 . 5 µM in chilled G-actin buffer-4 mM TRIS , 0 . 2 mM CaCl2 , 0 . 2 mM ATP , 1 . 0 mM DTT ) for 1 min at 37°C . They were then rapidly fixed using 0 . 5% glutaraldehyde dissolved in cytoskeletal buffer ( 10 mM MES , 3 mM MgCl2 , 138 mM KCl , 2 mM EGTA ) for 10 min . Following this , cells were washed and permeabilized using 0 . 5% X-100 dissolved in cytoskeletal buffer for 5 min , and then incubated with sodium borohydride ( 100 mM made fresh in PBS ) for 10 min . Subsequently , cells were processed for total F-actin staining , using Alexa488-phalloidin , mounted using Fluoromount , and visualized using TIRF microscopy . As described previously ( Sage et al . , 2012 ) , human dermal microvascular endothelial cells ( HDMVECs ) were cultured on fibronectin ( 10 mg/ml ) in complete EBM-2 MV media ( Lonza , MA ) . ECs were transfected with palmitoylated-YFP DNA construct ( ‘membrane-YFP’ , from Life Technologies , CA ) using Amaxa electroporation according to manufacturer's instructions . Prior to the experiments , ECs were activated for 48 hr with 100 ng/ml IFNγ and 12 hr with 50 ng/ml TNFα . ECs were eventually pulsed with 1 mg/ml SEB and TSST for 60 min ( sAg , Toxin technologies , FL ) , incubated with human peripheral T cells ( sAg expanded T cells , cultured in presence of IL-15 ) for the relevant time duration , and imaged using either an upright laser scanning confocal microscope ( Zeiss ) or spinning disc confocal microscopy ( Andor technology CT ) . For TIRF microscopy , imaging was carried out using a Nikon Eclipse Ti microscope with a 100× 1 . 49 NA objective , and an AndorDU897 back illuminated EMCCD camera . For spinning disc confocal microscopy , an Olympus IX70 microscope with a 150× 1 . 45 NA objective , CSU ( Solamere Technology , Salt Lake City , UT ) and a Hamamatsu Orca ER camera was used . Solid-state lasers ( Coherent , CA ) provided illumination at 488 , 561 and 641 nm , and narrow pass filters ( Chroma Technology , Burlington , VT ) were used for detection . All image processing was carried out using ImageJ and Metamorph software . For colocalization analysis , raw images were subjected to local background subtraction as follows . Raw images were subjected to a local 50% intensity rank ( median ) filter ( 1 . 6 × 1 . 6 microns box size , see Figure 1—figure supplement 1 , ) across the whole image in order to generate a ‘background image’ . This background image was then subtracted from the raw image . The resultant high-pass filtered image was thresholded , and used to assess co-localization using similar methodology to assess pixel overlap across overlaid thresholded images as described before ( Kumari et al . , 2008 ) , via ‘co-localization class’ plugins in ImageJ . The intensity of overlapping pixels for each cell were normalized to total intensity of the same cell in the thresholded images , and were plotted as ‘fraction co-localized’ per cell ( Figure 2—figure supplement 1 ) . To assess F-actin foci alone , raw images were subjected to 50% rank filtering ( 1 . 6 × 1 . 6 microns box size , subsample ratio 1 ) and resultant image was subtracted from original image ( Figure 1—figure supplement 1 ) . All the samples were subjected to identical filter settings prior to analysis . Cell intensities from the subtracted image were then calculated by marking regions around the cells , and using ‘regional measurements’ features of MetaMorph Software . In most graphs ‘intensity’ or ‘levels’ on Y-axis refers to integrated intensities in the cell synapse , unless otherwise indicated . The images are displayed at identical contrast settings , and all the images presented here are raw images , unless otherwise mentioned . All the experiments were repeated at least thrice unless otherwise mentioned . For the graphs showing F-actin foci intensity , the datasets were normalized to the average value from ‘control’ data for that experiment , and plotted as ‘normalized intensity’ unless otherwise mentioned . For statistical analysis , the Mann–Whitney unpaired t-test was performed , unless otherwise mentioned . The bars and error bars in all graphs represent mean ± SEM .	The immune system is made up of several types of cells that protect the body against infection and disease . Immune cells such as T cells survey the body and when receptors on their surface encounter infected cells , the receptors activate the T cell by triggering a signaling pathway . The early stages of T cell receptor signaling lead to the formation of a cell–cell contact zone called the immunological synapse . Filaments of a protein called F-actin—which are continuously assembled and taken apart—make versatile networks and help the immunological synapse to form . F-actin filaments have crucial roles in the later stages of T cell receptor signaling as well , but how they contribute to this is not clear . Whether it is the same F-actin network that participates both in synapse formation and the late stages of T cell receptor signaling , and if so , then by what mechanism , remains unknown . The answers came from examining the function of a protein named Wiscott-Aldrich Syndrome Protein ( WASP ) , which forms an F-actin network at the synapse . Loss of WASP is known to result in the X-linked Wiscott-Aldrich Syndrome immunodeficiency and bleeding disorder in humans . Although T cells missing WASP can construct immunological synapses , and these synapses do have normal levels of F-actin and early T cell receptor signaling , they still fail to respond to infected cells properly . Kumari et al . analyzed the detailed structure and dynamics of actin filament networks at immunological synapses of normal and WASP-deficient T cells . Normally , cells had visible foci of newly polymerized F-actin directly above T cell receptor clusters in the immunological synapses , but these foci were not seen in the cells lacking WASP . Kumari et al . found that the F-actin foci facilitate the later stages of the signaling that activates the T cells; this signaling was lacking in WASP-deficient cells . Altogether , Kumari et al . show that WASP-generated F-actin foci at immunological synapses bridge the early and later stages of T cell receptor signaling , effectively generating an optimal immune response against infected cells . Further work will now be needed to understand whether there are other F-actin substructures that play specialized roles in T cell signaling , and if foci play a related role in other cell types known to be affected in Wiscott-Aldrich Syndrome immunodeficiency .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "immunology", "and", "inflammation" ]	2015	Actin foci facilitate activation of the phospholipase C-γ in primary T lymphocytes via the WASP pathway
Group II introns are mobile ribozymes that are rare in bacterial genomes , often cohabiting with various mobile elements , and seldom interrupting housekeeping genes . What accounts for this distribution has not been well understood . Here , we demonstrate that Ll . LtrB , the group II intron residing in a relaxase gene on a conjugative plasmid from Lactococcus lactis , inhibits its host gene expression and restrains the naturally cohabiting mobile element from conjugative horizontal transfer . We show that reduction in gene expression is mainly at the mRNA level , and results from the interaction between exon-binding sequences ( EBSs ) in the intron and intron-binding sequences ( IBSs ) in the mRNA . The spliced intron targets the relaxase mRNA and reopens ligated exons , causing major mRNA loss . Taken together , this study provides an explanation for the distribution and paucity of group II introns in bacteria , and suggests a potential force for those introns to evolve into spliceosomal introns . Introns interrupt genes in all domains of life . Group II introns , which are found in bacterial and organellar genomes , are ribozymes that self-splice from pre-mRNA transcripts independent of protein catalysis ( Lambowitz and Belfort , 2015; Lambowitz and Zimmerly , 2011; Zimmerly and Semper , 2015 ) . Splicing of group II introns occurs by the same pathway as that of spliceosomal introns , of which group II introns are the presumed ancestors ( Lambowitz and Belfort , 2015; Novikova and Belfort , 2017; Zimmerly and Semper , 2015 ) . Group II introns are also mobile retroelements that transpose into DNA via an RNA intermediate ( Lambowitz and Zimmerly , 2011 ) . Both splicing and retromobility of group II introns in vivo require an intron-encoded protein ( IEP ) that is in complex with the intron ( Matsuura et al . , 2001; Matsuura et al . , 1997; Qu et al . , 2016; Wank et al . , 1999; Zimmerly et al . , 1995 ) . Group II introns have a highly conserved RNA structure consisting of six domains ( Qu et al . , 2016; Robart et al . , 2014; Toor et al . , 2008 ) . The largest domain , DI , contains the exon-binding sequences ( EBS ) , which interact through base pairing with the intron-binding sequences ( IBS ) in the RNA exons or DNA homing target , to define the sites for splicing and retromobility of the intron , respectively ( Lambowitz and Zimmerly , 2011 ) . DV is structurally the most conserved domain , which contains the ‘catalytic triad’ , nucleotide residues critical for catalysis ( Qu et al . , 2016; Robart et al . , 2014; Toor et al . , 2008 ) . DVI has a bulged adenosine , which is the nucleophile that initiates the splicing reaction , yielding the branch-point of an intron lariat resulting from splicing . DIV often accommodates an open reading frame ( ORF ) encoding the IEP , which is a reverse transcriptase ( RT ) with maturase activity that facilitates intron splicing ( Matsuura et al . , 2001; Matsuura et al . , 1997; Wank et al . , 1999; Zimmerly et al . , 1995 ) , and can promote retromobility of the intron in combination with the DNA endonuclease activity of the IEP ( Matsuura et al . , 1997; Qu et al . , 2016; Zimmerly et al . , 1995 ) . Distribution of group II introns is different in bacteria and eukaryotes . Only about one quarter of sequenced bacterial genomes contain group II introns and they are usually present in small numbers within one genome . Moreover , they are mostly located in intergenic regions or non-conserved genes that are rarely essential . In addition , bacterial group II introns frequently reside in various types of mobile DNAs , which include pathogenicity islands and virulence plasmids ( Candales et al . , 2012; Dai et al . , 2003; Toro and Martínez-Abarca , 2013 ) . This distribution suggests that bacterial group II introns might be deleterious to host gene expression ( Dai and Zimmerly , 2002a; Zimmerly and Semper , 2015 ) . In contrast , organellar introns are distributed more densely in mitochondrial and chloroplast genomes , and are almost exclusively in essential housekeeping genes . In addition , group II introns are absent in eukaryotic nuclear genomes although they are ancestrally closely related to spliceosomal introns ( Lambowitz and Belfort , 2015; Novikova and Belfort , 2017; Zimmerly and Semper , 2015 ) . Factors , including nucleus-cytosol compartmentalization , intracellular magnesium concentrations , and interactions between the intron and spliced mRNA , have been conjectured to be the evolutionary drivers for the evolution of group II introns into spliceosomal introns in eukaryotes ( Martin and Koonin , 2006; Qu et al . , 2014; Truong et al . , 2015 ) . Here we investigated the impact of group II introns on their host gene expression in bacteria , using Ll . LtrB , the group II intron that resides on a conjugative plasmid pRS01 and interrupts a relaxase host gene ltrB , in the bacterium Lactococcus lactis ( Mills et al . , 1996 ) . We demonstrate that this intron inhibits relaxase host gene expression and mobilization of the cohabiting conjugative element . This group II intron-promoted inhibition results from RNA-RNA interactions between the spliced intron and mRNA , reducing mRNA levels . These discoveries can explain the distribution of group II introns in bacteria and provide yet another example of strict control and silencing of conjugation ( Singh and Meijer , 2014; Zatyka and Thomas , 1998 ) . These data also suggest a driving force for the evolutionary transition of the group II introns to spliceosomal introns . The group IIA Ll . LtrB intron interrupts the ltrB relaxase gene , which resides on the conjugative plasmid pRS01 and is required for pRS01 conjugative transmission ( Belhocine et al . , 2004; Belhocine et al . , 2005; Mills et al . , 1996 ) . Both intron-containing ( Int+ ) and intron-less ( Int- ) ltrB genes , under control of a nisin-inducible promoter on a pCY20 plasmid , were expressed in the native host L . lactis strain IL1403 ( Figure 1A ) . Northern blotting analysis of RNA ( Figure 1B ) indicated that the relaxase mRNA abundance was much lower in the presence of the intron than in its absence ( 21% vs . 100% ) . Consistent with the mRNA difference , the LtrB relaxase determined by Western blotting also accumulated to lower levels in the Int+ cell ( Figure 1C , 36% vs . 100% ) . Notably , a previous study showed an even larger ( 20-fold ) difference in protein levels in Int- versus Int+ cells ( Chen et al . , 2005 ) . The dramatic decrease in the ltrB relaxase mRNA level in the presence of the intron was confirmed by performing quantitative reverse transcription PCR ( qRT-PCR ) . The difference in mRNA levels between the Int+ and the Int- cells was 23-fold ( Figure 1D , Figure 1—source data 1 , Figure 1—figure supplement 1 , 4 . 3% vs . 100% ) . The greater difference by qRT-PCR and Northern blotting , may be accounted for by the different reference RNAs used ( copA mRNA for qRT-PCR vs . 16S rRNA for Northern blot ) or by the inherent difference between the two techniques in determining RNA levels . By performing both reverse transcription and qRT-PCR analyses we determined that the decrease in mRNA was not simply due to group II intron-promoted reduction in transcription rate ( Figure 1—figure supplement 1 ) . We analyzed nascent ltrB primary transcripts in Int-/Int+ cells , utilizing a DNA primer that bound 50 nucleotides downstream from the 5′-end of the transcript ( Figure 1—figure supplement 1A–B ) . Because of the proximity of the primer to the 5′-end , we presumed that the yield of the cDNA or RT-PCR products generated was a reflection of transcription initiation , and indeed there was only a 20% difference in the amount of cDNAs between Int- and Int+ ( Figure 1—figure supplements 1A , 100% vs . 79%; 1B , 100% vs . 74% ) . In addition , comparison of the relative mRNA expression level in the Int- cell with that of total intron RNAs ( pre-mRNA + Intron ) in the Int+ cell , also indicated small transcription differences between Int-/Int+ cells ( Figure 1—figure supplement 1C , primers , 1D , relative levels , 20 . 04 ± 2 . 00 vs . 25 . 65 ± 6 . 47 ) . In regard to splicing , we used Northern blotting , primer extension and qRT-PCR to determine the relative abundance of the spliced intron . The data showed that the sum of intron-containing RNAs ( pre-mRNA + Intron ) was relatively abundant , in contrast to the spliced mRNA ( Figure 1B , Figure 1—figure supplement 1D ) . Notably , a previous analysis of Ll . LtrB group II intron expression in the natural pRS01 context also showed that the ltrB mRNA level was much lower than that of the total intron RNA ( Chen et al . , 2005 ) . When the intron splicing efficiency was defined as the ratio of spliced intron against the sum of the intron-containing RNAs ( pre-mRNA + Intron ) , the numbers ranged from 63% to 91% ( Figure 1B , 91%; Figure 1—figure supplement 1D , 73 . 8%; Figure 1—figure supplement 2B , 63% ) . Thus , it does not appear that a splicing deficiency alone resulted in the mRNA reduction . Furthermore , we compared the decay rate of mRNAs expressed in Int-/Int+ cells . After a 30 min induction of ltrB gene expression , the cells were treated with rifampicin to stop transcription , and mRNA level was monitored over time and analyzed by Northern blotting ( Figure 1—figure supplement 3A–B ) . Surprisingly , the decay rate of spliced mRNA for the Int+ cells was ¼ to ½ of the rate for the mRNA of the Int- cells . Because the rates apply to a small residual fraction of the mRNA , these results must be interpreted with caution . Nevertheless , they suggest that the reduction in mRNA is not due to increased degradation . We also investigated if gene expression was affected at the translation level . Comparison of mRNA polysome profiling showed that there was only a marginal difference in mRNAs enriched in the polysomal fractions between Int-/Int+ cells ( Figure 1—figure supplement 4 , 100% vs . 85% ) , thereby rendering unlikely the possibility that the spliced mRNA experienced translational repression . Additionally , although ltrB RNA levels can be condition-dependent ( Chen et al . , 2005 ) , the differential expression levels between Int+ and Int- genes were not influenced by stressors and environmental changes , such as temperature , pH , salt and redox ( data not shown ) . Taken together , we concluded that the group II intron-promoted inhibition in gene expression is likely to be mainly at the mRNA level . We also investigated if this substantial group II-intron mediated decrease in mRNA was independent of plasmids and promoters from which the ltrB relaxase gene was expressed ( Figure 2A ) . When the Int-/Int+ ltrB genes were expressed from different plasmids , including pDL278 and pAMJ328 , where gene expression was driven by either constitutive or pH-inducible promoters , respectively , we observed similar levels of mRNA reduction in the presence of the intron ( pDL: 100% vs . 28%; pAMJ: 100% vs . 45% ) ( Figure 2A ) . This demonstrates that group II intron-promoted inhibition was independent of the plasmid and promoter from which the ltrB relaxase gene was expressed . In addition , we extended this comparative study to other types of bacterial group II introns , including a typical IIB intron , EcI5 ( Dai and Zimmerly , 2002b ) and a IIC intron , BhI1 ( Candales et al . , 2012 ) . The Int-/Int+ constructs of these introns were fused to an enhanced GFP reporter at the N-terminus . When expressed in Escherichia coli , levels of the mRNA dropped substantially for both EcI5 and BhI1 in the presence of the intron ( Figure 2B , C ) . Also the GFP reporter was greatly reduced for EcI5 with the intron present ( Figure 2B bottom ) . Therefore , it appears that inhibiting gene expression in bacteria is a common property of group II introns . Because ltrB relaxase is required for initiation of plasmid conjugation ( Belhocine et al . , 2004; Belhocine et al . , 2005; Mills et al . , 1996 ) , and because ltrB gene expression is reduced in the presence of the intron , we wished to determine if the intron may inhibit conjugal transfer of pRS01 . To test this hypothesis , we measured conjugation frequency where the donor strain , IL1403 , co-hosted erythromycin resistant ( ermR ) conjugative plasmid pRS01 ltrB- ( ΔLtrB::tet ) and the pCY20 Int- or Int+ plasmid . After mating with fusidic acid-resistant ( FAR ) ILI403 the frequency of the ErmRFAR exconjugants was measured ( Figure 3 ) . Consistent with the LtrB relaxase protein levels ( Figure 1C ) , the conjugation frequency of the Int+ donor was 3- to 4-fold lower that of the Int- , at 2 . 0 ± 0 . 1 × 10−4 versus 7 . 3 ± 10−4 exconjugants per donor ( n = 2 , p<0 . 05 ) ( Figure 3 and data not shown ) . This result ties into a previous study ( Novikova et al . , 2014 ) suggesting that these two disparate mobile elements , the mobile intron and the conjugative pRS01 plasmid , have a functional interplay . The first crystal structure of a group II intron lariat has revealed that ligated exons ( mRNA ) remain bound to the spliced intron RNA ( Robart et al . , 2014 ) . The recently determined cryo-electron microscopy ( cryo-EM ) structure of the Ll . LtrB group II intron-IEP complex , that was isolated from its native host , also shows that after splicing the mRNA is retained within the ribonucleoprotein ( RNP ) ( Qu et al . , 2016 ) . To confirm the presence of this intron-mRNA interaction in the bacterial cell , we performed intron-RNA pull-down assays ( Qu et al . , 2014 ) . Briefly , a streptavidin-specific RNA aptamer was introduced into the Ll . LtrB intron in order to purify it over streptavidin resin ( Figure 4A ) . After purification , the intron and any associated RNAs , were isolated and analyzed by reverse transcription . With the aptamer-containing intron , we observed that while the full-length pre-mRNA and spliced intron were pulled down as expected , the small mRNA ( smRNA ) , generated from splicing of the aptamer-containing pre-mRNA with small exons ( S-smRNA ) , was co-isolated . However , the smRNA produced either from the Int- cell ( C-smRNA ) or from the aptamer-free Int+ cell ( S-smRNA ) were not co-isolated ( Figure 4B ) . These results demonstrate that the intron is interacting with the spliced mRNA in the bacterial cell . Next , we investigated the possibility that the intron-mRNA interaction might account for the group II intron-promoted mRNA reduction . Similar to a previous study , we developed an in trans , two-plasmid expression system ( Qu et al . , 2014 ) . We therefore co-expressed the intron-less ltrB relaxase from the pCY20 plasmid with either the Ll . LtrB intron-containing ( with flanking small exons ) , or intron-less ( small exons only ) second plasmid , pLNRK ( Figure 5A , left ) . Northern blotting analysis again showed that the level of relaxase mRNA dramatically dropped in the presence of the intron ( 21% vs . 100% ) . Additionally , an unexpected band smaller than the mRNA was detected ( we call this band RNA 3 , see below ) ( Figure 5A , right ) . Also , Western blotting indicated less accumulation of LtrB relaxase protein ( 24% vs . 100% ) . It was necessary to elucidate if mRNA loss in this case is due to retrohoming or RNA-RNA interaction alone . To this end , mutants of the intron’s IEP , LtrA , in which retrohoming was stopped due to inactivation of either the reverse transcriptase ( RT ) or DNA endonuclease ( EN ) , were created ( Figure 5—figure supplement 1A ) . RNA analysis showed that strong mRNA inhibition still persisted in these mutants ( Figure 5—figure supplement 1B ) , thus indicating that the inhibition of gene expression was due to the proposed RNA-RNA interaction rather than retrohoming . Using the trans-expression system , we validated that the intron-mRNA interaction requires EBS-IBS base-pairing , as shown previously in yeast ( Qu et al . , 2014 ) . Mutants with nucleotide substitutions either in the IBS of the relaxase mRNA , or in the EBS in the intron expressed in trans , were compared to the respective wild-type counterparts ( Figure 5B , C ) . The results showed that the relaxase mRNA level was recovered in the EBS mutant strain ( Figure 5C , 21% to 92% ) , and increased ~2-fold in the IBS mutant ( Figure 5D , 24% to 45% ) . Similar results were obtained with a different set of IBS-EBS mutants ( m* ) , with recovery of mRNA levels in the EBS and IBS mutants from 23% to 116% and 23% to 51% respectively ( Figure 5—figure supplement 2 ) . Importantly , relaxase mRNA level was again reduced when the IBS mutation in the mRNA complemented the mutated EBS from the intron in the cell ( Figure 5—figure supplement 2 , down to 18% ) . The observation that splicing ability of the intron is required for mRNA inhibition ( data not shown ) raised the possibility that the intron interaction with mRNA enables biochemical reactivity . We were also curious about the unexpected bands ( RNAs 3 and 4 ) observed in Figure 5C and D respectively . Biochemical reactivity was validated and the origin of the extraneous bands was examined with the in trans system through a series of Northern blotting assays with multiple probes and confirmed by Rapid Amplification of cDNA Ends ( RACE ) , which generates a cDNA copy for sequencing of the desired RNA . Four RNA molecules ( RNA 1-RNA 4 ) with identity clearly distinct from the RNAs that were predicted to be expressed , were revealed ( Figure 6A–C , Figure 6—figure supplement 1A–D , Figure 6—figure supplement 2 ) . Among the four RNA molecules , two of them appeared in the intron-specific probing ( Figure 6A; Figure 6—figure supplement 1A , lane 7 ) . The larger of the two , termed RNA 1 , was also revealed with both relaxase mRNA 5′- and 3′-exon probes ( Figure 6B–C; Figure 6—figure supplement 1C–D , lane 7 ) , and was identified as the intron-containing relaxase pre-mRNA . This was confirmed by probing a splicing-inactive catalytic triad mutant ( T ) , which produced only this pre-mRNA band , RNA 1 , and not RNAs 2–4 ( Figure 6—figure supplement 1A , C , D , lane 3 ) , and also by 5′-RACE followed by PCR amplification and sequencing of the cDNA ( Figure 6—figure supplement 2A ) . This result suggests that the intron expressed in trans , is reverse splicing into the relaxase mRNA ( Figure 6A ) . The smaller of the two intron-specific RNAs , termed RNA 2 , was also revealed with the probe for 3′-exon but was absent in the probing for the 5′-exon ( Figure 6B–C; Figure 6—figure supplement 1C–D , lane 7 ) . With selective Northern blotting , along with 5′-RACE followed by cDNA sequencing , this RNA was identified as the unspliced intron expressed in trans , with its small 3′-exon replaced by the full-length 3′-exon of the relaxase mRNA ( Figure 6A–B , Figure 6—figure supplement 1A , D , Figure 6—figure supplement 2A ) . This result suggests that there was shuffling of exons or RNA recombination , possibly between two reverse splicing reactions , that resulted in the formation of a chimeric precursor ( Figure 6B ) . The third RNA , RNA 3 , previously seen in Figure 5 , was observed in both the exon splice junction ( Figure 6—figure supplement 1B , lane 7 ) and the relaxase mRNA 3′-exon-specific probings ( Figure 6B; Figure 6—figure supplement 1D , lane 7 ) , but it was absent from the probing for 5′-exon of the relaxase ( Figure 6C; Figure 6—figure supplement 1C , lane 7 ) . This RNA was further defined by 5′-RACE followed by cDNA sequencing as a ligation product of the small 5′-exon expressed in trans and the 3′-exon of the ltrB relaxase mRNA ( Figure 6—figure supplement 2A ) . Identification of RNA 3 suggested that RNA 2 , the chimeric intron precursor resulting from reverse splicing and RNA recombination described above , could be subject to forward splicing in the cell ( Figure 6B ) . This result is in accord with splicing being favored over reverse splicing in vivo . Notably , RNA 3 appeared much more abundant than the relaxase mRNA ( Figure 6B; Figure 6—figure supplement 1B , D ) . Besides reverse splicing and its related RNA recombination , these experiments also suggested the occurrence of a spliced exons reopening ( SER ) reaction ( Jarrell et al . , 1988 ) ( Figure 6C ) . This was evidenced by identification of the fourth RNA product , RNA 4 , a free 5′-exon of the relaxase mRNA . RNA 4 was observed in the probing specific for the 5′-exon of the relaxase mRNA ( Figure 6C; Figure 6—figure supplement 1C , lane 7 ) but was absent with all other probes ( Figure 6A–B; Figure 6—figure supplement 1A , B , D , lane 7 ) . Additionally , its presence was validated by 3′-RACE analysis followed by PCR amplification and sequencing of the cDNA products ( Figure 6—figure supplement 2B ) . However , free 3′-exon of the relaxase mRNA , which should be produced simultaneously with the free 5′-exon from the SER reaction , was not detected either by Northern blotting or 3′-RACE ( Figure 6B; Figure 6—figure supplement 1D , lane 7; Figure 6—figure supplement 2B ) . This could be attributed to bacterial mRNAs usually having a triphosphate at the 5′ end that is more resistant to mRNA decay than monophosphate ( Celesnik et al . , 2007; Schoenberg , 2007 ) . Notably , this free 5′-exon RNA was also revealed when the intron was expressed in cis ( Figure 6C; Figure 6—figure supplement 1C , lane 2; Figure 6—figure supplement 2B ) . Thus , these results suggest the SER , followed by RNA decay of the opened 3′-exon , to be a mechanism resulting in mRNA loss in the cell . To further elucidate how these retargeting reactions are responsible for mRNA disappearance , we sought to examine mutants that were already created for this study . Initially we chose to identify the four RNA molecules in the intron_EBSm* and mRNA_IBSm* mutants ( Figure 5—figure supplement 2 ) . Using the in trans system , Northern blotting analyses showed that none of the four RNA products 1–4 , that appeared with the wild-type strain existed in the mutants ( Figure 7 , lanes 3 , 5 versus lane 2 ) . However , they reappeared when the EBS from the intron was mutated to complement the IBS within the relaxase mRNA in this mutant ( Figure 7 , lanes 6 ) . This result confirmed that all of the identified mRNA retargeting reactions rely on the intron-mRNA interaction that is based on EBS-IBS base pairing . Inspired by the recently determined group II intron RNP structure and by the distribution features of group II introns in bacterial genomes , we performed a comparative expression study of intron-containing and intron-less variants of otherwise identical genes , to investigate the impact of group II introns on host gene expression . We discovered that group II introns reduce gene expression by robustly decreasing the spliced mRNA level in bacteria ( Figure 1 ) . This group II intron-induced mRNA disappearance appears independent of plasmids and promoters from which the genes are expressed ( Figure 2 ) , and is independent of stresses and changes of environmental conditions ( data not shown ) . While this intron-induced reduction of mRNA levels was revealed mainly for the well-defined group IIA intron Ll . LtrB and its host ltrB relaxase gene in its native bacterial host Lactococcus lactis , it was also shown for two other types of bacterial group II introns , the group IIB EcI5 and group IIC BhI1 introns ( Figure 2 ) . Using the Ll . LtrB intron as the model , we also demonstrated that this mRNA reduction results from interaction between the intron with its spliced mRNA , via EBS-IBS base pairing ( Figures 4 , 5 and 7 ) . Since ltrB is part of an operon containing also ltrC , ltrD , ltrE and ltrF ( Chen et al . , 2005 ) , it will be interesting to determine how ltrB targeting by the group II intron affects levels of the entire transcript . Interestingly , group II introns can also cause gene silencing in eukaryotic cells ( Chalamcharla et al . , 2010; Zerbato et al . , 2013 ) . Here , EBS-IBS-based intron-mRNA interactions shut down nuclear gene expression ( Qu et al . , 2014 ) . Taken together , we conclude that group II introns are inherent and general inhibitors of gene expression . By using the in trans system , we revealed that through RNA-RNA interactions , the spliced intron can retarget the mRNA in the cell ( Figures 6–7 , Figure 6—figure supplement 1 and 2 ) . The intron’s ribozyme activities include reverse splicing into the spliced mRNA , and the SER reaction . Because forward splicing is usually favored in the cell , based on the high splicing efficiency shown in Figure 1B , Figure 1—figure supplements 1–2 , we speculate that reverse splicing accounts for a relatively small portion of mRNA loss . Additionally , a robust SER , which was also revealed in the in cis system ( Figure 6C ) , splits the 5′-exon from the mRNA , and the free 3′-exon appears to be degraded ( Figure 8A ) . Although group II introns commonly invade DNA sequences , targeting RNA in vivo has not been previously described . However , RNP invasion of DNA is a relatively unfavorable event ( Aizawa et al . , 2003 ) , which makes it tempting to speculate that the RNA SER and subsequent RNA degradation may favor DNA invasion and retromobility . The RNA decay study with rifampicin indicated slower mRNA degradation in the Int+ cell ( Figure 1—figure supplement 3 ) . However , this result may be misleading as the residual RNA that escapes intron targeting may be a more stable fraction . Alternatively , the intron may act as ribosomes do on bacterial mRNA , providing a ‘protective barrier’ by binding the mRNA that escapes degradative attack by the group II intron ( Deana and Belasco , 2005 ) . Additionally , enhanced stability may be a reflection of the reciprocal relationship between the stability and concentration of bacterial mRNAs ( Nouaille et al . , 2017 ) , such that lower mRNA levels in the Int+ cell , would lead to higher stability of the mRNA . Whichever mechanism underlies the phenomenon , the consequence is that a fraction of the mRNA is maintained for translation . The discovery that group II introns inhibit gene expression likely explains why they are located mostly outside of genes or in non-essential genes , and why they are maintained at low frequency in bacterial genomes ( Dai and Zimmerly , 2002a; Zimmerly and Semper , 2015 ) . Group II intron-promoted reduction of gene expression could be the major driver that selected for group II introns residing in intergenic or non-essential regions . It is therefore intriguing that there are rare cases of essential genes containing group II introns in bacterial genomes ( Candales et al . , 2012; Dai et al . , 2003; Dai and Zimmerly , 2002a; Zimmerly and Semper , 2015 ) . We speculate that , in these cases , either robust gene expression is not needed , that reduction in gene expression is modest or that the intron plays a regulatory role ( Belfort , 2017 ) . The effect of organellar group II introns that are commonly present in essential genes is largely unknown ( Dai and Zimmerly , 2002a; Zimmerly and Semper , 2015 ) . Perhaps many of these introns have lost the ability to target mRNA and cause its loss , or do so very inefficiently . Alternatively , a lower expression level might be required for proper functions of these host genes . Notably , favoring the latter notion , a recent study showed that the presence of self-splicing introns including group II introns in the mitochondrion of Saccharomyces cerevisiae was required for inefficient expression of host genes that was essential for maintaining proper mitochondrial function ( Rudan et al . , 2018 ) . Group II introns are also frequently found in mobile DNAs ( Candales et al . , 2012; Dai et al . , 2003; Dai and Zimmerly , 2002a; Zimmerly and Semper , 2015 ) . One of the examples is the Ll . LtrB group II intron that naturally resides in the conjugative plasmid pRS01 , interrupting the ltrB relaxase gene whose expression is required for the horizontal gene transfer ( HGT ) of the plasmid . A recent study indicated that pRS01 promotes retromobility of the intron by providing LtrB relaxase that stimulates both the frequency and diversity of retrotransposition ( RTP ) events with its off-target DNA nicking activity ( Novikova et al . , 2014 ) . Here our study showed that the Ll . LtrB group II intron reduces the conjugal transfer of pRS01 via inhibition of its host gene expression ( Figure 3 ) . Thus , these findings together suggest that the two mobile genetic elements ( MGEs ) functionally interact by exploiting ltrB relaxase gene expression . Whereas relaxase expression stimulates intron RTP and HGT of the conjugative element , inhibition of relaxase expression by the group II intron opposes the promotion of RTP and HGT ( Figure 8B ) , effects that are likely in equilibrium . In general , expression of conjugative transfer genes is tightly controlled to minimize the burden on the host ( Zatyka and Thomas , 1998 ) . Indeed conjugative plasmid gene expression is kept in a default ‘off’ state and is switched on only under conditions that are optimal for transfer of the conjugative element ( Singh and Meijer , 2014 ) . Silencing of the relaxase gene by its resident group II intron therefore represents a novel way in which a conjugative element is down-regulated . The demonstration that intron RNAs are not only able to transpose and cleave exogenous RNAs , but can also recombine RNAs from different sources , suggests that progenitors of group II introns could have processed RNAs in primitive genomes . These actions could have contributed to the generation of an evolved genome and a cell that has novel functions and evolutionary advantages . In regard to the ancestral relationship between group II and spliceosomal introns , expression of group II introns in nuclei causes gene silencing in eukaryotic cells ( Chalamcharla et al . , 2010; Zerbato et al . , 2013 ) . In this study , we demonstrated that group II introns also hinder gene expression in bacteria . Notably , a common mechanism underlying the group II intron-promoted inhibition of gene expression in bacteria and eukaryotes is based on interactions between intron and mRNA via EBS-IBS base pairing ( Qu et al . , 2014 ) . In contrast to group II introns , many spliceosomal introns are shown to enhance gene expression in plants and other organisms ( Le Hir et al . , 2003; Rose , 2008 ) . Interactions between the intron and mRNA no longer exist in the splicing of spliceosomal introns , where such pairings have been functionally substituted with interactions between U5 snRNA and mRNA ( Hang et al . , 2015; Qu et al . , 2016; Yan et al . , 2015 ) . We speculate that , through these changes , introns can promote mRNA levels and gene expression , or at very least not be inhibitory , and therefore provided a force that could have stimulated evolution of group II introns into spliceosomal introns in primitive nuclear genomes . Lactococcus lactis IL1403 was used for ltrB gene expression and RNA and protein analysis . IL1403 pRS01 ltrB- ( ΔLtrB::tet ) and IL1403 ( FAR ) were used as donor and recipient strains for conjugation , respectively . Cultures were grown in M17 media supplemented with 0 . 5% glucose ( GM17 ) in tightly-capped tubes or bottles at 30°C without shaking . For gene expression , cultures were grown to OD600 ~0 . 6 and nisin was added to a final concentration of 0 . 4 μg/ml , for 2–3 hr . Cultures were spun at 5000 x g and pellets were stored at −80°C . Where appropriate , the media contained spectinomycin ( Spec ) at 300 μg/ml , chloramphenicol ( Cam ) at 10 μg/ml , erythromycin ( Erm ) at 10 μg/ml , or fusidic acid ( FA ) at 25 μg/ml . Escherichia coli MC1061 ( DE3 ) was used for over-expression of EcI5 and BhI1 plasmids . Cultures were grown in LB media with 100 μg/ml ampicillin at 37°C with aeration . At OD600 ~0 . 3 , cultures were induced with 0 . 1 mM IPTG for 3 hr . Cultures were spun at 5000 x g and pellets were stored at −80°C . All plasmids created or used are listed in Supplementary file 1 , and DNA oligonucleotides used in this study are listed in Supplementary file 1 . Plasmids pCY20LtrB ( Int- ) and pCY20LtrB::Ll . LtrB ( Int+ ) were the vector pair used in the intron in cis system for comparative analysis of ltrB relaxase gene expression in the absence and presence of the Ll . LtrB intron . With the in trans system , pCY20LtrB was used for ltrB relaxase gene expression while plasmid pLNRK smEx::Ll . LtrB was used to express the intron . For an intron-less control in this system , pLNRK smEx::Ll . LtrB was replaced with plasmid pLNRK smEx , that contained only the small flanking exons . The ltrB IBS mutants , mRNA_IBSm and mRNA_IBSm* , and Ll . LtrB intron mutants including EBSm , EBSm* and Triad , and the IEP EN and RT mutants were created with the above pCY20 and pLNRK plasmids by site-directed mutagenesis ( SDM ) using QuickChange Lightning kit ( Agilent , Santa Clara , CA ) . More specifically , to create EBSm* , the IBS region of the small exons of EBSm were mutated to make the intron splicing competent . This was done with 2 rounds of SDM PCR , using the plasmid product from PCR1 as the template for PCR2 . All plasmids were confirmed by sequencing ( see oligonucleotides used in Supplementary file 1 ) . Construction of other plasmids used in this study are detailed below . For RNA analysis , 7 . 5 μg of total RNA was separated on a 1 . 2% agarose/formaldehyde gel and transferred to hybond XL membrane ( GE Healthcare , Pittsburgh , PA ) , which was then UV cross-linked and probed with 32P-labeled oligonucleotides using Rapid-hyb buffer ( GE Healthcare ) . For the Ll . LtrB intron , the membranes were hybridized for detection of mRNA exon splice junction ( IDT4685 ) , 5′-Exon ( IDT5374 ) , 3′-Exon ( IDT5012 ) , and Ll . LtrB intron ( IDT1073 ) . For the EcI5 intron , membranes were hybridized for the detection of mRNA ( IDT4972 ) and intron ( IDT4970 ) , and for BhI1 intron , membranes were hybridized for mRNA ( IDT4975 ) and intron ( IDT4973 ) . All membranes were probed for 16S rRNA ( IDT861 ) as a loading control . Images were exposed on a phosphor screen , scanned on a Typhoon Trio , and quantified using ImageQuant . For LtrB relaxase analysis , total cell lysate was separated on a 12% SDS-polyacrylamide gel and transferred to 0 . 2 μM Immuno-blot PVDF membrane ( Bio-Rad ) at 25V for 30 min . The membrane was blocked with 5% dry milk in TBS-T ( 20 mM Tris , 140 mM NaCl , 2% Tween ) , incubated with a 1/1500 dilution of primary anti-relaxase antibody for 1 hr , washed with TBS-T twice for 15 min , and incubated with a 1/10 , 000 dilution of secondary HRP-labeled anti-rabbit antibody ( Advansta , Menlo Park , CA ) for 1 hr . Chemiluminescent HRP substrate ( Advansta WesternBright ECL ) was used for detection . For lane normalization , total protein was visualized from a coomassie stained 12% SDS-polyacrylamide gel . All images were scanned using a Bio-Rad ChemiDoc MP . Relaxase bands and total protein were quantified using Bio-Rad Image Lab software . cDNA synthesis was performed in a 20 μl reaction with 10 ng of DNase treated ( Promega RQ1 DNase ) RNA template using iScript ( Bio-Rad , Hercules , CA ) , as per manufacturer’s protocol . Minus RT ( RT- ) controls were also performed . Total RNA quality and primer specificity were analyzed by gel electrophoresis prior to qPCR . qPCR reactions were done in a total volume of 10 μl , using iTaq Universal SYBR Green Supermix ( Bio-Rad ) and contained 2 μl of the cDNA reaction as template and 5 pmol of each primer . Reactions were run in technical triplicates and a no-template control was included in every run . Reactions were amplified with the following conditions: 95°C for 30 s , ( 95°C for 5 s , 60°C for 10 s ) x 40 cycles , melt curve 65°C to 95°C ( 0 . 5°C every 2 s ) , on a Bio-Rad CFX384 Touch Real-Time PCR Detection System . Three biological replicates were run for each sample . Amplification efficiencies ( E ) of all primers were calculated using a 10-fold dilution and standard curve . Efficiencies , Ct values and charts were obtained using Bio-Rad CFX Manager Software . Primers used , their percent amplification efficiencies and amplicon lengths are listed in Supplementary file 1 . Relative gene expression of each target ( tar ) was normalized to a reference gene ( ref ) , CopA , whose expression was demonstrated to be constant in a previous study ( Magnani et al . , 2008 ) , and calculated using the following equation , with subtraction of RT- background:[ ( Etar ) −Ct ( mean , RT+ ) − ( Etar ) −Ct ( mean , RT− ) ]/[ ( Eref ) −Ct ( mean , RT+ ) − ( Eref ) −Ct ( mean , RT− ) ] . To identify splicing products , primer extension was performed using SuperScript III Reverse Transcriptase ( Thermo Fisher Scientific , Waltham , MA ) , as per manufacturer’s protocol , with 4 μg of DNase-treated RNA and 0 . 4 pmol of 32P-labeled oligonucleotides . Oligonucleotide IDT4836 was used with the addition of ddTTP for detection of intron precursor and spliced intron . Products were separated on an 8% Urea-polyacrylamide sequencing gel . To measure 5′-end transcription levels of the mRNA and intron precursor , IDT4916 was used , and products were separated on a 10% Urea-polyacrylamide gel . To analyze RNAs that were pulled down with streptavidin resin , IDT5078 , IDT1073 and IDT5127 were used to probe the smRNAs , intron RNAs , and 6S non-coding RNA , respectively . To measure splicing efficiency of the intron_EBSm* mutant , oligonucleotide IDT1073 was used to detect the presence of intron precursor and spliced intron , and products were separated on a 10% Urea-polyacrylamide gel . Images were exposed on a phosphor screen , scanned on a Typhoon Trio , and quantified using ImageQuant . L . lactis IL1403 was grown to OD600 ~0 . 6 and nisin was added to a final concentration of 0 . 4 μg/ml for 0 . 5 hr . Rifampicin was then added to a final concentration of 0 . 2 mg/ml , and 10 ml of culture was removed for each aliquot and immediately spun down at 4°C . Total RNA was prepared and analyzed for Northern blots . To perform polysome profiling , 200 ml of L . lactis IL1403 Int+/Int- cells ( 1:20 dilution of saturated overnight culture ) were grown and induced for gene expression with nisin for 2–3 hr . 100 mg/ml chloramphenicol was added and the cultures were chilled on ice for ~30 min with intermittent whirling and then collected by centrifugation at 5 , 000 rpm for 10 min at 4°C . Cell pellets were resuspended in 500 μl of ice-cold lysis buffer ( 20 mM Tris-HCl , pH 8 . 0; 140 mM KCl; 40 mM MgCl2; 0 . 5 mM DTT; 100 μg/ml chloramphenicol; 1 mg/ml heparin; 20 mM EGTA; 1% Triton X-100 ) and washed twice with the same buffer . The cell pellets were resuspended again in 500 μl of lysis buffer and snap-frozen in liquid nitrogen . Then the cells were disrupted at 4°C in 15 ml falcon tubes with 500 μl of 0 . 1 mm ice-cold glass beads by rigorous vortexing ( 30 s for 20 times , with 1 min interval ) and then briefly spun down at 4 , 000 rpm for 5 min . Crude cell lysate was then gently mixed , transferred into 1 . 5 ml tubes on ice , and cleared by 14 , 000 rpm for 25 min at 4°C . 400 μl of the cleared lysates were then loaded onto prepared 10–50% sucrose gradients in lysis buffer without 0 . 5 mg/ml heparin and centrifuged at 36 , 000 rpm for 153 min at 4°C ( SW41 rotor ) . 30 fractions of ~400 μl each were collected for each gradient from top to bottom . RNAs were extracted from each fraction by using phenol/chloroform and analyzed by running gels and performing Northern blotting . To measure conjugation efficiencies , L . lactis IL1403 ( FAR ) was used as the recipient strain and was grown to an OD600 ~0 . 6 and then for an additional 3 hr . L . lactis IL1403 ΔLtrB::tet ( ermR ) was used as the donor strain , and contained either the Int- plasmid pCY20LtrB , or the Int+ plasmid pCY20LtrB::Ll . LtrB . Cultures were grown to OD600 ~0 . 6 and nisin was added to a final concentration of 0 . 4 μg/ml for 3 hr . Equal volume of donor and recipient were mixed , spotted on pre-incubated filters ( 0 . 45 μm pore size , Millipore ) on GM17 plates , and incubated ~18 hr at 30°C . The filters were removed from the plates , washed with 5 ml GM17 in a 50 ml conical tube by vortexing , and the wash was spotted or plated on GM17 plus 25 μg/ml fusidic acid and 10 μg/ml erythromycin for selection of transconjugants . Plates were incubated ~18 hr at 30°C . To calculate input donor , the donor strain was diluted to 10−6 and spotted or plated on GM17 plus 10 μg/ml erythromycin . Conjugation frequency was expressed as exconjugants per donor ( ErmRFAR/ErmR ) . Bacterial cells ( 100 ml ) were collected after 2–3 hr of nisin induction of RNA expression and disrupted in 800 μl of CB500 solution ( 20 mM Tris–HCl , pH 8 . 0 , 500 mM NaCl , 0 . 1 mM EDTA , 1 mM PMSF ) by vortexing ( 30 s , rest for 1 min , 24 cycles ) using 500 μl of 0 . 1 mm ice-cold glass beads ( Sigma ) . Lysates were cleared by centrifugation at 14 , 000 rpm for 25 min . To pull down RNAs , ~500 μl of cleared lysate was incubated with 100 μl of streptavidin resin ( Thermo Scientific ) that was equilibrated with CB500 . The resin was washed eight times with 1 ml CB500 buffer . RNAs bound to resin were eluted with 400 μl of 5 mM biotin for 1 hr and were extracted from the eluates . RNA identities were determined by using reverse transcription primer extensions . To further identify mRNA targeting products 1–4 , 5' and 3' RACE ( Rapid Amplification of cDNA Ends ) experiments were performed . Briefly , Int+ and Int- strains from cis and trans systems were grown and induced as previously described , and total RNA was prepared . RNAs were polyA-tailed using E . coli Poly ( A ) Polymerase ( New England BioLabs , Ipswich , MA ) . 5' and 3' RACE experiments were then done using Invitrogen 5' or 3' RACE kits ( Cat# 18374058 or 18373019 ) , or the Takara SMARTer 5'/3' kit ( Cat# 634859 ) , following the kit protocols . Reverse transcriptase was used to synthesize cDNAs , which were then used as template for PCR amplification using GSP ( gene specific primers IDT6074 and IDT6070 to ltrB 5' or 3' exon , respectively ) along with a kit universal primer annealing to the synthesized cDNA . 5' and 3' RACE PCR products were run out on a 1 . 2% agarose gel , and the bands were excised , gel purified , and sequenced directly ( Eton Bioscience , San Diego , CA ) or cloned into pGEM-T vector ( Promega ) , and then sequenced , to confirm their identity .	Proteins , the molecular workhorses of the cell , are encoded by genes within the DNA . A given gene is often formed of several portions of coding DNA , the exons , which are separated by sections of non-coding DNA called introns . When it is expressed , the entire gene , including the introns , is ‘copied’ into a molecule of pre-messenger RNA ( pre-mRNA ) . Enzymes then remove the introns and stitch the exons together to produce a mature messenger RNA ( mRNA ) that can serve as a template to create a protein . However , a particular type of intron , known as a group II intron , can remove or ‘splice’ itself out of a pre-mRNA without the help of additional enzymes . Once this is done , these introns can also insert themselves into DNA . Scientists think that group II introns originated in bacteria , yet only about a quarter of bacterial genomes sequenced so far contain this particular kind of intron . When these introns are present , their numbers are low , and most of the time they are restricted to genes that are not required for survival . It is not well understood why group II introns are so few and far between , and why they are often not associated with essential genes in bacteria . Qu et al . looked into a gene that contains a group II intron in the bacterium Lactococcus lactis , and discovered that the expression of this gene was dramatically low . This decrease was due to interactions between the group II intron and the exons within the mRNA . Once group II introns were spliced out , they could target their mRNAs and reopen the junction where the exons had been stitched together . These mRNAs were therefore lost , and the cell made less of the protein that the gene encoded . The findings by Qu et al . can help to explain why group II introns are so scarce , how they are excluded from essential genes , and also how cells could use these sequences . In particular , group II introns could have evolved to form the spliceosome , a complex structure that processes introns in higher organisms . Finally , group II introns could be used in the laboratory to artificially silence genes of interest .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology" ]	2018	Group II intron inhibits conjugative relaxase expression in bacteria by mRNA targeting
A defining feature of the brain is the ability of its synaptic contacts to adapt structurally and functionally in an experience-dependent manner . In the human cortex , however , direct experimental evidence for coordinated structural and functional synaptic adaptation is currently lacking . Here , we probed synaptic plasticity in human cortical slices using the vitamin A derivative all-trans retinoic acid ( atRA ) , a putative treatment for neuropsychiatric disorders such as Alzheimer’s disease . Our experiments demonstrated that the excitatory synapses of superficial ( layer 2/3 ) pyramidal neurons underwent coordinated structural and functional changes in the presence of atRA . These synaptic adaptations were accompanied by ultrastructural remodeling of the calcium-storing spine apparatus organelle and required mRNA translation . It was not observed in synaptopodin-deficient mice , which lack spine apparatus organelles . We conclude that atRA is a potent mediator of synaptic plasticity in the adult human cortex . The ability of neurons to express plasticity by responding to specific stimuli with structural and functional changes is critical for physiological brain function ( Citri and Malenka , 2008 ) . Over the past few decades , cellular and molecular mechanisms of synaptic plasticity have been extensively studied across various animal models ( Ho et al . , 2011 ) . However , direct experimental evidence for coordinated structural and functional synaptic changes in the adult human cortex is lacking ( Mansvelder et al . , 2019; Verhoog et al . , 2016 ) . It thus remains unclear whether the structural and functional properties of human cortical neurons adapt similarly to those in the rodent brain . Vitamin A ( all-trans retinol ) and its metabolites have recently been linked to physiological brain functions such as axonal sprouting , synaptic plasticity , and modulation of cortical activity ( Drager , 2006; Shearer et al . , 2012 ) . Specifically , all-trans retinoic acid ( atRA ) , which is used clinically in dermatology and oncology ( Dobrotkova et al . , 2018; Hu et al . , 2009 ) , has been studied for its neuroprotective and plasticity-promoting effects in animal models ( Chen et al . , 2014; Koryakina et al . , 2009 ) . Recent studies have evaluated the effects of atRA in patients with brain disorders associated with cognitive dysfunction , including Alzheimer’s disease , Fragile X syndrome , and depression ( Bremner et al . , 2012; Ding et al . , 2008; Zhang et al . , 2018 ) . For example , alterations in retinoic acid-mediated synaptic plasticity have been reported in neurons derived from inducible pluripotent stem cells ( iPSCs ) generated from Fragile X syndrome patients ( Zhang et al . , 2018 ) . However , direct experimental evidence for atRA-mediated effects on synaptic plasticity of principal neurons in the adult human cortex is lacking . Here , we used human cortical slices prepared from neurosurgical resections to assess atRA-mediated changes in the structural and functional synaptic properties of layer 2/3 pyramidal neurons . In this context , we also evaluated the role of the actin-modulating protein synaptopodin ( Mundel et al . , 1997 ) , an essential component of the spine apparatus organelle ( Deller et al . , 2003 ) , which is a key regulator of synaptic plasticity in the rodent brain ( Deller et al . , 2003; Segal et al . , 2010; Vlachos et al . , 2013 ) and has recently been linked to the cognitive trajectory in human aging ( Wingo et al . , 2019 ) . Cortical access tissue samples from eight individuals who underwent clinically indicated neurosurgical procedures , such as for tumors or epilepsy , were experimentally assessed in this study ( details provided in Supplementary file 1 ) . Acute cortical slices were treated for 6–10 hr with atRA ( 1 µM ) or vehicle-only , and superficial ( layer 2/3 ) pyramidal neurons were recorded in a whole-cell configuration ( Figure 1A , B ) . While no significant differences in active or passive membrane properties were detected between the two groups ( Figure 1C–E ) , a robust increase in the amplitudes of glutamate ( i . e . , α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid , AMPA ) receptor-mediated spontaneous excitatory postsynaptic currents ( sEPSCs ) was observed in the atRA-treated slices ( Figure 1F , G; see also Figure 1—figure supplement 1 ) . The mean sEPSC frequency was not significantly different between the two groups ( Figure 1G ) . These results demonstrate an atRA-mediated strengthening of excitatory neurotransmission onto human cortical pyramidal neurons . Since no major changes in active or passive membrane properties were detected in these initial experiments , we focused on the effects of atRA on excitatory synapses and dendritic spines . A positive correlation between excitatory synaptic strength , that is , sEPSC amplitude , and dendritic spine size has been demonstrated in various animal models ( Bosch and Hayashi , 2012; Matsuzaki et al . , 2004 ) . Hence , we wondered whether atRA also induces structural changes in dendritic spines from adult human cortical brain slices . To address this question , a set of recorded neurons was filled with biocytin and stained with Alexa Fluor-labeled streptavidin to visualize dendritic spine morphologies ( Figure 2A ) . No significant differences in spine density were observed between the two groups ( Figure 2B ) . However , marked increases in spine head sizes were evident in the atRA-treated group ( Figure 2C ) . These findings identify that excitatory synaptic strength , as indicated by sEPSC amplitudes , is positively correlated with dendritic spine size in human cortical pyramidal cells , thus revealing coordinated structural and functional changes of excitatory postsynaptic membranes in atRA-treated human cortical slices . Previous work revealed that the actin-modulating molecule synaptopodin ( Mundel et al . , 1997 ) is an essential component of the spine apparatus organelle ( Deller et al . , 2003 ) . This enigmatic cellular organelle is composed of stacked smooth endoplasmic reticulum and is found in subsets of dendritic spines ( Kulik et al . , 2019; Spacek , 1985 ) . Synaptopodin-deficient mice lack spine apparatus organelles and exhibit defects in synaptic plasticity and behavioral learning ( Deller et al . , 2003; Jedlicka et al . , 2009; Vlachos et al . , 2013 ) . In this context , we previously showed that synaptopodin promotes the accumulation of AMPA receptors at synaptic sites ( Vlachos et al . , 2009 ) . Given that atRA and synaptopodin have been linked to AMPA receptor-mediated synaptic plasticity ( Aoto et al . , 2008; Arendt et al . , 2015a; Maggio and Vlachos , 2018; Poon and Chen , 2008; Vlachos et al . , 2013; Vlachos et al . , 2009 ) , we asked whether atRA mediates its effects via synaptopodin and the spine apparatus organelle ( Figure 2D–J ) . Synaptopodin clusters were detected in a considerable number of dendritic spines of human superficial ( layer 2/3 ) pyramidal neurons: 74 ± 2% of all dendritic spines contained a synaptopodin cluster in a characteristic position , that is , in the base , neck , or head of the spine ( Figure 2D ) . Similar to previous reports ( Holbro et al . , 2009; Vlachos et al . , 2009; Yap et al . , 2020 ) , we also observed a positive correlation between synaptopodin cluster size and spine head size in the human cortex ( Figure 2E ) . Finally , synaptopodin-positive spines were significantly larger than their synaptopodin-negative neighbors ( Figure 2F ) . Hence , this study reveals that synaptopodin clusters are found in strategic positions in a subset of large dendritic spines of the adult human cortex . Systematic assessment of synaptopodin-positive and synaptopodin-negative spines revealed that atRA does not act specifically on synaptopodin-containing spines ( Figure 2F ) ; in fact , significant atRA-mediated increases in spine head size were observed in both synaptopodin-positive and -negative dendritic spines . Although we did not observe a significant difference in the number of synaptopodin-positive spines between the two groups ( control: 71 ± 4% in 14 dendritic segments; atRA: 77 ± 2% in 21 dendritic segments; p=0 . 25 , Mann–Whitney test ) , atRA caused a significant increase in the sizes of synaptopodin clusters ( Figure 2G; c . f . , Figure 2E ) . We conclude that remodeling of synaptopodin clusters accompanies atRA-mediated coordinated structural and functional changes in human dendritic spines . In an earlier study , we showed that remodeling of synaptopodin clusters reflects plasticity-related ultrastructural changes in spine apparatus organelles ( Vlachos et al . , 2013 ) . We therefore wondered whether atRA changes the ultrastructural properties of the spine apparatus organelle in human cortical neurons ( Figure 2H–J ) . After confirming that synaptopodin is a marker of the human spine apparatus organelle using pre-embedding immunogold labeling techniques ( Figure 2H ) , the ultrastructural properties of this organelle were assessed in cross-sectional transmission electron micrographs of 103 control and 114 atRA-treated asymmetric spine synapses from three independent samples ( Figure 2I , J ) . A marked increase in the cross-sectional area of the spine apparatus organelle was observed in the atRA-treated group , which is consistent with the atRA-induced increase in synaptopodin cluster size ( Figure 2J; positive correlations between spine apparatus cross-sections and spine cross-sections are shown in Figure 2—figure supplement 1 ) . Taken together , these results reveal a link between synaptopodin and the spine apparatus organelle in the human cortex . Specifically , these findings demonstrate that changes in the structural and functional properties of dendritic spines are accompanied by ultrastructural changes in spine apparatus organelles . We conclude that atRA is a potent mediator of coordinated ( ultra ) structural and functional synaptic changes in the adult human cortex . To further understand the relevance of synaptopodin and the spine apparatus organelle in atRA-mediated synaptic plasticity , we prepared acute medial prefrontal cortex ( mPFC ) slices from synaptopodin-deficient ( Synpo−/−; Deller et al . , 2003 ) and age-matched wild-type mice ( Synpo+/+ ) . Single-cell recordings of superficial ( layer 2/3 ) pyramidal neurons from Synpo+/+ slices showed that changes in excitatory neurotransmission are similar to what we observed in the human cortical slices: sEPSC amplitudes ( but not frequencies ) were significantly increased following atRA treatment ( Figure 3A , B ) . Notably , atRA also reduced the input resistance of these pyramidal cells ( Figure 3—figure supplement 1 ) . The resting membrane potentials and action potential ( AP ) frequencies were not affected by atRA , similar to what we observed in the human cortical slices ( Figure 3—figure supplement 1; c . f . , Figure 1 ) . We conclude that atRA exerts comparable effects on excitatory neurotransmission in both the adult mouse and human neocortex . In Synpo−/− preparations , no significant changes in sEPSC properties were observed following atRA treatment ( Figure 3C , D ) , thus demonstrating the relevance of synaptopodin in atRA-mediated synaptic plasticity . Additionally , a reduction in the input resistance was not observed in atRA-treated Synpo−/− preparations ( Figure 3—figure supplement 1 ) . Furthermore , the active and passive membrane properties of Synpo+/+ and Synpo−/− neurons were indistinguishable ( Figure 3—figure supplement 1 ) , despite significant increases in the baseline sEPSC properties of Synpo−/− preparations ( Figure 3—figure supplement 2 ) . To confirm and extend these findings , sEPSC recordings were carried out in acute slices prepared from Thy1-GFP/Synpo+/−× Synpo−/− mice . In this mouse model , the synaptopodin coding sequence is tagged with green fluorescent protein ( GFP ) and expressed under the control of the Thy1 . 2 promotor in the absence of endogenous synaptopodin ( Synpo−/− genetic background; Vlachos et al . , 2013 ) . In a previous study , we demonstrated that the transgenic expression of GFP/Synpo rescues the ability of Synpo−/− neurons to form spine apparatus organelles and to express synaptic plasticity ( Vlachos et al . , 2013 ) . Indeed , a significant increase in sEPSC amplitudes was observed in GFP/Synpo expressing pyramidal neurons following atRA treatment ( Figure 3E , F ) . Taken together , we conclude that the presence of synaptopodin is required for atRA-mediated synaptic strengthening . The biological effects of atRA occur at the levels of gene transcription and mRNA translation ( i . e . , protein synthesis; Drager , 2006; Poon and Chen , 2008 ) . Therefore , in a different set of acute cortical slices prepared from wild-type mice , actinomycin D ( 5 µg/ml ) was used to block gene transcription in the presence of atRA ( 1 µM; 6–10 hr; Figure 4A , B ) . To control for possible toxic effects , the basic active and passive membrane properties of cortical slices treated with actinomycin D or vehicle-only were recorded ( Figure 4—figure supplement 1 ) . Other than a reduction in AP frequencies at high-current injections , we did not observe any significant differences between the two groups ( Figure 4—figure supplement 1 ) . Moreover , baseline sEPSC properties were indistinguishable between actinomycin D- and vehicle-only-treated slices ( Figure 4A ) . However , we observed a significant increase in the amplitudes of AMPA receptor-mediated sEPSCs in the atRA-treated group in both vehicle-only and actinomycin D co-incubated slices ( Figure 4A , B ) , thus further confirming that atRA mediates synaptic strengthening . Notably , the atRA-mediated reduction in input resistance was not observed in these experiments ( Figure 4—figure supplement 1 ) . We next used anisomycin ( 10 µM ) to block mRNA translation during atRA treatment ( Figure 4C , D ) . Anisomycin incubation had no major effects on passive membrane properties , while reductions in AP frequencies were observed at high-current injections ( Figure 4—figure supplement 1 ) . In contrast to our experiments with actinomycin D ( Figure 4A ) , however , no significant changes in sEPSC amplitudes were detected in the presence of both atRA and anisomycin ( Figure 4C , D ) . Together , these results indicate that the effects of atRA on excitatory neurotransmission do not require major transcriptional changes , but depend on mRNA translation and protein synthesis . Based on the results we obtained in cortical slices from the murine brain , we returned to our human brain slice model to evaluate whether mRNA translation is required for atRA-mediated structural and functional synaptic plasticity in the human cortex ( Figure 5 ) . We recorded AMPA receptor-mediated sEPSCs once again from superficial ( layer 2/3 ) pyramidal neurons of atRA- and vehicle-treated human cortical slices in the presence of the pharmacologic mRNA translation inhibitor anisomycin ( 10 µM; 6–10 hr ) . No significant differences in sEPSC amplitudes were observed between the two groups ( Figure 5A , B ) . Consistent with these findings , anisomycin also blocked the observed atRA-mediated increases in dendritic spine head and synaptopodin cluster sizes , whereas the sizes of synaptopodin-positive and synaptopodin-negative spine heads remain significantly different ( Figure 5C , D ) . These results demonstrate that atRA-mediated plasticity requires mRNA translation to trigger coordinated changes in synaptic strength , spine head size , and synaptopodin cluster properties in cortical slices prepared from the adult human brain . Vitamin A and its metabolites bind to nuclear and cytoplasmic receptors and regulate important biological processes , such as cell growth , cell survival , and differentiation ( Al Tanoury et al . , 2013 ) . The pleiotropic effects of these molecules account for both the therapeutic ( e . g . , for treating promyelocytic leukemia ) and teratogenic effects of atRA ( Hu et al . , 2009 ) . Studies of the role of atRA in the central nervous system have primarily focused on embryonic and early postnatal development ( Luo et al . , 2004; Rataj-Baniowska et al . , 2015 ) . However , a growing body of literature indicates that physiological retinoid metabolism and signaling also occur in the adult murine brain ( Shearer et al . , 2012 ) . Several animal studies have revealed that retinoid receptors bound to specific cytoplasmic mRNAs control synaptic plasticity by regulating the synthesis , trafficking , and accumulation of synaptic proteins ( Aoto et al . , 2008; Groth and Tsien , 2008; Maghsoodi et al . , 2008; Poon and Chen , 2008 ) . Some of these findings have recently been replicated in neurons derived from human iPSCs ( Zhang et al . , 2018 ) . The results of the present study demonstrate the ability of atRA to induce synaptic plasticity in the adult human cortex . Consistent with recent reports on the importance of protein synthesis in synaptic plasticity ( Biever et al . , 2020; Hafner et al . , 2019; Sutton and Schuman , 2006 ) , our experiments revealed that atRA-mediated structural and functional synaptic plasticity in adult human cortical slices requires mRNA translation . At the mechanistic level , we identified synaptopodin as a target and mediator of atRA-induced synaptic plasticity . Synaptopodin is an actin-modulating protein ( Mundel et al . , 1997 ) that has been linked to the spine apparatus organelle ( Deller et al . , 2003 ) , a membranous extension of the smooth endoplasmic reticulum found in a subpopulation of telencephalic dendritic spines ( Gray , 1959; Spacek , 1985; Spacek and Harris , 1997 ) . A critical role of synaptopodin in the formation of the spine apparatus was reported in Synpo−/− mice ( Deller et al . , 2003 ) ; the neurons of these mice do not form spine apparatus organelles , and Synpo−/− animals exhibit defects in synaptic plasticity and memory formation ( Deller et al . , 2003; Jedlicka et al . , 2009; Korkotian et al . , 2014; Maggio and Vlachos , 2018; Vlachos et al . , 2013; Vlachos et al . , 2009 ) . The results of the present study demonstrate that ( 1 ) approximately 70% of dendritic spines in the superficial layers of the human cortex contain synaptopodin clusters; ( 2 ) synaptopodin is a marker of the human spine apparatus organelle; ( 3 ) synaptopodin clusters and spine apparatus organelles are found in large dendritic spines; and ( 4 ) a plasticity-inducing stimulus , that is , application of atRA , promotes remodeling of synaptopodin clusters , spine apparatus organelles , and dendritic spines in cortical slices prepared from the adult human brain ( c . f . , Figure 2 ) . The involvement of synaptopodin and the spine apparatus organelle in synaptic plasticity is still enigmatic; however , a role in local protein synthesis and the regulation of intracellular calcium dynamics have been suggested ( Fifková et al . , 1983; Pierce et al . , 2000 ) . Synaptopodin binds to actin and α-actinin and has been suggested to orchestrate spine dynamics via Rho-A signaling and long-term spine stability ( Asanuma et al . , 2006; Yap et al . , 2020 ) . More recently , synaptopodin and myosin V were reported to be associated ( Konietzny et al . , 2019 ) . Thus , synaptopodin may be relevant for the parallel functional and structural changes observed at excitatory synapses undergoing glutamate receptor-mediated synaptic plasticity ( Jedlicka and Deller , 2017; Vlachos et al . , 2009 ) . However , the strategic positioning of synaptopodin clusters in the head , neck , and base of murine and human dendritic spines indicates that this molecule may act primarily within or in association with the spine compartment . Here , synaptopodin may link spine actin and myosin to spine apparatus organelles , which have a more restricted localization ( Konietzny et al . , 2019 ) . In support of this hypothesis , atRA application in our study triggered changes in spine head size in both synaptopodin-positive and synaptopodin-negative spines from human cortical slices , suggesting that the rapid/initial atRA-mediated spine enlargement does not require the actual presence of synaptopodin clusters , that is , spine apparatus organelles ( Okubo-Suzuki et al . , 2008 ) . Considering that synaptopodin clusters are highly dynamic structures that can change their position within spines , and may appear in existing spines ( Vlachos et al . , 2009; Yap et al . , 2020 ) , it is interesting to speculate that atRA may trigger a molecular cascade that promotes rapid spine growth , even in spines that do not contain a spine apparatus organelle , eventually leading to the formation and enlargement of spine apparatus organelles . However , more work is required to identify the precise downstream signaling pathways through which atRA mediates synaptopodin- and protein synthesis-dependent synaptic plasticity in the adult human cortex . Similarly , the role of synaptopodin in atRA-mediated changes in intrinsic cellular properties , for example , reductions in the input resistance of principal neurons in the mouse cortex , warrants further investigation , particularly in the context of recent work on the distinct electrical properties of human and murine cortical neurons ( Gidon et al . , 2020 ) . Both atRA and synaptopodin have been linked to the expression of Hebbian and homeostatic synaptic plasticity in rodent brain tissue and are associated with Ca2+-dependent AMPA receptor synthesis and trafficking ( Arendt et al . , 2015b; Vlachos et al . , 2013; Vlachos et al . , 2009 ) . In turn , alterations in retinoid signaling and synaptopodin expression have been associated with synaptic plasticity defects in pathological brain states ( Bremner et al . , 2012; Maggio and Vlachos , 2014 ) . Considering that increased sEPSC amplitudes and frequencies were observed in the cortical neurons of Synpo−/− mice , it would be interesting to evaluate whether alterations in retinoid metabolism are present in this mouse model . Accordingly , alterations in synaptopodin expression and retinoid signaling have been observed in the brain tissue of patients with Alzheimer’s disease and cognitive decline ( Goodman , 2006; Goodman and Pardee , 2003; Mingaud et al . , 2008; Misner et al . , 2001; Reddy et al . , 2005 ) . Given that retinoids have been proposed as a potential therapeutic avenue for Alzheimer’s disease-associated cognitive decline ( Ding et al . , 2008; Endres et al . , 2014 ) , it is conceivable that atRA may act – at least in part – by modulating synaptopodin expression , thereby improving the ability of adult human cortical neurons to express synaptic plasticity . We are confident that the translational approach used in this study investigating both murine and human cortical slices will facilitate the identification of key mechanisms of synaptic plasticity in the adult human brain . After resection , cortical access tissue was immediately transferred to an oxygenated extracellular solution containing ( in mM ) 92 NMDG , 2 . 5 KCl , 1 . 25 NaH2PO4 , 30 NaHCO3 , 20 HEPES , 25 glucose , 2 thiourea , 5 Na-ascorbate , 3 Na-pyruvate , 0 . 5 CaCl2 , and 10 MgSO4 , ( pH = 7 . 3–7 . 4 ) at approximately 10°C ( NMDG-aCSF; Gidon et al . , 2020; Ting et al . , 2018 ) . Prior to slicing , cortical tissue was embedded in low-melting-point agarose ( Sigma–Aldrich , #A9517; 1 . 8% [w/v] in phosphate-buffered saline [PBS] ) . Tissue sections ( 400 µm ) were cut with a Leica VT1200S vibratome perpendicular to the pial surface in the same solution at 10°C under continuous oxygenation ( 5% CO2/95% O2 ) . Slices were transferred to cell strainers with 40 µm pores and placed in NMDG-aCSF at 34°C . Subsequently , sodium levels were gradually increased as previously described ( Ting et al . , 2018 ) . After recovery , slices were maintained for further experimental assessment at room temperature in an extracellular solution containing ( in mM ) 92 NaCl , 2 . 5 KCl , 1 . 25 NaH2PO4 , 30 NaHCO3 , 20 HEPES , 25 glucose , 2 thiourea , 5 Na-ascorbate , 3 Na-pyruvate , 2 CaCl2 , and 2 MgSO4 . Cortical slices from all human samples were macroscopically normal and showed no overt pathology . Adult mice ( C57BL/6J , B6 . 129-Synpotm1Mndl/Dllr [referred to as Synpo–/–] and B6 . Cg-Synpotm1MndlTg ( Thy1-Synpo/GFP ) 1Dllr/Dllr [referred to as Thy1-GFP/Synpo+/–× Synpo–/–]; 6–11 weeks old ) were used in this study . In Synpo–/– animals , the synaptopodin coding sequence is replaced by the lacZ sequence ( encoding β-galactosidase ) , thereby achieving a null synaptopodin allele ( Deller et al . , 2003 ) . This mouse strain was backcrossed onto the C57BL/6 genetic background for at least 10 generations . In experiments involving synaptopodin-deficient preparations , age-matched C57BL/6J animals served as controls , and experimental findings were confirmed using wild-type littermates from Synpo–/– mice . For preparing acute slices , animals were anesthetized with isoflurane and rapidly decapitated . Brains were rapidly removed , washed in chilled ( approximately 10°C ) NMDG-aCSF , and embedded in low-melting-point agarose ( Sigma–Aldrich #A9517; 1 . 8% w/v in PBS ) . Coronal sections of the mPFC were prepared using a Leica VT1200S vibratome in NMDG-aCSF with the brain tilted dorsally at a 15° angle . Slice recovery and maintenance prior to experimental assessment were performed as described above for acute human cortical slices . Acute cortical slices prepared from individual brain samples were randomly assigned to atRA or control ( vehicle-only ) treatment groups . Treatment with atRA was performed after slice recovery by adding atRA ( 1 µM , Sigma–Aldrich , #R2625 ) to the extracellular holding solution at a final concentration of 0 . 05% ( v/v in DMSO ) . The control group from the same set of slices was handled identically but treated with vehicle-only ( DMSO ) . Anisomycin ( 10 µM , Abcam , #ab120495 ) and actinomycin D ( 5 µg/ml , Sigma–Aldrich , #A9415 ) were added to the holding solution 10 min before the addition of atRA . Sections were treated for at least 6 hr before experimental assessment . Whole-cell patch-clamp recordings of superficial ( layer 2/3 ) cortical pyramidal neurons were carried out at 35°C in a bath solution containing ( in mM ) 92 NaCl , 2 . 5 KCl , 1 . 25 NaH2PO4 , 30 NaHCO3 , 20 HEPES , 25 glucose , 2 thiourea , 5 Na-ascorbate , 3 Na-pyruvate , 2 CaCl2 , and 2 MgSO4 . For experiments with acute mouse brain slices , superficial ( layer 2/3 ) pyramidal cells in the dorsomedial prefrontal cortex were visually identified using an LN-Scope ( Luigs and Neumann , Ratingen , Germany ) equipped with infrared dot-contrast and a 40× water-immersion objective ( numerical aperture [NA] 0 . 8; Olympus ) . For experiments with human cortical slices , superficial ( layer 2/3 ) pyramidal cells were visually identified on the pia-white matter axis at a distance of 500–1000 µm from the pial surface . Electrophysiological signals were amplified using a Multiclamp 700B amplifier , digitized with a Digidata 1550B digitizer , and visualized with the pClamp 11 software package . For sEPSC and intrinsic cellular property recordings , patch pipettes ( tip resistance: 3–5 MΩ ) contained ( in mM ) 126 K-gluconate , 4 KCl , 10 HEPES 4 MgATP , 0 . 3 Na2GTP , 10 PO-creatine , and 0 . 3% ( w/v ) biocytin ( pH = 7 . 25 with KOH; 285 mOsm/kg ) . For sEPSC recordings , pyramidal neurons were held at −70 mV in voltage-clamp mode . To record intrinsic cellular properties in current-clamp mode , a pipette capacitance of 2 . 0 pF was corrected , and series resistance was compensated using the automated bridge balance tool of the Multiclamp commander . Input–output ( I-V ) curves were generated by injecting 1-s square pulse currents starting at −100 pA and increasing in 10 pA increments ( sweep duration: 2 s ) . Series resistance was monitored , and recordings were discarded if the series resistance reached >30 MΩ . One superficial ( layer 2/3 ) cell in human neocortical slices ( Figure 1; atRA group ) with sEPSC amplitude = 38 . 4 pA and sEPSC frequency = 6 . 5 Hz showed interneuron characteristics and was therefore excluded from the analysis . The series resistance of one cell from a human cortical slice ( Figure 1; control group ) exceeded 30 MΩ during I-V curve recording . The respective I-V-curve was therefore excluded from further analysis . In five human superficial pyramidal neurons ( Figure 1 ) , the number of sweeps in the I-V curve recordings was lower compared to other recordings ( 40 sweeps vs . 60 sweeps ) . Thus , cells were excluded from further analysis of action potential frequency . Furthermore , one I-V-curve recording in the actinomycin-only treated group became unstable during the last sweeps and was consecutively excluded from action potential frequency analysis ( Figure 4—figure supplement 1 ) . In addition , one cell from a mouse neocortical slice ( Figure 3—figure supplement 1; Synpo+/+ , atRA group ) was excluded from further analysis because the signal displayed a marked electrical interference that caused disturbances in the baseline of the recordings . In the same data set , I-V curve recording from one cell ( Figure 3—figure supplement 1 , Synpo+/+ , control group ) was excluded due to I-V duplication in each sweep . Finally , one whole-cell patch-clamp recording of intrinsic cellular properties ( Figure 3—figure supplement 1; Thy1-GFP/Synpo , control group ) lost its integrity during the recording and was therefore excluded from further analysis . Cortical slices were fixed in 4% paraformaldehyde ( PFA ) ( prepared from 16% PFA stocks in PBS according to the manufacturer’s instructions; Thermo Scientific , #28908 ) at room temperature and stored at 4°C overnight in the same solution . After fixation , slices were washed in PBS and incubated for 1 hr with 10% ( v/v ) normal goat serum ( NGS; diluted in 0 . 5% [v/v] Triton X-100/PBS ) to reduce non-specific staining and increase antibody penetration . Subsequently , slices were incubated overnight at 4°C with rabbit anti-synaptopodin ( Synaptic Systems , #163002; 1:1000 ) or rabbit anti-NeuN ( Abcam , #ab104225 , 1:500 ) antibodies; both antibodies were diluted in 10% ( v/v ) NGS in 0 . 1% ( v/v ) Triton X-100/PBS . Sections were washed with PBS and incubated with goat anti-rabbit Alexa Fluor 488 or goat anti-rabbit Alexa Fluor plus 555 labeled secondary antibodies ( Invitrogen , #A-11034 and #A-32732 , respectively ) overnight at 4°C; both secondary antibodies were diluted 1:1000 in 10% ( v/v ) NGS in 0 . 1% ( v/v ) Triton X-100/PBS . For visualizing patched pyramidal cells , streptavidin-Alexa Fluor 488 ( Invitrogen , #S32354; 1:1000 ) was added during the secondary antibody incubation . Sections were washed again and incubated for 10 min with Sudan Black B ( 0 . 1% [w/v] in 70% ethanol ) to reduce autofluorescence . Sections were then incubated with DAPI for 10 min ( Thermo Scientific , #62248; 1:5000 in PBS ) to facilitate visualization of cytoarchitecture . After the final washing step , sections were transferred onto glass slides and mounted with a fluorescence anti-fade mounting medium ( DAKO Fluoromount ) . Confocal images were acquired using a Leica SP8 laser-scanning microscope equipped with a 20× multi-immersion ( NA 0 . 75; Leica ) , a 40× oil-immersion ( NA 1 . 30; Leica ) , and a 63× oil-immersion objective ( NA 1 . 40; Leica ) . Image stacks for dendritic spine and synaptopodin cluster analyses were acquired with a 63× objective at 6× optical zoom ( resolution: 1024 × 1024; z-step size: 0 . 2 µm; ideal Nyquist rate ) . Laser intensity and detector gain were set to achieve comparable overall fluorescence intensities throughout stacks between all groups . Confocal image stacks and single-plane pictures were stored as TIF files . Acute human cortical slices ( 400 µm ) were fixed with microwave irradiation ( Privileg 8020 E , 640 Watt , 2 . 45 GHz; Jensen and Harris , 1989 ) for 8 s on a petri dish filled with ice in 0 . 1% glutaraldehyde and 4% PFA ( dissolved in 0 . 1 M phosphate buffer [PB] and 0 . 05 M sucrose ) . The slices were kept in the same solution for 1 hr at room temperature and then transferred to 0 . 1 M PB . After 3 hr , 50 µm sections were prepared using a Leica VT1000S vibratome , washed for 30 min in 50 mM Tris-buffered saline ( TBS ) , and incubated for 1 hr with 20% NGS ( v/v ) in 50 mM TBS . Subsequently , sections were incubated with rabbit anti-synaptopodin ( Synaptic Systems , #163002; 1:100 in 2% NGS/50 mM TBS [v/v] ) at 4°C overnight . Sections were washed for 1 hr in 50 mM TBS and incubated with a suitable secondary goat anti-rabbit antibody ( Nanoprobes , #2004; 1 . 4 nM gold-coupled , 1:100 in 2% NGS/50 mM TBS [v/v] ) at 4°C overnight . After washing for 30 min in 50 mM TBS , sections were post-fixed with 1% glutaraldehyde/25 mM PBS ( w/v ) for 10 min . Sections were washed again in PBS , and silver intensification ( HQ Silver Enhancement Kit , Nanoprobes , #2012 ) was performed according to the manufacturer’s instructions . Subsequently , slices were incubated with 0 . 5% osmium tetroxide for 40 min , washed in graded ethanol ( up to 50% [v/v] ) for 10 min each , and incubated with uranyl acetate ( 1% [w/v] in 70% [v/v] ethanol ) for 35 min . Slices were then dehydrated in graded ethanol ( 80% , 90% , 95% , 2× 100% for 10 min each ) . Two 15 min washing steps in propylene oxide were performed prior to incubation with durcupan/propylene oxide ( 1:3 for 45 min followed by 3:1 for 45 min ) and durcupan ( overnight at room temperature ) . After slices were embedded in durcupan , ultra-thin sectioning ( 55 nm ) was performed using a Leica UC6 Ultracut . Sections were mounted onto copper grids ( Plano ) , at which point an additional Pb-citrate contrasting step was performed ( 3 min ) . Electron micrographs were captured using a Philips CM100 microscope equipped with a Gatan Kamera Orius SC600 ( magnification 5200x ) . Acquired images were stored as TIF files . After 6 hr of treatment with atRA or vehicle-only control , slices were fixed in 4% PFA ( w/v ) and 2 . 5% glutaraldehyde ( w/v; PBS ) overnight . After fixation , slices were washed for 4 hr in 0 . 1 M PB . Subsequently , slices were incubated with 1% osmium tetroxide for 45 min , washed in graded ethanol ( up to 50% [v/v] ) for 5 min each , and incubated overnight with uranyl acetate ( 1% [w/v] in 70% [v/v] ethanol ) overnight . Slices were then dehydrated in graded ethanol ( 80% , 90% , 98% for 5 min each , 2× 100% for 10 min each ) . Subsequently , two washing steps in propylene oxide for 10 min each were performed prior to incubation with durcupan/propylene oxide ( 1:1 for 1 hr ) and transfered to durcupan ( overnight at room temperature ) . Slices were embedded in durcupan , and ultra-thin sectioning ( 55 nm ) was performed using a Leica UC6 Ultracut . Sections were mounted onto copper grids ( Plano ) , at which point an additional Pb-citrate contrasting step was performed ( 3 min ) . Electron microscopy was performed using a Philips CM100 microscope equipped with a Gatan Orius SC600 camera at 3900× magnification . Acquired images were saved as TIF files and analyzed by an investigator blinded to experimental conditions . Electrophysiological data were analyzed using Clampfit 11 from the pClamp11 software package ( Molecular Devices ) . sEPSC properties were analyzed using the automated template search tool for event detection . Specifically , AP detection was performed using the input/output curve threshold search event detection approach , whereas AP frequency was determined based upon the number of APs detected during a given injection . Artifacts in electrophysiological recordings were excluded from further analysis . Immuno-labeled synaptopodin clusters in superficial ( layer 2/3 ) pyramidal cells of the human cortex were analyzed in image stacks of second- and third-order dendritic branches . Synaptopodin clusters that colocalized with dendritic spines ( either spine neck or head; Figure 2D ) were explored in further analyses . Single-plane images of both synaptopodin-positive and -negative clusters were extracted from image stacks at the point of maximum spine head cross-sectional area and stored as TIF files . Blinded analyses of spine head cross-sectional area and synaptopodin cluster size were performed manually using the ImageJ software package ( available at http://imagej . nih . gov/ij/ ) . Here , the outer borders of synaptopodin clusters and spine heads were marked independently of their overall fluorescence intensity . Data were transferred and stored in Excel files . Ultrastructural analysis of spine apparatus organelles was performed using single-plane images of human cortical excitatory synapses where pre- and postsynaptic structures could be readily identified . The cross-sectional areas of spine apparatus organelles were determined manually using the ImageJ software package , independent of their shape and internal structural organization . Statistical analyses were performed using the GraphPad Prism seven software package . Two-group comparisons were performed using a Mann–Whitney U test; U-values for statistically significant differences are reported in the figure legends . A Kruskal–Wallis test with Dunn’s multiple comparisons was used to compare more than two groups . Correlation of individual data points was visualized by a linear regression fit and analyzed by computing Spearman r . For statistical evaluation of XY-plots , we used an RM two-way ANOVA test ( repeated measurements/analysis ) with Sidak’s multiple comparisons . For the comparison of more than two groups in XY-plots , Tukey’s multiple comparisons were applied . For the in-sample analysis of human cortical slices ( paired experimental design ) , we used a Wilcoxon matched-pairs signed-rank test . p-values<0 . 05 were considered statistically significant ( p<0 . 05 , p<0 . 01 , *p<0 . 001 ) ; results that did not yield significant differences are designated ‘ns’ . Statistical differences in XY-plots were indicated in the legend of the figure panels ( ) when detected through multiple comparisons , irrespective of their localization and the level of significance . In the text and figures , values represent the mean ± standard error of the mean ( s . e . m . ) . Figures were prepared using Photoshop graphics software ( Adobe , San Jose , CA ) . Image brightness and contrast were adjusted .	The brain has an enormous capacity to adapt to its environment . This ability to continuously learn and form new memories thanks to its malleability , is known as brain plasticity . One of the most important mechanisms behind brain plasticity is the change in both the structure and function of synapses , the points of contact between neurons where communication happens . These sites of synaptic contact occur through microscopic protrusions on the branches of neurons , called dendritic spines . Dendritic spines are very dynamic , changing their shape and size in response to stimuli . Previous studies have shown that alterations in synaptic plasticity occur in various animal models of brain diseases . However , it remains unclear whether human cortical neurons express synaptic plasticity similarly to those in the rodent brain . Recently , a derivative of vitamin A has been linked to synaptic plasticity . In addition , several studies have evaluated the effects of this derivative in patients with cognitive dysfunctions , including Alzheimer’s disease , Fragile X syndrome , and depression . However , there is no direct experimental evidence for synaptic plasticity in the adult human cerebral cortex related to vitamin A signaling and metabolism . To investigate this , Lenz et al . used human cortical slices prepared from neurosurgical resections and treated them with a solution of the vitamin A derivative all-trans retinoic acid for 6-10 hours . Lenz et al . employed a variety of techniques , including patch-clamp recordings to measure neuron function as well as different types of microscopy to evaluate structural changes in dendritic spines . These experiments demonstrated that the derivative promoted the synaptic plasticity in the adult human cortex . Specifically , it increased the size of the dendritic spines and strengthened their ability to transmit signals . In addition , Lenz et al . found that the spine apparatus organelle – a structure found in some dendritic spines – was a target of the vitamin A derivative and promoted synaptic plasticity . These findings advance the understanding of the pathways through which vitamin A derivatives affect synaptic plasticity , which may aide the development of new therapeutic strategies for brain diseases . More generally , the results contribute to the identification of key mechanisms of synaptic plasticity in the adult human brain .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2021	All-trans retinoic acid induces synaptic plasticity in human cortical neurons
Inhibition plays a crucial role in neural signal processing , shaping and limiting responses . In the auditory system , inhibition already modulates second order neurons in the cochlear nucleus , e . g . spherical bushy cells ( SBCs ) . While the physiological basis of inhibition and excitation is well described , their functional interaction in signal processing remains elusive . Using a combination of in vivo loose-patch recordings , iontophoretic drug application , and detailed signal analysis in the Mongolian Gerbil , we demonstrate that inhibition is widely co-tuned with excitation , and leads only to minor sharpening of the spectral response properties . Combinations of complex stimuli and neuronal input-output analysis based on spectrotemporal receptive fields revealed inhibition to render the neuronal output temporally sparser and more reproducible than the input . Overall , inhibition plays a central role in improving the temporal response fidelity of SBCs across a wide range of input intensities and thereby provides the basis for high-fidelity signal processing . Dynamic processing in neural networks is controlled by an interplay of excitation and inhibition . In cortical processing , the dominant excitatory neurons interact reciprocally with inhibitory neurons , which serve key functions in shaping the responses ( reviewed in Isaacson and Scanziani , 2011 ) . In the auditory cortex , recent work has emphasized the role of inhibition in dynamically balancing excitation via a high degree of co-tuning ( e . g . Wehr and Zador , 2003; Renart et al . , 2010 ) that serves to shape and accelerate network dynamics . Similarly , in other modalities , inhibition was found to be co-tuned with excitation in the cortex , typically with a wider tuning , generating the well-described inhibitory sidebands ( auditory: Wang et al . , 2002; Wu et al . , 2008 , visual: Sohya et al . , 2007; Niell and Stryker , 2008; Liu et al . , 2009 , 2011; Katzner et al . , 2011 , olfactory: Poo and Isaacson , 2009 ) . Temporally , inhibition often follows excitation closely ( auditory: Wehr and Zador , 2003 , somatosensory: Wilent and Contreras , 2004 ) . In the auditory brainstem , the role of inhibition has also been studied , however , from a more fundamental perspective , without a focus on its functional role during complex stimulation . Various studies have shown prominent inhibitory influences on signal processing in the cochlear nucleus ( Caspary et al . , 1994; Kopp-Scheinpflug et al . , 2002; Gai and Carney , 2008 ) , in the medial and lateral superior olive ( Grothe and Sanes , 1993; Brand et al . , 2002; Myoga et al . , 2014 ) , and in the dorsal and ventral nuclei of the lateral lemniscus ( Yang and Pollak , 1994 , 1998; Burger and Pollak , 2001; Nayagam et al . , 2005; Pecka et al . , 2007; Spencer et al . , 2015 ) . The cochlear nucleus ( CN ) , the first stage of the central auditory system , is the starting point of distinct neuronal circuits involved in sound source localization . Spherical bushy cells ( SBC ) in the anteroventral division of the CN ( AVCN ) provide the temporally precise excitatory inputs to binaural neurons in the medial superior olive ( MSO ) , where interaural time differences are computed ( Yin and Chan , 1990 ) . These SBCs receive suprathreshold excitatory input from auditory nerve fibers ( ANF ) via large axosomatic terminals , the endbulbs of Held ( Brawer and Morest , 1975; Schwartz and Gulley , 1978; Ryugo and Sento , 1991; Nicol and Walmsley , 2002 ) . In addition , inhibitory inputs on SBCs have been reported , which provide surprisingly slow acoustically evoked inhibition mediated by glycine and GABA ( Wu and Oertel , 1986; Kolston et al . , 1992; Juiz et al . , 1996; Lim et al . , 2000; Mahendrasingam et al . , 2004; Xie and Manis , 2013 ) , with glycine dominating ( Nerlich et al . , 2014b ) . Due to the requirements of high-fidelity acoustic processing underlying sound localization , many studies focused on the fast and temporally precise signal transmission in auditory brainstem circuits . With respect to changes in temporal precision from ANF to neurons in the CN some studies – comparing population data – reported a general increase in temporal precision ( Joris et al . , 1994a , 1994b ) , while others found no change ( Bourk , 1976; Blackburn and Sachs , 1989; Winter and Palmer , 1990 ) , or reported decreased temporal precision at certain stimulation frequencies ( Paolini et al . , 2001; Fukui et al . , 2006 ) . More recent studies advanced the analysis to the single-cell level , by comparing the endbulb of Held evoked excitatory postsynaptic potentials ( EPSP ) with the action potentials ( AP ) of the SBCs allowing for a direct comparison of ANF input and SBC output ( Typlt et al . , 2010 ) . This enabled a direct assessment of the input-output function under the condition of acoustic stimulation , also in combination with pharmacological manipulations . The respective experiments revealed a slight increase in temporal precision of signal coding , attributed to the influence of acoustically evoked inhibition ( Dehmel et al . , 2010; Kuenzel et al . , 2011; Keine and Rübsamen , 2015 ) . It may be argued that the stimulus conditions employed were rather static and did not adequately reflect the challenge of processing the dynamics of spectrotemporal complex acoustic signals . While fast inhibition in T stellate cells has been attributed to a role in comodulation masking release ( Pressnitzer et al . , 2001 ) , the inhibitory dynamics in SBCs seem to be too slow for such an effect ( Xie and Manis , 2013 ) . Previous studies investigated inhibition using pure tone stimulation , and this is why the functional role of inhibition in signal processing at the ANF-SBC synapse during complex acoustic stimulation has not been fully resolved . In the present study , we set out to elucidate the functional role of acoustically evoked inhibition at the ANF-SBC synapse using combined in vivo loose-patch recordings with direct iontophoretic manipulation of inhibitory receptors and a detailed input-output signal analysis based on spectrotemporal receptive fields in responses to complex acoustic stimulation . Our results indicate a reliable co-tuning of inhibition with the main excitatory input . While we observed some sharpening of the response in time and frequency , our results suggest that inhibition functions as a gain control that renders the postsynaptic response sparser in time and more reproducible across trials . Temporal sparsity , i . e . a response restricted to fewer time-points , can increase the information per spike while reducing the energy expenditure . Reproducibility , i . e . a more consistent response to identical stimuli , can provide reliable stimulus encoding . These improvements are a consequence of the combined subtractive/divisive action of glycine ( Kuenzel et al . , 2011 , 2015 ) : The subtractive component enhances the temporal sparsity by raising the threshold for spiking . The divisive component acts primarily as a gain control , which - in conjunction with the co-tuning - maintains the SBC output rate in a smaller range across different stimulus levels . Together these two effects focus the SBC output onto well-timed stimulus events across a wide range of stimulus levels . Thus , inhibition improves the basis for the high-fidelity signal processing in downstream nuclei crucial for sound localization irrespective of the prevailing stimulus levels . The increased incidence of failures during acoustic stimulation has been attributed to the activation of inhibitory inputs ( Kopp-Scheinpflug et al . , 2002; Kuenzel et al . , 2011 , 2015; Keine and Rübsamen , 2015 ) . However , also in vitro experiments need to be considered , which showed strong depression at the ANF-SBC synapse ( Wang and Manis , 2008; Yang and Xu-Friedman , 2008 , 2009; Wang et al . , 2010 ) affecting SBC responsiveness for up to tens of milliseconds ( Yang and Xu-Friedman , 2015 ) . Such depression might also suppress SBC spiking in vivo and result in an increased failure fraction during acoustic stimulation . Still , in vivo the impact of depression was shown to be smaller , since ongoing spontaneous activity – completely absent in slice recordings – seems to keep the synapse in a chronically depressed state ( Hermann et al . , 2007; Lorteije et al . , 2009; Yang and Xu-Friedman , 2015 ) . Also , the in vivo calcium concentration was reported to be lower than in the artificial cerebrospinal fluid usually used in slice studies resulting in lower vesicle release probabilities and thus smaller depression ( Borst , 2010; Kuenzel et al . , 2011; Friauf et al . , 2015 ) . To determine the cause of altered reliability of synaptic transmission at the ANF-SBC synapse , and to dissect the effect of acoustically evoked inhibition from synaptic depression , we first quantified the dependence of the EPSP rising slope on the preceding spontaneous activity . As indicated above , the rising slopes of EPSPfail and EPSPsucc differ , but still show a considerable range of overlap . For each unit , the EPSP rising slopes were pooled for EPSPsucc and EPSPfail and binned . Then , the fraction of EPSPsucc was calculated for each bin , and a Boltzmann function was fitted to the EPSPsucc probability distribution ( Figure 2A left ) . The symmetric inflection point of this function indicates the threshold EPSP , i . e . the EPSP slope necessary to trigger an AP with >50% probability . EPSP rising slopes showed strong depression for inter-event intervals ( IEI ) < 2 ms resulting in high AP failure rates . But , already for IEIs > 5 ms , the preceding activity had only a minor influence on EPSP rising slopes ( Figure 2A , middle ) . Averaging the normalized EPSP slopes across cells showed a facilitating effect for IEI between 2 ms and 20 ms ( green markers , Figure 2A right ) . The threshold EPSP was increased for IEIs < 2 ms , but not for longer IEIs ( black line ) and the increase in threshold EPSP resulted in an increased failure fraction for IEIs < 2 ms ( orange histogram ) . While IEIs up to 20 ms resulted in increased EPSP slopes , the effect of IEIs on threshold EPSP and failure fraction was limited to short IEIs < 2 ms . These data are consistent with previous studies , suggesting the presence of short-term facilitation rather than depression of synaptic events . Considering only the last preceding IEI , however , disregards the potential impact of previous medium- and short-term afferent activity . Also , in vitro studies yielded the influence of short-term depression at the ANF-SBC synapse to extend well beyond the last IEI ( Yang and Xu-Friedman , 2015 ) . 10 . 7554/eLife . 19295 . 005Figure 2 . Preceding activity has only a minor , facilitating influence on EPSP rising slopes . ( A ) Left: Estimation of threshold EPSP for a representative cell: The EPSP rising slopes were binned ( 0 . 5 V/s bin size ) and the proportion of EPSPsucc calculated for each bin . A Boltzmann function was fit to these data . The symmetric inflection point of this function was considered the threshold EPSP and indicates the EPSP rising slope necessary to generate an AP with >50% probability . Middle: The inter-event-interval ( IEI ) between synaptic inputs had only a small influence on the EPSP rising slope , with small IEIs being correlated with moderately increased EPSP rising slopes . A more prominent difference was observed between EPSPfail ( gray ) which showed consistently smaller rising slopes than EPSPsucc ( blue ) and these differences prevailed over a wide range of IEIs . For IEI < 2 ms the SBCs relative refractoriness renders virtually all EPSPs unsuccessful in triggering a postsynaptic AP . The black line indicates the threshold EPSP . Right: Grand average of normalized EPSP slope , threshold EPSP ( left ordinate ) , and failure fraction ( right ordinate ) in dependence of preceding IEI pooled for EPSPsucc and EPSPfail ( n = 62 cells ) . The average EPSP slope ( green , left ordinate ) showed facilitation for IEIs between 2–20 ms ( error bars indicate standard deviation ) . The median threshold EPSP ( black line , left ordinate ) was elevated only for IEIs < 2 ms and well below average EPSP size for larger IEIs ( shaded area indicates first and third quartile ) . The elevated threshold EPSP resulted in an increased failure fraction for IEIs < 2 ms , while for longer IEIs the reliability of AP generation seemed not to be affected ( orange , right ordinate ) . ( B ) Consideration of a wider time span of preceding activity: Sketch of the quantification of preceding activity by exponentially weighting [W] all preceding EPSP rising slopes [S] ( ANF activity ) or AP amplitudes ( SBC activity ) depending on the distance to the event under investigation . ( C ) Left: EPSP rising slopes for the representative cell showed only minor dependence on previous ANF activity levels . Note that the EPSPfail ( gray ) showed consistently lower EPSP rising slopes ( histogram on the left ) ; still , the EPSPs slopes tend to increase during periods of high activity . The threshold EPSP ( black line ) increased as a function of ANF activity . Threshold EPSP was calculated for different levels of ANF activity ( bin size = 0 . 5 ) . Right: signal amplitudes of EPSPfail ( gray ) and APs ( blue ) as a function of preceding SBC activity showed decreasing AP amplitudes but increasing EPSP amplitudes . ( D ) Population data for 62 units ( n = 62 ) : While EPSP slopes tend to be elevated after periods of high activity ( left ) , AP amplitudes showed a negative correlation with preceding SBC activity ( p<0 . 001 , one-sample t-test against zero ) . Triangles indicate the data of the representative cell . Organization of the graph as described above . . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 00510 . 7554/eLife . 19295 . 006Figure 2—source data 1 . Correlation between EPSP slopes and signal amplitudes on preceding ANF activity . Shown are the Spearman correlation coefficient values obtained from the correlations of maximum EPSP slopes and signal amplitude vs . preceding ANF activity levels , separately for EPSPsucc and EPSPfail ( n = 62 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 006 To determine the impact of preceding activity on EPSP strength and AP generation in vivo , the preceding activity of each event was quantified as a weighted sum of all previous events , using an exponentially decaying kernel with a time constant of 60 ms , emphasizing temporally closer events over more distant ones ( Figure 2B ) . The analysis yielded only minor influences of preceding activity on EPSP rising slopes on both EPSPfail ( gray ) and EPSPsucc ( blue ) as shown in a representative cell in Figure 2C ( left and middle ) and also evidenced for the population of recorded units ( Spearman’s rho EPSPfail = 0 . 22 ± 0 . 16 vs . EPSPsucc = 0 . 24 ± 0 . 11 , Δ = 0 . 02 ± 0 . 12 , p=0 . 27 , paired t-test , n = 62 , U1 = 0 . 09 , Figure 2D , see also Figure 2—source data 1 ) . A small but consistent effect , seen in 61/62 recorded units ( 98% ) , was a positive correlation ( p<0 . 001 ) between preceding ANF activity and EPSP rising slopes indicating a facilitating rather than a depressive influence of higher activity levels in vivo . Postsynaptic spike depression may also contribute to the increase in postsynaptic spike failures ( EPSPfail ) . When analyzing the dependence of AP amplitude on preceding SBC spiking activity ( exemplary unit shown in Figure 2C right ) a significant negative correlation was observed in 92% of the cells ( 57/62 ) indicating smaller AP amplitudes after periods of higher SBC activity ( Figure 2D ) . The representative unit shown in Figure 2C ( right ) shows the respective change in AP amplitude and an inverse effect on the amplitudes of EPSPfail , consistent with the facilitating influence on EPSP rising slopes . These results are in agreement with previous reports ( Kuenzel et al . , 2011 , 2015 ) suggesting that endbulbs are mostly in a close-to-threshold state and show low synaptic depression in vivo . These results suggest that the increased failure fraction during acoustic stimulation in vivo is not explainable by endbulb depression evoked by high firing rates , highlighting the role of acoustically evoked inhibition on the input-output relationship at the ANF-SBC synapse . Frequency response areas ( FRA ) of SBCs show prominent inhibitory sidebands and reduced firing activity in the excitatory field compared to the ANF input ( Kopp-Scheinpflug et al . , 2002; Kuenzel et al . , 2011; Keine and Rübsamen , 2015 ) . Also , about half of the SBCs show pronounced non-monotonic rate-level functions pointing to an impact of inhibition ( Kopp-Scheinpflug et al . , 2002; Keine and Rübsamen , 2015; Kuenzel et al . , 2015 ) which has been further classified as ‘on-CF inhibition’ and ‘broadband inhibition’ ( Winter and Palmer , 1990; Caspary et al . , 1994; Kopp-Scheinpflug et al . , 2002 ) . In vitro and modeling studies showed that glycinergic inhibition can elevate the threshold EPSP for AP initiation ( Xie and Manis , 2013; Kuenzel et al . , 2015 ) . Thus the threshold EPSP can serve as a suitable indicator for the activation of inhibitory inputs . In the present loose-patch recordings , elevation in threshold EPSP ( Figure 3Aii/iii ) was observed throughout the FRAs ( Figure 3Ai/iii ) accompanied by an increase in failure fraction ( Figure 3Bii ) . The frequency profile of threshold EPSP elevation closely matched the one of increased failure fraction ( Figure 3Aii/Bii , Spearman correlation rs = 0 . 7 [0 . 4 , 0 . 76] , p<0 . 001 , Wilcoxon signed rank test , n = 62 , U1 = 0 . 98 , population data not shown ) . The FRA of threshold EPSP elevation was used to quantify the inhibitory influence , which was then compared to the SBC’s excitatory FRA . Both FRAs had similar CFs , defined as the stimulus frequency at which the lowest sound intensity resulted in a significant increase in ANF firing rate ( excitatory ) or threshold EPSP ( inhibitory ) ( 2 . 2 ± 0 . 6 kHz vs . 2 . 2 ± 0 . 9 kHz , respectively , Δ = 0 . 03 ± 0 . 83 kHz , p=0 . 77 , paired t-test , n = 62 , U1 = 0 . 05 , Figure 3Ci ) , but inhibitory FRAs exhibited higher thresholds ( excitatory = 4 . 8 ± 6 . 1 dB SPL vs . inhibitory = 19 . 8 ± 16 . 6 dB SPL , Δ = 15 ± 15 . 5 dB SPL , p<0 . 001 , paired t-test , n = 62 , U1 = 0 . 2 , Figure 3Cii , see also Figure 3—source data 1 ) . 10 . 7554/eLife . 19295 . 007Figure 3 . Inhibition at SBC is co-tuned with excitation and broadband , not off-CF and narrowband . ( A ) i: Representative frequency response area ( FRA ) of the excitatory ANF input ( EPSPfail and EPSPsucc ) characterized by a well-defined CF , the typical steep high-frequency flank , the formation of a low-frequency tail , and the absence of frequency-intensity domains of inhibition . ii: The same recording showed elevated threshold EPSPs throughout most of the excitatory response area and extending up to two octaves above CF . The frequency , where the lowest relative intensity caused elevated threshold EPSP , matched the units CF . iii: For the same unit , comparison of excitatory ( ANF , gray ) response area and frequency-intensity domain of inhibition ( threshold EPSP elevation , red ) . The inhibitory domain was symmetrically arranged around the unit’s CF . ( B ) i: FRA of the SBC output ( EPSPsucc ) shows a considerable reduction in firing activity compared to the ANF input . ii: Failure fraction , i . e . the proportion of EPSPfail . The increase in failure fraction is most prominent around the units CF . Note the similarity of the frequency-intensity domains of EPSP threshold increase and the respective domains with increased EPSPfail in Aii . iii: Rate-level functions of ANF input ( gray line , left ordinate ) and SBC output ( solid black line , left ordinate ) compared to threshold EPSP ( red , solid line , right ordinate ) . Increasing sound pressure levels result in a monotonic increase in ANF firing and correspondingly the threshold EPSP shows a monotonic increase . The SBC output is maximal at 20 dB SPL and declines towards higher stimulus intensities . ( C ) Population data: comparison of excitatory ( ANF , gray ) and inhibitory ( threshold EPSP , red ) FRA indicates ( i ) on-CF inhibition although ( ii ) with higher thresholds ( p<0 . 001 , paired t-test ) , which is ( iii ) broadly tuned ( Q10: p<0 . 01 , Q40: p<0 . 001 , two-way RM ANOVA ) , and ( iv ) shows a more symmetric tuning ( p<0 . 001 , paired t-test; the schematic drawing on the right indicates FRA shapes for different asymmetry indices ) . ( D ) Finally , the rate-level functions were shallower and showed a reduced gain in firing rate in the output compared to the input . SBC = spherical bushy cell , CF = characteristic frequency , EPSP = excitatory postsynaptic potential , ANF = auditory nerve fibers , FRA = frequency response area . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 00710 . 7554/eLife . 19295 . 008Figure 3—source data 1 . Tuning properties of excitatory and inhibitory inputs onto SBCs . Shown are the characteristic frequency ( in kHz ) , threshold ( in dB SPL ) , Q-values , asymmetry index , and rate-level gain for the excitatory ANF input and the inhibitory area determined by an elevation in threshold EPSP ( n = 62 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 008 The width of inhibitory and excitatory FRA was determined by calculating Q10 and Q40 values . Both measures were smaller for the inhibitory FRA compared to the excitatory FRA , i . e . inhibition showed reduced frequency selectivity compared to excitation ( Q10: excitatory = 2 . 5 ± 0 . 6 vs . inhibitory = 1 . 9 ± 1 . 1 , Δ = 0 . 5 ± 1 . 2 , p<0 . 01; Q40: excitatory = 0 . 96 ± 0 . 1 vs . inhibitory = 0 . 6 ± 0 . 3 , Δ = 0 . 4 ± 0 . 3 , p<0 . 001 , two-way RM ANOVA , Bonferroni-adjusted , n = 62 , η2 = 0 . 06 , Figure 3Ciii ) . Q-values provide information about the sharpness of tuning , but not about the actual shape of the FRA . The tuning shape was evaluated using the asymmetry index ( AI , see Materials and methods ) , with values of 0 indicating symmetric , <0 for low-frequency extended and >0 for high frequency extended tuning curves . While excitatory FRAs showed distinct low-frequency tails , typical for ANF , inhibitory FRAs were mostly symmetrically arranged around CF , partly covering high-frequency ranges above the excitatory response area ( AI excitation = –0 . 96 ± 0 . 42 vs AI inhibition = –0 . 26 ± 0 . 88 , Δ = 0 . 7 ± 0 . 97 , p<0 . 001 , paired t-test , n = 62 , U1 = 0 . 21 , Figure 3Civ ) . The rate-level function ( RLF ) of the SBC output was markedly flatter and thus less variable with respect to level than the rate-level function of the excitatory ANF input . The gain of the neuronal response across stimulus level was quantified as rate level gain ( RLG ) , defined as RLG=log10 ( FRmax−FRminFRspont ) , with FRmax and FRmin being the maximal and minimal firing rate in the RLF and FRspont , the spontaneous firing rate in the absence of acoustic stimulation ( see also supplementary Matlab code ) . This way , overall changes in firing rates are taken into account ( e . g . due to spontaneous failures ) . The output’s rate level function had a gain of 1 ± 0 . 4 which was significantly less than the input’s ( 1 . 4 ± 0 . 4 , Δ = 0 . 35 ± 0 . 3 , p<0 . 001 , paired t-test , n = 62 , U1 = 0 . 1 , Figure 3D ) . Taken together , these data demonstrate that inhibition in SBC is co-tuned with excitation and shows a broader and more symmetric frequency profile , which results in flatter rate-level functions and high-frequency inhibitory sidebands at the fringes of the tuning curve ( frequently observed in SBC output activity ) . These findings are consistent with previous reports ( Caspary et al . , 1994; Kopp-Scheinpflug et al . , 2002; Kuenzel et al . , 2011 ) . The lower rate-level dependence suggests a gain-normalization function of inhibition , discussed in detail below . As shown above , the tuning of the inhibitory input on SBCs largely matches the ANF excitation . In vitro studies and modeling suggested inhibition to prevent AP generation in SBC by elevating the threshold EPSP ( Xie and Manis , 2013; Kuenzel et al . , 2015 ) . Slice studies reported a predominately glycinergic inhibition with a smaller GABAergic contribution ( Nerlich et al . , 2014a , 2014b ) , and in vivo studies showed an effective , dose-dependent block of SBC spiking by iontophoretic application of glycine ( Keine and Rübsamen , 2015 ) . Consequently , we tested if the activation of glycinergic inputs can directly cause the observed elevation of threshold EPSP . Glycine was applied iontophoretically , mimicking the putative role of glycinergic inhibition , while monitoring the SBC’s spontaneous activity . Indeed , glycine caused an increase in the number of EPSPs that failed to trigger APs , and this specific effect could be blocked by simultaneous application of the glycine receptor antagonist strychnine ( Figure 4A ) . The iontophoretic current for glycine application was adjusted to cause an increase in the spontaneous failure fraction from 0 . 3 ± 0 . 17 to 0 . 64 ± 0 . 18 ( Δ = 0 . 34 ± 0 . 12 , p<0 . 001 , paired t-test , n = 11 , U1 = 0 . 5 ) to match the range observed under acoustic stimulation . This increase in failure fraction was accompanied by an elevation in threshold EPSP ( threshold EPSP spont = 6 . 1 ± 2 . 2 V/s vs threshold EPSP glycine = 8 . 4 ± 1 . 8 V/s , Δ = 2 . 3 ± 0 . 9 V/s , p<0 . 001 , paired t-test , n = 11 , U1 = 0 . 32 , Figure 4B , see also Figure 4—source data 1 ) . The application of the carrier alone had neither an effect on threshold EPSPs ( threshold EPSP spont = 6 . 8±1 . 6 V/s vs . threshold EPSP carrier = 6 . 8 ± 1 . 7 , Δ = 0 ± 0 . 3 , p=0 . 89 , paired t-test , n = 9 , U1 = 0 . 1 , data not shown ) nor on failure fraction ( failure fraction spont = 0 . 29 ± 0 . 17 vs . failure fraction carrier = 0 . 27 ± 0 . 16 , Δ = 0 . 03 ± 0 . 06 , p=0 . 79 , paired t-test , n = 9 , U1 = 0 . 11 , data not shown ) . 10 . 7554/eLife . 19295 . 009Figure 4 . Glycinergic inhibition elevates threshold EPSP and becomes activated during acoustic stimulation . ( A ) Representative recording of spontaneous activity with iontophoretically applied glycine to block SBC spiking ( red bar ) . This effect is suspended by strychnine application ( green bar ) . ( B ) ( Bi ) Left: During spontaneous activity , small EPSPs fail to generate APs ( gray = EPSPfail , blue = EPSPsucc , black line = threshold EPSP ) . Right: Iontophoretic application of glycine elevates the threshold EPSP ( solid red line ) for spike generation resulting in an increased failure fraction ( dashed black line shows threshold EPSP from control condition ) . ( Bii ) : Population data for 11 units showing the effect of glycinergic inhibition on the increase of threshold EPSP . ( C ) and ( D ) : Acoustically evoked FRAs while blocking glycinergic inhibition . ( Ci ) No effect on input FRA was observed when inhibition was blocked . ( Cii ) Population data confirming the lack of glycine effect on the input activity . ( Di ) SBC output FRA shows increased firing rates during the blockade of glycinergic inhibition . Note the absence of the inhibitory sideband after inhibition block . ( Dii ) Population data show a considerable increase in SBC firing after block of inhibition ( p<0 . 001 , paired t-test ) . ( Ei ) Left: Under control condition , t threshold EPSP is elevated during acoustic stimulation , indicating the presence of acoustically evoked inhibition . Right: This threshold elevation is absent when the glycinergic inhibition is blocked . ( Eii ) Population data showing the threshold EPSP during spontaneous activity and acoustic stimulation at the units’ CF for control condition ( gray , p<0 . 001 , two-way RM ANOVA ) and under inhibition block ( green ) . Note the absence of threshold EPSP elevation during acoustic stimulation under the inhibition block . Blocking glycinergic inhibition had no effect on threshold EPSP during spontaneous activity . Triangles in Bii–Eii denote representative cells from Bi–Ei . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 00910 . 7554/eLife . 19295 . 010Figure 4—source data 1 . Iontophoretic application of glycine and strychnine . Shown are the threshold EPSPs ( in V/s ) during control condition and during glycine application ( n = 11 ) . The data set also contains the ANF input firing rates , SBC output firing rates ( in Hz ) and threshold EPSPs during spontaneous activity and acoustic stimulation for both control conditions and during strychnine application ( n = 11 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 010 Next , the contribution of inhibition-mediated threshold EPSP elevation on spike failures during acoustic stimulation was tested . The specific glycine receptor antagonist strychnine was iontophoretically applied to block the acoustically evoked glycinergic inhibition . The effectiveness of the glycine block was tested before sound stimulation by simultaneously applying glycine and strychnine , with the application current for strychnine adjusted to block the effect of iontophoretically applied glycine . The ANF input firing rates were not influenced by the block of inhibition ( control = 282 ± 48 Hz vs . strychnine = 283 ± 50 Hz , Δ = 1 . 2 ± 26 . 7 Hz , p=0 . 88 , paired t-test , n = 11 , U1 = 0 . 18 , Figure 4C ) . The SBC output rates in the excitatory field , however , were substantially increased under glycine block ( control = 122 ± 49 Hz vs . strychnine = 192 ± 39 Hz , Δ = 70 ± 47 . 9 Hz , p<0 . 001 , paired t-test , n = 11 , U1 = 0 . 5 , Figure 4D ) . We next tested if the block of glycinergic inhibition differentially affects the threshold EPSP during spontaneous activity and during acoustic stimulation . When glycinergic inhibition was blocked , the threshold EPSP was only affected during acoustic stimulation , but not during spontaneous activity ( interaction drug × stimulus condition , p<0 . 01 , η² = 0 . 13 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 11 , Figure 4E ) . Acoustic stimulation at CF under control condition resulted in a significant threshold EPSP elevation ( threshold EPSP spont = 5 . 4 ± 1 . 6 V/s vs . stim = 8 . 8 ± 3 . 1 V/s , Δ = 3 . 5 ± 2 . 9 V/s , p<0 . 01 , two-way RM ANOVA , Bonferroni-adjusted , n = 11 , U1 = 0 . 41 ) and this shift was absent when the inhibition was blocked ( threshold EPSP spont = 5 . 9 ± 2 V/s vs . stim = 5 . 8 ± 1 . 8 V/s , Δ = 0 . 1 ± 1 . 3 V/s , p=0 . 82 , two-way RM ANOVA , Bonferroni-adjusted , n = 11 , U1 = 0 . 09 , Figure 4D ) . The effects observed under acoustic stimulation were very different from the respective manipulations performed during spontaneous activity . In the absence of acoustic stimulation , the block of glycinergic inhibition had no effect on output rates ( control = 51 ± 26 Hz vs . strychnine = 49 ± 20 Hz , Δ = 1 . 6 ± 18 . 5 Hz , p=0 . 79 , two-way RM ANOVA , Bonferroni-adjusted , n = 11 , U1 = 0 . 14 , data not shown ) , and threshold EPSP ( control = 5 . 5 ± 2 . 5 V/s vs strychnine = 5 . 6 ± 2 . 6 V/s , Δ = 0 . 03 ± 0 . 16 V/s , p=0 . 53 , two-way RM ANOVA , Bonferroni-adjusted , n = 11 , U1 = 0 . 09 , Figure 4Eii ) . Similar to acoustic stimulation , the input rates were not altered during inhibition block ( control = 85 ± 26 Hz vs . strychnine = 86 ± 22 Hz , Δ = 0 . 4 ± 16 . 6 Hz , p=0 . 94 , two-way RM ANOVA , Bonferroni-adjusted , n = 11 , U1 = 0 . 09 , data not shown ) . These data suggest a major role of glycinergic inhibition in acoustically evoked signal processing , but a negligible impact during spontaneous activity . Taken together , the data confirms previous reports of broadly tuned , predominantly glycinergic inhibition ( Kopp-Scheinpflug et al . , 2002; Kuenzel et al . , 2011 ) , which decreases and potentially normalizes SBC output firing across different stimulus conditions by an increase in threshold EPSP for spike generation . The results above suggest that acoustically evoked inhibition can considerably influence SBC spiking by increasing the threshold for AP generation . Previous studies directly comparing the ANF input and SBC output showed an increase in temporal precision which has been attributed to the impact of inhibition ( Dehmel et al . , 2010; Kuenzel et al . , 2011; Keine and Rübsamen , 2015 ) . These studies focused on the responses to static pure-tone stimulation leaving the question for a potential influence of acoustically evoked inhibition on signal transmission in a more complex , i . e . a more naturalistic acoustic environment unaddressed . Considering this issue , we first tested the responses of SBCs to sinusoidal amplitude-modulated ( SAM ) and frequency-modulated ( SFM ) acoustic stimuli . SAM stimuli were presented at the respective units’ CF 30 dB above the excitatory threshold with modulation frequencies between 50 Hz and 400 Hz ( modulation depth = 100% , Figure 5 ) . The discharge activity of the units showed different degrees of modulation congruent with the SAM for both the ANF input and the SBC output ( Figure 5B , C ) . The AP failure fraction increased from 0 . 27 ± 0 . 22 in the absence of acoustic stimulation to 0 . 43 ± 0 . 18 during SAM stimulation ( Δ = 0 . 16 ± 0 . 18 , p<0 . 01 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 14 , η² = 0 . 11 , data not shown ) and was independent of modulation frequency ( factor frequency: p=0 . 19 , η² < 0 . 01; interaction stimulus type × frequency: p=0 . 27 , two-way RM ANOVA , Greenhouse-Geisser corrected , η² < 0 . 01 , n = 14 ) . The temporal precision of ANF input and SBC output to SAM stimulation was estimated by calculating the vector strength ( VS ) at different modulation frequencies . The SBC output exhibited consistently higher VS compared to its ANF input ( Δ = 0 . 06 ± 0 . 04 , p<0 . 001 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 14 , η² = 0 . 09 , Figure 5D , see also Figure 5—source data 1 ) , and decreased for modulation frequencies above 200 Hz ( factor frequency: p<0 . 001 , η² = 0 . 09; interaction signal type × frequency: p<0 . 05 , η² < 0 . 01 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 14 ) . To estimate the degree of modulation of the neural response , the modulation depth was estimated by calculating the standard deviation of the first cycle of the normalized cross-correlation function . Modulation depth was considerably higher at the SBC output ( Δ = 0 . 04 ± 0 . 04 , p<0 . 001 , η² = 0 . 09 ) and decreased with modulation frequency ( factor frequency: p<0 . 001 , η² = 0 . 08; interaction signal type × frequency: p=0 . 12 , η² < 0 . 01 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 14 ) . 10 . 7554/eLife . 19295 . 011Figure 5 . Tone bursts with sinusoidal amplitude modulations ( SAM ) of different modulation frequencies were used to investigate the input-output function under the condition of dynamically altered amplitude profiles . Overall , SAM testing revealed higher temporal precision and reproducibility from ANF input to SBC output . ( A ) The upper panel ( black ) shows the stimulus and the lower panel the dot-raster plot of the discharges of a representative SBC to 200 stimulus presentations with a differentiation between EPSPsucc ( blue ) and EPSPfail ( gray ) . ( B ) Histogram of the discharge activity shown in A . Upper panel: blue = EPSPsucc , orange = ANF input , i . e . EPSPsucc+EPSPfail ) . Lower panel: The EPSPfail is also locked to the SAM , following the ANF input dynamics . ( C ) Period histograms of ANF input ( orange ) and SBC output ( blue ) to increasing modulation frequencies . For comparison , all histograms are centered to the maximum of the ANF input . Gray background indicates the stimulus modulation . ( D ) Trial-to-trial reproducibility and modulation depth were calculated from the cross-correlation between trials . Reproducibility was defined as the peak of the normalized cross correlation and modulation depths as the standard deviation of the first cycle . ( E ) Population data for 14 SBCs . Different measures of temporal precision and trial-to-trial reproducibility all revealed higher accuracy for the SBC output compared to its ANF input: The SBC output showed consistently higher vector strength ( left , p<0 . 001 , two-way RM ANOVA ) , increased modulation depth ( middle left , p<0 . 001 , two-way RM ANOVA ) , higher reproducibility ( middle right , p<0 . 01 , two-way RM ANOVA ) and higher representation of the stimulus envelope ( right , p<0 . 05 , two-way RM ANOVA ) throughout all modulation frequencies . Markers indicate mean ± standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 01110 . 7554/eLife . 19295 . 012Figure 5—source data 1 . Metrics of temporal precision and reproducibility during SAM stimulation . Shown are the vector strength , modulation depth , reproducibility and CorrNorm values for each modulation frequency separately for ANF input and SBC output . ( n = 14 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 012 The neuronal response to a given stimulus can vary between identical stimulus presentations . This trial-to-trial variability was quantified by calculating the within-cell , across-trial crosscorrelations separately for the ANF input and SBC output . The peak height of the crosscorrelation was termed reproducibility ( Joris et al . , 2006 ) . It provides a measure of how repeatable the neural response is across trials , given identical stimulus presentations . If the reproducible features of the response encode stimulus properties , e . g . certain salient events , then an increased reproducibility corresponds to more trustable encoding of stimulus information across trials . The analysis revealed higher reproducibility in the SBC output compared to the ANF input ( Δ = 0 . 4 ± 0 . 25 , p<0 . 001 , η² = 0 . 09 ) and also showed a systematic decrease with increasing modulation frequency ( factor frequency: p<0 . 001 , η² = 0 . 04; interaction signal type × frequency: p<0 . 01 , η² < 0 . 01 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 14 ) . To obtain a better understanding of how precisely the neuronal response reproduces the stimulus envelopes , the delay-adjusted period histograms were correlated to the stimulus envelope resulting in a CorrNorm between 0 . 82 and 1 ( see Materials and methods for explanation ) . The analysis revealed higher CorrNorm for the SBC output compared to the ANF input ( Δ = 0 . 01 ± 0 . 01 , p<0 . 05 , η² = 0 . 02 ) , which for both signal types increased with modulation frequency ( factor frequency: p<0 . 05 , η² = 0 . 08; interaction signal type × frequency: p=0 . 19 , η² < 0 . 01 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 14 ) . These analyses show that the increase in temporal precision observed during pure-tone stimulation is maintained during amplitude-modulated sounds across a wide range of modulation frequencies . In a next step , the modulation of unit discharges to periodic frequency modulations ( SFM ) was explored . For that purpose , the stimulus intensity was fixed at 30–40 dB above the unit’s threshold and the stimulus frequency modulated between one octave below and two octaves above the unit’s CF . This frequency range covers the whole excitatory area as well as the inhibitory sideband . Similar to the SAM stimulation , the SFM resulted in prominent modulations of the units’ firing rates ( Figure 6A ) and increased failure fractions ( Figure 6B ) ( spont = 0 . 34 ± 0 . 25 vs . stim = 0 . 6 ± 0 . 13 , Δ = 0 . 26 0 . 23 , p<0 . 001 , η² = 0 . 29 , data not shown ) . In contrast to SAM stimulation , SFM led to increased failure rates at higher modulation frequencies ( e . g . 0 . 52 ± 0 . 13 at 20 Hz vs . 0 . 65 ± 0 . 12 at 400 Hz modulation frequency , p<0 . 001 , two-way RM ANOVA , Greenhouse-Geisser corrected , η² = 0 . 02 , n = 19 , data not shown ) . For SFM stimulation – same as for SAM - the SBC output showed higher VS compared to their ANF input ( Δ = 0 . 14 ± 0 . 09 , p<0 . 001 , η² = 0 . 21 , see also Figure 6—source data 1 ) ( Figure 6D left ) ( factor frequency: p<0 . 001 , η² = 0 . 48; interaction signal type × frequency: p<0 . 001 , η² = 0 . 03 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 19 ) . Still , overall the VS of ANF input and SBC output decreased with increasing modulation frequency ( Figure 6D left ) . Notably , the VS of the SBC output deteriorated to a lesser degree than the ANF input . At modulation frequencies of 20 Hz , the output VS was not significantly different between ANF input and SBC output ( ANF input = 0 . 61 ± 0 . 06 vs . SBC output = 0 . 63 ± 0 . 1 , Δ = 0 . 04 ± 0 . 07 , p=0 . 17 , two-way RM ANOVA , Bonferroni-adjusted , n = 19 , U1 = 0 . 16 ) . For modulation frequencies of 400 Hz , however , the VS of the SBC output was considerably higher than the ANF input ( ANF input = 0 . 25 ± 0 . 08 vs . SBC output 0 . 4 ± 0 . 09 , Δ = 0 . 15 ± 0 . 06 , p<0 . 001 , two-way RM ANOVA , Bonferroni-adjusted , n = 19 , U1 = 0 . 5 ) . It has to be considered that the interpretation of VS values is difficult when the period histogram of the neuronal response shows multiple peaks ( Figure 6C ) . We , therefore , used a set of additional measures to describe the neuronal response to SFM stimuli when comparing ANF input and SBC output . The modulation depth was considerably higher for the SBC output than the ANF input ( Figure 6D midleft; Δ = 0 . 24 ± 0 . 16 , p<0 . 001 , η² = 0 . 28 ) and strongly depended on the modulation frequency ( factor frequency: p<0 . 001 , η² = 0 . 37; interaction signal type × frequency: p<0 . 001 , η² = 0 . 03 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 19 ) . The same holds for signal reproducibility ( Figure 6D midright; Δ = 1 . 9 ± 1 . 2 , p<0 . 001 , η² = 0 . 38 ) which also showed prominent frequency dependency ( factor frequency: p<0 . 001 , η² = 0 . 26; interaction signal type × frequency: p<0 . 001 , η² = 0 . 02 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 19 ) . Unlike the previous measures , the normalized correlation between SFM stimulus envelope and neural response revealed a lower reproducibility for the SBC output compared to the ANF input ( Figure 6D right; Δ = 0 . 02 ± 0 . 03 , p<0 . 001 , η² = 0 . 06 ) , and the difference also holds with respect to the effect of modulation frequency ( factor frequency: p<0 . 001 , η² = 0 . 43; interaction signal type × frequency: p<0 . 001 , η² = 0 . 03 , two-way RM ANOVA , Greenhouse-Geisser corrected , n = 19 ) . Overall , these data suggest that – same as for the SAM stimuli – the SBC output shows temporally increased precision and higher response reproducibility during SFM stimulation . 10 . 7554/eLife . 19295 . 013Figure 6 . Tone bursts with sinusoidal frequency modulations ( SFM ) of different modulation frequencies were used to investigate the input-output function under the condition of dynamically altered frequency profiles . Overall , SFM testing revealed improved temporal precision and across-trial reproducibility across the ANF-SBC synapse . ( A ) The upper panel ( black ) shows the SFM stimulus with a detail enlargement visualizing the dynamic frequency modulation . The dot-raster plot ( lower panel ) shows the activity of a representative SBC ( CF = 1 . 8 kHz ) to 200 stimulus repetitions with a differentiation between EPSPsucc ( blue ) and EPSPfail ( gray ) . ( B ) Histogram of the discharge activity shown in A . Upper panel: blue = EPSPsucc , gray = EPSPfail , orange = ANF input , i . e . EPSPsucc+EPSPfail . Lower panel: The EPSPfail is also locked to the SFM but showed reduced fine structure compared to the ANF input . ( C ) Period histograms for the same cell as in A and B at different modulation frequencies ( orange = ANF input , blue = SBC output ) . Design of the graph is identical to Figure 5C . Note the multiple peaks of the response in the period histogram . ( D ) Population data for 19 cells: Across all frequencies tested , the SBC output shows increased vector strength ( left; p<0 . 001 , two-way RM ANOVA ) , higher modulation depths ( mid left; p<0 . 001 , two-way RM ANOVA ) , and better across-trial reproducibility ( mid right; p<0 . 001 , two-way RM ANOVA ) compared to its ANF input . The stimulus reproduction ( CorrNorm ) was consistently lower at the SBC level ( right; p<0 . 001 , two-way RM ANOVA ) . Markers indicate mean ± standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 01310 . 7554/eLife . 19295 . 014Figure 6—source data 1 . Metrics of temporal precision and reproducibility during SFM stimulation . Shown are the vector strength , modulation depth , reproducibility and CorrNorm values for each modulation frequency separately for ANF input and SBC output . ( n = 19 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 014 Above , we demonstrated an improvement in temporal precision and reproducibility in response to SAM and SFM acoustic stimuli . In natural environments , however , the auditory system has to cope with simultaneous dynamic changes in both frequency and amplitude embedded in ambient background noise . To mimic such conditions , while preserving the possibility for a quantifying data analysis , dynamic acoustic stimuli composed of gamma-tones randomly placed in the spectrogram were used ( Figure 7A top , randomized gamma-tone sequence , RGS , see Materials and methods for details ) . The SBC activity can then be characterized using spectrotemporal receptive fields ( STRFs ) . In the present context , STRFs can also be used to quantify the spectrotemporal transformation of response properties across the ANF-SBC synapse , since the respective analysis can be performed for both the ANF input and SBC output . SBC activity was recorded while presenting 20–30 repetitions of identically structured RGS sequences of 30 s duration each . In Figure 7A , the upper panel shows a 200 ms-section of an RGS stimulus used for stimulation of an SBC with a CF of 2 kHz; the middle panel depicts the spike raster plot differentiating between EPSPsucc and EPSPfail . 10 . 7554/eLife . 19295 . 015Figure 7 . Input-output comparison of spectrotemporal receptive fields ( STRF ) indicates minor spectral sharpening and confirms broad , slow inhibitory action . ( A ) Top panel: Randomized gamma-tone sequence ( RGS , scaling of red color indicates stimulus levels with a maximum of 70 dB SPL , see Materials and methods for stimulus details ) were used to estimate STRFs of SBC output and its ANF input . The RGS spanned one octave below and two octaves above the unit’s CF; in the present example 2 kHz . Second panel: Dot raster of discharges of an exemplary SBC evoked by 30 repetitive RGS presentations ( blue = EPSPsucc , gray = EPSPfail ) . Third panel: PSTH of the recording shown above; the graph differentiates between the total of the ANF input ( EPSPfail + EPSPsucc , orange ) and SBC output ( SBC APs , blue ) . Fourth panel: From the same recording the histogram of the EPSPs that fail to trigger an SBC AP ( EPSPfail , gray ) . Note that EPSPs that elicited APs tended to be more prominent at the onset of excitatory response components . ( B ) STRF of the unit shown in A . Upper panel: Sketch of the two signal types , i . e . the totality of all EPSPs were considered to indicate the ANF input ( orange ) , while EPSPs that generate an AP defined the SBC output ( blue ) . Middle panel: corresponding STRFs . Note that there are clearly delineated areas of increased activity 2–3 ms after response-evoking stimulus components ( red ) which are distinct from areas with reduced activity . The spectrotemporal shape of the modulation at the ANF-SBC junction was quantified by the averaged difference-STRF . The STRFs of both ANF ( left ) and SBC ( right ) were computed separately and then subtracted ( bottom panel ) . Relative temporal alignment was achieved by time-locking both ANF input activity and SBC output on the respective timing of maximum EPSP slope . The difference reveals changes in stimulus responsiveness in spectrotemporal coordinates . Negative values indicate a reduction in responsiveness , most likely caused by local inhibition . ( C–G ) Population data for all recorded SBCs ( n = 34 ) ; triangles in the graphs indicate the respective values of the unit shown in A and B . ( C ) Stimulus-driven excitation was significantly reduced from the ANF input to SBC output , measured as the sum of all positive STRF bins ( p<0 . 001 , Wilcoxon signed rank test ) . ( D ) Stimulus-driven inhibition was significantly increased , measured by the negative sum of all negative STRF bins from the ANF input to SBC output ( p<0 . 001 , Wilcoxon signed rank test ) . ( E ) Spectral precision improved at the ANF-SBC junction , indicated by a reduced spectral half-width of the excitation ( p<0 . 001 Wilcoxon signed rank test ) . ( F ) Temporal precision , estimated as the temporal half-width , was not changed between ANF input and SBC output ( Wilcoxon signed rank test , p=0 . 16 ) . ( G ) The average difference-STRF ( n = 34 cells ) exhibited a prominent and broad ( > 2 octaves ) reduction around CF , which remained effective for ~10 ms ( black line indicates significant deviation , adjusted for a false discovery rate < 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 01510 . 7554/eLife . 19295 . 016Figure 7—source data 1 . Metrics of STRFs obtained with RGS stimulation . Shown are the excitation , inhibition , spectral half width ( in octaves ) , and temporal half width ( in ms ) obtained from the STRFs ( n = 34 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 016 The neural response to gamma-tones of both the ANF input and SBC output were temporally structured ( Figure 7A second and third panel ) . Failures of signal transmission ( Figure 7A gray dots in second panel and histogram in fourth panel ) were found to be increased following sequences of activation , suggesting a long-lasting action of inhibition ( e . g . Figure 7A fourth panel , where EPSPfail shows a considerable increase in the responses to the second of the first two peaks ) . Separate STRFs were computed for the ANF input ( Figure 7B middle left ) and the SBC output ( Figure 7B middle right ) . As expected , both STRFs showed common features , e . g . frequency domain of excitation above and below the unit’s CF ( in the present example 2 kHz , estimated from single tone tunings ) and also the response latency ( here 2 . 5 ms ) . Importantly , the reduction in the responses establishing a high-frequency sideband was already present in the ANF input to the SBCs and did not become more pronounced in SBC output . Since the ANF activity is not affected by acoustically evoked inhibition , the respective frequency-specific reduction observed in the ANF input to the SBCs likely reflects mechanical interactions in the cochlea , previously described as two-tone suppression ( Engebretson and Eldredge , 1968; Sachs and Kiang , 1968; Sellick and Russell , 1979 ) . To evaluate the signal processing at the ANF-SBC junction , the two STRFs were subtracted from each other after normalizing each by its standard deviation ( to compensate for overall firing rate differences , see Materials and methods for details; Figure 7B bottom ) . This normalization allows a quantification of changes in the tuning shape . The increase in EPSPfail in the STRF of the SBC output manifests itself as a broad field of negativity in the difference-STRF around CF extending up to ~10 ms after the onset of the effective signal components around 2 kHz . The respective differences between ANF input and SBC output were quantified in all recorded SBCs ( n = 34 ) separately for the positive ( red ) and negative ( blue ) regions in the STRF ( corresponding to influential spectrotemporal locations in the stimulus prior to the response ) . Summing all the positive regions revealed a significant reduction from ANFs to SBCs ( ANF = 0 . 38 [0 . 37 , 0 . 41] vs . SBC = 0 . 33 [0 . 31 , 0 . 36] , Δ = 0 . 05 [0 . 03 , 0 . 07] , p<0 . 001 , Wilcoxon signed rank test , n = 34 , U1 = 0 . 16 , Figure 7C , see also Figure 7—source data 1 ) . Similarly , the summed negative region in the SBC output was significantly larger in magnitude than the ANF input ( ANF = 0 . 18 [0 . 15 , 0 . 23] vs . SBC = 0 . 25 [0 . 18 , 0 . 29] , Δ = 0 . 04 [0 . 02 , 0 . 07] , p<0 . 001 , Wilcoxon signed rank test , n = 34 , U1 = 0 . 1 , Figure 7D ) . Together , this suggests an inhibitory influence acting broadly with respect to the neuron’s tuning . We further quantified changes in the shape of the main excitatory peak . The spectral tuning , measured as half-width of the excitatory region , was reduced at the SBC output compared to ANF input , suggesting a spectrally sharper tuning at the SBC output ( ANF = 1 . 2 [0 . 9 , 1 . 4] octaves vs . SBC = 1 [0 . 8 , 1 . 2] octaves , Δ = 0 . 1 [0 , 0 . 2] , p<0 . 001 , Wilcoxon signed rank test , n = 34 , U1 = 0 . 01 , Figure 7E ) . Temporal precision , measured correspondingly as the half-width of the excitatory region , was somewhat higher for the SBC output , but did not reach statistical significance ( ANF = 2 . 2 [1 . 8 , 3 . 4] ms vs . SBC = 2 . 0 [1 . 8 , 3 . 1] ms , Δ = 0 . 08 [-0 . 1 , 0 . 18] , p=0 . 16 , Wilcoxon signed rank test , n = 34 , U1 = 0 . 03 , Figure 7F ) . The overall shape of the difference-STRF of all units was studied by aligning all STRFs to the peak excitation and averaging them ( Figure 7G ) . As mentioned above , the reduction in the STRF outlasted the excitatory region for up to ~10 ms relative to the onset of the excitatory signal component . Significance was assessed point-wise using t-tests , followed by the Benjamini and Hochberg ( 1995 ) algorithm for multiple comparisons applied to the p-values of the t-tests . At a false discovery rate of 0 . 01 , the gray line shows the region of significant deviation . Overall , the STRF analysis confirmed the presence of inhibition co-tuned with excitation , exhibiting a longer-lasting time-course of about 10 ms with respect to the onset of the excitatory signal component . Consequently , also under dynamic broadband stimulation , inhibition is confirmed to only marginally act above or below the neuron’s excitatory receptive field and results in only a slight spectral sharpening of the SBC output . Next , we addressed the functional consequences of this co-tuned , prolonged inhibition . The functional consequences of co-tuned inhibition appear less evident than those of narrow , sideband inhibition . The latter can diversely shape the response properties , by reducing responses only for small , off-CF regions . Co-tuned inhibition , on the other hand , has been proposed to contribute to a precisely timed balancing of excitation to keep neurons within their dynamic ranges ( Renart et al . , 2010 ) . To test for such a mechanism , we quantified properties of the SBC output in comparison to the ANF input with respect to the temporal sparsity of the response and reproducibility across trials . Efficient neural codes have been proposed to show high sparsity , i . e . respond only rarely but then with high firing activity ( Field , 1994 ) . Again , the RGS stimulus was used to test the effect of acoustically evoked inhibition under complex acoustic conditions . The results yielded reduced mean firing rates of the SBC output compared to the ANF input ( Figure 8Ai , data from an exemplary SBC ) and increased sparsity in 28/32 cells ( units above line of equality , Figure 8Aii ) . Sparsity was calculated by relating the variance of the neuronal response to its mean firing rate . The population analysis revealed significantly larger temporal sparsity in the SBC output than in its ANF input ( ANF = 0 . 22 ± 0 . 08 vs . SBC = 0 . 31 ± 0 . 13 , Δ = 0 . 09 ± 0 . 08 , p<0 . 001 , paired t-test , n = 34 , U1 = 0 . 15 , Figure 8Aiii , see also Figure 8–source data ) . Sparsity was calculated by relating the variance of the neuronal response to its mean firing rate ( Rolls and Tovee , 1995; Willmore and Tolhurst , 2001 ) , but other measures for sparsity yielded qualitatively similar results ( see Materials and methods and Supporting Figure 8 ) . 10 . 7554/eLife . 19295 . 017Figure 8 . Inhibition renders the SBC responses sparse and increases across-trial reproducibility . ( A ) ( i ) Representative recording during RGS stimulation ( 2 s-section displayed ) shows significantly sparser SBC output activity ( blue ) than the ANF input activity ( orange ) . Marks on the right indicate the mean firing rate for ANF input and SBC output . ( ii ) Population data for all recorded SBCs ( n = 34 ) . Quantification of sparsity as the variance of the normalized firing rates shows that this relation holds for almost all units ( dots above line of equality; red mark indicates representative unit on the left ) and ( iii ) results in highly significant input-output differences ( p<0 . 001 , Wilcoxon signed rank test; triangle indicates the representative unit on the left ) . ( iv ) Blocking glycinergic inhibition in vivo by strychnine ( n = 12 ) deteriorated the improved sparsity of the SBC output and rendered it similar to the ANF input ( p<0 . 01 , Wilcoxon signed rank test ) , while the ANF input remained unchanged . ( B ) The reproducibility of the response improved from ANF input to SBC output . Reproducibility was calculated as the time-aligned correlation between the neuron’s responses to identical stimulus trials . High reproducibility indicates that the neural response is more constant across trials . ( i ) In the representative unit , higher reproducibility is seen for the SBC output ( blue ) compared to the ANF input ( orange ) . ( ii ) Population data ( n = 34 ) shows that the same relation holds for almost all units ( data point marked in red indicates the unit shown on the left ) , and ( iii ) the statistical analysis yielded a high significant input-output difference ( p<0 . 001 , Wilcoxon signed rank; triangles indicate the respective values from the exemplary unit ) . ( iv ) Application of strychnine impoverished reproducibility in the SBC output ( light blue ) significantly compared to the control condition ( dark blue; p<0 . 01 , Wilcoxon signed rank test ) . The reproducibility of the ANF input ( orange ) was not influenced by blocking the inhibition ( light orange ) . ( C ) The temporal dispersion for repetitive acoustic stimulation decreased from the ANF input to SBC output . The temporal dispersion was quantified as the half-width of the cross-correlation within each signal across trials ( i ) . Population analysis showed improved temporal precision , i . e . reduced half-width/dispersion in the SBC output compared to the ANF input in most of the tested cells ( ii , iii , same color coding as above , p<0 . 01 , Wilcoxon signed rank test ) . As above , blocking inhibition increases temporal dispersion of the SBC output to the level of the ANF input ( iv , p<0 . 01 , Wilcoxon signed rank test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 01710 . 7554/eLife . 19295 . 018Figure 8—source data 1 . Sparsity , reproducibility and temporal dispersion for ANF input and SBC output . Shown are the sparsity , reproducibility , and temporal dispersion ( in ms ) separately for ANF input and SBC output for both control condition ( n = 34 ) and during the block of glycinergic inhibition ( n = 12 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 01810 . 7554/eLife . 19295 . 019Figure 8—figure supplement 1 . Alternative measures of temporal sparsity lead to consistent results . Temporal sparsity can be quantified with different measures . While Sparsity was quantified using a popular variance-related measure ( Rolls and Tovee , 1995 ) in Figure 8 , other measures lead to consistent results ( see Materials and methods for details ) . ( A ) A classical measure – kurtosis of the firing rate distribution over time – is evaluated ( Field , 1994 ) , indicated as K in Ai for both ANF and SBC ) . As before the sparsity of SBCs is higher than for ANFs , for the large majority of cells , and thus highly significant on the population level ( p<0 . 001 , Wilcoxon signed rank test; triangle indicates example unit on the left ) . All conventions as in Figure 8 . ( B ) We defined an easily interpretable measure , the ‘Close-to-Silence-Index’ ( CSI ) , which simply quantifies which fraction of firing rate bins are above a certain threshold firing rate ( here: Threshold = 15 Hz was chosen ) . As for the other two measures , sparsity measured by CSI was significantly larger for SBCs than for ANF ( p<0 . 001 , Wilcoxon signed rank test ) . All conventions as in Figure 8 . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 019 The reproducibility of the temporal response pattern was quantified by computing across-trial cross-correlations ( Figure 8B ) . For this analysis , the obtained correlograms were divided by the product of the individual firing rates , rendering the results independent of absolute firing rates . Reproducibility was then calculated as the peak of the correlograms measured at 0 ms lag ( Figure 8Bi ) . In 97% ( 33/34 ) of all recorded cells , the SBC output exhibited a higher level of reproducibility ( units above line of equality , Figure 8Bii ) . Also , the population analysis yielded a significantly higher reproducibility of the SBC output than the ANF input ( ANF = 0 . 46 [0 . 37 , 0 . 65] vs . SBC = 0 . 83 [0 . 48 , 1 . 3] , Δ = 0 . 35 [0 . 12 , 0 . 73] , p<0 . 001 , Wilcoxon signed rank test , n = 34 , U1 = 0 . 18 , Figure 8Biii ) . To estimate the temporal precision across trials , the temporal dispersion was quantified as the half-width of the across-trial cross-correlation ( Figure 8Ci ) . Temporal precision across trials improved from ANF input to SBC output in two-thirds of the recorded SBCs ( 24/34 , units below line of equality , Figure 8Cii ) . Still , population analysis yielded a significant improvement in temporal precision ( temporal dispersion: ANF = 7 . 04 [5 . 25 , 7 . 93] ms vs SBC = 4 . 96 [3 . 47 , 6 . 9] ms , Δ = 1 . 1 [0 , 1 . 98] ms , p<0 . 01 , Wilcoxon signed rank test , n = 34 , U1 = 0 . 03 , Figure 8iii ) . Response reproducibility across trials renders the response more identifiable for downstream processing stages which rely on precisely timed inputs . The increased sparsity reduces the energy expense by removing spikes which reflect the constant part of the response . Temporal precision of encoding also improved , although this was only observed in about 70% of the cells . Finally , we directly tested whether the observed changes in response properties were indeed caused by acoustically evoked , glycinergic inhibition . Another set of 12 units were recorded under RGS stimulation , and glycinergic inhibition was blocked by iontophoretic application of strychnine ( Figure 8Aiv , Biv , Civ ) . Like in the experiments reported above , the analysis differentiated between the ANF input to SBCs and the respective SBC output . Under control conditions , cells showed the above-described increase in sparsity and reproducibility at the ANF-to-SBC transition . Blocking the glycinergic inhibition resulted in decreased sparsity of the SBC output ( Figure 8Aiv; control = 0 . 34 ± 0 . 1 vs . strychnine = 0 . 27 ± 0 . 07 , Δ = 0 . 07 ± 0 . 06 , p<0 . 01 , two-way RM ANOVA , Bonferroni–adjusted , n = 12 , U1 = 0 . 33 ) . Also , the block of inhibition caused a decrease in response reproducibility ( Figure 8Biv; control = 0 . 79 ± 0 . 39 vs . strychnine = 0 . 53 ± 0 . 19 , Δ = 0 . 26 ± 0 . 23 , p<0 . 05 , two-way RM ANOVA , Bonferroni-adjusted , n = 12 , U1 = 0 . 21 ) and an increase in temporal dispersion at the SBC output ( Figure 8Civ; control = 5 . 6 ± 1 . 6 ms vs . strychnine = 7 . 5 ± 1 . 7 ms , Δ = 1 . 9 ± 1 . 6 ms , p<0 . 01 , two-way RM ANOVA , Bonferroni-adjusted , n = 12 , U1 = 0 . 21 ) . In summary , the block of inhibition reduced the observed improvements from the ANF input to the SBC output , rendering both more similar . Importantly , the ANF input was not affected by the block of inhibition ( Figure 8Aiv , Biv , Civ ) ( sparsity: control = 0 . 18 ± 0 . 05 vs . strychnine = 0 . 18 ± 0 . 05 , Δ = 0 ± 0 . 01 , p=0 . 5 , U1 = 0 . 13; reproducibility: control = 0 . 32 ± 0 . 11 vs . strychnine = 0 . 32 ± 0 . 1 , Δ = 0 ± 0 . 03 , p=0 . 8 , U1 = 0 . 08; temporal dispersion: control = 8 . 5 ± 0 . 9 ms vs strychnine = 8 . 5 ± 1 ms , Δ = 0 ± 0 . 6 ms , p=0 . 99 , U1 = 0 . 08 , n = 12 , two-way RM ANOVA , Bonferroni-adjusted ) . In comparison with the pre-post data , the pharmacological dataset shows smaller variability across cells , which may be due to the lack of outliers in the latter , smaller dataset . In summary , these data directly show that glycinergic inhibition is a critical factor for the observed improvements from ANF input to the SBC output during complex acoustic stimulation . SBCs have been shown to be influenced by both hyperpolarizing and shunting effects of inhibition ( Kuenzel et al . , 2011 , 2015; Nerlich et al . , 2014a ) . While hyperpolarization has been attributed to a subtractive effect on firing rates ( Doiron et al . , 2001; Silver , 2010 ) , shunting inhibition has mainly divisive effects ( Mitchell and Silver , 2003; Prescott and De Koninck , 2003; Capaday and van Vreeswijk , 2006; Ly and Doiron , 2009 ) . We investigated the functional effect of either type on the response via a simple simulation: either a fixed fraction ( divisive , relative to the instantaneous firing rate ) or a fixed number ( subtractive ) of spikes was removed from the ANF spike trains , matching the experimentally observed SBC output rates . Purely divisive inhibition , corresponding to a scaling of the PSTH , does not improve sparsity , reproducibility or temporal precision ( Figure 9 , purple , Δsparsity = 0 ± 0 , p=0 . 99 , Δreproducibility = 0 . 01 ± 0 . 02 , p=0 . 25 , Δtemporal dispersion = 0 . 05 ± 0 . 41 , p=0 . 89 , n = 34 , one-way RM ANOVA , Bonferroni-adjusted , see also Figure 9—source data 1 ) . On the other hand , a purely subtractive inhibition matches the qualitative effects in the data well , i . e . improves all three properties ( Figure 9 , green , Δsparsity = 0 . 24 ± 0 . 08 , p<0 . 001 , Δreproducibility = 1 . 06 ± 0 . 73 , p<0 . 001 , Δtemporal dispersion = 1 . 8 ± 1 . 3 ms , p<0 . 001 , n = 34 , one-way RM ANOVA , Bonferroni-adjusted ) . Quantitatively , the simulated subtractive inhibition leads to larger improvements in sparsity and reproducibility than observed in the experimental data ( Figure 9 , blue , sparsity: data = 0 . 31 ± 0 . 13 vs . subtractive inhibition = 0 . 46 ± 0 . 12 , Δ = 0 . 15 ± 0 . 07 , p<0 . 001; reproducibility: data = 0 . 97 ± 0 . 61 vs . subtractive inhibition = 1 . 59 ± 0 . 9 , Δ = 0 . 62 ± 0 . 6 , p<0 . 001 , n = 34 , one-way RM ANOVA , Bonferroni-adjusted ) . We verified that similar relations hold for the SAM and SFM stimulation and the measures used in their analyses ( Figure 9 , Figure 9—figure supplements 1 and 2 , respectively ) . A temporally unspecific , subtractive effect of inhibition might , therefore , be sufficient to explain the improvement in sparsity , reproducibility , and temporal precision . When combined with the divisive , co-tuned gain control , this improvement generalizes to a wide range of stimulus levels . 10 . 7554/eLife . 19295 . 020Figure 9 . Subtractive inhibition , but not divisive inhibition can account for the improvement in sparsity , reproducibility , and temporal precision . ( A ) In response to the RGS stimulus , the SBC output ( blue ) showed a consistent increase in sparsity ( left ) , reproducibility ( middle ) and decreased temporal dispersion ( right ) . The simulated subtractive inhibition ( green ) showed similar improvements as the experimental data , while divisive inhibition ( purple ) had no effect on sparsity , reproducibility , and temporal dispersion . ( B ) These relations are also reflected in the population data , with significant changes in both the experimental data and the simulated subtractive inhibition ( p<0 . 001 , one-way RM ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 02010 . 7554/eLife . 19295 . 021Figure 9—source data 1 . Simulation of divisive and subtractive inhibition . Shown are the values of sparsity , reproducibility and temporal dispersion ( in ms ) obtained from the simulation of pure subtractive or pure divisive inhibition . The data set also contains the experimentally obtained data separately for ANF input and SBC output . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 02110 . 7554/eLife . 19295 . 022Figure 9—figure supplement 1 . Subtractive inhibition improves temporal precision and trial-to-trial reproducibility during SAM stimulation . To test , if the subtractive or divisive effect of inhibition are sufficient to explain the observed increase in temporal precision and reproducibility of the SBC output , the ANF input was compared to the experimental SBC output data and simulated SBC outputs caused by either subtractive or divisive inhibition . For the simulated SBC output , the ANF input spikes were removed to match the experimentally observed failure fraction . ( A ) Vector strength , modulation depth , reproducibility stimulus reproduction ( CorrNorm ) , and sparsity of the ANF input was compared to the experimental data ( blue ) and the simulated SBC output influenced by either divisive ( purple ) or subtractive ( green ) inhibition Pure divisive inhibition had no effect on temporal precision or reproducibility ( dots on line of equality , gray ) . Pure subtractive inhibition matched the experimental data well and showed an increase in vector strength , modulation depths , reproducibility , and sparsity of the SBC output ( dots above line of equality ) . ( B ) Population data showing the difference between ANF input and SBC output for the experimental ( blue ) and simulated divisive ( purple ) and subtractive inhibition ( green ) for different modulation frequencies . The experimental SBC output showed increased vector strength , modulation depth , reproducibility , and sparsity for all modulation frequencies . This change was also observed under simulated divisive inhibition , but absent for pure subtractive inhibition . Markers indicate mean ± standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 02210 . 7554/eLife . 19295 . 023Figure 9—figure supplement 2 . Subtractive inhibition , but not divisive inhibition can account for the improved temporal precision and reproducibility during SFM stimulation . Similar to the SAM analysis , the effect of an increased failure fraction during SFM stimulation was simulated as purely divisive or subtractive inhibition and compared to the experimental data . ( A ) Pooled data across all modulation frequencies . The experimental data showed an increase in vector strength , modulation depths , reproducibility , and sparsity of the SBC output ( blue dots above line of equality , gray ) . This change was present only for the simulated subtractive inhibition but absent for pure divisive inhibition . ( B ) Population data showing the difference between ANF input and SBC output for the experimental ( blue ) and simulated divisive ( purple ) and subtractive inhibition ( green ) for different modulation frequencies . The experimental SBC output showed increased vector strength , modulation depth and reproducibility for all modulation frequencies . This change was well matched by pure subtractive inhibition , but absent at pure divisive inhibition . The stimulus reproduction ( CorrNorm ) decreased in the experimental SBC output and the simulated subtractive inhibition but remained unchanged for pure divisive inhibition . Markers indicate mean ± standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 19295 . 023 The endbulb synapse depresses considerably during high-frequency firing ( Bellingham and Walmsley , 1999; Wang and Manis , 2008; Yang and Xu-Friedman , 2008 ) despite the large size of the presynaptic synaptic terminal and the reliable , suprathreshold excitation observed in slice recordings . The present in vivo recordings showed that the increased failure fraction during acoustic stimulation cannot be explained by synaptic depression alone . This was evidenced by an analysis of EPSP thresholds and furthermore confirmed by iontophoretic application of a glycine receptor agonist and antagonist . In conclusion , the elevation of the EPSP threshold has proven to be a reliable indicator for inhibitory action , leading to an increased failure fraction . These data are consistent with previous in vivo studies ( Kuenzel et al . , 2011 ) , as well as with slice and model studies demonstrating that an increase in inhibitory conductance can elevate threshold EPSP in bushy cells ( Xie and Manis , 2013; Kuenzel et al . , 2015 ) . In summary , the endbulb of Held–SBC synapse seems to operate close to AP threshold and shows variable reliability which is strongly influenced by acoustically evoked inhibition ( see also Kopp-Scheinpflug et al . , 2002; Kuenzel et al . , 2011; Keine and Rübsamen , 2015 ) . While the observed frequency response areas are consistent with previous reports , we did not observe two distinct types of inhibition as reported earlier , i . e . broadband vs . on-CF inhibition ( Caspary et al . , 1994; Kopp-Scheinpflug et al . , 2002 ) . Instead , the present data suggest that inhibition at SBCs is broadband and on-CF . SBCs receive inhibitory inputs both on their somata and dendrites ( Gómez-Nieto and Rubio , 2009 ) . Both glycine and GABA receptors were shown to be present , with the latter playing a secondary role as demonstrated in slice experiments ( Nerlich et al . , 2014a , 2014b ) . Therefore , we focused on the modulation of glycinergic inhibition . The applied dose was equated to match the acoustically evoked level of AP failures , keeping inhibition in the physiologically relevant range . This cautious approach will tend to underestimate the in vivo effect of glycine since the block of glycine receptors by local application of strychnine might be incomplete . Consistent with the slice data , the lack of threshold EPSP elevation during the block of glycinergic inhibition suggests only a minor influence of the GABAergic component during tone burst stimulation . The GABAergic inhibition might have an additional modulatory function or may only be activated during periods of high activity , as has been suggested for the glycinergic inhibition in the bird’s nucleus magnocellularis ( Fischl et al . , 2014 ) , the avian homolog of the AVCN . Overall , we find glycine to have a substantial influence in shaping transmission at the SBC junction . However , the increased SBC output rates during block of glycinergic inhibition might increase the influence of spike depression ( Lorteije et al . , 2009; Kuenzel et al . , 2011 ) and other factors such as spike threshold adaptation have to be taken into account ( Fontaine et al . , 2014; Huang et al . , 2016 ) . For a broad range of acoustic stimuli , we observed a consistently sparser and more reproducible response in the SBC output compared to the ANF input . Can a simple inhibition achieve these changes in signal representation ? Glycinergic inhibition has previously been demonstrated to be neither purely subtractive nor purely divisive ( Kuenzel et al . , 2011; 2015; Nerlich et al . , 2014b ) and act with a short delay of ~3 ms on time scales of ~10–15 ms . These properties may be sufficient for increasing sparsity and reproducibility under the assumption that large deviations of firing rate are the consequence of stimulus-elicited , high firing probabilities rather than noise ( see Figure 9 ) . The latter would be temporally unrelated to the stimulus , and its transmission would reduce the reproducibility , and probably also the usefulness of the transmitted information for further processing . In SBCs , this would translate to multiple , closely timed spikes for an individual input or across multiple excitatory inputs . The partially subtractive glycinergic inhibition would be strongly triggered by large instantaneous firing rates , and weaken subsequent inputs which do not occur closely to other inputs . On the PSTH level , a subtractive reduction thus almost inevitably increases sparsity ( see Figure 9 ) . Reproducibility will also be increased if the average temporal precision of high peaks is greater than the bulk of spikes at lower firing rates . Divisive inhibition typically leaves sparsity and reproducibility unchanged ( Figure 9 ) , although the influence on sparsity will partially depend on the measure used . Due to the inhibitory/excitatory co-modulation with level , the enhancement in sparsity and reproducibility can extend over a wide range of levels . These considerations do not rule out additional enhancements of temporal precision , via additional excitatory inputs , however , these would be influenced similarly by the inhibition . The considerations above are challenging to study since precise temporal control over multiple inputs would be required . Generally , if the information in the response is largely maintained , an increase in temporal sparsity can be advantageous for mainly three reasons . First , the information per spike is increased ( Barlow , 1972 , 2012; Olshausen and Field , 2004 ) . This is relevant for downstream cells , who can then combine information more efficiently with other inputs , to achieve higher precision , e . g . for estimating interaural time delays . Second , since the SBC’s response occurs on a reduced plateau of unmodulated firing than the ANF’s , changes in firing rate will lead to larger relative changes in firing rate , which should improve their detection in the target neurons . Avoiding saturating or strongly adapting postsynaptic responses also supports this improvement in change detectability ( Mitchison and Durbin , 1989; Graham and Willshaw , 1997 ) . Third , a reduction in the number of spikes reduces the energetic load on the system ( Levy and Baxter , 1996; Baddeley et al . , 1997; Attwell and Laughlin , 2001; Olshausen and Field , 2004; Graham and Field , 2007 ) . In the present case the failures of transmission may , therefore , more appropriately be contrasted with the spikes selected for transmission . SBCs provide the indispensable temporally precise excitatory inputs to the interaural time difference based sound localization in the MSO . Both , physiological and modeling studies suggest the neurons in the MSO act as coincidence detectors by primarily relying on the precise spike timing of binaural inputs ( Goldberg and Brown , 1969; Yin and Chan , 1990; Franken et al . , 2014 ) . While sparsity has predominantly been advocated as an advantageous coding principle due to its reduced energy demand ( Levy and Baxter , 1996; Graham and Field , 2007 ) , in the MSO there might be a different justification for sparsity: With the MSO neurons acting as coincidence detectors ( Couchman et al . , 2010; Plauška et al . , 2016 ) , sparse excitatory inputs would lead to an improved signal-to-noise ratio of the correlation , i . e . of its peak in relation to the common floor of correlation . Presently , this is shown by the reproducibility measure across trials , which – when just considering signal processing – can be regarded as the correlation between the activity of independent neurons from both sides ( Joris , 2003 ) . Equally important , the effective gain control achieved through the acoustically evoked inhibition will allow the MSO to retain a similar level of binaural sensitivity across a wide range of sound levels and stimuli . Importantly , this sparsening and increase in reproducibility were observed for the wide range of stimuli presently tested . The observed increase in reproducibility in SBC output compared to the ANF input is consistent with previously reported population data . Studies comparing ANF and AVCN neurons ( most likely spherical or globular bushy cells ) in the cat reported increased reproducibility in neurons of the AVCN ( Louage et al . , 2005; van der Heijden et al . , 2011 ) . Here , we directly studied the increase across trials within single ANF-SBC junctions and not across cells . We conjecture that reproducibility across cells always increases if both cells increase their reproducibility individually . However , this can only be assessed with the peak of the cross-correlation , if the cells are matched in their tuning , a condition which is typically assumed for bilateral inputs into the MSO . For non-matched cells , reproducibility should increase but would have to be assessed with a different measure , which measures the set of responses to a given stimulus ( for example Mutual Information ) . Presently , we show that this increase is directly achieved at individual ANF-SBC junctions through the postsynaptic interaction of acoustically evoked excitation and inhibition . Also , during both narrowband and broadband stimulation , the block of inhibition resulted in a considerable decrease in reproducibility , sparsity , and temporal precision compared to the control condition ( Figure 8 ) . Hence , our results show that acoustically evoked inhibition plays a major role in shaping the SBC response . Most models of pitch perception are based on or related to an autocorrelation of neuronal activity and rely on temporal precision and reproducibility ( Licklider , 1951; Meddis and Hewitt , 1991; Cariani and Delgutte , 1996; Yost , 1996; de Cheveigné , 1998; Denham , 2005; Joris , 2016 ) . The improvement of both characteristics at the ANF-SBC synapse might support the neuronal processing of pitch , but experimental evidence is lacking . Other transmitters such as acetylcholine ( Fujino and Oertel , 2001; Goyer et al . , 2016 ) or norepinephrine ( Kössl and Vater , 1989; Rothman and Manis , 2003 ) have modulatory effects on the SBC activity , but their effect on precisely-timed signal processing is not yet fully understood . The output of SBCs is also relevant for general processing of sounds , i . e . their representation for later analysis , e . g . auditory recognition , a process distinct from sound localization ( Clarke et al . , 1998; Maeder et al . , 2001 ) . The increase in reproducibility might be beneficial in this part of the pathway as well . On the other hand , sparsity may have adverse effects , if the overall level is unavailable or if low level sound information in the ANF response is not represented anymore in the SBC response . Natural stimuli tend to depart from the classical laboratory stimuli , in being spectrotemporally broad and diverse . We approximated this condition using the RGS stimulus in combination with distinct STRF estimation of both the pre- and the postsynaptic activity . STRFs were first introduced to study midbrain neurons in the grass frog ( Aertsen and Johannesma , 1980; Aertsen et al . , 1980 , 1981 ) , and have since been an essential tool for auditory neuroscience along many stations of auditory system , ranging from the auditory nerve ( Kim and Young , 1994 ) and the MNTB ( Englitz et al . , 2010 ) to the inferior colliculus ( Escabi and Schreiner , 2002 ) , and the auditory cortex ( Kowalski et al . , 1996; David et al . , 2012 ) . Recent developments in STRF estimation ( Theunissen et al . , 2001 ) make them a versatile tool to study the combined spectrotemporal stimulus selectivity of neurons for a wide range of acoustic stimuli . The present extension to pre- and postsynaptic activity is unique and provides a direct estimate of the spectrotemporal response modification occurring at a single endbulb synapse . It reveals inhibition to be co-tuned with excitation , but outlasting the latter , which – in total – leads to a slightly improved excitatory response precision . This response profile directly confirms the spectral properties gained from the pure-tone tuning curves and is in agreement with findings from of our previous studies on inhibition at the SBC ( Keine and Rübsamen , 2015 ) . The pre-post STRF analysis could provide a generally applicable tool for the investigation of a wider range of modulations of signal processing at the giant synapses in both MNTB and AVCN , for example the functional differences on synaptic transmission under the stimulation with natural acoustic stimuli or removal/application of specific neurotransmitters , e . g . GABA ( Nerlich et al . , 2014b ) or neuromodulators , e . g . acetylcholine ( Goyer et al . , 2016 ) . The RGS stimulus allows a robust estimation of STRFs while keeping the spectrum sparse ( unlike dense stimuli as the TORC stimulus , Klein et al . , 2000 ) . A sparse spectrum is advantageous for the present study of the potentially long-lasting inhibition ( Nerlich et al . , 2014b ) since it prevents a continuous activation of inhibitory inputs causing saturation . It remains to be addressed , whether the RGS is a sufficient model for natural stimuli , or whether the natural statistics lead to a more specific activation of the inhibition arriving at the SBCs . Overall level-dependence of STRF estimation was beyond the scope of this study and not investigated here in more detail . While the RGS stimulus contains different sound levels at different times and frequencies ( via the randomized placement and shape of each gamma-tone ) , the overall , average sound level was kept constant . We predict that SBC STRFs will exhibit a greater robustness to changes in level compared to ANF STRFs , based on the gain-modulating inhibitory input ( Figure 3 ) . Several nuclei have been suggested to provide the inhibitory inputs to SBCs . The present data suggest the inhibitory source to feature a broad , symmetrically shaped tuning , consistent with an integration over a wide set of primary tuned inhibitory cells . The integration would have to be weighted by the distance to the postsynaptic CF , in order to achieve the symmetry in inhibitory modulation . In a recent study , Campagnola and Manis ( 2014 ) showed directly that bushy cells receive symmetric inhibition from within the CN . Further , the ~1 ms delay of the onset of acoustically evoked inhibition compared to the onset of excitation ( Kuenzel et al . , 2011; Keine and Rübsamen , 2015 ) suggests a single additional synaptic relay . Together , both the broadly tuned D-stellate cells in the AVCN and the tuberculoventral cells in the DCN are candidate sources ( Wickesberg and Oertel , 1990; Saint Marie et al . , 1991; Campagnola and Manis , 2014 ) , as well as cells in the lateral nucleus of the trapezoid body ( Smith et al . , 1991; Schofield and Cant , 1992 ) . The data of Campagnola and Manis ( 2014 ) directly demonstrate the CN as a source of inhibition , but neurons in other areas may contribute in addition . While we considered here the effect of the inhibition on stimulus representation at the SBC , this broad integration provides the opportunity for additional integration in the source areas . In summary , while acoustically evoked inhibition on SBC renders the ANF-SBC junction less reliable with respect to signal transmission , it enables sparser and more reproducible SBC sound encoding , which might be of relevance for subsequent localization of sound sources . All experiments were performed at the Neurobiology Laboratories of the Faculty of Bioscience , Pharmacy and Psychology of the University of Leipzig ( Germany ) , approved by the Saxonian District Government , Leipzig ( TVV 06/09 ) , and conducted according to the European Communities Council Directive ( 86/609/EEC ) . Animals were housed in the animal facility of the Institute of Biology with a 12 hr light/dark cycle and with access to food and water ad libitum . Data were collected from 42 Mongolian gerbils ( Meriones unguiculatus ) of either sex aged 4 to 12 weeks ( P43 ± 13 ) . Before the experiment , animals were anesthetized by an intraperitoneal injection of a mixture of ketamine hydrochloride ( 140 µg/g body weight , Ketamin-Ratiopharm , Ratiopharm , Ulm , Germany ) and xylazine hydrochloride ( 3 µg/g body weight , Rompun , Beyer , Leverkusen , Germany ) . The surgical procedure was performed as described previously ( Keine and Rübsamen , 2015 ) . For multi- and single-unit recordings , the animal was tilted laterally by 12–18° . The recording electrode was lowered vertically by a step motor system into the anteroventral cochlear nucleus ( AVCN ) . Glass micropipettes ( GB150F-10 and GB150F-8P , Science Products , Hofheim , Germany ) were fabricated with a PC-10 vertical puller ( Narishige , Japan ) to have impedances of 3–5 MΩ when filled with the pipette solution ( in mM ) 135 NaCl , 5 . 4 KCl , 1 MgCl2 , 1 . 8 CaCl2 , 5 HEPES , pH adjusted to 7 . 3 with NaOH . At the beginning of each recording session , multiunit recordings were performed with low-impedance electrodes ( 1–3 MΩ ) to corroborate the stereotaxic coordinates of the rostral , low-frequency pole of the AVCN . The activity of SBCs was then acquired by loose-patch recordings ( Lorteije et al . , 2009; Kuenzel et al . , 2011 ) . For that , the recording electrode was lowered through the cerebellum aiming at the AVCN at a depth of about 5000 µm . When passing through non-auditory brain regions high positive pressure ( 200 mbar ) was applied to prevent the electrode from clogging , and the electrode was advanced at a speed of 50 µm/s . On entering the target region , indicated by multiunit activity triggered by broadband noise search-stimuli , the pressure was reduced to 30 mbar , and the electrode then advanced in 1 µm-steps . When approaching a neuron , indicated by a gradual increase in series resistance , the pressure was equalized or slightly negative pressure ( –5 mbar ) applied . To minimize the mechanical stress on the recorded neuron , the seal resistance was kept <40 MΩ ( Alcami et al . , 2012 ) . Single-units were recorded only when exhibiting a positive signal amplitude of more than 2 mV ( dataset: 4 . 2 ± 1 mV ) and a signal-to-noise ratio of at least 40 ( mean amplitude of the positive AP peak divided by the standard deviation of the baseline , dataset: 68 . 2 ± 14 . 9 ) . To study the impact of inhibition on signal processing at the ANF-SBC synapse , loose-patch recordings were combined with iontophoretic drug application of the glycine receptor agonist glycine ( Sigma-Aldrich , 100 mM , prepared in 0 . 9% NaCl , pH 6 , buffered with 10 mM HEPES ) and the glycine receptor antagonist strychnine ( strychnine hydrochloride , Sigma-Aldrich , 5 mM , same formula ) . Three-barreled piggy-back electrodes ( Havey and Caspary , 1980 ) were glued to the recording electrode and had the following steric configuration: tip diameter 4–8 µm , recording electrode protruding 20–40 µm ( 3GB120F-10 , Science Products ) . The iontophoretic current was applied using an iontophoresis amplifier ( EPMS-H-7 equipped with MVCS and MVCC modules , npi electronics ) with increasing current steps ( + 5 to +100 nA ) . To reduce potential unspecific effects of strychnine , the application current was adjusted for each cell: First , the minimum current necessary to block spontaneous activity with glycine application was determined . Then , the iontophoretic current for strychnine application was set to block the glycine effect . Control experiments were performed by iontophoretic application of the carrier alone ( 0 . 9% NaCl , pH 6 , buffered with 10 mM HEPES ) . Holding currents for each barrel was set to –20 nA and a channel filled with 0 . 9% NaCl was used for automatic capacitance compensation . All recordings were performed in a sound-attenuating and electrically isolated chamber ( Type 400 , Industrial Acoustics , Niederkrüchten , Germany ) on a vibration-cushioned table . Acoustic stimuli were generated by custom-written Matlab software ( MathWorks , Natick ) and digitized at a rate of 97 . 7 kHz . Signals were presented via a custom-made earphone ( DT48 , beyerdynamic , Heilbronn , Germany ) and delivered through a metal funnel ending just in front of the ear canal . The loudspeakers were calibrated using a condenser microphone ( Bruel and Kjaer 4133 ) and custom-written Matlab software . Total harmonic distortions ( ratio of the root-mean-squared ( RMS ) amplitude of higher harmonic frequencies to the RMS amplitude of the fundamental ) were below 0 . 02% across all frequencies tested ( 0 . 1–40 kHz ) . All acoustic stimuli were corrected for the loudspeaker’s impulse response prior to presentation . Frequency response areas ( FRA ) were obtained by pseudorandom presentation of pure tones ( 100 ms in duration , 5 ms cos2 ramps , 300 ms interstimulus interval ) derived from a predefined matrix consisting of 20 different frequencies equally spaced on a log scale and 10 different sound pressure levels ( SPL ) equally spaced on a linear scale . Each of these 200 frequency/intensity pairs was presented 5–10 times while continuously recording the unit’s discharge activity . The FRAs were used to detail each unit’s CF , ( the frequency at which the neuron is most sensitive ) , response thresholds , and – if present – the frequency-intensity domain of an inhibitory sideband . The rostral pole of the AVCN was targeted considering its tonotopic organization described previously ( Kopp-Scheinpflug et al . , 2002; Dehmel et al . , 2010 ) . Spherical bushy cells were recognized by their characteristic complex waveform ( Pfeiffer , 1966; Winter and Palmer , 1990; Englitz et al . , 2009; Typlt et al . , 2010 ) and their primary-like PSTH pattern ( Blackburn and Sachs , 1989 ) . The neurons’ voltage signals were pre-amplified and impedance-converted ( Neuroprobe 1600 ) , A-M Systems , Sequim , USA ) , noise-eliminated ( HumBug , Quest Scientific , North Vancouver , Canada ) , further amplified ( PC1 , Tucker-Davis Technologies ) , and digitized at a sampling rate of 97 . 7 kHz ( 24 bit , RP2 . 1 , Tucker-Davis Technologies ) . Signals were band-pass filtered between 5 Hz and 7 . 5 kHz using a zero-phase forward and reverse digital IIR filter and stored for offline analysis using custom-written Matlab software . Extracellularly recorded voltage signals of SBCs are typically composed of two ( PP-EPSP ) or three components ( PP-EPSP-AP , Figure 1 ) reflecting the respective discharge of the presynaptic endbulb of Held ( PP , prepotential ) , the postsynaptic EPSP , and the postsynaptically triggered AP ( Englitz et al . , 2009; Typlt et al . , 2010 ) . Signals were detected using a slope threshold for the rising flank of the EPSP . In this report , EPSPs that failed to trigger a postsynaptic AP were termed EPSPfail while EPSPs that successfully trigger an AP were termed EPSPsucc . The separation between EPSPsucc and EPSPfail was based on the maximum falling slope following the detection time point . The APs following EPSPsucc exhibited a considerably faster-falling flank than EPSPfail enabling a clear separation of both signal types ( Figure 1 ) . The detection thresholds were kept fixed for each recorded unit , but varied between units to account for different signal-to-noise ratios . For comparison of timing between EPSPsucc and EPSPfail , all events were time-stamped on their respective maximum EPSP slope . Previous studies showed that the maximum slope of the EPSP is a reliable measure of EPSP strength ( Kuenzel et al . , 2011 ) . To determine the threshold EPSP , the maximum EPSP slopes of EPSPsucc and EPSPfail were binned ( bin size = 0 . 5 V/s ) , and the fraction of EPSPsucc calculated for each bin . Then , a Boltzmann function of the form ϕ ( x ) =1/ ( 1+ed−xa ) was fitted to the data with each bin weighted relative to the number of events in that bin . The symmetric inflection point d indicates the threshold EPSP , i . e . the EPSP slope necessary to yield a > 50% probability of triggering a postsynaptic AP ( see also supplementary Matlab code ) . Earlier studies showed the influence of preceding neural activity on EPSC/EPSP and AP amplitude ( Englitz et al . , 2009; Lorteije et al . , 2009; Yang and Xu-Friedman , 2015 ) . While usually the preceding inter-event-interval ( IEI ) is used as a measure of previous activity , recent studies showed that short-term plasticity can extend well beyond the last IEI in vitro ( Yang and Xu-Friedman , 2015 ) . Therefore , a weighted average of all preceding events was used , with the impact dependent on the distance and EPSP slope of the respective events . The weighting was implemented as a single-exponentially decaying kernel , emphasizing temporally close events over more distant ones ( Sonntag et al . , 2011 ) . For an EPSP slope at t0 the preceding activity was computed as PreAct ( t0 ) =1median ( Si ) ∑i=−1−∞Sie ( ti−t0 ) /τ , with Si indicating the EPSP slopes and ti the temporal distance of the i-th preceding event . The time constant τ was set to 60 ms , as this was shown to be the time window of influence in slice studies ( Yang and Xu-Friedman , 2015 ) . In addition , the calculation was also performed with time constants of 10 ms and 100 ms yielding qualitatively the same results . The influence of preceding spiking activity on AP amplitude was performed in a similar manner , but limiting the preceding events to successful APs . To evaluate the shapes of the inhibitory and the excitatory FRAs , asymmetry indices ( AI ) were calculated for both . AI was defined as lnlog2 ( FU/CF ) log2 ( CF/FL ) , where CF indicates the neuron’s characteristic frequency and FL and FU the respective low and high border-frequency of the FRA 40 dB above threshold ( see also supplementary Matlab code ) . An AI of 0 indicates symmetric tuning curves , whereas negative and positive values describe asymmetric FRAs extending to lower or higher frequencies , respectively . For sinusoidally modulated tones ( SAM and SFM ) the first 20 ms of every repetition were discarded from analysis to reduce the influence of onset effects , and analysis was constrained to complete periods to avoid unequal sampling . The temporal precision of spikes throughout the stimulus period was assessed by calculating the vector strength ( Goldberg and Brown , 1969 ) . The significance of phase-locking was tested based on the Rayleigh approximation ( Mardia , 1972 ) . Vector strength is an inadequate measure if the units firing rate reproduces the stimulus modulation . Therefore , we calculated the stimulus reproduction ( CorrNorm ) as the normalized cross-correlation between the stimulus modulation and the respective response PSTH , adjusted for the latency of the neural response ( see Tolnai et al . , 2008 ) for a detailed explanation ) . Note that CorrNorm is constrained to the positive range [0 , 1] and an unmodulated response would yield a value of 0 . 82 . High CorrNorm values ( >0 . 9 ) are only obtained when the response shape follows the stimulus envelope . The reproducibility of the neuronal responses was estimated by measuring the central peak of the shuffled autocorrelation across identical stimulus presentations ( Joris et al . , 2006 ) . The modulation depth of the neural response to the 100% amplitude modulation was estimated by calculating the standard deviation of the first cycle of the normalized cross-correlation function . Spectrotemporal receptive fields ( STRFs ) represent the neural tuning in the dimensions of the spectrogram ( time and frequency ) and help to identify stimulus properties that control spiking at high temporal resolution . Specifically , we used STRFs to study the time course of inhibition during ongoing , spectrally dispersed stimulation . Estimation of STRFs was performed using generalized reverse correlation , as described elsewhere ( Theunissen et al . , 2000; Englitz et al . , 2010 ) . STRFs were estimated for both the input to the SBCs ( EPSPfail + EPSPsucc ) , as well as the SBC output ( EPSPsucc ) . Both input and output were aligned to their maximum EPSP slope , to enable a subtraction of the two STRFs using congruent reference points . STRFs were individually normalized to their standard deviation ( positive peak lead normalization to very similar results ) , in order to allow an evaluation of tuning shape , removing the gain from overall firing rate . Without normalization , the result remains qualitatively the same . However , the inhibitory effect is then dominated by the difference at the STRF peak , which stems from the overall firing rate difference between ANF and SBC . The resulting difference-STRF indicates the translation in spectrotemporal sensitivity from ANF input to the SBC output ( see Figure 7 ) . We interpret this difference to be a consequence of three factors ( i ) local inhibition , together with ( ii ) postsynaptic processes of spike-frequency adaptation and ( iii ) Na-channel inactivation . We evaluated multiple measures of synaptic responsiveness to quantify the input-to-output signal processing at these second order neurons of the ascending central auditory pathway . All measures were computed separately for the ANF input and SBC output and then compared . Data sets were tested for Gaussianity and equality of variance using the Shapiro-Wilk test ( Shapiro and Wilk , 1965 ) and Levene’s test ( Levene , 1960 ) , respectively . Comparison between two independent groups was performed using student's two-tailed t-test or Wilcoxon rank sum test as appropriate . Aggregated data are reported as mean ± standard deviation or median [first quartile , third quartile] , respectively . Within-subject comparisons were performed by paired t-test , Wilcoxon signed rank test , or by multi-factorial repeated-measures ( RM ) ANOVA after testing for sphericity using the Mauchly test ( Mauchly , 1940 ) . If the assumption of sphericity was violated , Greenhouse-Geisser correction was applied . Bonferroni correction was applied to all multiple comparisons ( Bonferroni , 1936 ) . Correlation between quantities was assessed by Spearman’s rank correlation ( Spearman , 1904 ) to cover linear and nonlinear relationships . For interpretation of all results , a p-value less than 0 . 05 was deemed significant . Significance thresholds are abbreviated in figure panels as asterisks , with * , , * , corresponding to p<0 . 05 , p<0 . 01 , p<0 . 001 , respectively . The effect size was calculated using the MES toolbox in Matlab ( Hentschke and Stüttgen , 2011 ) and reported as eta-squared ( η² ) for RM ANOVA and Cohen’s U1 for two-sample comparisons . No statistical methods were used to pre-determine sample sizes .	In humans and other animals , small differences in the time at which a sound arrives at each ear are crucial for determining the location of the sound . Neurons in the first processing station of the brain – the cochlear nucleus – receive information about sounds ( or “inputs” ) from the ears . They then produce electrical signals that relay this information to other areas of the brain . Some of these inputs increase the activity of the neurons and so are known as “excitatory” inputs , while other “inhibitory” inputs decrease the activity of the neurons . The balance between these two inputs determines what information is passed to other parts of the brain , but it is not clear how these inputs interact . Keine , Rübsamen and Englitz studied electrical activity in the brains of Mongolian gerbils while being exposed to sounds with more natural properties than previously studied . The experiments reveal that inhibitory inputs play an important role in controlling the activity of neurons in the cochlear nucleus . By decreasing the neurons’ activity , inhibitory inputs allow these cells to respond to many different levels of sound , from very loud to very quiet . The experiments also show that excitatory and inhibitory inputs are triggered by similar sounds so that the two processes quickly balance each other . This means that the brain is equally able to work out where a sound is coming from regardless of whether it is loud or quiet . Further work is now needed to understand responses to natural sounds and to determine how experimentally removing the inhibitory inputs affects hearing .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	Inhibition in the auditory brainstem enhances signal representation and regulates gain in complex acoustic environments
Assessing the behavioral relevance of the hippocampal theta rhythm has proven difficult , due to a shortage of experiments that selectively manipulate phase-specific information processing . Using closed-loop stimulation , we triggered inhibition of dorsal CA1 at specific phases of the endogenous theta rhythm in freely behaving mice . This intervention enhanced performance on a spatial navigation task that requires the encoding and retrieval of information related to reward location on every trial . In agreement with prior models of hippocampal function , the behavioral effects depended on both the phase of theta and the task segment at which we stimulated . Stimulation in the encoding segment enhanced performance when inhibition was triggered by the peak of theta . Conversely , stimulation in the retrieval segment enhanced performance when inhibition was triggered by the trough of theta . These results suggest that processes related to the encoding and retrieval of task-relevant information are preferentially active at distinct phases of theta . Theta oscillations ( 4–12 Hz ) are one of the most prominent rhythms in the mammalian brain ( Vanderwolf , 1969; Buzsáki et al . , 1983; Colgin , 2013 ) . Theta is a distributed oscillation , which is broadly coherent between the left and right hippocampi , the entorhinal cortex , the medial septum , and various other cortical and subcortical recipients of hippocampal projections ( Buzsáki , 2002 ) . Neural activity is highly structured within each cycle of theta , with the firing rates of genetically defined cell types peaking at different phases ( Klausberger et al . , 2003 , 2004 , 2005 ) . Spatially selective principal cells in the hippocampus fire at progressively earlier phases of theta as animals traverse their individual place fields , a phenomenon known as phase precession ( O'Keefe and Recce , 1993; Schmidt and Lipson , 2009 ) . Thus , the information content of hippocampal outputs changes throughout each cycle ( Mehta et al . , 2000 , 2002 ) . The organization of activity relative to theta appears to be important for behavior , since the degree to which other regions synchronize to the hippocampal theta rhythm is correlated with spatial decision-making performance ( Jones and Wilson , 2005; Sigurdsson et al . , 2010 ) . But the specific role of theta in guiding behavior remains unclear , due to a lack of studies employing causal interventions with adequate temporal precision to selectively disrupt or enhance activity within this rhythm . Here , we employed a closed-loop approach to target an optogenetic manipulation to particular phases of endogenously generated theta oscillations . Closed-loop control is an under-utilized strategy for interrogating neural circuits , as it facilitates the testing of hypotheses that would be difficult or impossible to address through correlative methods ( Fetz , 1969; Rolston et al . , 2010; Newman et al . , 2012; Ngo et al . , 2013; Paz et al . , 2013; Wallach , 2013 ) . One specific hypothesis about the mnemonic role of theta is that it partitions processes related to the encoding of new information and the retrieval of stored information ( Hasselmo et al . , 2002 ) . In order to carry out their roles in spatial navigation , the hippocampus and related structures must be able to distinguish activity that tracks the current state of the world from activity that reflects prior experience . Theta could serve to coordinate cell assemblies such that encoded and retrieved information are less likely to interfere ( Hasselmo et al . , 2002; Hasselmo , 2005; Colgin and Moser , 2010 ) . There is abundant correlative evidence that inputs to the hippocampus vary as a function of theta phase . Input from the entorhinal cortex ( EC ) , the major source of cortical projections to the hippocampus , is highest at the trough of theta waves recorded at the hippocampal fissure ( Brankack et al . , 1993; Kamondi et al . , 1998 ) . Because it conveys information about the outside world , this input is likely associated with encoding of the current state of the environment ( Hasselmo , 2005 ) . At this same phase , the hippocampus is more susceptible to long-term potentiation ( Hyman et al . , 2003; Kwag and Paulsen , 2009 ) , consistent with the idea that this phase is optimized for encoding new information . At the 180° phase offset ( the peak of fissure theta ) , CA1 cells receive greater input from upstream cells in CA3 ( Hasselmo , 2005 ) . At this phase , stimulation of Schaffer collateral or temporoammonic inputs induces long-term depression ( Hyman et al . , 2003; Kwag and Paulsen , 2009 ) , which could suppress information storage during memory retrieval . As a result of these phase-specific physiological changes , hippocampal networks can regulate the behavioral impact of different types of information as a function of task context . Simultaneously recording in CA1 , CA3 , and EC reveals that oscillations in the high gamma range ( 60–100 Hz ) are a signature of enhanced coordination between CA1 and EC , whereas low gamma ( 25–50 Hz ) indicates enhanced coordination between CA3 and EC . Furthermore , these oscillations occur at different phases of theta , and typically on different cycles ( Colgin et al . , 2009 ) . Taken together , these results indicate that the balance between the relative influence of the outside world ( via EC ) and internal states ( via CA3 ) on hippocampal outputs is strongly modulated as a function of theta phase . Other studies have observed task-dependent modulation of hippocampal firing that is consistent with encoding of novel stimuli and retrieval of stored memories being biased to different phases ( Manns et al . , 2007; Lever et al . , 2010; Douchamps et al . , 2013; Newman et al . , 2013 ) . Computational modeling studies provide further support for this hypothesis ( Hasselmo et al . , 2002; Hasselmo and Eichenbaum , 2005; Kunec et al . , 2005 ) . These results do not imply that new information is encoded and stored information is retrieved on every cycle of theta ( ∼8 Hz ) . When encoding and retrieval do occur , though , they may be preferentially active at different phases , to take advantage of the temporal structure of activity within the hippocampal–EC loop ( Mizuseki et al . , 2009; Colgin and Moser , 2010 ) . Thus , a manipulation that targets a specific phase of theta could , on average , selectively modulate one process or the other . If encoding and retrieval processes are most active at different times within the theta cycle , the consequences of a phase-specific intervention should depend on the behavioral context . Manipulations that alter hippocampal outputs at one phase of theta may have a strong impact on behavior if they occur while information is being encoded , but no effect ( or the opposite effect ) if they occur while information is being retrieved . Conversely , manipulations that occur with a 180° phase offset may have their behavioral impact limited to intervals in which retrieved information is used to guide behavior , but have no effect ( or the opposite effect ) at times when task-relevant information is being encoded . In this study , we used millisecond-timescale optogenetic control of intrinsic inhibition to gate hippocampal outputs at specific phases of the ongoing theta cycle . Although previous studies have used optogenetic interventions to highlight the role of inhibition in phase precession ( Royer et al . , 2012 ) and theta resonance ( Stark et al . , 2013 ) , here the goal was to suppress firing of CA1 in a phase-specific manner . We directly activated parvalbumin-positive interneurons , which deliver fast and powerful inhibition to the cell bodies of pyramidal neurons in the hippocampus ( Bartos et al . , 2007 ) , at either the falling phase or rising phase of theta recorded in the local field potential ( LFP ) . We performed LFP-phase-triggered optogenetic feedback in the context of a spatial navigation task , in which mice must encode and retrieve location information on every trial ( Jones and Wilson , 2005 ) . Our stimulation occurred relative to the phase of locally recorded theta on the trigger electrodes , rather than the phase at the hippocampal fissure , to which much of the previous literature uses as a landmark ( Brankack et al . , 1993; Kamondi et al . , 1998; Hasselmo et al . , 2002 ) . However , post-hoc analysis revealed that light pulses were delivered at similar absolute phases across animals . Using closed-loop optogenetics to intervene on the timescale of theta oscillations is a powerful approach . It allows us to alter hippocampal outputs relative to ongoing theta rhythms on a trial-by-trial basis , providing within-animal controls for all stimulation conditions . We found that triggering inhibition on the peak of theta improved performance when it occurred in the encoding segment of the task , but had no effect in the retrieval segment . Triggering inhibition on the trough of theta had the opposite effects: it enhanced retrieval performance , but did not affect encoding processes . We trained mice on a navigation task that required encoding and retrieval of reward location on individual trials . Mice were placed on an H-shaped track , which consisted of two choice points separated by a central arm ( Jones and Wilson , 2005 ) ( Figure 1A ) . At one junction , a movable barrier forced mice to make a left or right turn in order to arrive at the start location . At the other junction , mice were free to turn in either direction . A food reward was delivered only if mice chose the arm closest to the most recent start location . 10 . 7554/eLife . 03061 . 003Figure 1 . Overview of the behavioral task . ( A ) Scale drawing of the end-to-end T-maze used in all experiments . On each trial , mice navigate through the ‘retrieval segment’ in the direction of the solid arrow and must choose between one of two reward sites . Reward is delivered for trajectories that involve two turns in the same direction ( e . g . , the ‘left/left’ trajectory shown ) . Once the reward site is reached , mice must travel back to one of two start locations in order to initiate the next trial . A movable barrier determines the start location for that trial , and hence which reward site will contain the food pellet . The barrier is repositioned randomly after each visit to a reward site ( whether correct or incorrect ) . A second barrier ( not shown ) prevents mice from navigating between reward sites after a decision has been made . The maze is surrounded by 10 cm walls made of clear acrylic , through which distal cues are visible . ( B ) Fraction of correct trials in each session leading up to the start of optogenetic stimulation for N = 4 individual mice ( open shapes ) and the mean ± SEM . across all subjects ( 5-day running average ) . In the 5 days before the start of optogenetic stimulation ( shaded region ) , all mice perform significantly above chance ( p<0 . 05 , based on p . d . f . of the binomial distribution with probability of 0 . 5 ) . ( C ) Trials per minute for the same sessions as in C . Mean for last 5 days ( shaded region ) is 36 . 1 ± 18 . 3 trials per session per mouse . DOI: http://dx . doi . org/10 . 7554/eLife . 03061 . 003 In order to perform the task above chance , mice must update their knowledge of reward location on a trial-by-trial basis . During the encoding segment of the task ( start arms ) , environmental cues signal the location of the upcoming reward . During the retrieval segment of the task ( central arm ) , information about the start arm is no longer present , and thus activity that drives decision-making must be generated internally . This task makes it possible to dissociate the effects of theta phase–specific inhibition on encoding and retrieval processes by separating the cues to reward location from the time and location of the mouse's decision . Four mice were trained on this task over the course of 2 to 4 weeks . All mice expressed the gene for Cre-recombinase in parvalbumin-positive cells , to allow us to target expression of channelrhodopsin to these cells later in the experiment . In the last 5 days of training , all mice performed at levels significantly above chance ( p<0 . 05 , based on p . d . f . of binomial distribution with chance level of 0 . 5 ) , with an average probability of correct response of 0 . 61 ± 0 . 05 ( Figure 1B ) . In addition to improving their accuracy , mice also increased the speed at which they performed the task , to 1 . 49 ± 0 . 69 trials per minute during the last 5 days of training ( Figure 1C ) . After at least 8 days of training , mice were implanted with a multielectrode array that targeted movable tetrodes and stationary fiber optic cables to hippocampus bilaterally . Two fiber optic cables ( one per hemisphere ) were implanted to a depth of 0 . 9 mm at the time of surgery . In the same procedure , we injected 1 . 0 µl of an adeno-associated virus carrying the gene for channelrhodopsin-2 ( Nagel et al . , 2003 ) into both sides of the brain , centered on CA1 approximately 1 mm posterior to the septal pole of hippocampus . Expression spread at least 2 mm along the septotemporal axis , covering most of dorsal CA1 as well as overlying cortex ( Figure 2A ) . 10 . 7554/eLife . 03061 . 004Figure 2 . Direct recruitment of fast-spiking inhibition with light . ( A ) Expression of ChR2-EYFP throughout the dorsal hippocampus . Note the strong labeling in stratum pyramidale , indicative of dense PV+ projections in this layer . Bilateral fiber optic lesions are marked with white rectangles , centered at ∼2 mm posterior to bregma and ∼1 . 75 mm lateral to the midline . ( B ) Projection plot of peak heights from a CA1 electrode containing a well-isolated fast-spiking unit ( blue ) and a well-isolated regular-spiking unit ( yellow ) . ( C ) Mean waveforms ( with SD ) for each tetrode channel for the same units as in panel B . ( D ) Raw , broadband trace for a single trial , aligned to the 10 ms light pulse . Four light-evoked spikes from the fast-spiking unit are clearly identifiable . ( E ) Peri-stimulus time histogram for the fast-spiking unit in B , C , and D , aligned to the start of each light pulse ( N = 1106 pulses from one session ) . This unit responds with 3–4 spikes per stimulus , then remains silent for a period of ∼15 ms following light offset . ( F ) Peri-stimulus time histogram for the regular-spiking unit in B and C , aligned to the start of each light pulse ( N = 1106 pulses from one session ) . This unit is silenced for ∼25 ms following light onset . DOI: http://dx . doi . org/10 . 7554/eLife . 03061 . 004 We waited at least 2 weeks for ChR2 expression levels to increase , during which we lowered electrodes toward the hippocampus and continued to train animals to criterion . During test sessions , we used a 465 nm LED light to drive parvalbumin-positive interneurons , which are primarily fast-spiking , soma-targeting basket and chandelier cells in the hippocampus ( Pawelzik et al . , 2002 ) . All light pulses lasted 10 ms and had an irradiance of 50 mW/mm2 ( ∼2 . 5 mW from a 250 micron fiber optic cable ) . Individual pulses reliably elicited up to four spikes from well-isolated fast-spiking units ( peak rate of 400 Hz , Figure 2B–E ) . Nearby regular-spiking units were inhibited for a period of 25 ms following light onset ( Figure 2F ) , consistent with the known time constant of fast-spiking inhibition ( Bartos et al . , 2007 ) . We combined optogenetic stimulation with closed-loop feedback in order to trigger inhibition at specific phases of theta . In each mouse ( N = 4 ) , an electrode with high theta power in the local field potential was chosen as the ‘trigger’ channel . The signals from these electrodes were filtered between 4 and 12 Hz in software . When the signal reached a local maximum or minimum , the software triggered a 10 ms light pulse delivered simultaneously to both implanted fiber optic cables ( Figure 3A , B ) . A light pulse was delivered once per theta cycle as long as the mouse remained in the stimulation zone . The same trigger channel was used throughout the course of the experiment . 10 . 7554/eLife . 03061 . 005Figure 3 . Properties of theta-triggered stimulation . ( A ) Schematic of steps involved in delivering closed-loop feedback . An event occurs in the brain ( bottom ) , which is detected and digitized by the Open Ephys recording hardware ( left ) , and sent to software for analysis ( top ) . When the target event is detected , the software activates an LED ( right ) which delivers light to brain via implanted fiber optic cables ( bottom ) . ( B ) Examples of raw and theta-bandpassed LFP during baseline trials ( top ) , peak-triggered stimulation trials ( middle ) , and trough-triggered stimulation trials ( bottom ) . Vertical blue bars indicate the time at which 10 ms light pulses occur on each cycle . ( C ) Distribution of delays between detection of the actual theta peak ( purple ) or trough ( teal ) and the time of stimulus delivery . ( D ) Distribution of actual theta phases at which stimulation occurred , for both peak ( purple ) and trough ( teal ) trials . Peak-triggered stimulation tends to occur during the falling phase of theta , whereas trough-triggered stimulation occurs around the actual trough and rising phase . Phase was calculated for data filtered offline between 4 and 12 Hz , to eliminate the phase delays inherent in online filtering . ( E ) Distribution of pulses per trial for the retrieval segment of the track . ( F ) Same as E , but for encoding segments of the track . ( G ) Occupancy times in different segments of the track for trials with retrieval-segment stimulation . Values for peak-triggered and trough-triggered stimulation are shown in purple and teal , respectively ( mean ± SD for N = 4 mice; = occupancy time decreased significantly for one mouse , p<0 . 005 , Wilcoxon rank sum test with Bonferroni correction ) . ( H ) Occupancy times in different segments of the track for trials with encoding-segment stimulation . Values for peak and trough-triggered stimulation are shown in purple and teal , respectively ( mean ± SD for N = 4 mice; = occupancy time decreased significantly for one mouse , p<0 . 005; † = occupancy time increased significantly for one mouse , and decreased significantly for a different mouse , p<0 . 05; Wilcoxon rank sum test with Bonferroni correction ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03061 . 005 Within an individual session , stimulation was confined to the retrieval ( middle arm ) or encoding ( sample arms ) segments of the track ( Figure 1A ) . In the retrieval segment , mice run toward the choice point . Stimulation in this region may affect the retrieval of stored information about reward location , but not encoding of information directly relevant for task performance . In the encoding segments , mice explore one of two sample arms . In these regions , stimulation could affect the encoding of available information about reward location . In both cases , however , the behavioral readout is the same: whether or not the mouse turned in the correct direction to retrieve the reward for that trial . Individual trials were classified as one of three types: baseline ( no stimulation ) , peak-triggered stimulation , or trough-triggered stimulation ( Figure 3B ) . All trials types were randomly interleaved and occurred with equal probability . Three mice experienced the retrieval stimulation condition first , followed by the encoding stimulation condition . One mouse experienced the conditions in the opposite order . Analysis was limited to the first 150 trials for each condition ( encoding or retrieval ) . The properties of our closed-loop stimulation were as follows: the mean delay between the trigger event ( peak or trough of theta reached ) and the onset of the light pulse was 21 . 7 ± 7 . 2 ms for peak-triggered stimulation and 21 . 3 ± 7 . 4 ms for trough-triggered stimulation ( Figure 3C ) . This is equivalent to approximately 1/6 of a 125 ms theta cycle . The mean phase of stimulation ( based on offline-filtered theta with no phase delay ) was 96 ± 54° for peak-triggered stimulation and −131 ± 63° for trough-triggered stimulation ( Figures 3E and 0° = peak ) . The phase targeting for trough-triggered stimulation was less precise , as the rising phase of theta is shorter than the falling phase ( Belluscio et al . , 2012 ) . The mean number of pulses per trial in the retrieval-stimulation condition ( middle arm ) was 8 . 8 ± 3 . 3 for peak-triggered stimulation and 8 . 3 ± 8 . 3 for trough-triggered stimulation . For the encoding-stimulation condition , the mean number of pulses was 46 . 1 ± 57 . 2 for peak-triggered stimulation and 39 . 6 ± 39 . 6 for trough-triggered stimulation ( Figure 3D–F ) . Stimulation did not generally alter occupancy time in different segments of the track ( Figure 3G–H ) . On each trial , mice spent the majority of time in the encoding segment ( average of 2–3 s for inbound trajectories and 5–8 s for outbound trajectories ) . Once they left the sample arm , they ran quickly toward the goal , spending 1–2 s in the retrieval segment and a similar amount of time running toward the reward location after making their decision . The addition of optogenetic feedback only changed occupancy times significantly for one mouse in the retrieval and outbound encoding segments , and a second mouse in the outbound encoding segment . Otherwise , all occupancy times were similar ( p>0 . 05 , Wilcoxon rank sum test with Bonferroni correction for two tests per segment , N ≥ 44 trials per segment per mouse ) . Optogenetic feedback altered the average power spectrum across trigger channels , for example by increasing the peak frequency and amplitude of theta during the peak-triggered stimulation condition ( Figure 4A ) . There was also an increase in power in the low-gamma band ( 25–35 Hz ) for both peak and trough stimulation , but this was associated with a much stronger peak in the beta band ( 16–25 Hz ) , which may have affected the low-gamma band via spectral leakage . Based on the shape of the evoked response to each optogenetic stimulus , it appears that these effects are due to the frequency content of the average waveform , rather than non-phase-aligned induced power in different frequency bands ( Figure 4B ) . Aligning the local field potential to the start of each light pulse revealed a large deflection , 200–400 µV in amplitude . The shape of the average response accounts for both the shifts in theta frequency ( based on the location of the subsequent peak ) , and the beta-range power increases ( due to ∼50 ms deflections ) . Individual pulses affected the amplitude of subsequent cycles of theta , as evidenced by the difference in the mean LFP between −100 and −75 ms for actual ( purple and teal ) vs dummy ( gray ) stimulation conditions . 10 . 7554/eLife . 03061 . 006Figure 4 . Electrophysiological changes induced by theta-triggered stimulation . ( A ) Mean power spectra for baseline , peak-triggered stimulation , and trough-triggered stimulation trials , while mice were in the retrieval segment heading toward the reward arm ( left ) or the encoding segment prior to entering the trial start location at the end of the sample arm ( right ) ( N = 4 electrodes from four mice used for triggering online feedback ) . Theta , low gamma , and high gamma frequency bands are highlighted . ( B ) Average light-evoked LFP response from N = 3 hippocampal electrodes for peak and trough-triggered stimulation trials ( purple and teal traces , respectively ) , for both encoding and retrieval epochs ( mean ± SEM ) . Gray traces indicate the average theta waveforms for baseline trials , aligned to the time a stimulus would have occurred , but for which no actual light pulse was present . ( C ) Locations of trigger electrodes ( yellow ) and passive recording electrodes ( white ) for four mice used in this experiment . The location of each lesion is indicated by red circles superimposed over histological sections ( DAPI stain , grayscale image of blue channel ) . Next to each of the images is a histogram of peak and trough stimulation phases , relative to the peak of high gamma power on that electrode for baseline ( no stimulation ) trials ( indicated by 0° ) . High gamma power ( a signature of synchronization between hippocampus and medial entorhinal cortex ( Colgin et al . , 2009 ) , provides an absolute indication of theta phase , against which the time of our optogenetic stimulation can be compared . In all electrodes ( except for the one trigger electrode in cortex , where high gamma was not measured ) , trough stimulation occurs after the peak of high gamma power , while peak stimulation occurs before the peak of high gamma . DOI: http://dx . doi . org/10 . 7554/eLife . 03061 . 006 Optogenetic stimulation was always aligned to the relative peak or trough of the 4–12 Hz bandpassed signals on each trigger electrode ( Figure 3D ) . To permit meaningful interpretation of the analysis of our behavioral results , it was necessary to measure the time of stimulation relative to an absolute indicator of theta phase . We chose high gamma ( 60–80 Hz ) power , which showed strong phasic modulation across all hippocampal electrodes , and has been previously shown to occur at a consistent phase of theta ( Colgin et al . , 2009 ) . Therefore , the peak of high gamma on baseline trials served as a landmark within each cycle of theta . In 3/4 mice , we measured stimulation times relative to the peak of high gamma for both the trigger electrode and a neighboring electrode that was passively recording signals ( Figure 4C ) . Although post-mortem analysis of electrolytic lesions revealed different locations for each electrode , all electrodes indicated that peak-triggered stimulation occurred just after the trough of high gamma , whereas trough-triggered stimulation occurred around or after the high gamma peak . In one mouse , we could not measure absolute stimulation phase , due to the trigger electrode's location in L5/6 of cortex overlying hippocampus . This electrode expressed high theta power , presumably volume-conducted from hippocampus , which was used to trigger stimulation . However , it lacked associated high gamma power , which occurs more locally . The consistency of this result indicates that , despite variations in electrode location , absolute stimulation phase was similar across animals . Although we did not measure CA1–MEC synchronization directly , previous studies have shown high gamma power to be a reliable indicator of enhanced coordination between these regions ( Colgin et al . , 2009; Yamamoto et al . , 2014 ) . Therefore , we hypothesize that trough-triggered stimulation resulted in optogenetic stimulation occurring during phases of theta in which CA1–MEC coordination was high , thereby providing CA1 with access to information about the current state of the world ( Hasselmo et al . , 2002; Hasselmo and Eichenbaum , 2005; Colgin et al . , 2009; Colgin and Moser , 2010 ) . Peak stimulation , on the other hand , targeted stimulation to phases in which CA1 and CA3 are most active ( Mizuseki et al . , 2009 ) , during which information from the hippocampus can drive downstream structures . The effects of closed-loop optogenetic feedback on behavior depended on both the phase of theta used to trigger stimulation and the region of the track in which the stimulation occurred . On individual trials , 10 ms light pulses were triggered on either the peak or trough of theta ( Figure 5A , phase relative to theta at the hippocampal fissure ) . When stimulation occurred in the retrieval segment , performance did not differ between baseline and peak-triggered stimulation for 4/4 mice ( mean of 57 . 3 ± 10 . 0% correct for baseline vs 57 . 8 ± 10 . 5% correct for peak , individual results in Table 1 ) . For trough-triggered stimulation , however , performance improved significantly in 4/4 mice ( 71 . 0 ± 8 . 2% correct for trough; significance determined by the p . d . f . of the binomial distribution , with baseline accuracy for each mouse used as the ‘chance’ level ) . The opposite effects were observed for stimulation in the encoding segment . In this condition , performance during trials with trough-triggered stimulation did not differ from baseline in 3/4 mice ( mean of 59 . 1 ± 2 . 4% correct for baseline vs 58 . 7 ± 10 . 5% correct for trough; 1 mouse had significantly impaired performance in the trough-stimulation condition ) . For peak-triggered stimulation , performance improved significantly in 3/4 mice ( mean of 69 . 6 ± 14 . 7% correct; 1 mouse showed no difference from baseline ) . 10 . 7554/eLife . 03061 . 007Figure 5 . Behavioral modulation depends on both theta phase and task segment . ( A ) Illustration of the two manipulations performed in this experiment . On any given ‘non-baseline’ trial , stimulation was triggered by the peak ( purple phase ) or trough ( teal phase ) of the 4–12 Hz theta rhythm . The resulting light pulses recruited inhibition for ∼25 ms , or approximately 1/5 of the 125 ms theta cycle . ( B ) Accuracy relative to baseline for four mice in four conditions: optogenetic stimulation triggered at the peak ( purple ) or trough ( teal ) of theta , in either the retrieval ( left ) or encoding ( right ) segments of the track . Mean ± SEM , with results for each mouse overlaid . Individual results in the gray regions are significantly different from baseline ( p<0 . 05 , p . d . f . of binomial distribution with probability equal to baseline accuracy ) . ( C ) Same data as in b , but represented on the same axes . Note that peak-triggered stimulation in the encoding segment consistently improves performance more than the same type of stimulation in the retrieval segment ( points above diagonal line ) . The opposite effects are seen for trough-triggered stimulation . ( D ) Schematic of all possible ‘double-dissociation’ scenarios used for establishing bootstrap significance levels of the actual result . ( E ) Performance on baseline ( no stimulation ) trials for four different trial types: ( 1 ) mice are cued to switch arms after a correct choice ( correct/switch ) , ( 2 ) mice are cued to return to the same arm after a correct choice ( correct/stay ) , ( 3 ) mice are cued to switch arms after an incorrect choice ( incorrect/switch ) , and ( 4 ) mice are cued to return to the same arm after an incorrect choice . Trials are grouped by retrieval stimulation or encoding stimulation conditions . For both conditions , changing trial type has a significant effect on performance: retrieval stimulation , χ2 = 8 . 4 , p=0 . 038; encoding stimulation , χ2 = 8 . 1 , p=0 . 044; Friedman test ( nonparametric , repeated-measures ANOVA ) . ( F ) Change in performance with the addition of closed-loop optogenetic stimulation for the four trial types in E . DOI: http://dx . doi . org/10 . 7554/eLife . 03061 . 00710 . 7554/eLife . 03061 . 008Table 1 . Results for individual miceDOI: http://dx . doi . org/10 . 7554/eLife . 03061 . 008RetrievalEncodingBaselinePeakTroughBaselinePeakTroughMouse 10 . 550 . 490 . 710 . 580 . 480 . 43p=0 . 09P=0 . 02p=0 . 07P=0 . 02Mouse 20 . 520 . 570 . 640 . 600 . 750 . 66p=0 . 09P=0 . 02P=0 . 01p=0 . 06Mouse 30 . 500 . 530 . 670 . 620 . 730 . 64p=0 . 10P=0 . 01P=0 . 03p=0 . 11Mouse 40 . 720 . 730 . 820 . 560 . 820 . 62p=0 . 12P=0 . 04P=0 . 0001p=0 . 08Probability of a correct response for four mice under six conditions: baseline ( no stimulation ) , peak-triggered stimulation , and trough-triggered stimulation in both the retrieval and encoding segments . p-values computed from the p . d . f . of the binomial distribution with chance levels equal to the baseline performance for that mouse . Significant changes ( p<0 . 05 ) are highlighted in bold . On average , trough-triggered stimulation resulted in a 13 . 7% improvement in accuracy for the retrieval condition , while peak-triggered stimulation resulted in a 10 . 5% improvement in accuracy for the encoding condition ( Figure 5B ) . The effects were consistent across individual mice , with trough-triggered stimulation improving performance more for the retrieval segment than the encoding segment in 4/4 mice , and peak-triggered stimulation improving performance more for the encoding segment than the retrieval segment for 3/4 mice ( Figure 5C ) . Such effects represent a double-dissociation , as phase-specific optogenetically recruited inhibition reversed its behavioral impact depending on the region of stimulation . To estimate the probability that these results could have occurred by chance , we had to consider all outcomes in which a double dissociation was present . Our initial hypothesis was only that the effects of stimulation would depend on both task phase and theta phase , not that performance would be specifically impaired or improved . We used a bootstrap procedure with 10 , 000 repetitions to determine the probability that any of the possible double dissociations shown in Figure 5D could have occurred by chance . We randomized the labels for all trials ( baseline , peak-triggered , and trough-triggered ) and looked for the presence of significant changes relative to baseline in any of the conditions in the 2 × 2 square . If 3/4 mice showed the same behavior ( enhancement , impairment , or no change ) , we considered that a ‘consistent’ quadrant . The probability that the same effects would be seen along any diagonal was 0 . 0013 . The probability that the same effects were seen along both diagonals ( what we observed in the actual data ) was 0 . 0001 . To better understand the source of these behavioral effects , we analyzed the types of mistakes made by the mice , and how activating inhibition at the appropriate phase serves to correct them . It is clear that the outcome of the previous trial has a strong effect on decision-making: mice are much more likely to make a correct choice if they are cued to switch arms after failing to receive reward ( 82 ± 14% correct ) or return to the same arm after receiving reward ( 72 ± 18% correct; mean response across encoding and retrieval conditions , baseline trials only ) . They are less likely to switch arms after a correct decision ( 32 ± 9% correct ) or to return to the same arm after an incorrect choice ( 40 ± 21% correct ) . Thus , mice exhibit a bias toward a ‘win-stay , lose–switch’ strategy . They favor returning to arms in which they just received reward , or switching to the opposite arm if there was no reward on the previous trial ( Figure 5E ) . Adding optogenetic stimulation on certain trials allows mice to overcome this inherent bias . The strongest effects were seen for trials in which mice were cued to enter the opposite reward arm after making a correct choice . On average , trough-triggered stimulation in the retrieval segment improved performance on these trials by 19 . 5 ± 7 . 3% , with improvement seen in 4/4 mice ( versus 5 . 1 ± 14 . 7% for peak-triggered stimulation ) . Peak-triggered stimulation in the encoding segment improved performance on these trials by 25 . 0 ± 10 . 3% , again with improvement in 4/4 mice ( versus 8 . 7 ± 14 . 6% for trough-triggered stimulation ) . The effects of phase-specific stimulation on other types of errors were less pronounced , but there was no evidence for reduced performance by the ‘optimal’ stimulation phase for any trial type ( Figure 5F ) . These results provide new evidence for a hypothesis that was previously supported by correlational studies and computational models: processes related to encoding new information and retrieving stored information occur preferentially at different phases of the theta oscillation ( Hasselmo et al . , 2002; Hasselmo , 2005; Colgin et al . , 2009; Colgin and Moser , 2010 ) . We have shown that interventions targeting the falling or rising phases of theta have different effects depending on the behavioral context . When environmental cues to reward location are available ( as in the encoding segment of the task ) , triggering hippocampal inhibition on the peak of theta enhanced navigational accuracy . When behavioral guidance must be based on internal signals alone ( as in the retrieval segment of the task ) , triggering hippocampal inhibition on the trough of theta increased the probability of a correct choice . What is the neural basis these effects ? Our favored explanation is that phase-specific inhibition serves to reduce the response to task-irrelevant inputs . Figure 6 , which was inspired by a similar diagram in Hasselmo et al . ( 2002 ) , illustrates the mechanism by which this could occur . On average , the influence of CA3 and entorhinal cortex inputs to CA1 changes as a function of theta phase ( Hasselmo et al . , 2002 ) . Under baseline conditions , the relative influence of CA3 and EC is ‘balanced’ . With the addition of closed-loop optogenetic feedback , excess inhibition reduces spike activity either during EC-dominant or CA3-dominant periods of the theta cycle . Although the parvalbumin-positive interneurons recruited by this manipulation are typically active during the CA3-dominant phases of theta ( Lasztóczi and Klausberger , 2014 ) , causal control allows us to activate them synchronously at arbitrary times during the theta cycle . Under the proposed mechanism , inhibiting CA1 during EC-dominant cycles in the retrieval segment improves task performance by increasing the relative influence of CA3 . Conversely , inhibiting CA1 during CA3-dominant cycles in the encoding segment improves task performance by increasing the relative influence of EC or by suppressing retrieval of interfering cross-trial information . In both cases , enhanced navigational accuracy could result from suppression of task-irrelevant information , rather than the enhancement of task-relevant information . 10 . 7554/eLife . 03061 . 009Figure 6 . Proposed mechanism . This diagram illustrates how the relative influence of CA3 vs entorhinal cortex ( EC ) inputs to CA1 could explain the experimental results . At the top , a sine wave indicates the phase of theta . Below , the purple and green histograms show the fluctuating influence of CA3 and EC on CA1 on each cycle . Levels represent averages; on individual cycles , one or the other may dominate ( Colgin et al . , 2009 ) . When optogenetic inhibition is triggered on the trough of theta ( T ) , it tends to reduce firing rates in CA1 during periods of high EC influence . This tips the balance in favor of CA3 , thereby improving performance during periods of retrieval ( R ) . When optogenetic inhibition is triggered on the peak of theta ( P ) , it tends to reduce firing rates in CA1 during period of high CA3 influence . This tips the balance in favor of EC , thereby improving performance during periods of encoding ( E ) . On the whole , our closed-loop manipulation may improve performance by reducing the influence task-irrelevant inputs as a function of both theta phase and maze region . DOI: http://dx . doi . org/10 . 7554/eLife . 03061 . 009 Our data supports the presence of strong local inhibition in CA1 at specific phases of theta ( Figures 2F and 3D ) . The duration of this inhibition ( ∼25 ms ) is equivalent to approximately 1/5 of a 125 ms ( 8 Hz ) theta cycle , long enough to impact encoding or retrieval functions , but precise enough to avoid disrupting the entire cycle . This suggests a simple , CA1-specific mechanism could be sufficient to explain our behavioral results . According to our proposed mechanism , suppression of inputs carrying information about the current state of the world improves performance in the retrieval segment , whereas suppression of inputs carrying information about past states improves performance in the encoding segment . Although a local mechanism can provide the simplest explanation of our results , we cannot rule out a mechanism that involves changes in inter-regional coupling strength without simultaneous recordings from the relevant downstream areas . The navigation task used in this study must engage a wide network of brain regions , including those that receive monosynaptic projections from CA1 . Besides projections back to the entorhinal cortex , the major cortical output of dorsal CA1 is retrosplenial cortex ( Cenquizca and Swanson , 2007 ) . Retrosplenial cortex is known to be important for spatial working memory ( Vann and Aggleton , 2004; Keene and Bucci , 2009; Vann et al . , 2009 ) , and changes in coupling between hippocampus and this region could affect performance on the present task . There is also a projection from CA1 to prefrontal cortex , although this originates primarily from the ventral regions ( Swanson , 1981; Cenquizca and Swanson , 2007 ) . It is possible that our optogenetic manipulation is affecting prefrontal-dependent decision-making via ventral CA1 ( Colgin , 2011; Schmidt et al . , 2013; O'Neill et al . , 2013 ) , subicular projections ( Jay and Witter , 1991 ) , or multisynaptic pathways . Our manipulation could also be exerting long-range effects through the actions of projecting interneurons , a small fraction of which are parvalbumin-positive ( Jinno , 2009 ) . Coupling strength between CA1 and medial prefrontal cortex is known to depend on task phase ( Jones and Wilson , 2005 ) , and it is therefore possible that phase-specific inhibition of CA1 could have a differential impact on CA1–mPFC synchrony during the encoding and retrieval segments . However , if a change in inter-regional coupling strength is at play , we might expect the impact of our optogenetic manipulation to be most pronounced during the retrieval segment , when cortical regions involved in behavioral guidance are likely to access the hippocampal representation of space . In this case , it is not clear that phase-specific stimulation of the hippocampus in the encoding segment should also affect performance . Although our observation of a double-dissociation suggests that the main mechanism is local to the hippocampus , disruption of CA1–mPFC coupling during the encoding segment could minimize cross-trial interference , also leading to enhanced performance . The influence of alternative behavioral strategies must also be considered . Mice could employ multiple strategies for completing the task , such as an egocentric , hippocampus-independent strategy based on turn direction or an allocentric , hippocampal-dependent strategy based on an internal map ( Packard et al . , 1989; Floresco et al . , 1997 ) . It is possible that inhibiting the hippocampus biases the mouse toward an egocentric strategy , which happens to improve performance . If this were the case , again , we would not expect to see such a striking double-dissociation effect in our results . We would instead predict that the same phase of stimulation would enhance performance in both the encoding and retrieval segments . Further evidence against a ‘strategy switching’ mechanism could come from an experiment in which inhibition was activated at the optimal phase for both the encoding and retrieval segments on individual trials . If stimulation is merely invoking a change in strategy , we would not expect to see additive effects; that is , stimulation in the encoding phase would be sufficient to reach peak performance . However , if combined encoding and retrieval stimulation improved performance more than one or the other in isolation , it would suggest that specific effects on encoding and retrieval operations are at play . The analysis of types of errors in Figure 5E makes it clear that encoding and retrieval are continuous processes . They are not necessarily confined to the ‘encoding’ and ‘retrieval’ segments specific to this task . The mice are actually performing at least two tasks concurrently , one which involves entry into the cued reward arm ( the trained task ) and one which involves acting based on the outcome of the previous trial ( an untrained task ) . Could optogenetic stimulation merely serve to ‘reset’ the system , reducing interference between trials ? We consider this possibility unlikely , due to the fact that the effects of stimulation at different phases depended on the track segment in which it occurred . This double-dissociation indicates that stimulation at the appropriate phase allowed mice to more accurately update and retrieve knowledge of the upcoming reward location , rather than simply suppress the influence of the previous trial . In addition , optogenetic stimulation did not impair performance for ‘easy’ trials , in which the cued reward location was consistent with animals' tendency toward a ‘win–stay , lose–switch’ strategy ( Figure 5F ) . If stimulation brought mice back to a naïve ‘baseline’ state , we would expect them to make more errors when inhibition is recruited on these trials . Overall , this analysis highlights the fact that encoding and retrieval cannot be considered discrete states that depend on the task at hand , and are instead occurring continuously as animals explore their environment . Our results indicate that optogenetic inhibition of CA1 serves to improve performance across all mice tested . Our initial hypothesis was that the effects of stimulation would depend on both the task segment and phase , but we were unsure if they would be beneficial or punitive . Given that we are recruiting inhibition , and thereby suppressing CA1 output , one might expect the behavioral impact on a hippocampal-dependent task to be negative . Recruiting inhibition during the ‘retrieval’ phase should impair performance in the retrieval segment , whereas recruiting inhibition during the ‘encoding’ phase should impair performance during the encoding segment . The fact that we instead observed only enhanced performance ( rather than enhancement for one phase of stimulation and impairment for the other ) may be explained by a floor effect . Baseline performance was modest at the start of testing ( Figure 1A , Table 1 ) , which makes it unlikely that we could observe a significant impairment , even if it did exist . Mice were strongly influenced by the outcome of the previous trial ( Figure 5E ) , which explains why their accuracy on the trained task is only slightly ( but significantly ) above chance . Our phase-specific optogenetic intervention helps them overcome this bias , especially in the case of trials in which they are required to switch arms after receiving reward ( Figure 5F ) . However , even for trials in which the reward location was consistent with animals' intrinsic biases , stimulation did not interfere with performance . It is possible that higher light intensities , alternate fiber placements , or a different target phase could have created the conditions necessary to negatively impact behavior . Revealing a convincing mechanistic explanation for the behavioral effects seen in this study will require further investigation . The present results justify more extensive inquiry along these lines by providing evidence that processes related to encoding new information and retrieving stored information are most active at different phases of theta . This hypothesis was originally based on a computational model ( Hasselmo et al . , 2002 ) , which was later supported by correlative evidence ( Manns et al . , 2007; Colgin et al . , 2009; Douchamps et al . , 2013; Newman et al . , 2013 ) . We advance this line of investigation through the use of a closed-loop optogenetic intervention that allowed us to interact with the hippocampus at specific phases of theta on a trial-by-trial basis . As the tools for closed-loop control become more accessible , experiments that couple precise stimulation to internal state variables have the potential to enhance our understanding of a wide range of topics related to the study of neural systems . All mice were male parvalbumin-Cre ( PV-Cre ) heterozygotes , derived from PV-Cre BAC transgenics back-crossed into a C57BL/6J line ( Jackson Laboratory strain B6; 129P2-Pvalbtm1 ( cre ) Arbr ) . Mice were 8–12 weeks old at the start of training ( mean age = 10 . 8 ± 1 . 5 weeks ) and 10–15 weeks old at the time of surgery ( mean age = 13 . 5 ± 2 . 1 weeks ) . Animals were individually housed and maintained on a 12-hr light/dark cycle ( lights on at 7:00 AM ) . All experimental procedures and animal care protocols were approved by the Massachusetts Institute of Technology Institutional Animal Care and Use Committees and were in accordance with NIH guidelines . The task was adapted from that used in a previous study ( Jones and Wilson , 2005 ) . The track consisted of two T-mazes placed end-to-end to form an ‘H’ shape , with movable gates at both choice points . When running toward the sample arms , the location of the gate forced the mouse in one direction or the other . No reward was delivered in the sample arm , but mice were required to reach the end of it in order to initiate a new trial . When running in the opposite direction , mice could choose between one of two ‘free choice’ arms . Reward was only delivered if the mouse entered the free choice arm closest to the most recent sample arm it had visited . Rewards consisted of one 14 mg sugar pellet ( Bio-Serv , Flemington , NJ; product #F05684 ) , and were always preceded by a 2 kHz tone lasting 250 ms , triggered by entry into the reward zone . The track was made from laser-cut acrylic , with transparent walls and a black floor . Distal cues were provided by three large black curtains with high-contrast patterns in the center and the experimenter's body , which remained in a consistent location across days . IR sensors were used to monitor entry and exit from different regions of the track . An Arduino sent information about IR beam breaks to a computer running custom software written in Processing ( https://github . com/jsiegle/t-maze ) . The experimenter manually moved the ‘forced choice’ gates at the start of each trial according to a sequence generated randomly by the behavior computer . If mice were biased toward one reward arm , the probability of reward appearing in the opposite arm increased according to the following equation: P ( reward in left arm ) = P ( mouse chose right arm during last 12 trials ) . Prior to the start of training , mice were restricted to 2–3 g of dry food per day , with unlimited access to water . Training began with 4–6 days of habituation , during which mice freely explored the track while receiving reward in both the choice and sample arms . Next , a period of ‘forced choice’ training began , in which a gate always forced the mice in the correct direction at each choice point . After 5–6 days of forced choice training , a ‘free choice’ condition was added , in which mice were allowed to make incorrect decisions . Subsequent sessions typically consisted of 10–15 min of forced choice training , followed by 15–20 min of free choice training . Mice received free choice training for 0–10 days before surgery , and 14–26 days prior to the start of behavioral testing . We used AAV-5 viral vectors containing double-floxed , inverted , open-reading-frame ChR2 ( H134R variant ) coupled to EYFP and driven by the EF1α promoter . Implants were constructed according to the procedure described in Voigts et al . ( 2013 ) . Design files can be found on GitHub ( https://github . com/open-ephys/flexdrive ) , and assembly instructions are hosted on the Open Ephys wiki ( https://open-ephys . atlassian . net/wiki/display/OEW/flexDrive ) . The base of the drive consisted of two stainless steel cannulae with their centers 3 . 6 mm apart . Each cannula held four electrodes spaced in a ring around a central fiber optic cable ( 240 micron core diameter , 0 . 51 NA , Edmund Optics , Barrington , NJ; part #02-531 ) . The fiber optic cables protruded 0 . 9 mm past the end of each cannula . Electrodes were made from 12 . 5 μm polyimide-coated nichrome wire ( Kanthal , Hallstahammar , Sweden ) , twisted and heated to form tetrodes ( Nguyen et al . , 2009 ) . Individual electrodes were gold plated to an impedance of 200–400 kΩ . Mice were anesthetized with isofluorane gas anesthesia ( 0 . 75–1 . 25% in 1 l/min oxygen ) and secured in a stereotaxic apparatus . The scalp was shaved , wiped with hair removal cream , and cleaned with iodine solution and alcohol . Following IP injection of Buprenex ( 0 . 1 mg/kg , as an analgesic ) , the skull was exposed with an incision along the midline . After the skull was cleaned , six steel watch screws were implanted in the skull , one of which served as ground . Next , a ∼1 . 5 mm-diameter craniotomy was drilled over left hippocampus ( 2 . 0 mm posterior to bregma and 1 . 8 mm lateral to the midline ) and the dura was removed . Virus was delivered through a glass micropipette attached to a Quintessential Stereotaxic Injector ( Stoelting , Wood Dale , IL ) . The glass micropipette was lowered through the center of the craniotomy to a depth of 1 . 2 mm below the cortical surface . A bolus of 1 μl of virus ( see details above ) was injected at a rate of 0 . 05 μl/min . After the injection , the pipette was held in place for 10 min at the injection depth before being fully retracted from the brain . The same procedure was then repeated for the opposite hemisphere . The fiber optic–electrode implant ( see details above ) was aligned with the two craniotomies , and lowered until the cannulae were flush with the cortical surface . This placed the two fiber optic cables just above the CA1 region of hippocampus ( depth of ∼0 . 9 mm ) . Once the implant was stable , a small ring of black dental acrylic was placed around its base . A drop of surgical lubricant ( Surgilube , Fougera Pharmaceuticals , Melville , NY ) prevented dental acrylic from contacting the cortical surface . Adhesive luting cement ( C&B Metabond , Parkell , Edgewood , NY ) was used to further affix the implant to the skull . Once the cement was dry , the scalp incision was closed with VetBond ( 3M , Saint Paul , MN ) , and mice were removed from isoflurane . Following 2–4 days of recovery , electrodes were lowered to their final location over the course of 2–3 weeks . Once stimulation began , electrodes were not adjusted . On testing days , the track was wiped with an anti-static liquid ( Staticide , ACL , Chicago , IL ) and cleared of all debris . Electrophysiological data was recorded with the Open Ephys platform ( http://open-ephys . org ) , an open-source data acquisition system based on Intan amplifier chips ( http://www . intantech . com ) . Tetrode signals were referenced to ground , filtered between 1 and 7500 Hz , multiplexed , and digitized at 30 kHz on the headstage ( design files available at https://github . com/open-ephys/headstage/tree/master/1x32_Omnetics_Standard ) . Digital signals were transmitted over a 12-wire cable counter-balanced with a system of pulleys and weights . Mouse location was determined via IR gates at behaviorally relevant points along the track and an overhead camera monitoring a red LED mounted on the headstage . Online feedback was delivered using the Open Ephys GUI ( full source code available at https://github . com/open-ephys/GUI ) . The trigger channel was filtered between 4 and 12 Hz ( 2nd-order Butterworth ) and sent to a ‘Phase Detector’ module . When the mouse entered the stimulation segment ( either one of two sample arms for ‘encoding’ sessions or the central arm on forward and reverse trajectories for ‘retrieval’ sessions ) , the Phase Detector emitted trigger events when the signal reached a local maximum ( ‘peak’ ) or local minimum ( ‘trough’ ) . Trials of each type ( ‘peak’ , ‘trough’ , or ‘blank’ ) were randomly interleaved with equal probability . Stimulation was triggered via a USB connection to a Pulse Pal ( https://sites . google . com/site/pulsepalwiki/ ) , and consisted of 10 ms light pulses from a Plexon PlexBright LED ( 465 nm , ∼50 mW/mm2 ) . At the end of training , electrodes sites were lesioned with 15 μA of current for 10 s . Mice were transcardially perfused with 100 mM PBS followed by 4% formaldehyde in PBS . Brains were post-fixed for at least 18 hr at 4°C . 60 μm sections were mounted on glass slides with Vectashield ( Vector Laboratories , Burlingame , CA ) , coverslipped , and imaged with an upright fluorescent microscope . Viral expression was confirmed by observing EYFP expression beneath the fiber optic lesions in CA1 of all animals . Expression spread ∼2 mm along the length of the dorsal hippocampus , primarily in CA1 , but also in the lateral portion of CA3 . Labeling was strongest in the hippocampal cell layer , where parvalbumin-positive cells have the densest projections . There was also expression in overlying cortex . No expression was observed in the dentate gyrus . In all animals , the lesion corresponding to the electrode used to trigger stimulation was identified . All data analysis was performed using custom Matlab scripts ( https://github . com/open-ephys/analysis-tools ) . Spike activity was extracted offline by thresholding the 300–6000 Hz bandpassed signal . Units were clustered with Simple Clust software ( https://github . com/moorelab/simpleclust ) , based on peak heights and regression coefficients for individual waveforms . Spikes were aligned to light pulses using event timestamps , or to the phase of LFP theta . To determine the actual phase of theta without the phase shift associated with online filtering , we filtered the wideband , full-sample-rate data offline using the Matlab ‘filtfilt’ function ( 2nd-order Butterworth , 4–12 Hz bandpass ) . We used the angle of the Hilbert-transformed signal to compute the phase in degrees ( −180°–180° , peak at 0° ) . Spectral analysis was performing using the Chronux toolbox ( http://www . chronux . org ) , using multitaper methods ( time–bandwidth product = 2 , number of tapers = 3 ) . Behavioral analysis was limited to the first 150 trials performed in each condition ( encoding stimulation vs retrieval stimulation ) . Trials were grouped by stimulation type ( blank , peak-triggered , or trough-triggered ) and the responses ( 0 = correct choice , 1 = incorrect choice ) were averaged . p-values for individual mice were computed using the probability density function of the binomial distribution , with N = the number of trials of a given type and p = baseline accuracy . The probability that a double-dissociation would occur by chance was computing using a bootstrap method with randomized trial labels ( 10 , 000 iterations ) .	Around 15 years ago , an imaging study compared the brains of London taxi drivers—who need to know their way around one of the biggest cities in the world—with those of the general public , and found that a structure called the hippocampus was routinely larger in the taxi drivers . This finding was consistent with previous studies from rats , which showed that anatomical changes occur in the hippocampus after animals learn to navigate through various mazes . Together , these results suggest that the hippocampus is important for spatial awareness in both humans and rodents . The hippocampus—which takes its name from the Greek for ‘seahorse’ due to its shape—consists mostly of cells called pyramidal neurons , which communicate with one other using an excitatory molecule called glutamate . However , it also contains cells that suppress the activity of the pyramidal neurons , using an inhibitory molecule called GABA . When electrodes are used to record the combined electrical activity of many cells in the hippocampus—including both excitatory and inhibitory cells—the resulting pattern resembles a wave with peaks and troughs that repeat roughly eight times per second . Although this activity , known as the theta rhythm or cycle , has been observed in countless experiments , it has been difficult to pin down how it is relevant to behavior . Siegle and Wilson now show that the theta cycle may help the brain to keep incoming information separate from information stored in memory . This conclusion is based on the results of experiments on mice with hippocampi that had been modified to make them sensitive to light: in particular , light was needed to activate the neurons that suppress the activity of the pyramidal neurons . This meant that it was possible to reduce the overall level of activity in the hippocampus by shining light on certain neurons . The mice were trained to perform a spatial memory task that consisted of an encoding stage—where they learned the location of a reward—and a retrieval stage , in which they recalled this location from memory . On certain trials , pulses of light could be delivered to the brain at specific points in the theta cycle . Delivering light near the peak of the cycle during the encoding stage resulted in improved memory performance , as did delivering light near the trough of the cycle during the retrieval stage . These results suggest that the hippocampus preferentially encodes and retrieves information at different stages of the theta cycle . Specifically , activity just after the peak of the theta cycle is biased towards retrieval , meaning that reducing hippocampal activity at this time point will make it easier to form new memories . By contrast , reducing activity just after the trough of the theta cycle—when the hippocampus is biased towards encoding—will enhance memory retrieval .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Enhancement of encoding and retrieval functions through theta phase-specific manipulation of hippocampus
Glycosyltransferases ( GTs ) are prevalent across the tree of life and regulate nearly all aspects of cellular functions . The evolutionary basis for their complex and diverse modes of catalytic functions remain enigmatic . Here , based on deep mining of over half million GT-A fold sequences , we define a minimal core component shared among functionally diverse enzymes . We find that variations in the common core and emergence of hypervariable loops extending from the core contributed to GT-A diversity . We provide a phylogenetic framework relating diverse GT-A fold families for the first time and show that inverting and retaining mechanisms emerged multiple times independently during evolution . Using evolutionary information encoded in primary sequences , we trained a machine learning classifier to predict donor specificity with nearly 90% accuracy and deployed it for the annotation of understudied GTs . Our studies provide an evolutionary framework for investigating complex relationships connecting GT-A fold sequence , structure , function and regulation . Complex carbohydrates make up a large bulk of the biomass of any living cell and play essential roles in biological processes ranging from cellular interactions , pathogenesis , immunity , quality control of protein folding and structural stability ( Varki and Gagneux , 2019 ) . Biosynthesis of complex carbohydrates in most organisms is carried out by a large and diverse family of Glycosyltransferases ( GTs ) that transfer sugars from activated donors such as nucleotide diphosphate and monophosphate sugars or lipid linked sugars to a wide range of acceptors that include saccharides , lipids , nucleic acids and metabolites . Nearly 1% of protein coding genes in the human genome , and more than 2% of the Arabidopsis genome , are estimated to be GTs . GTs have undergone extensive variation in primary sequence and three-dimensional structure to catalyze the formation of glycosidic bonds between diverse donor and acceptor substrates . However , an incomplete understanding of the relationships connecting sequence , structure , function and regulation presents a major bottleneck in understanding pathogenicity , metabolic and neurodegenerative diseases associated with abnormal GT functions ( Ryan et al . , 2019; Day et al . , 2012 ) . Structurally , GTs adopt one of three folds ( GT-A , -B or -C ) with the GT-A Rossmann like fold being the most common ( Figure 1 , Figure 1—source data 1 ) . The GT-A fold is characterized by alternating β-sheets and α-helices ( α/β/α sandwich ) found in most nucleotide binding proteins ( Breton et al . , 2012 ) . The majority of GT-A fold enzymes are metal dependent and conserve a DxD motif in the active site that helps coordinate the metal ion and the nucleotide sugar . Currently , 110 GT families have been catalogued in the Carbohydrates Active Enzymes ( CAZy ) database ( accessed in February 2020 ) ( Lombard et al . , 2014 ) . These families can be broadly classified into two categories based on their mechanism of action and the anomeric configuration of the glycosidic product relative to the sugar donor , namely , inverting or retaining ( Figure 1 ) . Inverting GTs generally employ an SN2 single displacement reaction mechanism that results in inversion of anomeric configuration for the product . In contrast , retaining GTs are believed to employ a dissociative SNi-type mechanism , where the anomeric configuration of the product is retained ( Moremen and Haltiwanger , 2019; Lairson et al . , 2008 ) . While the sequence basis for inverting and retaining mechanisms is not well understood , most inverting GT-As have a conserved Asp or Glu within a xED motif that serves as the catalytic base to deprotonate the incoming nucleophile of the acceptor , and initiate nucleophilic attack with direct displacement of the phosphate leaving group ( Lairson et al . , 2008; Gloster , 2014 ) . Retaining GT-As bind the sugar donor similarly to the inverting enzymes , but shift the position of the acceptor nucleophile to attack the anomeric carbon from an obtuse angle using a phosphate oxygen of the sugar donor as the catalytic base and employ a dissociative mechanism that retains the anomeric linkage for the resulting glycosidic product ( Moremen and Haltiwanger , 2019 ) . Such mechanistic diversity of GTs is further illustrated by recent crystal structures of GTs bound to acceptor and donor substrates which show that different acceptors are accommodated in the active site through variable loop regions emanating from the catalytic core ( Moremen and Haltiwanger , 2019; Ramakrishnan and Qasba , 2010a; Kadirvelraj et al . , 2018; Gordon et al . , 2006 ) . However , whether these observations hold for the entire super-family is not known because of the lack of structural information for the vast number of GTs . The wealth of sequence data available on GTs provides an opportunity to infer underlying mechanisms through deep mining of large sequence datasets . In this regard , the CAZY database serves as a valuable resource ( Lombard et al . , 2014 ) for generating new functional hypotheses by classifying GT enzymes into individual families based on overall sequence similarity . However , a broader understanding of how these enzymes evolved to recognize diverse donor and acceptor substrates requires a global comparison of diverse GT-A fold enzymes . Such comparisons are currently a challenge due to limited sequence similarity between families and the lack of a phylogenetic framework to detect evolutionary events associated with GT functional specialization . Previous efforts to investigate GT evolution have largely focused on individual families or pathways ( Taujale and Yin , 2015; Lombard , 2016 ) and have not explicitly addressed the challenge of mapping the evolution of functional diversity across families . Here through deep mining of over half a million GT-A fold-related sequences from diverse organisms , and application of specialized computational tools developed for the study of large gene families ( Kannan et al . , 2007; Kwon et al . , 2019 ) , we define a common core shared among diverse GT-A fold enzymes . Using the common core features , we generate a phylogenetic framework for relating functionally diverse enzymes and show that inverting and retaining mechanisms emerged independently multiple times during evolution . We identify convergent modes of substrate recognition in evolutionarily divergent families and pinpoint sequence and structural features associated with functional specialization . Finally , based on the evolutionary and structural features gleaned from a broad analysis of diverse GT-A fold enzymes , we develop a machine learning ( ML ) framework for predicting donor specificity with nearly 90% accuracy . We predict donor specificity for uncharacterized GT-A enzymes in diverse model organisms and provide testable hypotheses for investigating the relationships connecting GT-A fold structure , function and evolution . To define common features shared among diverse GT-A fold enzymes , we generated a multiple sequence alignment of over 600 , 000 GT-A fold related sequences in the non-redundant ( NR ) sequence database ( Pruitt et al . , 2007 ) using curated multiple-aligned profiles of diverse GTs . The alignment profiles were curated using available crystal structures ( Materials and methods ) ( Neuwald , 2009 ) . The resulting alignment revealed a GT-A common core consisting of 231 aligned positions . These aligned positions are referred to throughout this analysis and are mapped to representative structures in Supplementary file 2 . The common core is defined by eight β sheets and six α helices , including three β sheets and α helices from the N-terminal Rossmann fold ( Figure 2A , B ) . Quantification of the evolutionary constraints imposed on the common core reveal twenty residues shared among diverse GT-A fold families . These include the DxD and the xED motif residues involved in catalytic functions , and other residues not typically associated with catalysis ( Figure 2A ) such as the conserved glycine at aligned position 151 ( G335 in 2z87 ) in the flexible G-loop and a histidine residue ( H386 in 2z87 ) in the C-terminal tail at aligned position 207 , henceforth referred to as the C-His . Residues from the G-loop in some families , such as the blood ABOs ( GT6 ) and glucosyl-3-phosphoglycerate synthases ( GpgS; GT81 ) , contribute to donor binding ( Patenaude et al . , 2002; Empadinhas et al . , 2011 ) . The C-His , likewise , coordinates with the metal ion and contributes to catalysis in a subset of GTs , such as polypeptide N-acetylgalactosaminyl transferases ( ppGalNAcTs; GT27 ) and lipopolysaccharyl-α−1 , 4-galactosyltransferase C ( LgtC; GT8 ) ( Fritz et al . , 2004; Persson et al . , 2001 ) . The conservation of these residues across diverse GT-A fold enzymes suggest that they likely perform similar functional roles in other families as well . The remaining core conserved residues include fourteen hydrophobic residues that are dispersed in sequence , but spatially cluster to connect the catalytic site and the Rossmann fold . Eleven out of the fourteen residues ( highlighted in yellow in Figure 2D ) are shared by other Rossmann fold proteins ( Figure 2—figure supplement 1 ) suggesting a role for these residues in maintaining the overall fold . Three hydrophobic residues ( V249 , F340 , F365; shown in red surface in Figure 2D ) , however , are unique to GT-A fold enzymes , and structurally bridge the αF helix ( containing the xED motif ) , the αD helix and the Rossmann fold domain . Although the functional significance of this hydrophobic coupling is not evident from crystal structures , in some families ( GT15 and GT55 ) the hydrophobic coupling between αF and the Rossmann fold domain is replaced by charged interactions ( Figure 2—figure supplement 2 ) . The structural and functional significance of these family specific variations are discussed below . Our broad evolutionary analysis also reveals three hypervariable regions ( HVs ) extending from the common core . These include an extended loop segment connecting β3 strand and αC helix ( HV1 ) , a segment longer than 28 amino acids connecting β6 and β7 strand ( HV2 ) and a C-terminal tail extending from the β8 strand ( HV3 ) in the common core . These HVs , while conserved within families , display significant conformational and sequence variability across families ( Figure 2A , Figure 2—figure supplement 3 ) and encode family-specific motifs that contribute to acceptor specificity in individual families , as discussed below . Having delineated the common core , we next sought to generate a phylogenetic tree relating diverse GT-A fold families using the core alignment . Because of the inherent challenges in the generation and visualization of large trees ( Sanderson and Driskell , 2003 ) , we used a representative set of GT-A fold sequences for phylogenetic analysis by first clustering the ~600 , 000 sequences into functional categories using a Bayesian Partitioning with Pattern Selection ( BPPS ) method ( Neuwald , 2014 ) . The BPPS method partitions sequences in a multiple sequence alignment into hierarchical sub-groups based on correlated residue patterns characteristic of each sub-group ( Materials and methods ) . This revealed 99 sub-groups with distinctive patterns . Representative sequences across diverse phyla from these sub-groups ( 993 sequences , Figure 3—source data 2 ) were then used to generate a phylogenetic tree ( Figure 3 ) . Based on the phylogenetic placement of these sequences , we broadly define fifty-three major sub-groups , thirty-one of which correspond to CAZy-defined families ( Figure 3—source data 1 ) . The remaining sub-groups correspond to sub-families within larger CAZy families . In particular , we sub-classified the largest GT family in the CAZy database , GT2 , into ten phylogenetically distinct sub-families . Likewise , GT8 and GT31 were classified into seven and five sub-families , respectively . These sub-families are not explicitly captured in CAZy and are annotated based on overall sequence similarity to functionally characterized members . For example , ‘GT2-LpsRelated’ corresponds to a sub-family within GT2 most closely related to the bacterial β−1–4-glucosyltransferases ( lgtF ) involved in Lipopolysaccharide biosynthesis ( Figure 3 , Figure 3—figure supplement 1 ) . Such a hierarchical classification captures the evolutionary relationships between GT-A fold families/sub-families while keeping the nomenclature consistent with CAZy . GT-A fold families and sub-families can be further grouped into clades based on shared sequence features and placement in the phylogenetic tree ( Figure 3 ) . For example , clade one groups four GT2 sub-families ( GT2-CeS , GT2-CWR , GT2-Chitin-HAS and GT2-Bre3 ) with GT84 and GT21 with high confidence , as determined by bootstrap values ( see Figure 3 legends ) . Members of these six families are all involved in either polysaccharide or glycosphingolipid biosynthesis . Additionally , the pattern-based classification identified a conserved [QR]XXRW motif in the C-terminal HV3 ( Figure 3—figure supplement 2 ) which is unique to members of this clade . The [QR]XXRW motif residues coordinate with the donor and acceptor in a bacterial cellulose synthase ( from GT2-CeS family ) ( Morgan et al . , 2013 ) and mutation of these residues in bacterial cyclic β−1 , 2-glucan synthetase ( Cgs , GT84 ) abrogates activity ( Ciocchini et al . , 2006 ) , suggesting a critical role of this motif in functional specialization of clade 1 GT-As . The GT8 sub-families form sub-clades within the larger clade 9 . For example , GT8 sequences involved in the biosynthesis of pectin components group together in the GT8-GAUT and GT8-GATL families ( Figure 3 ) . The human LARGE1 and LARGE2 GTs are multi-domain enzymes with two tandem GT-A domains . Their N-terminal GT-A domains fall into the GT8-Lrg subfamily that groups closely with GT8-xylosyltransferase ( GT8-XylT ) subfamily enzymes and places all the GT8 xylosyltransferases into a single well supported sub clade . The lipopolysaccharide α-glucosyltransferases ( GT8-LpsGlt ) group with the glucosyltransferases of the GT24 family , suggesting a common ancestor associated with glucose donor specificity . On the other hand , the GT8-Glycogenin sub family , which also includes members that transfer a glucose , is placed in a separate sub-clade , possibly indicating an early divergence for its unique ability to add glucose units to itself ( Alonso et al . , 1995 ) . Clade nine members also share common sequence features associated with substrate binding that includes a lysine residue within the commonly shared KPW motif in HV3 that coordinates with the phosphate group of the donor ( e . g . bacterial LgtC GT8-LpsGlt and other structures of clade nine members ) ( Figure 3—figure supplement 2 ) . We noticed that three out of four MGAT GT-A families responsible for the branching of N-glycans ( GT13 MGAT1 , GT16 MGAT2 and GT54 MGAT4 ) fall in the same clade ( clade 6 ) , as expected ( Figure 3 ) . In contrast , the fourth family , GT17 MGAT3 , which adds a bisecting GlcNAc to a core β-mannose with a β−1 , 4 linkage , is placed in a separate clade with GT14 and GT82 ( clade 7 ) , while a fifth MGAT member creating β−1 , 6-GlcNAc linkages ( GT18 MGAT5 ) is a GT-B fold enzyme ( Nagae et al . , 2018 ) . We further note that fifteen out of fifty-three GT-A families are found in both prokaryotes and eukaryotes . These fifteen families fall on different clades throughout the tree . GT-A families present only in prokaryotes , like GT81 , GT82 and GT88 , are also spread out in different clades ( Figure 3 ) . Similarly , other GT-A families that are present within restricted subsets of taxonomic groups ( like GT40 and GT60 present only in prokaryotes and protists ) are also scattered throughout the tree . These observations suggest that the divergence of most GT-A families predates the separation of prokaryotes and eukaryotes . To obtain insights into the evolution of catalytic mechanism , we annotated the phylogenetic tree based on known mechanisms of action ( inverting or retaining ) . Inverting GTs are colored in blue in the phylogenetic tree , while retaining GTs are colored in red ( Figure 3 ) . The dispersion of inverting and retaining families in multiple clades suggests that these catalytic mechanisms emerged independently multiple times during GT-A fold evolution . We find that natural perturbations in the catalytic base residue , an important distinction between the inverting and retaining mechanisms , correlates well with these multiple emergences across the tree . The residue that acts as a catalytic base for inverting GTs ( aspartate within the xED motif , xED-Asp ) is variable across the retaining families consistent with its lack of role in the retaining SNi mechanism ( Moremen and Haltiwanger , 2019 ) . In the inverting families , the xED-Asp is nearly always conserved and appropriately positioned to function as a catalytic base ( Figure 4A ) , though some exceptions have been noted ( Moremen and Haltiwanger , 2019; Gandini et al . , 2017 ) . Out of the five clades grouping inverting and retaining families , inverting families in three of these clades do not conserve the xED-Asp ( GT2-DPs , GT2-LpsRelated and GT43 ) . The heterogeneous nature of this residue in these families suggests that change of the catalytic base residue could be a key event in the transition between inverting and retaining mechanisms . Unlike families that conserve the xED-Asp , these families achieve inversion of stereochemistry through alternative modes that may relieve the constraints necessary to conserve the xED-Asp . For example , in GT43 , the Asp base is replaced by a glutamate residue , which shifts the reaction center by one carbon bond ( Moremen and Haltiwanger , 2019 ) . Further , the dolichol phosphate transferases ( DPMs and DPGs ) in the GT2-DP family , which lack the xED-Asp entirely , transfer sugars to a negatively charged acceptor substrate ( a phosphate group ) and thus do not need a catalytic base to initiate nucleophilic attack ( Gandini et al . , 2017 ) . Other GT-A inverting families lacking the xED-Asp ( GT12 , GT14 , GT17 , GT49 and GT82 ) are grouped into separate monophyletic clades segregating them from inverting families with the conserved xED-Asp ( Figure 3 ) . Out of these , only GT14 has representative crystal structures where a glutamate serves as the catalytic base ( Briggs and Hohenester , 2018 ) . For other inverting families with a non-conserved xED-Asp , residues from other structural regions may serve as a catalytic base . On the other hand , retaining families like GT64 conserve the xED-Asp , yet do not use it as a catalytic base . Thus , there may be multiple ways in which inverting and retaining mechanisms diverge , with one path being mutation of the xED-Asp catalytic base . One strongly supported clade that includes both inverting and retaining families is clade two that groups inverting GT-A family members that transfer sugars to phosphate acceptors ( GT2-DPs ) with three retaining GT-A families that also have phosphate-linked acceptors ( GT55 , GT78 and GT81 ) . This placement is further supported by the observation that these families share structurally equivalent conserved residues in the HV2 region that coordinate the phosphate group of the acceptor . In the GT2-DP subfamily , R117 , R131 and S135 ( Figure 5A ) in HV2 coordinate with the acceptor phosphate groups . The conservation of these residues in GT55 and GT81 suggests that they likely perform similar interactions in these latter subclades . Indeed , in the crystal structure of M . tuberculosis GpgS ( GT81 ) , HV2 adopts a conformation similar to GT2-DPs and the shared residues G184 , R185 and T187 ( equivalent to R117 , R131 and S135 ) form similar interactions with the phosphate group of the acceptor ( Figure 3—figure supplement 2 ) . Clade five places the inverting GT7 and GT2-CHS with the retaining GT27 and GT60 families ( Figure 3 ) . This supports the evolution of these families from a close common ancestor through gene duplication and divergence , which has been suggested through structural similarities between GT7 and GT27 ( Ramakrishnan and Qasba , 2010b ) . After this initial divergence in mechanism within clade 5 , the subclades group the β−1 , 4-GalNAc transferase domains of bacterial and protist chondroitin polymerases ( involved in the elongation of glycosaminoglycan chondroitin ) ( GT2-CHS ) with the GT7 family . The GT7 family includes the higher organism counterparts of the β−1 , 4-GalNAc transferase domains of chondroitin synthases , along with β−1 , 4-Gal transferases . The close placement of GT60 and GT27 families in this clade is also directly supported by previous literature indicating that these families share a conserved mode of polypeptide Ser/Thr O-glycosylation ( Heise et al . , 2009 ) . Clade five thus consolidates previous independent findings and suggests a shared ancestor , potentially extending the common ancestry of GT2-CHS and GT7 to include GT27 and GT60 , with an ancestral divergence in mechanism . Analysis of the patterns of conservation and variation in the common core indicates that each residue position within the core has been mutated in some context during the course of evolution , highlighting the tolerance of the GT-A fold to extensive sequence variation . While some of these variations are confined to specific clades or families , such as replacement of DxD motif with DxH motif in GT27 and GT60 , other variations are found independently across distal clades ( Figure 4A ) . For example , GT14 and prokaryotic members of GT6 that fall on different clades , have independently lost the DxD motif and no longer require a metal ion for activity ( Briggs and Hohenester , 2018; Pham et al . , 2014 ) . The C-His is also lost independently in multiple clades ( Figure 4A ) . In order to investigate how the loss of metal binding C-His is compensated , we analyzed the C-His-metal ion interactions across all available crystal structures . Structural alignment of GT-A families lacking the C-His such as GT13 , GT6 and GT64 families revealed a water molecule coordinating the metal ion in a manner similar to the C-His sidechain ( Figure 4B ) . In other families , such as GT24 , we found that the C-His is substituted by an aspartate ( D1427 ) , which coordinates with the metal ion similar to C-His ( Figure 4B , bottom panel ) . Likewise , the conserved hydrophobic coupling between αF helix and the Rossmann domain is replaced by charged interactions ( R388 and E274 , respectively ) in some retaining GTs such as GT15 and GT55 ( Figure 2—figure supplement 2 ) . These substitutions point to the ability of GT-As to accommodate changes , even in conserved positions at the core , through compensatory mechanisms . The HV regions show significant variability across GT-A families and extend from the common core to perform various roles from substrate binding to large conformational changes that position the donor and acceptor substrates for the enzymatic reaction ( Jamaluddin et al . , 2007; Tsutsui et al . , 2013; Albesa-Jové et al . , 2017 ) . Mutations within these HV regions , for example , at aligned position 126 in the HV2 region ( Y177A , G in 4lw6 , GT7 ) , have also been shown to induce a shift in acceptor specificity ( Tsutsui et al . , 2013 ) . Despite significant sequence variability , we find that these HV regions in fact conserve family specific residues that contribute to acceptor specificity . For example , a distinctive arginine ( R117 ) and aspartate ( D154 ) along with R131 and serine S135 within the HV2 of DPM1 ( GT2-DP sub-family ) contribute to specificity towards a dolichol phosphate acceptor by creating a charged binding pocket for the phosphate group ( Figure 5A ) . Likewise , family-specific residues ( R198 , H221 and E224 in 5vcm ) within the HV1 of MGAT2 ( GT16 ) form a unique scaffold for recognizing the terminal GlcNAc of the N-glycan acceptor ( Figure 5B ) . Similarly , the C-terminal GT64 domain of the multidomain EXTLs contain specific residues in HV2 ( R181 and Y193 ) and HV3 ( H289 and R293 ) that form a unique binding pocket for the tetrasaccharide linker acceptor used to synthesize glycosaminoglycans ( Figure 5C ) . Together these examples illustrate the ability of HVs to evolve family specific motifs to recognize different acceptors . As discussed above , the conserved catalytic residues dictate the mechanism of sugar transfer and metal binding while the extended HVs use family specific motifs to dictate acceptor specificity . We also find some clade specific features ( such as the conserved Lys in clade 9 , and QXXRW in clade 1 ) and G-loop residues involved in donor binding , however , the overall framework that dictates donor sugar specificity in GTs is largely unknown . Sequence homology alone is insufficient to predict donor specificity because evolutionarily divergent families can bind to common substrates , and sometimes even two closely related sequences bind to different donors ( Figure 6—figure supplement 1; Patenaude et al . , 2002 ) . For a subset of GT-B fold families , ML methods have been successfully applied towards predicting substrate specificities ( Yang et al . , 2018 ) . Our global analysis provides a comparative basis to expand such methods and contrast sequences that bind different donors across all GT-A families . To test whether evolutionary features gleaned from this global analysis can be used to better predict donor substrate specificity , we employed a ML framework that learns from the specificity-determining residues of functionally characterized enzymes to predict specificity of understudied sequences . In brief , using an alignment of a well curated set of 713 GT-A sequences ( Figure 6—source data 1 , Figure 6—figure supplement 2—source data 1 , Figure 6—figure supplement 2 ) with known donor sugars , we derived five amino acid properties ( hydrophobicity , polarity , charge , side chain volume and accessible surface area ) from each aligned position within the common core . These properties were then used as features to train multiple ML models . Among the seven methods used , the gradient-boosted regression tree ( GDBT ) model achieved the best prediction performance ( accuracy ~90% ) based on a 10-fold cross validation ( CV ) using 239 contributing features ( Figure 6A , B , Figure 7—source data 1 ) . This model adds an ensemble boosting to tree based learners used for predicting GT1 substrate specificities ( Yang et al . , 2018 ) . To further validate the model , we tested its performance on a validation set of 64 sequences that were not used to train the ML model but have known sugar specificities . The GDBT classifier correctly predicted donor substrates for 92% of these sequences , 89% of which were predicted with high confidence ( blue rows in Figure 6—source data 2 ) . The GDBT model was then used to predict donor sugars for GT-A domains with unknown specificities from five organisms: H . sapiens , C . elegans , D . melanogaster , A . thaliana and S . cerevisiae ( Figure 6—source data 2 ) . Each prediction is associated with a confidence level derived from the probability for each of the six donor classes ( Materials and methods ) . Nearly 77% of the predictions have high and moderate confidence levels and present good candidates for further investigation ( Figure 6D ) . The remaining 23% of the predictions are low confidence . This likely reflects their promiscuity for donor preferences , as seen across many GT-As ( Empadinhas et al . , 2011; Blixt et al . , 1999 ) , or non-catalytic GT-As like C1GALT1C1 ( Cosmc ) ( Aryal et al . , 2012 ) . Our predictions assign putative donors for 10 uncharacterized human GT-A domains ( Figure 6—source data 2 ) . B3GNT9 is predicted to employ UDP-GlcNAc with high confidence like other GT31 β−3-N-acetylglucosaminyltransferases ( B3GNTs ) in humans ( Togayachi et al . , 2010 ) . The two procollagen galactosyltransferases in humans ( COLGALT1 and COLGALT2 ) are multidomain proteins with two tandem GT-A domains . While their respective C-terminal domains catalyze β-Gal addition to hydroxylysine side-chains in collagen ( Schegg et al . , 2009 ) , our predictions assign a putative GlcNAc and Glc transferase role for their N-terminal GT domains , respectively . More interestingly , GLT8D1 , a GT8 GT with an unknown function implicated in neurodegenerative diseases ( Cooper-Knock et al . , 2019 ) , is predicted to have a glucosyltransferase specificity . In other organisms , the GT2 sequences in A . thaliana ( mostly involved in plant cell wall biosynthesis ) are predicted to bind glucose and mannose substrates , the primary components of the plant cell wall ( Figure 6—source data 2 ) . We also identify a novel N-acetylglucosyltransferase function for a GT25 enzyme in C . elegans . These predictions can guide characterization of new GT sequences with unknown functions . We next sought to identify features that contribute most to substrate ( donor ) prediction . To do this , we rank ordered the 239 features based on their contribution to predicting a donor subtype using a six way classification ( six donors ) ( Materials and methods ) . This revealed that the most contributing features of the GDBT model also contribute significantly to at least one specific donor type prediction , thereby enabling new inferences to be drawn between residue properties and donor sugar specificity ( Figure 7A , Figure 7—source data 1 ) . As expected , some of the most contributing features include residues directly involved in substrate binding and catalytic functions such as the Asp within the DxD motif , residues in the G-loop , the catalytic base and the C-His ( Patenaude et al . , 2002; Empadinhas et al . , 2011; Gandini et al . , 2017 ) . Additionally , multiple residues from the alpha-C helix ( aligned position 65–72; Y217-N224 in 2z87 ) immediately following HV1 are also identified as key specificity determining residues . The C-helix is positioned close to the donor sugar binding pocket and many residues from this region have been shown to play roles in donor binding ( Gagnon et al . , 2018; Schuman et al . , 2010; McArthur and Chen , 2016 ) . For example , Ramakrishnan et . al . showed that mutation of a single residue at position 67 in bovine β−1 , 4-galactosyltransferase T1 ( R228K in 1o0r , GT7 ) resulted in relocation of the catalytic base and a change of donor specificity from Gal to Glc ( Figure 7B; Ramakrishnan et al . , 2005; Hancock et al . , 2006 ) . Our analysis identifies volume , polarity and accessible surface area of the residue at position 67 as an important contributor to donor specificity ( Figure 7 ) . In addition , our analysis identifies residue volume at position 149 as an important determinant of Gal specificity . Consistent with this observation , mutation of Y289 ( position 149 in the consensus sequence ) by a leucine broadens the specificity from Gal to GalNAc by creating additional space for accommodating the N-acetyl moiety ( Hancock et al . , 2006; Ramakrishnan and Qasba , 2002 ) . While some of the highly ranked features are directly involved in donor binding , many others ( such as aligned position 77 , 88 , 155 and 159 , green sticks in Figure 7B , D ) are distal from the donor binding site and are not directly involved in donor binding . An example of allostery has been observed in the human GT6 blood ABO α−1 , 3-galactosyltransfearse where mutation of a proline at position 117 ( P234S in 5c4c ) results in an alternative conformation of a methionine at position 150 ( M266 in 5c4c ) allowing for the accommodation of GalNAc instead of Gal ( Hancock et al . , 2006; Marcus et al . , 2003 ) . Further , a Random forest model trained using features from only the donor binding residues performs with an accuracy of only 75% , indicating the importance of features other than those directly involved in donor binding . Thus , despite only a few residues being directly involved in donor interactions , additional contributions to donor specificity come from residues more distal from the active site . Contributions from these peripheral secondary shell features surrounding the donor binding site ( Figure 7B–E ) highlight the potential role of higher order ( allosteric ) interactions in determining donor substrate specificity . Prior studies on the evolution of GTs have generally focused either on distinct GT subfamilies or biosynthetic pathways with additional structural classifications of GTs into one of three distinct protein fold superfamilies ( Moremen and Haltiwanger , 2019; Taujale and Yin , 2015; Lombard , 2016 ) . In our present work we focused on the analysis of the largest of the GT superfamilies , those that comprise a GT-A protein fold characterized by an extended Rossmann domain with associated conserved helical segments . These enzymes generally employ the Rossmann domain for nucleotide sugar donor interactions and extended loop regions for acceptor glycan interactions ( Moremen and Haltiwanger , 2019 ) . Using an unbiased profile search strategy , we assembled a total of over 600 , 000 GT-A fold related sequences from all domains of life for deep evolutionary analysis . To support this profile-based assembly , we leveraged structural alignments on GT-A fold enzymes in PDB and secondary structure predictions when no crystal structures were available . The resulting alignment allowed the definition of a common structural core shared among the diverse GT-A fold enzymes and defined positions where hypervariable loop insertions were elaborated to provide additional functional diversification ( Figure 2 ) . In cases where data was available for enzyme-acceptor complexes these latter loop insertions generally contribute to unique , family specific acceptor interactions . Thus , a structural framework is presented for GT-A fold enzyme evolution . Since the common core is present across all kingdoms of life , it presumably represents the minimal ancestral structural unit for GT-A fold catalytic function by defining donor substrate interactions and minimal elements for acceptor recognition and catalysis . In fact , we find several archaeal and bacterial sequences that closely resemble this common core consensus sequence ( Supplementary file 3 ) . Based on our studies , we propose a progressive diversification of GT function through evolution of donor specificity by accumulation of mutations in the common core region and divergence in acceptor recognition through expansion of the hypervariable loop regions . Consistent with this view , we find conserved family-specific motifs within the hypervariable regions that confer unique acceptor specificities in various families . These expansions likely contributed to the evolution of new GT functions and catalyzed new glycan diversification observed in all domains of life . A surprising finding from our studies is the dispersion of inverting and retaining catalytic mechanisms among families in the GT-A fold evolutionary tree ( Figure 3 ) . Recent models indicate that distinctions between inverting and retaining catalytic mechanisms arise from differences in the angle of nucleophilic attack by the acceptor toward the anomeric center of the donor sugar ( Moremen and Haltiwanger , 2019 ) . Inverting mechanisms require an in-line attack and direct displacement by the nucleophile relative to the departing nucleotide diphosphate of the sugar donor and a conserved placement of the xED-Asp carboxyl group as catalytic base at the beginning of the αF helix . In contrast , retaining enzymes generally alter the angle of nucleophilic attack by the acceptor , use a donor phosphate oxygen as catalytic base , and employ a dissociative mechanism for sugar transfer ( Moremen and Haltiwanger , 2019 ) . The fundamental differences in these catalytic strategies would suggest an early divergence of enzymes employing these respective mechanisms . However , the GT-A fold phylogenetic tree strongly suggests that inverting and retaining mechanisms evolved independently at multiple points in the evolution of GT-A families ( Figure 3 ) . Since the main difference in these mechanisms is the change in position of the nucleophilic hydroxyl and catalytic base , substitutions at the catalytic base may have served as a catalyzing event in switching between mechanisms . The xED-Asp carboxyl group is highly conserved in the inverting enzymes and is appropriately placed for acceptor deprotonation . Variants of this motif either lack the residue entirely , as seen in many retaining enzymes , or use compensatory modes to accommodate changes at this position , as seen for the inverting enzymes in GT43 , GT2-DPs , and GT2-LPSRelated . In fact , in each of the latter cases the respective inverting GT family is clustered with closely related GT families employing a retaining catalytic mechanism . Thus , inverting enzyme variants that accommodate changes to the xED motif group may represent examples of transitional phases in evolution between inverting and retaining catalytic mechanisms . Other inverting enzymes harboring variants in the xED motif segregate into separate clades and could represent outlier families that have developed alternative ways to compensate for the loss of xED-Asp . This ability to evolve distinct catalytic strategies , in some cases through presumed convergent evolution , could allow each family to evolve independent capabilities for donor and acceptor interactions as well as for anomeric linkage of sugar transfer , while retaining other essential aspects of protein structural integrity through the use of a conserved and stable Rossmann fold core . In an effort to define the sequence constraints for the respective catalytic mechanisms we also employed a ML framework for prediction of the mechanism for unknown sequences and were able to assign the donor sugar nucleotide for a test set of enzymes with high accuracy . Our model expands on the approaches used in previous ML efforts focused on the GT1 family of GT-B fold GTs ( Yang et al . , 2018 ) . The phylogenetic and comparative framework presented here enables expansion of such models across all GT-A fold families with improved prediction accuracies . As additional functional data on GTs become available , the proposed ML framework can be extended to predict acceptor specificity and catalytic mechanisms , as described for the GT1 family . Surprisingly , the contributing features for accurate donor prediction include residues involved in donor binding as well as positions that are distal to the active site that likely contribute through secondary shell effects or allosteric interactions . Due to their indirect involvement , such positions are generally difficult to pinpoint using structural studies alone emphasizing the need for complementary ML-based approaches in investigating GT functional specialization . Numerous additional insights into GT function were also revealed through inspection of the aligned sequences and the phylogenetic tree . For example , the clustering of mammalian N-glycan GlcNAc branching enzymes ( MGAT1 ( GT13 ) , MGAT2 ( GT16 ) , and MGAT4 ( GT54 ) ) in the same clade suggests a common origin for these enzymes , while placement of MGAT3 ( GT17 ) in a separate clade could point to its unique role in adding a bisecting GlcNAc to the N-glycan core thereby regulating N-glycan extension ( Ikeda et al . , 2014 ) . In contrast , MGAT5 ( GT18 ) involved in N-glycan β1 , 6-GlcNAc branching is a GT-B fold enzyme with a clearly distinct evolutionary origin . While most clades are well resolved , bootstrap support values for nodes at the base of the tree are low and need to be interpreted with caution . This low resolution results from high divergence between families and possibly other events like horizontal gene transfer and convergent evolution . However , trees generated using alternative strategies support the overall topology ( Figure 3—figure supplement 3 ) and clades are congruent with clusters obtained using an orthogonal Bayesian classification scheme , which adds confidence to the phylogeny ( Figure 3—source data 1 ) . For some GT-A fold enzymes variations in the catalytic site can also be accommodated by other compensatory changes . An example is the use of the C-His motif for coordination of the divalent cation in most GT-A fold enzymes in contrast with enzyme variants that employ water molecules to compensate for the loss of this residue ( Figure 4B ) . Similarly , some inverting GTs dispense with the use of the divalent cation and the DxD motif and substitute interactions with the sugar donor through use of basic side chains ( e . g . GT14 ) . A further extreme is the duplication , divergence and pseudogenization within the GT31 family . Human C1GALT1C1 ( GT31 , COSMC ) shares a high sequence similarity to another GT31 member , C1GALT1 ( T-synthase ) , yet COSMC has lost both the DxD and the xED motifs and has no catalytic activity . Instead , COSMC acts as an important scaffold and chaperone for the proper assembly and catalytic function of T-synthase ( Aryal et al . , 2012 ) . The ability of GT-As to harbor such structural variations that allow them to develop new functions make them well-suited to evolve rapidly and facilitate the synthesis of a diverse repertoire of glycans across all living organisms . Our unbiased , top-down sequence-based analysis suggests new and unanticipated evolutionary relationships among the GT-A fold enzymes . Prior suggestions of such relationships have been inferred by the clustering of GT sequences into families in the CAZy database . However , the CAZy database of GT sequences does not provide access to the broader sequence relationships among the GT-A fold enzymes or how a general model of a core conserved GT-A fold scaffold can serve as a progenitor catalytic platform for binding sugar donors and facilitating glycan extension . The sequence assembly , phylogenetic tree , and placement within the framework of known GT-A fold structures in the present studies provide key insights into conserved elements of the hydrophobic core , linkage to the DxD motif for cation and sugar donor interactions , and the conserved αF helix harboring the xED catalytic base . Additional hypervariable extensions at defined positions from this conserved core were then progressively recruited to confer unique modes of acceptor interactions to develop new specificities and evolve new functions . Thus , the core of the protein scaffold can be maintained to facilitate protein stability while rapid evolution of the hypervariable loops can develop new glycan synthetic functionalities through presentation of novel acceptors to the catalytic site . Variation in the location of the acceptor hydroxyl nucleophile relative to the donor sugar anomeric center presents the opportunity for distinctions in catalytic mechanism and anomeric outcome for sugar transfer . The result is a rapidly evolving set of GT enzymatic templates as the biosynthetic machinery for diverse glycan extension on cell surface and secreted glycoproteins and glycolipids . In such contexts the resulting glycoconjugates confer potential functional selective advantages at the cell surface , but also act as ligands and pathogen entry points for negative evolutionary pressure . These positive and negative selective pressures which force organisms to constantly adapt to an ever-changing environment is known as the Red Queen Hypothesis . These red queen effects on glycan synthesis have led to the remarkable diversity in GT enzymes and their resulting glycan structural products . We anticipate that the sequence and structural principles that drive GT-A fold evolution will also likely extend to GT-B and GT-C fold enzymes and represent a common theme for the elaboration of diverse glycan structures in all domains of life . A select representative set of structures were collected from all Rossmann-fold containing protein domains using the SCOP database ( Andreeva et al . , 2014 ) . mTM-align ( Dong et al . , 2018 ) was used to align these structures with a subset of GT-A structures ( Figure 2—figure supplement 1 ) . A representative subset of 24 , 650 GT-A sequences were generated from the ~600 , 000 putative GT-A sequences by using a family-wise sequence similarity filtering ( only keep <70% similar sequences;<50% for GT2 and GT8 families ) . This sequence set was then used to apply the Optimal multiple-category Bayesian Partitioning with Pattern Selection ( omcBPPS ) scheme ( Neuwald , 2014 ) . omcBPPS identifies patterns of column-wise amino acid conservation and variation in the multiple sequence alignment . The resulting family specific positions were then used as statistical measures to classify the GT-As into 99 unique sets that correspond to the 53 families described in this study ( Figure 3—source data 1 ) . omcBPPS also identified aligned positions that are conserved across all GT-A fold families . This revealed the 20 conserved positions within the core component , that were also verified by calculating conservation scores using the Jensen-Shannon divergence score as described and implemented by Capra and Singh ( 2007 ) ( used in Figure 2A ) . The GT-A sequences were first classified into pattern-based groups using omcBPPS . Based on the placement of representative sequences from these groups in the phylogenetic tree , they were merged into GT-A families and sub-families . The correspondence between the 53 GT-A families and subfamilies with the 99 pattern-based groups are provided in Figure 3—source data 1 . Sequences from some families did not form any distinct pattern-based groups due to either a low number of sequences for a statistically significant grouping ( GT78 ) or a lack of distinguishing patterns within the aligned positions ( GT25 , GT88 ) . Representative sequences for these families were collected from the seed alignments for these families as described above . We also identified the N-terminal GT2 domain of the multidomain chondroitin polymerase structure from E . coli ( Pdb Id: 2z87 ) as the prototypic GT-A structure to use as a comparative basis for structural analyses . This sequence was selected based on the lowest E-value and highest similarity score of a BLAST search of all pdb structures against the GT-A consensus sequences . Weblogos for the conserved active site residues were derived for each GT-A subfamily using Weblogo 3 . 6 . 0 ( Crooks et al . , 2004 ) . A sequence similarity network was generated to evaluate the general predictability of a donor substrate based on sequence homology alone . Using an Edge-Weighted Spring Embedded Layout ( Kamada and Kawai , 1989 ) in Cytoscape ( Shannon et al . , 2003 ) with the same sequence dataset as the ML data , we produced a series of homologous networks constrained by an E-value cut-off of 0 . 05 . The ancient archaeal and bacterial GT-A domain sequences listed in Supplementary file 3 represent sequences with the minimal GT-A domains ( no long inserts , no additional domains ) in prokaryotic organisms . These sequences could represent the most ancient progenitors of the GT-A fold families . These were selected based on closest homology to a consensus sequence derived from the GT-A profile alignments . First , a BLAST ( Altschul et al . , 1990 ) search was conducted with the consensus sequence on the NCBI nr database . The 500 hits with an e-value better than 1e-10 were selected . These hits were then aligned to the GT-A profiles using mapgaps . Using this alignment , these sequences were further filtered to include hits that had a ) more than 140 aligned positions to remove fragmentary hits , b ) less than 20 insert positions throughout the GT-A domain alignment , c ) no more than 100 amino acids in the N-terminal region and d ) no more than 200 amino acids in the C-terminal region .	Carbohydrates are one of the major groups of large biological molecules that regulate nearly all aspects of life . Yet , unlike DNA or proteins , carbohydrates are made without a template to follow . Instead , these molecules are built from a set of sugar-based building blocks by the intricate activities of a large and diverse family of enzymes known as glycosyltransferases . An incomplete understanding of how glycosyltransferases recognize and build diverse carbohydrates presents a major bottleneck in developing therapeutic strategies for diseases associated with abnormalities in these enzymes . It also limits efforts to engineer these enzymes for biotechnology applications and biofuel production . Taujale et al . have now used evolutionary approaches to map the evolution of a major subset of glycosyltransferases from species across the tree of life to understand how these enzymes evolved such precise mechanisms to build diverse carbohydrates . First , a minimal structural unit was defined based on being shared among a group of over half a million unique glycosyltransferase enzymes with different activities . Further analysis then showed that the diverse activities of these enzymes evolved through the accumulation of mutations within this structural unit , as well as in much more variable regions in the enzyme that extend from the minimal unit . Taujale et al . then built an extended family tree for this collection of glycosyltransferases and details of the evolutionary relationships between the enzymes helped them to create a machine learning framework that could predict which sugar-containing molecules were the raw materials for a given glycosyltransferase . This framework could make predictions with nearly 90% accuracy based only on information that can be deciphered from the gene for that enzyme . These findings will provide scientists with new hypotheses for investigating the complex relationships connecting the genetic information about glycosyltransferases with their structures and activities . Further refinement of the machine learning framework may eventually enable the design of enzymes with properties that are desirable for applications in biotechnology .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "computational", "and", "systems", "biology" ]	2020	Deep evolutionary analysis reveals the design principles of fold A glycosyltransferases
Recent hypotheses have posited that orbital frontal cortex ( OFC ) is important for using inferred consequences to guide behavior . Less clear is OFC’s contribution to goal-directed or model-based behavior , where the decision to act is controlled by previous experience with the consequence or outcome . Investigating OFC’s role in learning about changed outcomes separate from decision-making is not trivial and often the two are confounded . Here we adapted an incentive learning task to mice , where we investigated processes controlling experience-based outcome updating independent from inferred action control . We found chemogenetic OFC attenuation did not alter the ability to perceive motivational state-induced changes in outcome value but did prevent the experience-based updating of this change . Optogenetic inhibition of OFC excitatory neuron activity selectively when experiencing an outcome change disrupted the ability to update , leaving mice unable to infer the appropriate behavior . Our findings support a role for OFC in learning that controls decision-making . In order to examine state-dependent control over decision-making , we examined how changes in state alter action control independent of changes in action contingency or direct changes in outcome sensory and motivational properties . We adapted an incentive learning task previously used in rats ( Balleine and Dickinson , 1998; Wassum et al . , 2009 ) to mice . In brief , mice were trained to press a left lever under a random ratio schedule ( up to RR4 or RR8 ) to gain access to a right lever . One subsequent right lever press ( fixed ratio 1 ) resulted in 20% sucrose solution delivery to outcome port connected to a lickometer located in between the two levers . The left—then—right lever press chain used produces a distal seeking response relative to outcome delivery that is less sensitive to general motivational state changes than the subsequent more proximal taking response ( Balleine and Dickinson , 2005 ) . After acquisition , we altered the motivational state and mice were given a re-exposure period to non-contingent deliveries of the same sucrose solution . In this re-exposure phase , mice had the opportunity to experience the sucrose in a changed motivational state and undergo incentive learning . The next day , we assess whether the new motivational state induced a change in sucrose valuation by examining seeking left lever presses in a brief ( 5 min ) non-reinforced session . We used daily time spent under food restriction to induce different motivational states . We first trained adult ( >7 weeks ) male and female mice under very minimal levels of food restriction; rodent chow was removed from animals for 2 hr beginning between 1 . 5 and 3 hr into their light cycle ( Vollmers et al . , 2009 ) . Immediately at the end of the 2 hr food restriction , mice were placed into the operant chamber for instrumental training . After training , mice were returned to their home-cage with rodent chow readily available . Mice showed a slight increase in body weight across training ( male: baseline weight = 24 . 55 ± 0 . 69 SEM , last day of training weight = 26 . 85 ± 0 . 75 SEM; paired t-test; t7 = 5 . 33 , p=0 . 001 ) ( female: baseline weight = 17 . 79 ± 0 . 53 , last day of training weight = 19 . 49 ± 0 . 39; paired t test , t7 = 4 . 66 , p=0 . 002 ) . This suggests that the 22 hr mice had access to home-cage rodent chow was a sufficient time to maintain and increase weight . Importantly , mice had unlimited access to water except during their training sessions in the operant box . Not surprisingly , given the minimum level of food restriction , a moderate percentage of mice failed to show evidence of lever-press acquisition for a sucrose solution . We applied a minimum response rate >0 . 25 left lever presses per minute ( minimum 15 left lever presses in one session ) on the last two days of acquisition to ensure access to right lever , and we confirmed that mice had indeed pressed the right lever for sucrose delivery . This resulted in a 31% subject loss ( n = 7/22 ) . The remaining subjects showed clear acquisition of left lever presses ( last day of training average; left lever presses = 46 ± 6 . 5; response rate = 0 . 58 ± 0 . 08 ) and earned on average 4 . 3 ± 0 . 85 sucrose deliveries on the last day of training ( Figure 1—figure supplement 1A–C ) . Mice were then divided into two groups matched for response rates on the last two days of training . One group was maintained at 2 hr food restriction ( Group 2–2 ) ( n = 6 ) , while the other group underwent a 16 hr food restriction ( Group 2–16 ) ( n = 11 ) ( rodent chow removed 2–4 hr prior to dark cycle onset ) . At the end of each restriction period , mice were placed into the operant chamber where sucrose was delivered non-contingently on a random time schedule ( RT120 ) , equating to on average one sucrose delivery every 2 min for an average of 30 sucrose deliveries across the 60 min session . We performed lick analyses on the observed anticipatory and consummatory licking behavior ( Figure 1B–J ) . Changes in licking behavior reflect palatability changes induced by the food restriction state change ( Berridge , 1991 ) . An increase to 16 hr food restriction produced an increase in total licks ( unpaired t-test: t14 = 2 . 39 , p=0 . 03 ) ( Figure 1B ) that was observed across the entire duration of the session ( 2-way repeated measures ANOVA ( Group x Time block ) : no interaction; trend toward main effect of Group F ( 1 , 14 ) =3 . 28 , p=0 . 09; trend toward main effect of Time block F ( 5 , 70 ) =2 . 26 , p=0 . 05 ) ( Figure 1C ) . Mice that had undergone 16 hr of food restriction trended towards being faster to initiate licking behavior ( t14 = 2 . 14 , p=0 . 05 ) ( Figure 1D ) . While all mice organized their licks into bouts of licking instead of isolated licks ( p>0 . 5 ) with similar inter-lick intervals within a bout ( p>0 . 5 ) , and similar bout durations ( p>0 . 1 ) , Group 2–16 contained more licks in a bout ( t14 = 5 . 05 , p=0 . 0002 ) ( Figure 1E–H ) . The above analyses did not discriminate between anticipatory and consummatory licking patterns . To examine whether a motivational state change induced a different pattern of licking during sucrose consumption , we isolated the first lick burst following a sucrose delivery . In the 2–16 group , we found a trend towards an increase in burst duration following outcome delivery ( t14 = 1 . 79 , p=0 . 09 ) ( Figure 1I ) , as well as a trend towards an increase in the number of licks within that first burst ( t14 = 1 . 98 , p=0 . 07 ) ( Figure 1J ) . Together , these data suggest that an increase in food-restriction resulted in a state change that produced an increase in palatable licking behavior . Hence , the motivational state produced by increased food restriction increased the palatable value of sucrose . We next examined whether the state-dependent updated incentive value was retrieved and used to control decision-making . Mice were maintained in the assigned food-restriction state and given a brief non-reinforced test session where we measured the rate of left lever pressing . State-dependent increases in sucrose value increased the outcome value as indexed by a higher response rate ( Figure 1K ) . There was a trend toward Group 2–16 mice to have a higher percent of baseline response rate than Group 2–2 mice ( unpaired t-test: t14 = 2 . 10 , p=0 . 05 ) . One-sample t-test against 100% to assess changes from baseline showed that mice kept at 2 hr restriction had a similar response rate to that observed on the last two days of acquisition ( t4 = 1 . 19 , p>0 . 2 ) . However , mice that underwent 16 hr food restriction showed an increase in response rate from baseline ( one-sample t-test against 100%: t10 = 3 . 41 , p=0 . 007 ) . To assess the contribution of context and the necessity of sucrose re-exposure to the behavioral effects observed , we performed an additional positive incentive learning experiment in naïve mice where we manipulated access to the previously trained context or sucrose during the re-exposure session ( Figure 1L ) . Mice were trained on the incentive task under 2 hr food restriction . Prior to re-exposure session , all mice underwent 16 hr of food restriction , and kept at 16 hr food restriction during the test . As seen previously in rats ( Balleine et al . , 1995 ) , the ability of the updated value to control decision-making was dependent on re-exposure to sucrose in the new motivational state , as re-exposure to the context alone ( no sucrose delivered ) was insufficient to change action control ( Figure 1M ) . Further , the training context contributed minimally to sucrose re-exposure , as sucrose re-exposure in a novel context was sufficient to update sucrose value and increase seeking response rates ( Figure 1M ) . A two-way ANOVA ( Sucrose x Context ) did not show a significant interaction ( F ( 1 , 29 ) =0 . 13 , p=0 . 7 ) or a significant effect of Context ( F ( 1 , 29 ) = 0 . 8 , p=0 . 38 ) . However , a main effect of Sucrose was observed ( F ( 1 , 29 ) = 4 . 96 , p=0 . 03 ) . Hence , mice readily show positive incentive learning , whereby a change in action control following an increase in motivational state requires state-dependent experience to update value representations later used for goal-directed decision-making . While we readily observed positive incentive learning in mice , it may be that an increase in value ( in this case induced by an increase in hunger ) exerts more control over decision-making than a decrease in value . However , outcome devaluation experiments used to probe goal-directed control suggests that mice are indeed sensitive to decreases in outcome value ( e . g . , Gourley et al . , 2016; Gremel et al . , 2016 ) . In incentive learning , experiencing food in an altered motivational state drives the relative change in outcome value , while in outcome devaluation tests sensory-specific outcome satiation or direct aversive conditioning to the outcome representation ( lithium chloride pairings with outcome ) is commonly used to directly change outcome value . Effects of sensory-specific outcome devaluation are often compared to a control state where subjects are pre-fed a control outcome to control for general effects of satiation ( e . g . Gremel and Costa , 2013; Gremel et al . , 2016 ) and tested in a non-reinforced session without any re-exposure to the trained outcome . Incentive learning would arise if those subjects in the sated control condition ( Valued day ) were given a reinforced session , where subjects would lever press and experience the outcome in the sated state and undergo incentive learning ( Balleine and Dickinson , 2005 ) . We next examined the capacity for negative incentive learning in mice and asked whether a state-dependent decrease in value will also alter action control . Mice underwent 16 hr daily food restriction across lever press acquisition . Food was removed from cages 3–4 hr prior to dark cycle onset , and mice were trained and tested 2–3 hr into their next light cycle . Mice were able to maintain and increase their baseline weight across acquisition ( male: prior to training = 21 . 8 g ± 0 . 36; last training day = 24 . 5 g ± 0 . 51; paired t-test: t7 = 14 . 74 , p<0 . 0001 ) ( female: prior to training = 15 . 9 ± 0 . 51; last training day = 18 . 5 g ± 0 . 61; paired t-test: t6 = 11 . 02 , p<0 . 0001 ) . Hence , the 6–7 hr where food was present was sufficient to maintain baseline bodyweight . In contrast to positive incentive learning , mice readily acquired lever press training ( only 1/17 removed for low response rate ) ( Figure 2—figure supplement 1A–C ) . We then induced a change in motivational state . One group of mice was kept on 16 hr food restriction ( Group 16–16 ) ( n = 8 ) , while for the other group food restriction was reduced to only 2 hr ( Group 16–2 ) ( n = 7 ) prior to the sucrose re-exposure session . The reduced duration of food restriction affected appetitive and consummatory licking behaviors measured during the re-exposure session . Reducing food restriction to 2 hr led to a decrease in the total number of licks measured ( unpaired t-test: t13 = 3 . 93 , p=0 . 0017 ) ( Figure 2B ) that was present across the first half of the session ( two-way repeated measures ANOVA ( Group x Time block ) : interaction F ( 5 , 65 ) =6 . 95 , p<0 . 001; main effect of Group F ( 1 , 13 ) =15 . 41 , p=0 . 001; main effect of Time block F ( 5 , 65 ) =15 . 68 , p<0 . 0001 ) ( Figure 2C ) . Groups showed a similar latency to the first lick in a session ( p>0 . 1 ) ( Figure 2D ) . However , Group 16–2 mice organized fewer of their licks into bursts ( t13 = 2 . 82 , p=0 . 02 ) , and those bursts were shorter in duration ( t13 = 3 . 06 , p=0 . 01 ) , contained fewer licks ( t13 = 6 . 86 , p=0 . 0002 ) , and had longer inter-lick intervals within a burst ( t13 = 3 . 2 , p=0 . 01 ) ( Figure 2E–H ) . We next examined consummatory licking behavior tied to sucrose delivery , we found that bursts following outcome delivery were shorter in duration ( t13 = 3 . 29 , p=0 . 006 ) , and contained fewer licks ( t13 = 3 . 55 , p=0 . 006 ) ( Figure 2I , J ) . Together , this data suggests that shortening of food restriction from 16 hr to 2 hr resulted in a decreased motivational state that reduced the palatability of sucrose , with mice showing less appetitive and consummatory licking behaviors . To examine whether the state-dependent decrease in sucrose value would be retrieved and used to control decision-making , we performed a non-reinforced test session the following day in both groups of mice . Experiencing sucrose in a decreased motivational state induced incentive learning as assessed by the decreased seeking response observed during testing ( Figure 2K ) . Response rates between groups differed significantly ( unpaired t-test: t13 = 4 . 5 , p=0 . 0015 ) . Group 16–2 mice reduced their response rate from baseline ( one-sample t-test against 100%: t6 = 3 . 1 , p=0 . 02 ) , while Group 16–16 showed an increase from baseline ( one-sample t-test against 100%: t7 = 3 . 66 , p=0 . 008 ) . Hence , mice also readily show negative incentive learning where a change in action control following a decrease in motivational state requires state-dependent experience to update value representations later used for goal-directed decision-making . While positive incentive learning reflects a state-dependent increase in sucrose value , negative incentive learning reflects a relative decrease in sucrose value , with the state-dependent decrease in sucrose value supporting less seeking behavior . Previous work has found that OFC neurons encode action value ( Gremel and Costa , 2013 ) , choice value ( Rich and Wallis , 2016 ) , value estimates ( Padoa-Schioppa and Assad , 2006; Kennerley and Wallis , 2009; Padoa-Schioppa and Assad , 2008; Padoa-Schioppa , 2009Padoa-Schioppa , 2009; Kennerley et al . , 2011Kennerley et al . , 2011; McDannald et al . , 2014 ) , and sensory attributes of value ( Rolls et al . , 1989; Critchley and Rolls , 1996; Pritchard et al . , 2008; Gremel and Costa , 2013 ) . These findings suggest that a state-dependent increase in palatable value of sucrose could require OFC neuron encoding . However , it could also be that OFC neuron activity is necessary for updating value representations independent of any direct representation of sucrose value or cached value representation . The sucrose re-exposure day in the incentive learning task provides a unique design with which to disambiguate these two hypotheses . The first hypothesis would predict that OFC neuron activity is necessary to produce the increase in sucrose palatability , while the latter would predict that OFC activity during sucrose re-exposure is responsible for the updating of the increased value representations subsequently used to control decision-making during testing . To probe whether OFC activity functionally contributes to the two above hypotheses , we took a chemogenetic approach to selectively attenuate OFC projection neuron activity and thereby disrupt OFC neuron encoding during sucrose re-exposure ( Figure 3A ) . We took a rigorous approach and used two methods to restrict hM4Di receptors to OFC excitatory projection neurons . First , B6 . 129S2-Emx1tm1 ( cre ) Krj/J mice ( Emx1-Cre ) backcrossed onto C57BL/6J mice for several generations , were given bilateral lateral OFC injections of AAV5-hSyn-DIO-hM4D ( Gi ) -mCherry ( DIO-H4 ) ( 100 nl per side; UNC Vector Core ) . Since the Emx1-Cre line expresses Cre-recombinase in excitatory projection neurons , this manipulation restricted DREADD expression to OFC excitatory projection neurons . Second , C57BL/6J mice were given bilateral lateral OFC injections of AAV5-CamKIIa-GFP-Cre ( CamKII-Cre ) ( 100 nl per side; UNC Vector Core ) and AAV5-hSyn-DIO-hM4D ( Gi ) -mCherry ( 100 nl per side; UNC Vector Core ) . This also limited Cre recombinase and DREADD expression to excitatory CamKIIa-expressing projection neurons in OFC . Additional Emx1-Cre and C57BL/6J mice were given injections of AAV5-hSyn-DIO-mCherry ( DIO-mCherry ) or AAV5-CamKIIa-GFP-Cre ( 100 nl per side; UNC Vector Core ) with DIO-mCherry to control for any effects of surgery , AAV infection , and CNO administration . We found no differences between strains in control measures across task acquisition ( positive or negative incentive learning ) , re-exposure licking , or percent of baseline responding on the non-rewarded test day ( see Supplementary file 1 , Table 1 ) and strains were combined for the remaining analyses . To confirm function of our manipulation , we conducted whole-cell current clamp recordings in identified OFC projection neurons expressing mCherry from infusions of AAV5-hSyn-DIO-hM4Di-mCherry ( Figure 3B ) . Bath application of CNO ( 10 µM ) resulted in a significant decrease in excitability ( two-way repeated measures ANOVA ( Current x CNO ) , interaction: F ( 10 , 70 ) =13 . 75 , p<0 . 001; main effect of CNO and current ( Fs > 11 . 80 , ps <0 . 003 ) ( Figure 3C , n = 8 ) . Following surgical procedures , all mice underwent incentive learning task training . Prior to the sucrose re-exposure session , mice were given injections of 0 . 9% saline ( 10 ml/kg ) or CNO ( 1 mg/kg , 10 ml/kg ) . We used viral and drug controls in each experiment and did not see differences between controls injected with saline or CNO ( Supplementary file 1 , Table 2 ) ; therefore , we collapsed across controls for ease of presentation . Attenuating OFC activity during the sucrose re-exposure session did not alter the appetitive or consummatory licking behavior observed . Increasing food restriction led to more licks independent of OFC attenuation ( 2-way repeated measures ANOVA ( food restriction Group x Treatment ) : no interaction F = 2 . 6 , p=0 . 11; main effect of Group: F ( 1 , 43 ) =26 . 21 , p<0 . 0001; no main effect of Treatment: F = 1 . 13 , p=0 . 29 ) ( Figure 3E ) . When we examined licking rate in 10 min blocks across the 60 min session , we found that food restriction groups differed across the entire session independent of treatment ( 3-way repeated measures ANOVA ( Session block x Treatment x Group ) ; no 3-way interaction: F = 0 . 47 , p>0 . 79; no interaction of Session block x Treatment: F = 0 . 72 , p=0 . 61; interaction of Session block x Group: F ( 5 , 225 ) =6 . 98 , p<0 . 001; main effect of session block: F ( 5 , 225 ) =21 . 06 , p<0 . 001 ) ( Figure 3F ) . In addition , the decrease in average latency to the first lick following an increase in food restriction was similar between Treatments ( main effect of Group: F ( 1 , 43 ) =10 . 88 , p=0 . 002; no interaction or main effect of Treatment: Fs <0 . 5 , ps >0 . 48 ) ( Figure 3G ) . Food restriction groups differentially organized their licks into bursts ( no interaction: F = 0 . 04 , p=0 . 83; main effect of Group: F ( 1 , 43 ) =5 . 81 , p=0 . 02; no main effect of Treatment: F = 0 . 65 , p=0 . 42 ) ( Figure 3H ) . All mice showed similar burst durations ( no interactions or main effects , Fs <4 , ps >0 . 05 ) ( Figure 3I ) , and the same number of licks in a burst independent of food restriction treatment ( no interaction or main effects , Fs <3 . 3 , ps >0 . 07 ) ( Figure 3J ) . All groups maintained their average inter-lick intervals within a burst ( no interaction or main effects , Fs <0 . 97 , ps >0 . 33 ) ( Figure 3K ) . We next examined burst licking behaviors following the delivery of sucrose to examine consumption-related licking and found no effect of OFC attenuation on consumption behaviors . OFC attenuated mice at a higher food restriction showed similar burst durations in comparison to controls groups ( main effect of Group: F ( 1 , 43 ) =16 . 46 . p=0 . 002; no interaction or main effect of Treatment: Fs <2 . 8 , ps >0 . 10 ) , and showed a similar increase in licks within a burst following sucrose ( main effect of Group: F ( 1 , 43 ) =17 . 31 , p=0 . 0001; no interaction or main effect of Treatment ( Fs <2 . 92 , ps >0 . 09 ) ( Figure 3L , M ) . Together , our data show that OFC attenuation during re-exposure to sucrose in a changed motivational state did little to alter the increased palatability of sucrose . To examine whether increased OFC projection neuron activity was necessary to update state-dependent value representations during the sucrose re-exposure , we subsequently tested all mice the next day when OFC activity was intact . Attenuating OFC projection neuron activity during sucrose re-exposure disrupted incentive learning subsequently used to infer what actions to take . A two-way ANOVA revealed a significant interaction between Food Restriction and Treatment Group ( F ( 1 , 43 ) =5 . 54 , p=0 . 02 ) ( no main effects , ps > 0 . 1 ) ( Figure 3N ) . Post hoc Bonferroni-corrected comparisons within each Treatment found that Control 2–16 mice that underwent a state-dependent increase in motivation had significantly higher response rates than Control mice kept at 2 hr food restriction ( 2-2 ) ( p<0 . 05 ) . In contrast , OFC 2–16 mice that were shifted from 2 hr to 16 hr food restriction and had OFC attenuated during sucrose re-exposure showed similar response rates to mice with OFC attenuated but kept in the training motivational state ( OFC 2–2 ) ( p<0 . 05 ) . Indeed , only Control 2–16 mice showed a significant increase from baseline response rates ( one-sample t-test against 100%: t13 = 2 . 65 , p=0 . 02 ) , while Control 2–2 , OFC 2–2 , and OFC 2–16 mice did not ( ps > 0 . 2 ) . This suggests that OFC projection neuron attenuation prevented the experience-based updating of state-dependent increases in sucrose value . The lack of increased response rate during subsequent testing suggests that OFC 2–16 mice did not have an updated value representation to retrieve , and instead relied on the representation learned during acquisition to control decision-making . We then gave a subset of mice the opportunity to update sucrose value representations in a test session the next day where lever presses were rewarded ( i . e . they earned sucrose deliveries ) . With OFC intact , OFC 2–16 mice showed increased response rate during the rewarded test compared to the non-rewarded test session ( t28 = 3 . 10 , p=0 . 004 ) ( Figure 3O ) . Control 2–16 mice had already updated the state-dependent increase in sucrose value during the re-exposure session , with their response rate consistent between non-rewarded and rewarded test sessions ( t22 = 0 . 02 , p=0 . 97 ) . Together , these data suggest that OFC projection neuron activity was necessary for mice to learn about relative increases in value which they subsequently used to infer how valuable their actions would be , but does not contribute to state-dependent changes in value perception itself . Our above results implicate a necessity for OFC projection neuron activity for positive incentive learning . Given prior findings suggesting OFC activity positively correlates with multiple aspects of value ( Padoa-Schioppa , 2009 ) , one could make the hypothesis that some pattern of OFC activity is necessary for relative decreases in value . However , previous findings examining cue-outcome associations failed to find evidence of OFC neurons decreasing firing rate when delivered outcomes were less than expected ( Takahashi et al . , 2009; 2013 ) . It may be that OFC activity is necessary for incentive learning processes in general following a state change , be that relative increases or decreases in value . To examine whether increases in OFC projection neuron activity are also necessary to update state-dependent decreases in sucrose value , we injected Emx1-Cre and C57BL/6J mice with the necessary combinations of DIO-H4 or CamKII-Cre and DIO-H4 , or Control mice with mCherry into lateral OFC . Mice were then trained on the incentive learning task to lever press for sucrose ( Figure 4A ) . Prior to the re-exposure session , mice were given an injection of CNO ( 1 mg/kg , 10 ml/kg ) or 0 . 9% saline ( 10 ml/kg ) . Similar to what we observed with negative incentive learning ( Figure 2 ) , we found that a decrease in motivational state produced by a reduction from 16 hr to 2 hr food restriction resulted in reduced appetitive and consummatory licking behaviors . A two-way ANOVA of food restriction Group x Treatment show a similar reduction in total number of licks between Treatment groups following a decrease in food restriction ( main effect of food restriction Group , F ( 1 , 63 ) =28 . 21 , p<0 . 0001 ) ( Figure 4B ) , that was similar across the session duration ( no interaction of Session block x Treatment x food restriction Group , F = 1 . 76 , p=0 . 12; no effect of Session block x Treatment , F = 1 . 20 , p=0 . 31; interaction of Session block x Group , F ( 5 , 315 ) = 3 . 79 , p=0 . 002; Main effect of Session duration , F ( 5 , 315 ) = 9 . 534 , p<0 . 001 . ) ( Figure 4C ) . OFC attenuated mice showed a similar latency to start licking with the re-exposure session ( main effect of Group: F ( 1 , 63 ) =21 . 79 , p<0 . 0001; no interaction or main effect of Treatment: Fs <3 . 12 , ps >0 . 08 ) ( Figure 4D ) . OFC attenuation also did not affect the organization of licks into bursts ( no interaction or main effects: Fs <2 . 2 , ps >0 . 14 ) , or the average burst duration ( no interaction or main effects: Fs <2 . 57 , ps >0 . 11 ) ( Figure 4E–F ) . While decreases in the duration of food restriction did reduce the number of average number of licks within a burst ( main effect: F ( 1 , 63 ) =7 . 87 , p=0 . 006 ) , there was no effect of OFC attenuation ( no interaction or main effect of Treatment: Fs <2 . 60 , ps >0 . 11 ) , nor was there any effect on the inter-lick interval within a burst ( no interaction or main effects , Fs <1 . 48 , ps >0 . 22 ) ( Figure 4G–H ) . Examining potential effects of OFC attenuation on consummatory licking patterns , we examined licks within the first burst following sucrose delivery . Once again , a decrease in food restriction reduced burst duration ( main effect of Group: F ( 1 , 63 ) =16 . 16 , p=0 . 0002 ) and the number of licks within a burst ( main effect of Group: F ( 1 , 63 ) =20 . 08 , p<0 . 001 ) , but neither were altered by OFC attenuation ( no interactions or main effects of Treatment: Fs <0 . 28 , ps >0 . 59 ) ( Figure 4I–J ) . Together , these data suggest that OFC attenuation during anticipatory and consummatory licking did not prevent a downshift in sucrose palatability following a decrease in motivational state . We next examined whether OFC attenuation during re-exposure would disrupt the ability to subsequently infer the decrease in value to control decision-making . Attenuating OFC activity during sucrose re-exposure prevented the updating of a state-dependent decrease in sucrose value . A two-way ANOVA ( food restriction Group x Treatment ) revealed a significant interaction ( F ( 1 , 63 ) =4 . 16 , p=0 . 04 ) ( Figure 4K ) . Planned comparisons within each Treatment found that Control 16–2 mice that underwent a state-dependent decrease in motivation had significantly lower response rates than Control mice kept at 16 hr food restriction ( 16-16 ) ( p=0 . 01 ) . In contrast , OFC 16–2 mice that were shifted from 16 hr to 2 hr food restriction and had OFC attenuated during sucrose re-exposure , showed similar response rates to mice with OFC attenuated but kept in the training motivational state ( OFC 16–16 ) ( p=0 . 41 ) . Control 16–2 mice showed a significant reduction from baseline ( one-sample t-test against 100%: t15 = 5 . 53 , p<0 . 001 ) and control 16–16 mice did not ( p=0 . 15 ) . While OFC 16–16 and OFC 16–2 groups of mice had similar response rates during testing , OFC 16–16 mice showed a slight but significant reduction from baseline ( one-sample t test against 100%: t15 = 3 . 17 , p=0 . 006 ) . Importantly , OFC attenuation in OFC 16–2 mice prevented any shift from baseline ( p=0 . 33 ) . Hence , attenuating OFC activity during sucrose re-exposure in a decreased motivational state prevented the updating of a decreased sucrose value . Indeed , during a subsequent reward test session with OFC intact , OFC 16–2 mice appeared to decrease responding from non-rewarded to rewarded tests and Control 16–2 mice stayed the same , although this difference was not significant ( ps >0 . 2 ) ( Figure 4L ) . In addition , OFC attenuation did not alter sucrose perception . Prior to a two-bottle choice test , mice free fed for at least 4 days , were pretreated with CNO or saline for 20–30 min , and then had access to bottles of 20% sucrose and water for one hour . Both OFC inhibited and control mice were able to discriminate between 20% sucrose and water , showing a similar preference for sucrose ( 2-way-ANOVA , no effect of Treatment , no interaction , main effect of Sucrose concentration: F ( 1 , 36 ) =24 . 47 , p<0 . 0001 ) ( Figure 4—figure supplement 1A ) . The same cohort also underwent a 2-bottle choice test in which mice similarly discriminated 4% from 20% sucrose ( 2-way-ANOVA: no interaction; main effect of Sucrose concentration: F ( 1 , 36 ) =11 . 11 , p=0 . 002 ) ; main effect of Treatment: F ( 1 , 36 ) =8 . 421 , p=0 . 006 ) ( Figure 4—figure supplement 1B ) . While OFC inhibited mice consumed slightly less overall during a 1 hr period ( main effect of treatment ) , this effect was driven by one replication . Overall , these data suggest that altered sucrose perception does not contribute to the pattern of current findings following OFC inhibition . So far , our data suggests that OFC activity is generally necessary to update value representations . However , general OFC attenuation across the entire state-task space in which mice are showing anticipatory and consummatory licking behavior in a trained context following a state change does not provide information about the temporal specificity of when OFC activity is necessary to update value changes . It could be that updated value representations accrue across time in a diffuse manner , derived from changes in anticipatory as well as consummatory behavior . Thus , OFC inhibition at any point within re-exposure session would be sufficient to disrupt value updating . However , given recent findings showing OFC encoding of sensory information ( Wikenheiser et al . , 2017 ) and our observation that chemogenetic activation of OFC did not increase value ( Figure 3—figure supplement 1 ) , OFC may use information gained during the sucrose consumption period to update relative changes in value representations . Thus , we hypothesized that OFC inhibition while consuming sucrose would disrupt value updating . To directly examine when OFC activity is necessary for value updating , we took an optogenetic approach with the goal of inhibiting OFC projection neuron activity selectively during sucrose consumption or randomly throughout the re-exposure session . To inhibit OFC projection neurons , we relied upon a commonly used approach of expressing channelrhodopsin in parvalbumin ( PV ) interneurons such that light activation would induce a PV inhibitory clamp of projection neuron activity ( e . g . Li et al . , 2016 ) ( Figure 5B ) . B6;129P2-Pvalbtm1 ( cre ) Arbr/J ( PV-Cre ) mice were given an injection of AAV Ef1a-DIO-hChR2 ( H134R ) -eYFP ( ChR2 ) bilaterally into the OFC , and optic fiber ferrules were targeted to lateral OFC ( Figure 5A ) . To inhibit OFC projection neurons selectively during sucrose consumption , we employed closed-loop feedback during the re-exposure session such that the first lick after a sucrose delivery would result in light activation of PV interneurons ( delay ~10–20 ms ) ( Figure 5D ) . To target our light delivery to encompass the duration of sucrose consumption , we use a delivery of light at 20 Hz , 5 ms pulse width for 5 s ( longer than our average burst durations following outcome delivery ) . We verified our ability to induce an inhibitory clamp on OFC projections ex vivo , where in whole-cell recordings we saw optogenetic activation of PV interneurons inhibited projection neuron spiking for 5 s , with no evidence of an increase in rebound firing post light activation ( Figure 5C ) . Thus , we were able to selectively inhibit OFC projection neuron activity during sucrose consumption . Mice with channelrhodopsin expressed and ferrules implanted were trained on our negative incentive-learning task . We used a modified negative incentive learning task to reduce subject loss and increase response rates . Mice were chronically food restricted to 90% of their baseline free-feeding weights throughout lever-press training . Following acquisition , mice were assigned to one of two groups that were matched for acquisition response rates . The evening before the re-exposure session , mice were given access to ad libitum food in their home-cage for their dark cycle . During the subsequent re-exposure session , our experimental group ( ChR2 ) had light delivery paired with the first lick following sucrose delivery in the re-exposure session ( Figure 5D ) . Each ChR2 mouse had a Yoked control mouse ( matched for response rates during acquisition ) , whose light delivery was yoked to the ChR2 mice and independent of any behavior exhibited by the Yoked mouse . Thus , we had two groups with different conditions of light delivery; for one group light delivery was tied to the direct sucrose consumption experience and the other light was delivered randomly throughout the session . Light delivery and the subsequent inhibition of OFC projection neurons had little effect on anticipatory and consummatory licking behaviors exhibited across the re-exposure session . While it appeared ChR2 mice made fewer licks overall than Yoked mice , it was not significant ( unpaired t-test: t16 = 1 . 54 , p=0 . 14 ) ( Figure 5E ) . When we looked at the licking rate across the session , we did see an interaction ( 2-way repeated measures ANOVA ( Treatment x Time ) : F ( 5 , 80 ) =2 . 54 , p=0 . 035 ) and a main effect of Time ( F ( 5 , 80 ) =3 . 52 , p=0 . 006 ) but not of Treatment ( p=0 . 14 ) ( Figure 5F ) , suggesting that ChR2 mice may have a different pattern of licking across the session . However , there were no group differences in other lick measures as assessed through paired comparisons . There were no differences seen in the latency to make the first lick within the session ( p=0 . 89 ) , the organization of licks into bursts ( p=0 . 36 ) , the duration of bursts ( p=0 . 18 ) , the average licks within a burst ( p=0 . 09 ) , or the inter-lick interval within a burst ( p=0 . 75 ) ( Figure 5G–K ) . When we evaluated consumption licking behavior ( i . e . after an outcome delivery ) , we did see an effect of light delivery on the duration of the licking burst ( unpaired t-test: t16 = 2 . 211 , p=0 . 041 ) ( Figure 5L ) . However , we saw no difference in the number of licks in a burst ( p=0 . 07 ) following outcome delivery ( Figure 5M ) . Our data suggests that light delivery during sucrose consumption did not grossly alter licking behavior in comparison to our yoked controls . It appears OFC inhibition during consumption leaves sucrose palatability largely intact . However , when we assessed whether mice had updated the decrease in outcome value through a non-rewarded test session the next day , we found that OFC projection neuron inhibition while mice were directly consuming sucrose appeared to disrupt value updating ( Figure 5N ) . Yoked mice significantly reduced test response rates from acquisition baseline ( one-sample t-test against 100% baseline: t7 = 3 . 07 , p=0 . 018 ) , while ChR2 mice did less so ( t7 = 2 . 20 , p=0 . 06 ) . Six out of eight pairs of ChR2 mice and their yoked control showed a decreased response rate in yoked versus ChR2 mice , although the group difference was not significant ( paired t-test: t5 = 1 . 67 , p=0 . 14 ) . Our data suggest that OFC inhibition during sucrose experience decreases value updating . Importantly , Yoked mice showed evidence that they had updated a decrease in outcome value corresponding to experiencing sucrose in a decreased motivational state , and ChR2 mice did not . While we attempted to deliver light throughout the normal duration of sucrose consumption by targeting the first licking bout after an outcome delivery , we cannot rule out that some sucrose consumption happened outside this time frame in our ChR2 group . Further , it could be that the inhibition overlaid sucrose consumption in our Yoked group . While our data does not speak to how much sucrose experience is necessary to be able to update value changes , our data does suggest that OFC inhibition overlapping with sucrose experience interferes with the ability to update , while OFC inhibition not explicitly paired with sucrose consumption does not . Here we find that the OFC is involved in updating value changes subsequently used to control decision-making , highlighting the contribution of OFC to model-based action control . We used an incentive learning task as a tool to differentiate the ability to learn or update value representations from deploying use of those representations . We found OFC projection neurons are necessary to learn relative value changes that are subsequently used to inform goal-directed behavior . Importantly , this is independent of the valence of the motivational state change , as OFC activity was necessary to encode both decreases and increases in relative value . The latter finding contradicts previous hypotheses suggesting a role for OFC in behavioral disinhibition . Further , our finding for the necessity of OFC activity when directly experiencing value changes suggests that OFC performs computations based on sensory information directly gained from the consumption experience . In light of OFC’s prior implications in direct sensory and gustatory processing ( Grabenhorst et al . , 2008; Lara et al . , 2009; Ohla et al . , 2012 ) , we hypothesize OFC updates value changes from an experienced aspect derived from sensory information . Model-based learning requires the ability to readily update associative information then used to infer the most appropriate action . Most often , devaluation tasks have been used to examine model-based or goal-directed control , where devaluation procedures are immediately followed by tests assessing to what degree the expected outcome value controls responding . Manipulations have largely affected both devaluation and testing periods , therefore making it difficult to separate contributions of updating value from inference control during performance , or changes to action outcome contingency and motivation . In addition , compensatory brain mechanisms that arise across long durations of OFC attenuation may obscure contributing roles ( Murray et al . , 2015 ) . Previous work has shown that optogenetic activation of lOFC only during testing selectively increases goal-directed actions ( Gremel and Costa , 2013 ) , while precise optogenetic OFC inhibition prevented value updating in a Pavlovian task ( Takahashi et al . , 2013 ) . Our present data adds to this literature by showing the lOFC activity is necessary for value updating in a decision-making task characterized as model-based control . These data suggest that mechanisms within lOFC support both model-based learning as well as the ability to infer such changes to support decision-making . We used a task where the sensory properties of the outcome , in our case a sucrose solution , stayed the same across the entire experiment . Hence , any change in value came not from changes in outcome size , identity , or flavor , but from changes to the hunger state of the mouse . In addition , the choice of whether or not to press the lever stayed the same , and there was no manipulation to the action-outcome contingency ( the same lever press-outcome relationships were in place ) . Mice needed to infer the consequence of their potential action and adjust their action frequency appropriately . Experiencing sucrose in an increased hunger state led to an increase in the palatability of sucrose and an increase in action frequency ( Figure 1 ) . In contrast , experiencing sucrose in a decreased hunger state led to a decrease in sucrose palatability and decreased response rates ( Figure 2 ) . Thus , in both instances of incentive learning , the representation of sucrose value was updated and subsequently used to decide whether to press the lever or not . The absence of such change in control mice was apparent when a change in motivational state was not accompanied by experience ( Figure 1 ) . This separation of palatability from incentive learning processes provided a unique task with which to investigate OFC’s hypothesized role in inferring value ( Stalnaker et al . , 2015; Schuck et al . , 2016 ) . While our data supports a crucial role for OFC in updating value representations supporting model-based behavior , our data also adds to recent findings that show OFC is not necessary for perceiving changes in palatability ( Gardner et al . , 2017 ) . Broadly attenuating OFC projection neuron activity or temporally inhibiting OFC projection neuron activity did little to change appetitive and consummatory licking behaviors ( Figures 3–5 ) . Mice that had not undergone a motivational state change , but had OFC attenuated during the re-exposure session , showed licking behaviors similar to mice that had OFC function intact . This suggests that mice with OFC attenuated were still able to retrieve the sucrose representation . However , it is also possible that OFC attenuation would block retrieval of any sucrose representation independent of a motivational state change . This contrasting hypothesis suggests sucrose delivery on its own , independent of the animal’s learned behavior , would elicit innate frequencies of anticipatory and consummatory behaviors subject to modulation by motivational state . While this could conceivably be another explanation for the observed palatability changes , there were no differences in the initial latency to lick between groups . The lack of a difference suggests that OFC attenuation did not affect retrieval of the sucrose representation used to initiate the first appetitive lick ( Figures 3G and 4D ) . The present incentive learning data suggests that updating value representations under novel circumstances can be used to infer consequences for appropriate decision-making , as re-exposure happening outside of the training context was sufficient to update the representation ( Figure 1M ) . This raises the question of what information gained during re-exposure is used to update value representations . We examined this question via optogenetically inhibiting OFC ( Figure 5 ) , where we time-locked OFC inhibition to the initial lick response following sucrose delivery in our experimental mice and compared responding to that of yoked mice given the exact same stimulation independent of their behavior . Optogenetic inhibition reduced the ability to update value in our experimental mice and not in the yoked controls . However , given the lack of a significant group difference , some caution should be taken in the interpretation . We do know that given the precise timing of the inhibition manipulation , OFC activity was intact the majority of the session . OFC inhibition did not prevent any additional motivational processes that might occur after the direct consummatory experience ( e . g . , lasting pleasantness ) from contributing to the sucrose value representation . Further , the yoked group had the same amount of inhibition , but not tied to any behavior . Hence , it is possible that OFC inhibition in the yoked group overlapped with sucrose consumption . In addition , we cannot be 100% confident that all sucrose consumption in our experimental mice happened within the duration of OFC inhibition . There was a small but significant difference in the duration of a licking bout following sucrose delivery , which could indicate a less consumption during light delivery . Thus , our data does not speak to how much experience with the outcome is necessary to update value representations following a motivational state change . Our data does suggest that accrued aspects of the sucrose consumption experience may be used to guide value updating processes . Here we have shown that OFC is necessary for the updating of value under both positive and negative motivational states . To our knowledge , this is the first experiment demonstrating OFC’s role in retrieving and updating the value of an outcome and storing it for retrieval under a previously learned action contingency . OFC has been hypothesized to function as an unobservable task state space ( Wilson et al . , 2014; Schuck et al . , 2016; Bradfield et al . , 2015 ) . Our data offer support for this hypothesis in model-based behavior . We show that in addition to the well-documented role OFC has in inferring value to control decision-making , mice also need OFC to update outcome representations . During the re-exposure session , the unobservable component is the action portion of the action-outcome association . Our data show that OFC activity is necessary to update the outcome value used for inferring the value of the action . In conclusion , these studies provide insight into the role of orbitofrontal cortex in decision-making by showing a role for OFC in updating experienced-based changes in value . Male and female C57BL/6J , B6 . 129S2-Emx1tm1 ( cre ) Krj/J ( Emx1-Cre ) , and B6;129P2-Pvalbtm1 ( cre ) Arbr/J ( PV-Cre ) mice were housed 2–5 per cage under a 14/10 hr light/dark cycle with access to food ( Labdiet 5015 ) and water ad libitum unless stated otherwise . Emx1-Cre ( obtained from breeding homozygous B6 . 129S2-Emx1tm1 ( cre ) Krj/J ( Jackson ) x C57BL/6J in house ) , C57BL/6J ( The Jackson Laboratory , Bar Harbor , ME , USA ) , and PV-Cre ( obtained from breeding homozygous B6;129P2-Pvalbtm1 ( cre ) Arbr/J x C57BL/6J mice in house ) mice were at least 6 weeks of age prior to intracranial injections and at least 7 weeks of age prior to behavioral training . All surgical and behavioral experiments were performed during the light portion of the cycle . The Animal Care and Use Committee of the University of California San Diego approved all experiments and experiments were conducted according to the NIH guidelines . Mice received stereotaxically guided bilateral injections via Hamilton syringe into lateral OFC ( coordinates from bregma: A , 2 . 7 mm; M/L , 1 . 65 mm; V , 2 . 65 mm ) . To limit our manipulations to OFC projection neurons , we used two approaches . In the first , C57BL/6J mice were given co-injections of AAV5-CamKIIa-Cre-GFP ( 100 nl per side; UNC Vector Core ) and AAV5-hSyn-DIO-hM4D ( Gi ) -mCherry ( 100 nl per side; UNC Vector Core ) in order to express inhibitory DREADD only in CamKIIa-expressing excitatory neurons . In the second approach Emx1-Cre mice , in which Cre expression is limited to projection neurons ( Gorski et al . , 2002 ) were given bilateral injections of AAV5-hSyn-DIO-hM4D ( Gi ) -mCherry ( 100 nl per side; UNC Vector Core ) . To excite OFC projection neurons , we injected AAV5-hSyn-DIO-hM3D ( Gq ) -mCherry into Emx1-Cre mice . Based on our previous work examining in vivo the time course to observe circuit suppression ( Gremel and Costa , 2013 ) mice were given an intraperitoneal injection with either Clozapine-n-oxide ( CNO ) ( 10 ml/kg 1 mg/kg dose ) or 0 . 9% isotonic saline 20–30 min prior to testing . After testing , mice were euthanized and brains extracted and fixed in 4% paraformaldehyde . Viral spread was qualified by examining in 100–150 μm thick brain slices under a macro fluorescence microscope ( Olympus MVX10 ) . Mice received stereotaxically guided bilateral injections via Hamilton syringe into lateral OFC ( coordinates from bregma: A , 2 . 7 mm; M/L , 1 . 65 mm; V , 2 . 65 mm ) . To limit our manipulations to OFC Parvalbumin inhibitory neurons , we used a PV-Cre line ( Hippenmeyer et al . , 2005 ) and injected 200 nl Ef1a-DIO-hChR2 ( H134R ) -eYFP ( Chr2 ) . Following injections , mice were trained on the instrumental task . The last 4 days of schedule training , mice were lightly anaesthetized and had optical fibers coupled to their ferrule implants using a ceramic sleeve to accustom the mice to moving with the fibers on their heads . Prior to re-exposure to sucrose , mice had the fibers coupled to their implants and were allowed 30 min to recover from effect of anesthesia . Mice received 5 ms pulses of 470 nm light at 20 Hz ( 1–3 mW ) for 5 s triggered by their first lick after each sucrose delivery . We chose this duration of light based on our previously collected data on burst duration after a reinforce delivery . High-powered 470 nm LEDs ( Thor Labs ) were controlled with custom Arduino Script ( https://github . com/gremellab/arduinoLEDcontrol [Baltz and Gremel , 2017]; copy archived at https://github . com/elifesciences-publications/arduinoLEDcontrol ) . Viral spread and ferrule placement were assessed under a macro fluorescence microscope ( Olympus MVX10 ) . Coronal slices ( 250 μm thick ) containing the OFC were prepared using a Pelco easiSlicer ( Ted Pella Inc . , Redding , CA ) . Mice were anesthetized by inhalation of isoflurane , and brains were rapidly removed and placed in 4°C oxygenated ACSF containing the following ( in mM ) : 210 sucrose , 26 . 2 NaHCO3 , 1 NaH2PO4 , 2 . 5 KCl , 11 dextrose , bubbled with 95% O2/5% CO2 . Slices were transferred to an ACSF solution for incubation containing the following ( in mM ) : 120 NaCl , 25 NaHCO3 , 1 . 23 NaH2PO4 , 3 . 3 KCl , 2 . 4 MgCl2 , 1 . 8 CaCl2 , 10 dextrose . Slices were continuously bubbled with 95% O2/5% CO2 at pH 7 . 4 , 32°C , and were maintained in this solution for at least 60 min prior to recording . Whole-cell current clamp recordings were made in pyramidal cells of the OFC . Pyramidal cells that expressed hM4Di were identified by the fluorescent mCherry label using an Olympus BX51WI microscope mounted on a vibration isolation table and a high-power LED ( LED4D067 , Thor Labs ) . Recordings were made in ACSF containing ( in mM ) : 120 NaCl , 25 NaHCO3 , 1 . 23 NaH2PO4 , 3 . 3 KCl , 0 . 9 MgCl2 , 2 . 0 CaCl2 , and 10 dextrose , bubbled with 95% O2/5% CO2 . ACSF was continuously perfused at a rate of 2 . 0 mL/min and maintained at a temperature of 32°C . Picrotoxin ( 50 µM ) was included in the recording ACSF to block GABAA receptor-mediated synaptic currents . Recording electrodes ( thin-wall glass , WPI Instruments ) were made using a PC-10 puller ( Narishige International , Amityville , NY ) to yield resistances between 3–6 MΩ . Electrodes were filled with ( in mM ) : 135 KMeSO4 , 12 NaCl , 0 . 5 EGTA , 10 HEPES , 2 Mg-ATP , 0 . 3 Tris-GTP , 260–270 mOsm ( pH 7 . 3 ) . Access resistance was monitored throughout the experiments . Cells in which access resistance varied more than 20% were not included in the analysis . Recordings were made using a MultiClamp 700B amplifier ( Molecular Devices , Union City , CA ) , filtered at 2 kHz , digitized at 10 kHz with Instrutech ITC-18 ( HEKA Instruments , Bellmore , NY ) , and displayed and saved using AxographX ( Axograph , Sydney , Australia ) . A series of fixed current injections ( 20 pA increments from −80 to 300 pA ) were used to elicit action potential firing and the number of spikes were counted at each current step . For verification of DREADD function , current injections were done prior to ( baseline ) and 15–30 min after CNO ( 10 µM ) bath application . For PV inhibition of OFC firing , ChR2 expressing PV neurons were optically stimulated using 470 nm blue light ( 5 ms ) , delivered via field illumination using a high-power LED ( LED4D067 , Thor Labs ) . Optical stimulation was done at 20 Hz for 5 s during current injections . Data from each neuron within a treatment group was combined and presented as mean ± SEM . We adapted an incentive learning task previously used in rats ( Balleine and Dickinson , 1998; Wassum et al . , 2009 ) . Mice were trained in standard operant chambers containing two levers situated around a food magazine containing a fluid well with contact lickometers and a house light on the opposite wall within sound-attenuating boxes ( Med-Associates ) . Regular food pellets and water were freely available prior to the start of training . In brief , mice underwent either a 2 hr ( positive incentive learning ) or 16 hr ( negative incentive learning ) food restriction each day but had unlimited access to water in their home cages . Mice were trained under a chain schedule of lever presses for a sucrose delivery ( 20–30 μL of 20% solution per sucrose delivery ) . The incentive value of sucrose was manipulated during test days by maintaining , increasing or decreasing the length of food restriction and then providing a chance for sucrose revaluation . One subject in the 16–16 Ctl group in negative incentive learning was excluded for a percent of baseline response rate more than 3 SD from the norm . Mice were trained on a 2 hr food restriction and then switched to a 16 hr restriction prior to re-exposure and incentive learning test sessions . For these experiments , the availability and location of the sucrose during the re-exposure session was manipulated . This produced four distinct groups: 1 ) mice were either placed in a home cage with no access to sucrose , 2 ) placed in a home cage and had ad libitum access to a bottle of sucrose for 1 hr , 3 ) placed in the operant context for a 1 hr session but with no sucrose deliveries , or 4 ) placed in the operant context for a 1 hr RT120 session with sucrose deliveries . Following these re-exposure sessions , mice were given a five-minute non-reinforced test as described above . Two cohorts of OFC-H4 mice underwent 2-bottle choice tests under ad libitum food access ( at least 4 days of no food restriction ) . Mice were given pretreatment of CNO ( 1 mg/kg , 10 ml/kg ) or 0 . 9% saline 30 min prior to testing . Mice were placed in an empty cage with grating on top to hold two small test tubes of liquid . Bottle sides were counterbalanced and sham bottles were used to estimate the amount of spillage and evaporation . Bottles were measured immediately before and immediately after an hour had elapsed . In the first test , mice had access to bottles of 20% weight/volume sucrose and water for 60 min . Final group n’s were OFC-H4 n = 20 , Control n = 18 . For the second test , bottles of 20% sucrose and 4% sucrose were available . Final group n’s were OFC-H4 CNO n = 19 , Saline + H4 n=11 , mCherry + CNO n=8 . We excluded data from mice for obvious spillage of the bottles . The alpha level was set at 0 . 05 for all experiments . Data were analyzed using Prism 6 ( GraphPad ) , JASP ( open-source statistical package ) , and custom Matlab ( Mathworks ) scripts . Data are presented as mean ± SEM ( standard error of the mean ) and averaged across counterbalanced conditions . For acquisition data , effects of Treatment ( Control vs . OFC hM4Di ) , food restriction Group ( 2 vs . 16 hr ) and Day were analyzed with repeated-measures ANOVA on lever-presses , response rates , head entries , and licking were examined .	As we go about our daily lives , we do not simply react to the world around us . Instead we build up mental representations of the world and use these to guide our behavior . For example , we know that if we are hungry we can go into the kitchen to get a slice of our favorite cake . But we also adapt our behavior when circumstances change . If you have just eaten an entire box of cookies you are unlikely to go looking for cake . Which parts of the brain help us to adapt our decisions to reflect our circumstances ? To find out , Baltz et al . trained mice to press levers in order to receive sugar water . In initial experiments the mice completed the training while not hungry . Afterward , some of the mice were placed on a diet that made them hungry . A re-exposure period then occurred where the mice could taste more sugar water without having to press the lever . Finally , the next day , they were given the opportunity to press the levers again . Mice that were hungry during the re-exposure period pressed the levers more than mice that had been re-exposed while full . Further experiments showed that this was true regardless of how hungry the mice were when they first learned the task . The mice updated how much they valued the sugar water – and so changed how eagerly they tried to obtain it – based on how hungry they were during the re-exposure period . Baltz et al . repeated the experiments , but this time blocked the activity of a brain region called the orbitofrontal cortex in the mice during the re-exposure period . This prevented the mice from updating how much they valued the sugar , and so they did not adjust their behavior accordingly . If hungry mice had performed the first training stage when they were full , they pressed the levers less often than expected after the re-exposure period . Likewise , full mice who had trained when they were hungry pressed on the lever more times , as if they were still hungry . This suggests that the orbitofrontal cortex helps to update the values that guide decision-making . There are many disorders that can impair decision-making and prevent people from adjusting their behavior when circumstances change . These include addiction , in which affected individuals also show altered activity in their orbitofrontal cortex . This raises the possibility that in the future we may be able to treat disorders like addiction by restoring normal activity in this region of the brain .	[ "Abstract", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2018	Orbital frontal cortex updates state-induced value change for decision-making
Membrane trafficking is essential to fundamental processes in eukaryotic life , including cell growth and division . In plant cytokinesis , post-Golgi trafficking mediates a massive flow of vesicles that form the partitioning membrane but its regulation remains poorly understood . Here , we identify functionally redundant Arabidopsis ARF guanine-nucleotide exchange factors ( ARF-GEFs ) BIG1–BIG4 as regulators of post-Golgi trafficking , mediating late secretion from the trans-Golgi network but not recycling of endocytosed proteins to the plasma membrane , although the TGN also functions as an early endosome in plants . In contrast , BIG1-4 are absolutely required for trafficking of both endocytosed and newly synthesized proteins to the cell–division plane during cytokinesis , counteracting recycling to the plasma membrane . This change from recycling to secretory trafficking pathway mediated by ARF-GEFs confers specificity of cargo delivery to the division plane and might thus ensure that the partitioning membrane is completed on time in the absence of a cytokinesis-interphase checkpoint . In post-Golgi membrane trafficking , cargo proteins are dynamically distributed between trans-Golgi network ( TGN ) , various endosomes , lysosome/vacuole and plasma membrane ( Surpin and Raikhel , 2004 ) . In contrast to animals , the TGN also functions as an early endosome in plants and is a major trafficking hub where secretory , endocytic , recycling and vacuolar pathways intersect ( Viotti et al . , 2010; Reyes et al . , 2011 ) . Therefore , it has been notoriously difficult to functionally delineate the recycling vs secretory pathways in plants . Sorting of cargo proteins occurs during the formation of transport vesicles , involving activation of small ARF GTPases by ARF guanine-nucleotide exchange factors ( ARF-GEFs ) and recruitment of specific coat proteins ( Casanova , 2007 ) . Arabidopsis ARF-GEFs are related to human large ARF-GEFs , GBF1 or BIG1 . Whereas the three GBF1-related members GNOM , GNL1 and GNL2 have been characterised in detail ( Geldner et al . , 2003; Richter et al . , 2007 , 2012 ) , of the 5 BIG1-related ARF-GEFs only BIG5 has been analysed so far and implicated in pathogen response ( MIN7 ) and endocytic traffic ( BEN1 ) ( Nomura et al . , 2006 , 2011; Tanaka et al . , 2009; Tanaka et al . , 2013 ) . Here , we show that ARF-GEFs BIG1-4 play a crucial role in post-Golgi traffic , which enables us to dissect the regulation of secretory and recycling pathways in interphase and cytokinesis . Up to three of ARF-GEFs BIG1 to BIG4 ( BIG1-4 ) were knocked out without recognisable phenotypic effect except for big1 , 2 , 3 , which was retarded in growth because BIG4 is predominantly expressed in root and pollen ( Figure 1A , Figure 1—figure supplement 1A ) . Other triple mutants were growth-retarded only if the activity of the respective fourth gene was reduced to 50% . No quadruple mutants were recovered because BIG1-4 were essential in male reproduction , sustaining pollen tube growth ( Figure 1B , Figure 1—figure supplement 1B ) . BIG1-4 functional redundancy would be consistent with the occurrence of BIG1-4-like single-copy or closely related sister genes in lower plants ( Figure 1—figure supplement 1C ) . Although large ARF-GEFs are often inhibited by the fungal toxin brefeldin A ( BFA ) , the SEC7 domain of BIG3 ( At1g01960; formerly named BIG2 in Nielsen et al . , 2006; see nomenclature used by Cox et al . , 2004 ) displayed BFA-insensitive GDP/GTP exchange activity in vitro ( Nielsen et al . , 2006 ) . BFA treatment of big3 mutants impaired seed germination and seedling root growth , in contrast to wild-type ( Figure 1D , E ) . We engineered a BFA-resistant variant of the naturally BFA-sensitive ARF-GEF BIG4 by replacing amino acid residue methionine at position 695 with leucine , as previously described for the recycling ARF-GEF GNOM ( Geldner et al . , 2003 ) . Engineered BFA-resistant BIG4-YFP rescued BFA-inhibited seed germination of big3 ( Figure 1F ) . The rescue activity of BFA-resistant BIG4 was comparable to that of BIG3 when both were expressed from the ubiquitin 10 ( UBQ10 ) promoter whereas BFA-sensitive BIG4 did not at all rescue BFA-inhibited primary root growth of big3 mutant seedlings ( Figure 1—figure supplement 1D , E ) . Thus , BFA treatment of big3 single mutants effectively causes conditional inactivation of BIG1-4 ARF-GEF function , providing us with a unique tool for studying BIG1-4-dependent trafficking in an organismic context . 10 . 7554/eLife . 02131 . 003Figure 1 . BIG1 – BIG4 act redundantly at TGN and are involved in several physiological processes . ( A ) big1 , 2 , 4 ( big1 big2 big4 ) , big2 , 3 , 4 ( big2 big3 big4 ) , big1 , 3 , 4 ( big1 big3 big4 ) and big1 , 2 , 3/+ , 4 ( big1 big2 big3/BIG3 big4 ) mutant plants without obvious phenotype but big1/+ , 2 , 3 , 4 ( big1/BIG1 big2 big3 big4 ) , big1 , 2/+ , 3 , 4 ( big1 big2/BIG2 big3 big4 ) and big1 , 2 , 3 ( big1 big2 big3 ) were dwarfed ( yellow arrowheads ) . Scale bar , 2 cm . ( B ) F1 of reciprocal crosses between wild-type ( Col ) and big1 big2 big3/BIG3 big4 ( 1 , 2 , 3/+ , 4 ) mutants: 0% or 48% big3 heterozygous seedlings derived from mutant male or female gamete , respectively . ( C ) BFA inhibited primary root growth of big3 mutant seedlings with or without BFA-resistant GNOM ( GNR big3 ) . Numbers of analysed seedlings are indicated ( B and C ) . ( D-H ) BFA treatment did not prevent seed germination in wild-type ( Col; D ) and BFA-resistant GN ( GNR; G ) but did so in big3 mutants without ( E ) or with BFA-resistant GNOM ( GNR big3; H ) . This defect was suppressed by BFA-resistant BIG4 ( UBQ10::BIG4R-YFP big3; F ) . Scale bar , 5 mm . ( I-L ) Live imaging of BIG4-YFP ( I ) and TGN marker VHA-a1-RFP ( J ) revealed co- localization ( K; L , intensity–line profile ) . ( M–P ) Immunolocalization of BIG4 ( UBQ10::BIG4-YFP; M ) and Golgi-marker γCOP ( N ) indicated no co-localization ( O; P , intensity–line profile ) . ( I–K , M–O ) Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 00310 . 7554/eLife . 02131 . 004Figure 1—figure supplement 1 . Expression and phylogeny of BIG ARF-GEFs . ( A ) Expression profiles of BIG1–BIG5 and GNOM . Data from AtGenExpress ( Schmid et al . , 2005 ) ( http://jsp . weigelworld . org ) . Note preferential expression of BIG4 in roots and floral organs . ( B ) Analysis of in vitro pollen tube growth of pollen from wild-type ( Col ) and big124 3/+ plants . Note approximately 50% of germinated pollen from mutant plants produced short tubes ( red column ) ; asterisks: red , short; green , long pollen tubes ) . ( C ) Phylogenetic tree ( ClustalW ) of BIG ARF-GEFs from flowering plants ( dicots Arabidopsis [At] , poplar [Pt] and grapevine [Vv]; monocots rice [Os] and Brachypodium [Bd] ) , gymnosperm Picea abies ( Pa ) , lower plants ( lycopod Selaginella [Sm] , moss Physcomitrella [Pp] , and algae Chlamydomonas [Cr] and Volvox [Vc] ) , and outgroups ( human [Hs] , Saccharomyces cerevisiae [Sc] ) ( grey ) . Three distinct subclades ( BIG1/4 [blue] , BIG2/3 [red] and BIG5 [green] ) are present in angiosperms . However , only BIG5 is distinct in all plant species whereas lower plants have a single subclade BIG1-4 corresponding to the two subclades BIG1/4 and BIG2/3 in angiosperms ( orange ) . Accession numbers and source of data are listed in the table . ( D and E ) BFA-resistant BIG4 and BIG3 expressed from the Ubiquitin10-promoter ( UBQ10::BIG4R/BIG3-YFP ) can partially rescue the BFA-inhibited primary root growth of big3 mutants . ( E ) Percentage of root growth of BFA-treated seedlings shown in ( D ) relative to untreated controls . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 00410 . 7554/eLife . 02131 . 005Figure 1—figure supplement 2 . BIG3 and BIG4 localize at the TGN . ( A–N ) Live imaging of YFP-tagged BIG3 ( A and I ) , expressed from its own promoter , and BIG4 ( E and L ) , expressed from the UBQ10 promoter , in seedling roots counterstained with endocytic tracer FM4-64 ( B , F , J , M ) . Both BIG3 and BIG4 co-localized with FM4-64 , which visualized the TGN after 10 min of incubation ( C , G; D , H , intensity profiles of lines numbered in C , G ) . ( I–N ) After BFA treatment , both BIG3 and BIG4 co-localized with FM4-64 in BFA compartments ( K and N ) . Scale bars , 5 μm . ( O–R ) Co-localization of BIG4-YFP fluorescence ( O ) with immunofluorescence labeling of the TGN marker ARF1 ( P ) in 350 nm thin cryosections ( = high axial resolution ) revealed by overlay ( Q ) and image frames shifted by 5 pixels ( R ) . Blue , DAPI-stained nuclei . Scale bar ( O ) , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 00510 . 7554/eLife . 02131 . 006Figure 1—figure supplement 3 . Ultrastructural localization of BIG4-YFP and ultrastructural abnormalities in BFA-treated big3 mutant seedling root cells . ( A and B ) Anti-GFP immunogold labeling of YFP-tagged BIG4 in root epidermal cells of UBQ10::YFP-BIG4 transgenic ( A ) and Col-0 wild-type control ( B ) seedlings ( thawed cryosection labeling ) . ( C–F ) Ultrastructural TEM analysis of BFA-treated big3 ( C and D ) and Col-0 wild-type ( E and F ) root epidermal cells . Note Golgi stacks are abnormally shaped ( C and G ) or reduced ( D , arrowheads ) near endosomal BFA aggregates ( C and D ) , in contrast to well-formed Golgi stacks in wild-type ( E and F ) . ( F ) Higher magnification of boxed area in ( E ) . g , Golgi stack; m , mitochondrion; mvb , multivesicular body; tgn , trans-Golgi network; v , vacuole . Scale bars , 500 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 006 BIG4-YFP co-localized with TGN markers vacuolar H+-ATPase ( VHA ) subunit a1 and ARF1 GTPase ( Figure 1I–L , Figure 1—figure supplement 2O–R; Dettmer et al . , 2006; Stierhof and El Kasmi , 2010 ) but not with Golgi marker COPI subunit γCOP ( Figure 1M–P; Movafeghi et al . , 1999 ) . TGN localization of BIG4-YFP was confirmed by immunogold labeling on EM sections ( Figure 1—figure supplement 3A , B ) . BIG3-YFP and BIG4-YFP co-localized with endocytic tracer FM4-64 , labeling TGN after brief uptake ( Figure 1—figure supplement 2A–H; Ueda et al . , 2001; Dettmer et al . , 2006 ) . BIG3 and BIG4 also accumulated together with FM4-64 in BFA-induced post-Golgi membrane vesicle aggregates ( ‘BFA compartments’ ) , consistent with ultrastructural abnormalities in these aggregates and Golgi stacks in BFA-treated big3 mutant ( Figure 1—figure supplement 2I–N , 3C–F ) . Together , these data suggest a role for BIG1-4 in post-Golgi membrane trafficking . To identify trafficking routes regulated by BIG1-4 , pathway-specific soluble and membrane-associated cargo proteins were analysed in BFA-treated wild-type and big3 mutant seedlings ( for a list of markers used , see Supplementary file 1; Figure 2—figure supplement 1S , T ) . Secretory GFP ( secGFP ) ( Viotti et al . , 2010 ) , which is normally secreted from the cell , and plasma membrane ( PM ) -targeted syntaxin SYP132 were trapped in BFA compartments and did not reach the plasma membrane of big3 seedlings , in contrast to wild-type , suggesting a role for BIG1-4 in late secretory traffic , that is from the TGN to the plasma membrane ( Figure 2A–D ) . There was a slight retention of SYP132 in the BFA compartments of wild-type seedling roots , which probably reflects slowed-down passage of newly-synthesized proteins through the TGN . This becomes apparent upon BFA treatment because of TGN aggregation into BFA compartments , as has been reported earlier for HS::secGFP ( Viotti et al . , 2010 ) . Vacuolar cargo proteins also pass through the TGN via multivesicular bodies ( MVBs ) to the vacuole ( Reyes et al . , 2011 ) . Soluble RFP fused to phaseolin vacuolar sorting sequence AFVY accumulated in BFA compartments in big3 mutant , in contrast to wild-type ( Scheuring et al . , 2011; Figure 2E–J , Figure 2—figure supplement 1A–F ) . Endocytosed PM proteins are delivered to the vacuole for degradation , for example boron transporter BOR1 in response to high external boron concentration ( Takano et al . , 2005; Figure 2K–N ) . BFA treatment prevented boron-induced trafficking of BOR1 to the vacuole in big3 mutant , but not in wild-type ( Figure 2L , N ) . BOR1 was rapidly turned over in the vacuole of wild-type , leaving no trace of GFP ( Figure 2L ) . As expected , ARF-GEF BIG4 and its putative cargo BOR1 co-localized in BFA compartments ( Figure 2—figure supplement 1G–I ) . Thus , BIG1-4 mediate both late secretory and vacuolar trafficking from the TGN . 10 . 7554/eLife . 02131 . 007Figure 2 . BIG1 – BIG4 regulate secretory and vacuolar trafficking by recruiting AP-1 adaptor complex . ( A and B ) BFA inhibited secretion of heat shock ( HS ) -induced secGFP in big3 mutants ( B ) but not in wild-type ( Col; A ) . ( C and D ) BFA inhibited trafficking of estradiol ( Est ) -induced YFP-SYP132 to the plasma membrane in big3 mutants ( D ) but not in wild-type ( Col; C ) . ( E–J ) BFA inhibited trafficking of soluble cargo AFVY-RFP to the vacuole ( v ) , labeled by FM1-43 ( F and I ) , in big3 mutants ( H–J ) but not in wild-type ( Col , E–G ) . ( K–N ) Live imaging of BOR1-GFP localization . Without boron ( −B ) , BOR1-GFP localized at the plasma membrane in wild-type ( K ) and big3 mutants ( M ) . After BFA and boron treatment ( +B ) , BOR1-GFP was degraded in the vacuole of wild-type ( L ) but accumulated in BFA compartments of big3 mutants ( N ) . ( O–T ) Immunostaining of 3xHA-tagged muB2 subunit of AP-1 complex ( AP1M2; O , R ) and COPI subunit γCOP ( P and S ) in BFA-treated seedlings . AP1M2 accumulated in BFA compartments surrounded by γCOP in wild-type ( Col; Q ) . In big3 mutants , γCOP was still recruited to Golgi membranes whereas AP1M2 was cytosolic ( R–T ) . Blue , DAPI-stained nuclei . Scale bars , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 00710 . 7554/eLife . 02131 . 008Figure 2—figure supplement 1 . BIG1 – BIG4 regulate trafficking of secretory and vacuolar cargo by recruiting AP-1 complex . ( A–F ) Live imaging of vacuolar cargo AFVY-RFP ( A and D ) and vacuolar membrane marker Wave9Y ( Wave9-YFP/YFP-VAMP711; B , E; Geldner et al . , 2009 ) in root cells of big3 mutant seedlings . ( C and F ) Overlays . Traffic of AFVY-RFP to the Wave9-labeled vacuole ( v; A–C ) was blocked by BFA treatment , with AFVY-RFP accumulating in small compartments distinct from vacuoles ( D–F ) . ( G–I ) Live imaging of BOR1-GFP in boron ( +B ) and BFA-treated big3 mutants expressing BIG3-promoter driven BIG4-RFP ( BIG3::BIG4-RFP ) . BOR1-GFP was not transported to the vacuole but co-localized with BFA-sensitive BIG4 in BFA compartments . ( J–O ) Immunolocalization of HA-tagged AP1M2 and TGN marker SYP61 in BFA-treated seedling roots . AP1M2 co-localized with SYP61 in BFA compartments in wild-type ( J–L ) but was cytosolic in big3 , in contrast to TGN-associated SYP61 ( M–O ) . ( P–R ) Immunostaining of UBQ10 promoter-driven BIG4-YFP and SYP61 in untreated big3 mutant seedling roots . BIG4 ( P ) co-localized with SYP61 ( Q ) at TGN ( R ) . ( S and T ) YFP-SYP132 ( S ) and PIN1-RFP ( T ) expressed from the estradiol-inducible promoter localize at the plasma membrane in an unpolar or polar fashion , respectively . Scale bars , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 008 ARF-GEFs activate ARF GTPases , resulting in recruitment of vesicular coat proteins to the respective endomembrane compartment , such as COPI complex to Golgi stacks or adaptor protein ( AP ) complexes to post-Golgi compartments ( Robinson , 2004 ) . Like BIG1-4 , AP-1 complex subunit muB2-adaptin ( AP1M2 ) localizes to SYP61-labeled TGN and is required for late secretory and vacuolar trafficking ( Park et al . , 2013; Teh et al . , 2013; Wang et al . , 2013; Figure 2—figure supplement 1P–R ) . AP1M2 also co-localized with TGN marker SYP61 in BFA compartments ( Figure 2—figure supplement 1J–L ) . In BFA-treated big3 mutant , however , AP1M2 was cytosolic whereas SYP61 was still TGN-associated ( Figure 2O , R; Figure 2—figure supplement 1J–O ) . In contrast to AP1M2 , Golgi association of COPI subunit γCOP , which is mediated by BFA-resistant ARF-GEF GNL1 ( Richter et al . , 2007 ) , was not affected in BFA-treated big3 mutant ( Figure 2O–T ) . Thus , BIG1-4 specifically mediate AP-1 recruitment to the TGN . Another ARF-GEF in post-Golgi traffic , GNOM regulates polar recycling of auxin-efflux carrier PIN1 to the basal plasma membrane ( Geldner et al . , 2003 ) . BFA treatment of wild-type and big3 mutant seedlings inhibited recycling of PIN1 , which accumulated in BFA compartments , and this defect was suppressed by engineered BFA-resistant GNOM ( Figure 3A–D ) . Thus , BIG1-4 did not play any obvious role in PIN1 recycling . PIN1 is a stable protein such that most protein detectable at the plasma membrane is delivered via the recycling but not the secretory pathway ( Geldner et al . , 2001 ) . In order to analyse the behavior of newly-synthesized PIN1 protein , we generated transgenic plants expressing estradiol-inducible PIN1 . In contrast to recycling PIN1 , newly-synthesized PIN1 protein was trapped in BFA compartments of big3 mutant , regardless of BFA-resistant GNOM ( Figure 3E–H ) . In conclusion , secretory ARF-GEFs BIG1-4 and recycling ARF-GEF GNOM regulate different post-Golgi trafficking pathways to the plasma membrane that function independently of each other . 10 . 7554/eLife . 02131 . 009Figure 3 . Secretion and recycling to the plasma membrane are regulated by different ARF-GEFs . ( A–D ) PIN1 localization in interphase cells of BFA-treated seedlings; apolar at the plasma membrane ( PM ) and in BFA compartments in wild-type ( Col; A ) and big3 mutants ( B ) ; at the basal PM in BFA-resistant GN in wild-type ( GNR , C ) or big3 mutant background ( GNR big3 , D ) . Blue , DAPI-stained nuclei . ( E–H ) After BFA treatment , estradiol ( Est ) -induced PIN1-RFP was trafficked to the PM in wild-type ( E ) and BFA-resistant GN seedlings ( GNR , G ) but not in big3 mutants without ( F ) or with expression of BFA-resistant GN ( GNR big3; H ) . Scale bars , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 009 Gravitropic growth response of the seedling root relies on GNOM-mediated PIN1 recycling ( Geldner et al . , 2003 ) . We tested whether BIG1-4 are also required , using DR5::NLS-3xGFP expression to visualise auxin response ( Weijers et al . , 2006 ) . BFA-induced inhibition of auxin response in wild-type and big3 mutant was overcome by BFA-resistant GNOM , suggesting that BIG1-4 mediated secretion plays no role in gravitropic growth response ( Figure 4A–D ) . GNOM-dependent PIN1 recycling is also required for lateral root initiation ( Geldner et al . , 2003 ) . Surprisingly , BFA-resistant GNOM failed to initiate lateral root primordia in BFA-treated big3 mutant in spite of stimulation by NAA , in contrast to seedlings that expressed both BIG3 and BFA-resistant GNOM ( Figure 4E–L ) . big3 mutants displayed binucleate cells , suggesting an essential role for secretory traffic in cytokinesis required for lateral root initiation ( Figure 4M–T ) . For comparison , the BFA-induced defects in seed germination and primary root growth of big3 were not rescued by engineered BFA-resistant GNOM , thus depending on secretory traffic rather than recycling ( Figure 1C , E , H ) . 10 . 7554/eLife . 02131 . 010Figure 4 . BIG1-4 in response to auxin application . ( A–D ) Visualization of auxin distribution by DR5::NLS-3xGFP ( green ) in BFA-treated seedlings after gravistimulation . Arrows , gravity vector . Cell walls were stained by propidium iodide ( PI; magenta ) . Wild-type ( A ) and big3 mutant seedling roots ( B ) did not respond to gravity ( open asterisks ) , in contrast to BFA-resistant GN either in wild-type ( GNR , C ) or big3 mutant background ( GNR big3 , D ) . Asterisks , auxin response in epidermal cell layer on lower side ( C and D ) . ( E–H ) NAA and BFA treatment led to proliferation of pericycle cells ( arrows ) in wild-type ( E ) but not big3 mutants without ( F ) or with BFA-resistant GN ( H ) . Normal lateral root primordia only formed in BFA-resistant GN ( GNR , G ) . Scale bars , 25 µm . ( I–L ) Bright-field microcopy of developing lateral root primordia in NAA-treated seedlings; genotypes: wild-type ( Col; I ) , big3 ( J ) , BFA-resistant GN ( GNR; K ) and BFA-resistant GN in big3 mutant background ( GNR big3; L ) . ( M–T ) Live imaging of DR5::NLS-3xGFP of seedling roots after NAA and BFA treatment . DR5::NLS-3xGFP signals ( left panels M , O , Q , S ) overlaid with Nomarski images ( right panels N , P , R , T ) . Pericycle cells proliferated in wild-type ( M and N ) but became binucleate ( asterisks ) in big3 ( O and P ) and GNR big3 ( S and T ) mutants . Normal lateral root primordia were only formed in BFA-resistant GN ( GNR; Q , R ) mutant . Scale bars , 25 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 010 In plant cytokinesis , which is assisted by a dynamic microtubule array named phragmoplast , both newly-synthesized and endocytosed proteins traffic to the plane of cell division on post-Golgi membrane vesicles that fuse with one another to form the partitioning cell plate ( Samuels et al . , 1995 ) . This raises the problem of coordinating different trafficking routes in the brief period of mitotic division ( Reichardt et al . , 2011 ) . Cell-plate formation requires cytokinesis-specific syntaxin KNOLLE , newly synthesized during late G2/M phase ( Lauber et al . , 1997; Reichardt et al . , 2007 ) . In contrast to wild-type , KNOLLE targeting to the division plane was inhibited in BFA-treated big3 mutants , with KNOLLE accumulating in BFA compartments together with BIG4-YFP ( Figure 5A–F , Figure 5—figure supplement 1A–D ) . Cell-plate formation was disrupted , resulting in binucleate cells , which sometimes displayed cell-wall stubs ( Figure 5—figure supplement 2A–C ) . We used the non-cycling plasma-membrane syntaxin SYP132 expressed from the strong mitosis-specific KN promoter as another secretory marker for trafficking to the cell–division plane ( Reichardt et al . , 2011 ) . SYP132 also accumulated , together with KN , in BFA compartments of BFA-treated big3 mutants , in contrast to BFA-treated wild-type ( Figure 5—figure supplement 1E–J ) . We also analysed endocytosed plasma-membrane proteins PEN1 and PIN1 for BFA-sensitive trafficking to the cell plate in big3 mutants . PEN1 syntaxin involved in non-host immunity accumulates at the pathogen entry site by GNOM-dependent relocation following endocytosis from other regions of the plasma membrane ( Collins et al . , 2003; Nielsen et al . , 2012 ) . PEN1 continually cycles between plasma membrane and endosomes in interphase and accumulates at the cell plate in cytokinesis ( Reichardt et al . , 2011 ) . To make sure that we were only looking at endocytosed PEN1 , PEN1 was expressed from a histone H4 expression cassette that limits protein synthesis to S phase ( Reichardt et al . , 2011 ) . In wild-type , BFA treatment inhibited PEN1 recycling to the plasma membrane but not its trafficking to the cell plate ( Reichardt et al . , 2011; Figure 5G–I ) . In contrast , in BFA-treated big3 mutants , endocytosed PEN1 was not trafficked to the cell division plane but accumulated , together with KNOLLE , in BFA compartments ( Figure 5J–L , asterisks ) . Endocytosed PIN1 trafficked , like KNOLLE , to the cell plate in BFA-treated wild-type but both PIN1 and KNOLLE were trapped in BFA compartments of big3 mutants ( Figure 5M–R ) . Expression of engineered BFA-resistant GNOM did not overcome the trafficking block to the division plane but rather diverted PIN1 to the basal plasma membrane ( Figure 5S–X; compare Figure 5X with Figure 5R ) . Careful analysis of mitotic cells revealed polar accumulation of PIN1 at the plasma membrane of BFA-resistant GNOM seedling roots throughout mitosis while additional PIN1 accumulates at the forming and expanding cell plate , suggesting that trafficking to the plane of division and polar recycling to the plasma membrane occur simultaneously ( Figure 5—figure supplement 3 ) . Thus , both endocytosed and newly-synthesized plasma-membrane proteins require secretory ARF-GEF function BIG1-4 for trafficking to the plane of cell division . 10 . 7554/eLife . 02131 . 011Figure 5 . Trafficking to the plane of cell division is mediated by BIG1 – BIG4 . ( A–F ) Immunolocalization of KNOLLE ( KN; A , D ) and tubulin ( B and E ) in cytokinetic root cells of BFA-treated seedlings ( 50 µM for 3 hr ) . ( A–C ) KN was located at the cell plate ( A ) flanked by tubulin-positive phragmoplast ( B ) in wild-type . ( D–F ) In big3 mutants , KN accumulated in BFA compartments separated from tubulin-positive phragmoplast , resulting in a binucleate cell . ( G–L ) Co-localization of GFP-tagged KN and endocytosed RFP-PEN1 ( H4::RFP-PEN1 ) in BFA-treated seedlings . KN and PEN1 co-localized at the cell plate and in BFA compartments of wild-type ( G–I ) but only in BFA compartments in big3 mutants ( J–L ) . ( M–X ) Immunostaining of GFP-KN and PIN1 in cytokinetic root cells of BFA-treated seedlings . ( M–R ) PIN1 localized apolarly at the plasma membrane ( PM ) and co-localized with KN in BFA compartments and at the cell plate in wild-type ( M–O ) but only in BFA-compartments in big3 mutants ( P–R ) . ( S–U ) In GNR , PIN1 localized polarly at the plasma membrane ( T ) and co-localized with KN ( S ) at the cell plate ( U ) . ( V–X ) Although PIN1 localized polarly at the PM ( W ) in GNR big3 , neither PIN1 ( W ) nor KN ( V ) was located at the cell plate . Blue , DAPI-stained nuclei . Asterisks label nuclei of binucleate cells ( F , L , R , X ) . Scale bars , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 01110 . 7554/eLife . 02131 . 012Figure 5—figure supplement 1 . BIG4 and cargo proteins trapped in BFA compartments of dividing cells in BFA-treated big3 mutant seedlings . ( A–C ) Immunostaining of UBQ10 promoter-driven BIG4-YFP and KN in BFA-treated big3 mutant seedlings . BFA-sensitive YFP-tagged BIG4 ( A ) co-localized with its cargo KN ( B ) in BFA aggregates . ( C ) Overlay . ( D ) Intensity profile of line numbered in ( C ) indicated overlapping signals . ( E–J ) Immunostaining of Myc-tagged SYP132 expressed from the KN promoter ( KN::Myc-SYP132; E , H ) and KN ( F and I ) in BFA-treated cytokinetic cells of wild-type and big3 mutant seedlings . ( E–G ) SYP132 localized at the plasma membrane and KN-labeled cell plate in wild-type whereas both SYP132 and KN were trapped in BFA compartments in big3 mutant ( H–J ) . Asterisks , binucleate cells ( C and J ) . Scale bars , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 01210 . 7554/eLife . 02131 . 013Figure 5—figure supplement 2 . Ultrastructural appearance of cryofixed , freeze-substituted and resin-embedded big3 seedling root tips treated with BFA . ( A–C ) Binucleate cells in big3 seedling roots treated with BFA . ( A ) Overview of ultrastructural TEM analysis . ( B ) Higher magnification of boxed area in ( A ) . Note absence of cell-wall remnants or membrane vesicles between the daughter nuclei ( B ) . ( C ) Another cell showing cell wall remnants ( stubs , blue asterisks ) . Nuclei ( n ) have been false-colored; second nucleus in ( C ) is in a different focal plane . cw , cell wall; er , endoplasmic reticulum; m , mitochondrion; n , nucleus; v , vacuole . Scale bars , 2 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 01310 . 7554/eLife . 02131 . 014Figure 5—figure supplement 3 . PIN1 recycling in mitotic cells . BFA-treated seedlings expressing PIN1-GFP ( green ) were counterstained with tubulin ( magenta ) and DAPI ( blue chromatin ) . ( A–I ) In wild-type , PIN1-GFP localizes apolarly at the plasma membrane and at the cell plate in different stages during cytokinesis ( BFA 50 μM 3h ) . ( J–U ) In BFA-resistant GNOM lines ( GNR ) , PIN1-GFP localizes polarly at the plasma membrane and at the cell plate at the same time ( BFA 50 μM 3h ) . ( V–X ) Even after prolonged BFA-treatment ( 6h ) of BFA-resistant GNOM seedlings , PIN1 polarity and cell-plate localization are maintained . Mitotic stages: arrowhead , metaphase ( panels F and O ) ; asterisk , anaphase ( panels C , L , X ) ; circle , telophase ( panels C , I , R , U , X ) ; cross , late cytokinesis ( panels F , I , O , R ) . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 01410 . 7554/eLife . 02131 . 015Figure 5—figure supplement 4 . Highly schematic model of secretory and recycling trafficking pathways in interphase and cytokinesis . In interphase , proteins are synthesized at the ER and are transported via the Golgi to the TGN . The TGN serves as a sorting station . From there , cargo can be transported to the plasma membrane ( PM; secretion , green ) and this pathway requires the ARF-GEFs BIG1-BIG4 . When plasma membrane localized proteins are endocytosed ( endocytic pathway , red ) , they can return to the plasma membrane via the GNOM-dependent recycling pathway ( orange ) . During cytokinesis , newly synthesized proteins are transported from the ER to the Golgi/TGN and from there to the cell plate ( CP ) in BIG1-BIG4 dependent fashion . Not only newly synthesized cargo but also endocytosed proteins follow this secretory route to the cell plate . PIN1 appears to be exceptional among endocytosed proteins , being recycled to the basal plasma membrane in a GNOM-dependent manner during cytokinesis ( ? ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02131 . 015 It is a particularity of Arabidopsis and some other flowering-plant species that the secretory pathway of membrane traffic is comparatively insensitive to BFA treatment whereas endosomal recycling of endocytosed plasma-membrane proteins is rather sensitive ( Geldner et al . , 2001 , 2003; Teh and Moore , 2007; Richter et al . , 2007 ) . The BFA insensitivity of the secretory pathway depends on the BFA resistance of ARF-GEF GNL1 , which mediates COPI-vesicle formation in retrograde Golgi-ER traffic ( Teh and Moore , 2007; Richter et al . , 2007 ) , and also requires another BFA-resistant ARF-GEF acting in post-Golgi traffic to the plasma membrane . Here we show that ARF-GEFs BIG1-4 act at the TGN to mediate secretion of newly synthesized proteins to the plasma membrane in interphase but not recycling of endocytosed plasma-membrane proteins , and that BIG3 is BFA-resistant , unlike GNOM involved in recycling to the plasma membrane . Thus , there are two distinct trafficking pathways from the TGN to the plasma membrane in interphase . This is best illustrated by the trafficking of auxin-efflux carrier PIN1 - whereas newly synthesized PIN1 requires BIG1-4 on the late secretory pathway for non-polar delivery to the plasma membrane , polar PIN1 recycling to the basal plasma membrane solely depends on ARF-GEF GNOM ( see model in Figure 5—figure supplement 4 ) . Like newly synthesized proteins , endocytosed proteins are targeted to the division plane during cytokinesis ( Reichardt et al . , 2011 ) . Proteins that cycle between endosomes and the plasma membrane in interphase accumulate , preferentially or even exclusively , at the cell plate ( Reichardt et al . , 2011 ) . In general , recycling to the plasma membrane appears to be switched off during cytokinesis . Here we show that secretory ARF-GEFs BIG1-4 are essential for protein trafficking to the plane of cell division , regardless of proteins being newly synthesized or endocytosed from the plasma membrane ( see model in Figure 5—figure supplement 4 ) . Although trafficking to the plane of cell division appears to override recycling of endocytosed proteins to the plasma membrane , we noticed one clear exception—auxin-efflux carrier PIN1 , which accumulates polarly at the plasma membrane in interphase and during cell division when both BFA-resistant BIG3 and engineered BFA-resistant GNOM were expressed . Rather than substituting for BIG1-4 in traffic to the plane of cell division , recycling ARF-GEF GNOM appeared to counteract that process by promoting PIN1 recycling to the basal plasma membrane . Of course , the critical question is whether both processes occur at the same time or whether GNOM-dependent PIN1 recycling only sets in after trafficking to the cell plate has come to an end . Although there are no time-course studies , which would be difficult to perform because the process is very fast , detailed analysis of dividing cells at different mitotic stages revealed that polar recycling mediated by BFA-resistant GNOM occurs throughout mitosis and cytokinesis . Furthermore , only in the absence of both BFA-resistant BIG3 and BFA-resistant GNOM is PIN1 trapped in BFA compartments . If then BFA-resistant GNOM is expressed PIN1 is not delivered to the plane of division but rather polarly recycled to the plasma membrane , again suggesting that the latter pathway is a direct route bypassing the cell plate . PIN1 might be exceptional because continuous recycling of PIN1 is required for maintaining the polar transport of auxin across tissues ( Geldner et al . , 2003 ) . If PIN1 recycling were shut down during cytokinesis this would disrupt the polar auxin transport required in specific developmental situations such as forming lateral root primordia when essentially all cells proliferate ( Geldner et al . , 2004 ) . Another problem in auxin flow arises from cell division when the partitioning membrane has physically separated the two daughter cells: one daughter suddenly has PIN1 located at opposite ends . Obviously , PIN1 has to be removed from the wrong end in order to sustain polar auxin transport . This seems to be a fast process and has been studied for the related auxin-efflux carrier PIN2 in detail ( Men et al . , 2008 ) . Animal and plant cytokinesis differ in the way the partitioning membrane is laid down . In animals , secretory and recycling pathways contribute to the ingrowth of the plasma membrane mediated by a contractile actomyosin ring and to the subsequent abscission of the daughter cells ( Schiel and Prekeris , 2013 ) . In plants , a massive flow of membrane vesicles from TGN/early endosome to the plane of cell division sustains , by fusion , the rapid formation and outward expansion of the partitioning cell plate ( Samuels et al . , 1995 ) . This process is orchestrated by a specialised cytoskeletal array termed phragmoplast that delivers those membrane vesicles to the division plane . Phragmoplast-assisted trafficking might be required for completing the partitioning membrane on time , in the absence of a cytokinesis-interphase checkpoint , and would thus effectively rule out recycling of endocytosed proteins to the plasma membrane . However , our results make clear that this is not the case because recycling to the plasma membrane is not switched off during cytokinesis . Rather , endocytosed proteins enter the late-secretory pathway to reach the division plane at the expense of being recycled to the plasma membrane , which requires the late-secretory ARF-GEFs BIG1-4 . In conclusion , our results raise the possibility that in general , different ARF-GEFs have different specificity of action during vesicle formation such that the same cargo protein can be delivered to different destinations . Plants were grown on soil or agar plates in growth chambers under continuous light conditions at 23°C . big mutant lines: big1 ( GK-452B06 ) and big2 ( GK-074F08 ) T-DNA lines were from GABI-KAT ( http://www . gabi-kat . de ) , big3 ( SALK_044617 ) and big4 ( SALK_069870 ) T-DNA lines from the SALK collection ( http://signal . salk . edu/cgi-bin/tdnaexpress ) . big3 mutant lines were selected on MS plates using kanamycin . The following transgenic marker lines were used: H4::RFP-PEN1 ( Reichardt et al . , 2011 ) ( expressed from HISTONE4 ( H4 ) promoter during S phase ) , KN::Myc-SYP132 ( Reichardt et al . , 2011 ) ( expressed during lateG2/M phase ) , HS::secGFP ( Viotti et al . , 2010 ) ( expressed from heat shock promoter ) , GFP-KN ( Reichardt et al . , 2007 ) , BOR1-GFP ( Takano et al . , 2005 ) , DR5::NLS-3xGFP ( Weijers et al . , 2006 ) , VHA-a1-RFP ( Viotti et al . , 2010 ) , AP1M2-3xHA ( Park et al . , 2013 ) . Primers used to test for big1 heterozygosity: 5′GCAAGATCAGGGAAGACG 3′ and 5′ACCAGAGGAAGGTGCTTCTTC 3′ Primers used to test for big1 homozygosity: 5′TCGTCCCATCTTCTTCATTTG 3′ and 5′ACCAGAGGAAGGTGCTTCTTC 3 Primers used to test for big2 heterozygosity: 5′GCAAGATCAGGGAAGACG 3′ and 5′TTGAGGGGTTCATATGACAGC 3′ Primers used to test for big2 homozygosity: 5′TTTCCCACTTTTTCCACTGTG 3′ and 5′TTGAGGGGTTCATATGACAGC 3′ Primers used to test for big3 heterozygosity: 5′AAACTCTCCACTGGCTAAGCC 3′ and 5′ATTTTGCCGATTTCGGAAC 3′ Primers used to test for big3 homozygosity: 5′AAACTCTCCACTGGCTAAGCC 3′ and 5′GCAAGTTTTCTTGCGCAATAC 3′ Primers used to test for big4 heterozygosity: 5′ATTTTGCCGATTTCGGAAC 3′ and 5′CTATCTTGCGCTGGAGACAAC 3′ Primers used to test for big4 homozygosity: 5′TCCTCTTCAAACTCGTCAACG 3′ and 5′CTATCTTGCGCTGGAGACAAC 3′ Genomic BIG4 was amplified and introduced into pDONR221 ( Invitrogen , Darmstadt , Germany ) and afterwards into UBQ10::YFP destination vector ( Grefen et al . , 2010 ) . For generation of BFA-resistant UBQ10::BIG4R-YFP , methionine at position 695 was exchanged with leucine by site-directed mutagenesis . BIG3 promoter was amplified and introduced into pUC57L4 via KpnI and SmaI restriction sites . Multistep gateway cloning was performed using pUC57L4-BIG3-promoter , pEntry221-BIG4 and R4pGWB553 ( Nakagawa et al . , 2008 ) yielding BIG3::BIG4-RFP . Cloning the CDS from BIG3 into pGREENII via ApaI and SmaI restriction sites generated pGII-BIG3 . The 1 kb BIG3 promoter was amplified and introduced into pGII-BIG3 via ApaI . 1 kb of 3′UTR was amplified and introduced into pGII-BIG3::BIG3 via SmaI and SpeI . C-terminal YFP was inserted via SmaI and SpeI . AFVY-RFP was amplified from 35S::AFVY-RFP ( Scheuring et al . , 2011 ) and introduced into pDONR221 ( Invitrogen ) generating a pEntry clone . Afterwards , LR reaction was performed introducing AFVY-RFP into the estradiol-inducible destination vector pMDC7 ( Curtis and Grossniklaus , 2003 ) . PIN1 cDNA was cloned into pGem-T ( Promega , Mannheim , Germany ) . RFP was inserted in PIN1 via the XhoI site . PIN1-RFP was amplified and introduced first into pDONR221 and then into pMDC7 . YFP-SYP132 was amplified and introduced into pDONR221 and then into pMDC7 . All constructs were transformed into big3 mutants and BFA-resistant GN ( GNR ) in big3 mutant background . T1 plants of UBQ10::BIG4-YFP , UBQ10::BIG4R-YFP and BIG3-YFP were selected by spraying with Basta . T1 seeds of estradiol-inducible lines and BIG3::BIG4-RFP were selected with hygromycin . Experiments were performed using T2 or T3 seedlings . At least three independent lines were analysed . 5 days old seedlings were incubated in 1 ml liquid growth medium ( 0 . 5x MS medium , 1% sucrose , pH 5 . 8 ) containing 50 µM BFA ( Invitrogen , Molecular Probes ) for 1 hr or 3 hr at room temperature in 24-well cell-culture plates . Seedlings treated with 50 µM BFA for ( a ) 1 hr or ( b ) 3 hr , respectively , were used for the following immunolocalisation studies: ( a ) AP1M2 vs γCOP , AP1M2 vs SYP61 , PIN1; ( b ) KNOLLE vs Tubulin , KNOLLE vs PIN1 , H4::RFP-PEN1 vs GFP-KN and KN::Myc-SYP132 vs KN . Incubation was stopped by fixation with 4% paraformaldehyde in MTSB . Immunofluorescence staining was performed as described ( Lauber et al . , 1997 ) or with an InsituPro machine ( Intavis , Cologne , Germany ) ( Müller et al . , 1998 ) . Antibodies used: mouse anti-MYC ( Santa Cruz Biotechnology , Heidelberg , Germany ) 1:600 , mouse anti-HA 1:1000 ( BAbCO , Richmond , CA , USA ) , rat anti-tubulin 1:600 ( Abcam , Cambridge , UK ) , rabbit anti-PIN1 1:1000 ( Geldner et al . , 2001 ) , rabbit anti-γCOP 1:1000 ( Agrisera , Vännäs , Sweden ) , rabbit anti-KNOLLE 1:2000 ( Reichardt et al . , 2007 ) and rabbit anti-SYP61 1:700 ( Park et al . , 2013 ) . Alexa-488 or Cy3-conjugated secondary antibodies ( Dianova , Hamburg , Germany ) were diluted 1:600 . Live-cell imaging was performed with 2 µM FM4-64 or FM1-43 ( Invitrogen , Molecular Probes ) or propidium iodide ( 10 µg/ml ) . Estradiol induction was performed using 10 or 20 µM estradiol . BFA incubation ( 25 µM ) was done together with estradiol for 6 hr . Heat-shock inducible secGFP ( HS::secGFP ) lines were first incubated for 30 min at 37°C in MS at pH8 . 1 . BFA treatment ( 50 µM ) in MS at pH8 . 1 followed for 4 hr at plant room conditions . Analysis of BOR1 degradation was performed according to Takano et al . ( 2005 ) . In addition , we treated the seedlings with BFA , 5 µM , for 1 hr together with boron . For ultrastructural analysis , root tips were high-pressure frozen ( Bal-Tec HPM010; Balzers ) in hexadecene ( Merck Sharp and Dohme , Haar , Germany ) , freeze-substituted in acetone containing 2 . 5% osmium tetroxide , washed at 0°C with acetone , and embedded in Epon . For immunogold labeling of ultrathin thawed cryosections , root tips were fixed with 8% formaldehyde ( 2 hr ) , embedded in gelatin , and infiltrated with 2 . 1 M sucrose in PBS as previously described ( Dettmer et al . , 2006 ) . Thawed ultrathin sections were labeled with rabbit anti-GFP antibodies ( 1:300; Abcam ) and silver-enhanced ( HQ Silver , 8 min; Nanoprobes , Yaphank , NY , USA ) goat anti-rabbit IgG coupled to Nanogold ( no . 2004; Nanoprobes ) . Antibodies and markers were diluted in blocking buffer ( PBS supplemented with 0 . 5% BSA and 1% milk powder ) . Fluorescence images were acquired at 512 × 512 or 512 × 256 pixels with the confocal laser scanning microscope TCS-SP2 or TCS-SP8 from Leica , using the 63x water-immersion objective and Leica software . All images were processed with Adobe Photoshop CS3 only for adjustment of contrast and brightness . Intensity line profile was performed with Leica software . Pollen medium was prepared as described ( Boavida and McCormick , 2007 ) . Pollen germinated over night or for 5 hr before microscopic analysis . To investigate primary root growth , 5–6 days old seedlings were transferred to plates with 10 µM BFA and analysed after 5–7 additional days using ImageJ . DR5::NLS-GFP expressing seedlings analysed for lateral root formation were treated with 5 µM NAA or 5 µM NAA plus 10 µM BFA over night . Roots were cleared according to Geldner et al . ( 2004 ) . Gravitropic response was investigated by transferring 5 days old seedlings , expressing DR5::NLS-GFP , to BFA plates ( 5 µM ) . Seedlings were grown vertically for 1 hr on BFA plates before rotated by 135° for 4 hr . For analysis of seed germination , seeds were sown out on MS medium containing 5 µM BFA . Images were taken after 5 days of growth . Full-length protein sequence of BIG3 was used to search for related sequences from different plant species with sequenced genomes that are available at the phytozome homepage ( http://www . phytozome . net/ ) . ARF-GEFs from different species were aligned by ClustalW ( www . ebi . ac . uk/clustalw ) and the phylogenetic tree was drawn with Dendroscope ( Huson et al . , 2007 ) .	Cells are surrounded by a plasma membrane , and when a cell divides to create two new cells , it must grow a new membrane to keep the two new cells apart . Animal cells and plant cells tackle this challenge in different ways: in animal cells the new membrane grows inwards from the surface of the cell , whereas the new membrane grows outwards from the centre of the cell in plant cells . The materials needed to make the plasma membrane are delivered in packages called vesicles: most of these materials arrive from a structure within the cell called the trans-Golgi network , but some materials are recycled from the existing plasma membrane . In plants the formation of the new cell membrane is orchestrated by scaffold-like structure that forms in the plant cell called the ‘phragmoplast’ . It is widely thought that this structure guides the vesicles bringing materials from the trans-Golgi network , but the details of this process are not fully understood . Now , Richter et al . have discovered four proteins , called BIG1 to BIG4 , that control the formation of the new cell membrane in the flowering plant Arabidopsis thaliana , a species that is routinely studied by plant biologists . These four proteins belong to a larger family of proteins that control the trafficking of vesicles within a cell . Richter et al show that a plant cell can lose up to three of these four proteins and still divide , as the plant can still grow and develop as normal . Thus , BIG1 to BIG4 appear to perform essentially the same role in the plant . Richter et al . also show that , when a plant cell is not dividing , these proteins are involved in controlling the delivery of new materials to surface membrane , and not the recycling of material . However , when a cell is dividing , these proteins switch to regulate both processes , but direct all the material to a new destination—the newly forming membrane , instead of the established surface membrane . Richter et al . suggest that this switch is important to stop any recycling to the plasma membrane that might move material away from the new membrane . The next challenge will be to identify the molecular signals and mechanisms that enable the proteins BIG1 to BIG4 to re-route the recycling of membrane material during cell division .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology", "cell", "biology" ]	2014	Delivery of endocytosed proteins to the cell–division plane requires change of pathway from recycling to secretion
Despite recent improvements in microscope technologies , segmenting and tracking cells in three-dimensional time-lapse images ( 3D + T images ) to extract their dynamic positions and activities remains a considerable bottleneck in the field . We developed a deep learning-based software pipeline , 3DeeCellTracker , by integrating multiple existing and new techniques including deep learning for tracking . With only one volume of training data , one initial correction , and a few parameter changes , 3DeeCellTracker successfully segmented and tracked ~100 cells in both semi-immobilized and ‘straightened’ freely moving worm's brain , in a naturally beating zebrafish heart , and ~1000 cells in a 3D cultured tumor spheroid . While these datasets were imaged with highly divergent optical systems , our method tracked 90–100% of the cells in most cases , which is comparable or superior to previous results . These results suggest that 3DeeCellTracker could pave the way for revealing dynamic cell activities in image datasets that have been difficult to analyze . Imaging cells to reveal the dynamic activities of life has become considerably more feasible because of the remarkable developments in microscope hardware in recent years ( Frigault et al . , 2009; Ahrens and Engert , 2015; Bouchard et al . , 2015; Weisenburger and Vaziri , 2018 ) . In addition , multiple software platforms for processing 2D/3D still images and 2D + T images have been developed ( Eliceiri et al . , 2012 ) . However , processing cells in 3D + T images has remained difficult , especially when cells cannot be clearly separated and/or their movements are relatively large , such as cells in deforming organs . For processing objects in 3D + T images , the following two steps are required: ( 1 ) segmentation: segmenting the regions of interest in each 3D image into individual objects ( Figure 1—figure supplement 1 , left ) and ( 2 ) tracking: linking an object in a particular volume to the same object in the temporally adjacent volume ( Figure 1—figure supplement 1 , right ) . For segmenting and tracking ( hereafter collectively called ‘tracking’ ) cells of deforming organs in 3D + T images , programs optimized for processing images in particular conditions have been developed ( Schrödel et al . , 2013; Toyoshima et al . , 2016; Venkatachalam et al . , 2016; Nguyen et al . , 2017 ) . However , these methods cannot be used under conditions other than those for which they were designed , at least without a loss in processing efficiency . In other words , 3D + T images , especially those obtained under challenging conditions , can be efficiently processed only when customized software is developed specifically for those images . One reason for this is that many parameters must be optimized to achieve good results—For example , even in 2D image processing , changes in lighting could require the re-optimization of parameters for segmentation and tracking ( Egnor and Branson , 2016 ) . One way to solve this problem is to optimize the parameters automatically using machine learning , especially deep learning . Deep learning methods use an artificial neural network with multiple layers , that is , a deep network , to process complex data and automatically optimize a number of parameters from training data , which allows users to easily apply a single method to images obtained under different conditions . In addition to this flexibility , deep learning methods have outperformed conventional methods in some image processing tasks such as image classification ( LeCun et al . , 2015; Krizhevsky et al . , 2012 ) . Nevertheless , deep learning methods are used mostly for segmentation and/or object detection , but not for tracking because of the difficulty in preparing training data ( Moen et al . , 2019 ) : Correctly tracking a number of objects manually to obtain training data , especially from 3D + T images recorded over long period of time , is extremely challenging . In addition , designing a multiple-step pipeline for segmenting cells and tracking their positions to work under various conditions has been difficult . In this study , we developed 3DeeCellTracker , a new pipeline utilizing deep learning techniques in segmentation and , for the first time to our knowledge , in tracking of cells in 3D + T images . We solved the problem of training data preparation for tracking by creating a synthetic dataset ( see below ) . We also designed a multiple-step pipeline to achieve accurate tracking . 3DeeCellTracker was implemented on a desktop PC to efficiently and flexibly track cells over hundreds of volumes of 3D + T images recorded under highly divergent optical or imaging conditions . Using only one volume for training and one initial correction , 3DeeCellTracker efficiently tracked 100–150 neurons in the brains of semi-immobilized or ‘straightened’ ( special pre-processing step based on the worm's posture ) freely moving Caenorhabditis elegans roundworms from four image datasets , obtained using spinning disk confocal systems in three different laboratories . With a few modifications , 3DeeCellTracker also tracked ~100 cells in the naturally beating heart of a zebrafish larva monitored using swept confocally aligned planar excitation ( SCAPE ) microscopy , a novel oblique light sheet microscope system for very rapid 3D + T imaging ( Bouchard et al . , 2015; Voleti et al . , 2019 ) . Furthermore , 3DeeCellTracker tracked ~900 cells in a cultured 3D tumor spheroid , which were imaged with a two photon microscope system . Our pipeline provided robust tracking results from the above-mentioned real datasets as well as from degenerated datasets , which differed in terms of signal-to-noise ratios , cell movements , and resolutions along the z-axis . Notably , 3DeeCellTracker's performance was better in terms of the tracking results than those from recently developed 2D/3D tracking software running on a desktop PC , and comparable to software running on a high-performance computing cluster ( Toyoshima et al . , 2016; Nguyen et al . , 2017; Bannon et al . , 2018 ) . Furthermore , by using the positional information of the tracked cells , we extracted dynamics of the cells: the worm's neurons exhibited complex activity patterns in the brain , the zebrafish heart cells exhibited activities synchronized with heart chamber movement , and the tumor spheroid cells exhibited spontaneous activities without stimulation . These results indicate that , 3DeeCellTracker is a robust and flexible tool for tracking cell movements in 3D + T images , and can potentially enable the analysis of cellular dynamics that were previously difficult to investigate . We developed a new pipeline , 3DeeCellTracker , which integrates novel and existing techniques ( Figure 1A ) . After preprocessing ( see Materials and methods ) , it performs automatic segmentation of cells in all 3D + T images using 3D U-Net for classifying individual voxels into cell or non-cell categories ( Ronneberger et al . , 2015; Çiçek et al . , 2016 ) . Continuous regions of ‘cell’ voxels are separated into individual cell regions using the watershed method ( Beucher and Meyer , 1993 ) , and then numbered . The segmented cells are manually corrected only in the first volume of 3D images . In the following 3D tracking step , we considerably increased the efficiency by introducing a deep learning technique , feedforward network ( FFN ) , to predict cell positions based on spatial patterns of cells maintained between previous and current images . The predicted positions are corrected with a non-rigid point set registration method called PR-GLS ( Ma et al . , 2016 ) and by our custom method to obtain precise cell locations , which are critical for extracting the accurate dynamics of cellular signals . The 3D U-Net and the FFN are pre-trained using manually confirmed data or synthetic data ( Figure 1B , C ) from a single 3D image volume . The tracking results were visually inspected by comparing the locations of tracked cells with the corresponding raw images ( Figure 1—figure supplement 2 ) . The keys to our method are the use of simulation to produce large amounts of training data for the FFN and the carefully designed post-processing methods for the FFN , which result in the flexible and robust tracking of moving cells in very different 3D + T datasets . In the following two sections , we describe details of the segmentation and tracking method . For segmentation , cell-like regions in an image should be segmented into individual cells which may differ in intensities , sizes , shapes , and textures . Cell segmentation in 2D images using deep networks has been previously reported ( Ronneberger et al . , 2015; Van Valen et al . , 2016; Bannon et al . , 2018 ) . In this study , we utilized a deep network called 3D U-Net to segment cells in 3D images to predict the class labels ( cell or non-cell ) of individual voxels based on information contained in neighboring voxels ( Figure 2A and Figure 2—figure supplement 1; Ronneberger et al . , 2015; Çiçek et al . , 2016 ) . The U-Net can generate precise segmentation under diverse imaging conditions and can be trained with very few annotated images ( e . g . only one 3D image volume in this study; see Figure 1B ) . The pre-processed images of the first volume are used to train the 3D U-Net ( Figure 1B ) , and the trained U-net is then used for the segmentation of cells in all the following volumes . Once trained on one dataset , the 3D U-Net can be directly reused for different datasets obtained under similar optical conditions . The cell-like regions detected using 3D U-Net are grouped and separated into individual cells using the watershed method ( see Figure 2A and Materials and methods ) . For tracking cells , two major strategies can be considered . One strategy is to utilize the information contained in each cell region , for example , local peaks or distributions of intensities ( Toyoshima et al . , 2016 ) . Using this information , a program can update the position of each cell by searching for its new position in nearby regions in the next volume . However , the obvious drawback of this strategy is that cells can be mistracked if their movements are comparable to or greater than the distances between adjacent cells ( see below ) . Another strategy is to represent cells by their center points , ignoring the information in each cell region and treating the cells as a set of points . In this strategy , the new positions of a cell set in the next volume can be estimated based on the patterns of their spatial distributions , and cells with large movements can also be tracked based on the global trend of the movements . Previously reported methods utilizing this strategy , called point set registration ( Ma et al . , 2016; Jian and Vemuri , 2005; Myronenko et al . , 2006 ) , used the spatial distributions and coherency of movements to track points within artificial datasets . However , the spatial patterns are conventionally characterized by features designed by experts , an approach that did not work well for the cell images used in this study ( refer to the results below concerning Fast Point Feature Histograms [FPFH; Rusu et al . , 2009] versus FFN ) . Another problem with this strategy is that the estimated positions are not always accurate because the information contained in each cell region is ignored ( see the results below concerning our method for accurate correction ) . In order to obtain more robust and accurate tracking results , we integrated the spatial pattern ( i . e . point sets ) and local cell region strategies and used a deep learning technique , FFN , for the former . We used FFN to match temporally adjacent cells based on the distance pattern between each cell and its 20 surrounding cells ( Figure 2B ) . Using the pattern obtained by the FFN , all the cells at t1 are compared with all the cells at t2 , and the most similar ones are regarded as the same cells at t1 and t2 , a process we call initial matching . Although deep network techniques are expected to produce superior cell tracking results , the method had not been used previously because it requires large amounts of training data . These data are difficult to prepare manually , especially for 3D + T images , as validating and correcting large numbers of cell positions over time by switching multiple layers along the z axis is virtually impossible . To solve this difficulty in preparing training data for FFN , we generated >500 , 000 , 000 synthetic training data points by simulating cell movements ( see Figure 2—figure supplement 2 and Materials and methods ) . In the pipeline , the center points of cells are extracted from the cell regions segmented by 3D U-Net and the watershed method , and a pre-trained FFN ( Figures 1C and 2B ) is applied to the cell points to generate the initial matching from volumes t to t+1 ( Figure 2C , panels 2–2 and 4–1 ) . To improve the initial matching , a non-rigid point set registration method ( PR-GLS ) ( Ma et al . , 2016 ) is used to generate a coherent transformation , that is , neighboring cells should have similar movements ( Figure 2C , panel 4–2 ) . Originally , PR-GLS was used in conjunction with the FPFH method to predict cell positions ( Ma et al . , 2016 ) ; however , our FFN generates more accurate initial matchings than FPFH ( Figure 2—figure supplement 3A and B ) , and our combination of FFN + PR-GLS generates more accurate predictions of cell positions than the FPFH + PR-GLS or the classic affine alignment ( Myronenko et al . , 2006 ) does ( Figure 2—figure supplement 3C and D ) . Nevertheless , our method sometimes generates subtle errors because some cells may show slightly different movements from those of their neighboring cells , which can accumulate over time . To overcome these errors , the estimated positions are accurately corrected to compensate for their differences by utilizing information from local cell regions contained in the 3D U-Net output ( Figure 2C , panel 5 , Figure 2—figure supplement 4 , and Materials and methods ) . To test the performance of 3DeeCellTracker , we analyzed 3D + T images of neurons in the deforming brain of C . elegans . C . elegans has been used as a model for imaging all neuronal activities in the brain ( ‘whole brain imaging’ ) owing to its small brain ( ~40 µm3 in an adult ) in a transparent body , the complete description of all connections of 302 neurons , feasibility in the use of genetically encoded calcium indicators ( GECI ) , and the capability of perception , memory , and decision-making in its small brain ( de Bono and Maricq , 2005 ) . Whole brain imaging of C . elegans has been reported from several laboratories , and the most popular optical system currently is the spinning disk confocal system , for which each laboratory has developed their own tracking software ( Schrödel et al . , 2013; Toyoshima et al . , 2016; Venkatachalam et al . , 2016; Nguyen et al . , 2017 ) . We established our own spinning disk confocal system for whole-brain imaging , OSB-3D ( see Materials and methods for details ) , and obtained 3D + T images of whole brain activity from a strain established in our laboratory and from the one used in a previous study by Nguyen et al . , 2016 ( datasets worm #1 and #2 , respectively ) . In addition , we analyzed the whole-brain images published previously using a different optical system ( Toyoshima et al . , 2016; dataset worm #3 ) . In all datasets , the worms were semi-immobilized either by anesthetization ( worm #1 and #2 ) or by constriction ( worm #3 ) , and red fluorescent proteins and GECI were used to monitor cell positions and cell activities , respectively . After segmentation , we manually confirmed 164 neurons in the head in the first volume of 3D + T images , in which the distribution of cell signals was mostly , but not completely , separated from the background signals ( dataset worm #1a; see Figure 3A–C ) . While the worm was anesthetized , its head deformed and the cells moved during imaging . Our method tracked all the neurons in all 171 volumes except for those that moved out of the field of view of the camera ( Figure 3D–F , Figure 3—video 1 ) . To evaluate the difficulty in tracking cells with large movements , we calculated a score for the relative movements ( RM ) , which is the movement of a cell divided by the distance from that cell to its closest neighboring cell . When RM is small , searching for the closest cell is the simplest method to find the new position of a given cell in the next volume . However , such a simple approach may lead to tracking errors when RM ≥0 . 5 ( Figure 4 and Materials and methods ) . Although most of the cell RM values were ≤0 . 5 in the worm #1 dataset ( Figure 3B , bottom ) , other datasets had cells with RM ≥0 . 5; nevertheless , these were also successfully tracked by our program ( see below ) . We also tested worm #1b dataset , obtained from the same strain with worm #1a ( Figure 3—figure supplement 1 ) , in which we again achieved 100% tracking accuracy without changing any parameters aside from the noise level ( see Tables 1 and 2 ) . The positional information of cell nuclei in both datasets was then used to extract the calcium dynamics of the neurons based on GECI intensities , which reflect spontaneous activities of neurons in the worm's brain ( Figure 5 , Figure 5—figure supplement 1 and Figure 5—video 1 ) . The same method was applied to 3D + T images obtained using the same OSB-3D system in our laboratory but from the neurons of a worm strain used in a previous whole brain imaging study ( Nguyen et al . , 2016; AML14 strain , dataset worm #2; Figure 3—figure supplement 2A ) . Our method again achieved a 100% accuracy of tracking and extracted calcium signals from all 101 neurons even though this dataset differs considerably from the worm #1 dataset in terms of nuclear marker intensity and movement patterns ( Figure 3—figure supplement 2A–F , Figure 3—video 2 and Figure 5—video 2 ) . It should be noted that more neurons were detected using our method than using the original method ( approximately 90 neurons or less ) ( Nguyen et al . , 2016 ) . We also applied our method to a publicly available dataset , which was obtained using a different strain , a different microscopy setup , and different imaging conditions from worm #1 and #2 datasets ( worm #3 , Figure 3—figure supplement 3A; Toyoshima et al . , 2016 ) . In this case , the worm was loosely constrained in a microfluidic device and not anesthetized , thus exhibited larger head deformations and cell displacements between volumes ( by a factor of approximately three times ) compared to the worm #1 and #2 datasets ( Figure 3—figure supplement 3A and B ) . In addition , this dataset had a lower resolution ( half the resolution in the x and y directions ) . Nevertheless , with a few parameter modifications ( Table 2 and Materials and methods ) , our method correctly tracked 171/175 ( =98% ) of the neurons ( Figure 3—figure supplement 3C–F and Figure 3—video 3 ) . Our result was comparable to that of the original report , in which 171 out of 198 neurons were tracked without error ( Toyoshima et al . , 2016 ) . These results indicate that our method can flexibly process 3D + T images of neurons in a semi-immobilized worm's brain obtained under different conditions . To reveal the relationships between neuronal activities and animal behavior , Nguyen et al . , developed a method to track neurons in a freely moving worm , in which the deformation of the brain and the movements of neurons are considerably larger than those occurring in semi-immobilized worms ( Nguyen et al . , 2016; Nguyen et al . , 2017 ) . After ‘straightening’ the worm and segmentation of its neurons , they made a set of reference volumes to which each neuron was matched to assign its identity . Although this method is powerful , it requires a high-performance scientific computing cluster running on up to 200 cores simultaneously . We therefore tested whether the 3DeeCellTracker implemented on a desktop PC could process such a challenging dataset ( worm #4 ) . With a few modifications ( see Materials and methods ) , we used 3DeeCellTracker to segment and track 113 cells in the initial 500 volumes from the worm #4 dataset . Here we analyzed the images preprocessed by the straightening method ( Nguyen et al . , 2017 ) , which is necessary for our current method . Even after the straightening , the cell movements were quite large , with many comparable to the distances between cells , that is , RM ≥0 . 5 ( Figure 6A and B ) . We visually inspected the tracking of 90 cells while the remaining 23 were not checked because of difficulties arising from weak intensities/photobleaching and/or the cells being in dense regions . Note that 70 cells were inspected in the original report ( Nguyen et al . , 2017 ) . We found that 66 out of 90 cells ( 73% ) were correctly tracked without errors ( Figure 6D–F , single mode ) , which is acceptable but not ideal . To further improve our method for the larger cell movements , we developed a new mode , in which multiple predictions of cell positions are made from different previous time points , and the final prediction is taken as the average of these predictions . We call this ‘ensemble mode’ and the previous method ‘single mode’ , wherein the predictions of cell positions at time t are derived from the results at t-1 ( Figure 6C ) . When applying the ensemble mode , we again found 66 cells correctly tracked without errors , and that the remaining 24 cells were correctly tracked in most volumes ( 94%–98 . 2% ) . This was a substantial improvement over the single mode , in which errors at t are maintained until the end of tracking ( Figure 6D–F , ensemble mode; Figure 6—video 1 ) . In total , the ensemble mode correctly tracked 44 , 905 out of 45 , 000 cell movements ( 99 . 8% ) , a result at least comparable to that in the original report ( Nguyen et al . , 2017 ) . From the trajectories of two example cells , we found that large movements , including ones along the z-axis ( across ~30 layers ) , occurred frequently ( Figure 6G ) , demonstrating the excellent performance of our method on a desktop PC in this challenging dataset containing considerably large scale movements . However , because the ensemble mode requires longer times for the tracking than the single mode ( 10 versus 2 min/volume , respectively ) , we used the single mode in the following analyses . To test the general applicability of the 3DeeCellTracker , we applied our method to the 3D + T images of a naturally beating heart in a zebrafish larva obtained at 100 volumes per second using a substantially different optical system , the SCAPE 2 . 0 system ( Voleti et al . , 2019; Figure 7A ) . The speed of image acquisition of this system is extraordinary relative to that of the spinning confocal system , which generally allows for the recording of ≤10 volumes/s . This dataset includes both large cells with stronger intensities that are easy to segment and small cells with weaker intensities that are difficult to segment ( Figure 7B , top and middle ) . The photobleaching in the dataset made it challenging to detect and track the weak intensity cells in the later part of the imaging , because of the substantial overlap between the small cell signals and background signals ( Figure 7B; Figure 7—figure supplement 1 ) . This weak intensity is unavoidable because of the extremely quick scanning rate of the system ( 10 , 568 frames per second ) . In addition , the rapid beating of the heart caused relatively large movements of all the cells in the x-y-z directions ( Figure 7B , bottom; Figure 7—video 1; see below ) , making cell tracking more challenging than that for worm #1–3 , which predominantly moved in the x-y plane . After segmentation , we manually confirmed 98 cells in the first volume and tracked them automatically ( Figure 7C and D; Figure 7—video 2 ) . We found that among the 30 larger cells ( size: 157–836 voxels ) with higher intensities , 26 of them ( 87% ) were correctly tracked in all 1000 volumes . Even though the smaller cells with lower intensities were more difficult to track , when including them , we still correctly tracked 66 out of 98 cells ( 67% ) ( Figure 7E and F ) . The tracked movements of two example cells showed regular oscillations in 3D space ( x , y , and z axes; Figure 7G ) , consistent with the regular beating movement of the heart . It should be noted that we did not optimize the pipeline procedures for the zebrafish data , except for a few parameter changes ( Table 2 ) . Our results indicate that 3DeeCellTracker is also capable of analyzing images with rapid and dynamic movements in 3D space , obtained from a substantially different optical system . We then used the tracking results to answer a biological question: What is the relationship between the intracellular calcium signals and the beating cycles of the heart ? We extracted the calcium dynamics of the correctly tracked 66 heart cells , which co-express GECI as in the worm datasets , and analyzed the phases of activities in these cells relative to the beating cycles of the ventricular and atrial regions , respectively . As shown in Figure 7H and I , the calcium signals were largely synchronized with the beating cycles . Although a portion of the heart cells ( 32/98 total ) was not correctly tracked and therefore not analyzed , this result is still remarkable because the tracked cells show the relationships between calcium dynamics and the natural heartbeats in vivo . Observation of this relationship was only made available by the developments of a state-of-the-art microscope system that can monitor 100 volumes per second and of our software pipeline that can correctly track large portions of the corresponding cell movements in 3D space . We also tested our method on a dataset more related to biomedical application , namely a dataset of ~900 cells in a 3D multicellular tumor spheroid ( MCTS ) imaged with a two-photon microscope . 3D MCTS is increasingly being used for drug screening because of its similarity to tumor cells in vivo ( Riedl et al . , 2017 ) . Therefore , the measurement of individual cell activities in 3D MCTS has become necessary , although tracking the movements of large numbers of cells in 3D + T images of 3D MCTS has presented a considerable challenge . We obtained 3D + T images of a 3D MCTS cells expressing FRET type ERK sensor , EKAREV-NSL ( Komatsu et al . , 2011 ) using a two-photon microscope ( see Materials and methods ) . This dataset shows normal distributions of intensities and movements but includes a much larger number of cells than either the worm brain or the zebrafish heart dataset ( Figure 8A and B; Figure 8—video 1 ) . Furthermore , cell division and death occurred during the imaging . Our method segmented and tracked 901 cells , of which we visually inspected the tracking results of 894 cells ( the remaining seven cells were found to have segmentation errors in volume #1 ) . We excluded the cells that experienced cell death or cell division after such events occurred , and found that 869 out of 894 cells ( 97% ) were correctly tracked ( Figure 8C–E , Figure 8—video 2 ) . Using the tracking results , we extracted the ERK activity from the FRET signal . In three representative cells with cross-layer movements , we confirmed that the extracted signals correctly reflected intensity changes in the cells ( Figure 8F ) . We also found that the spheroid as a whole moved downwards , although each cell moved in a different direction ( Figure 8G ) . In the assessments described in the preceding sections , we successfully analyzed multiple types of datasets that differ in terms of image resolution , signal-to-noise ratio , and types of cell movement etc . We then systematically evaluated the performance of our method ( single mode ) under a variety of conditions using a series of degenerated datasets obtained by modifying the worm #3 dataset . For a fair comparison , we used the same pre-trained 3D U-Net and the same manually corrected segmentation at t = 1 used on the original worm #3 dataset . A general difficulty in segmentation arises from images with low signal-to-noise ratios . Excessive noise can obscure the real cell signal , leading to incorrect segmentation and ultimately incorrect tracking . We tracked cells in three degenerated datasets with different levels of Poisson noise added to the original images: sd = 60 , sd = 100 , and sd = 140 ( Figure 9A ) . The noise level in the original images in the non-cell regions was sd = 4 . 05 and the median intensity was 411 , whereas the median intensity of cell regions was 567 with a 95% confidence interval of 430–934 , indicating a tiny overlap between non-cell and cell regions ( Figure 3—figure supplement 3B ) . In the sd = 60 condition , the intensities of non-cell and cell regions overlapped to a greater extent ( Figure 9B ) , and image quality appeared poorer than that of the original image ( Figure 9A ) . Nevertheless , our method achieved a low error rate ( 6/175 = 3%; Figure 9C–E ) . Even in the sd = 140 condition , in which the intensities overlapped extensively ( Figure 9B ) and the image quality was quite poor ( Figure 9A ) , our method achieved an acceptable error rate ( 16/175 = 9% , that is , 91% correct; Figure 9C–E ) . Note that the tracking error rate was much lower than the segmentation error rate ( 16/175 vs 57/175 , respectively; Figure 9—figure supplement 1 ) , indicating that our FFN-based tracking method can compensate for errors in cell segmentation . Considering that the overlap of intensity distributions between cell regions and background in the degenerated datasets were much more severe than the overlap in the real datasets ( Figure 9B and panel B in Figures 3 , 6 , 7 and 8 ) , these results suggest that our method can robustly track cells in 3D + T images with severe noise . Another difficulty comes from large displacements of cells between volumes during tracking . We enhanced this effect by removing intermediate volumes in the worm's 3D + T images , and then tested the three datasets with 1/2 , 1/3 , and 1/5 of the volume of the original dataset ( Figure 9F ) . As expected , when more volumes were removed , the movements of cells and the number of tracking errors increased ( Figure 9G–L ) . Nevertheless , the error rate in the 1/3 vol condition was acceptable ( 14/175 = 8% , i . e . 92% correct ) , while the error rate in the 1/5 vol condition was relatively high ( 25/175 = 14% , i . e . 86% correct ) . We then tested whether our method can track cell movements along the z-axis , in which the resolution is much lower than that in the x-y plane . In such conditions , 3D + T tracking is more challenging along the z-axis than in the x-y 2D plane . In the deforming worm and the tumor spheroid datasets , the resolution along the z-axis was approximately 1/10-1/5 of that in the x-y plane ( panel A in Figure 3 , Figure 3—figure supplements 1 , 2 and 3 , and Figure 8 ) . We confirmed that 5 , 16 , 25 , and 668 cells in worm #1a , #1b , and #3 , and tumor spheroid datasets , respectively , showed cross-layer movements along the z-axis . Still , our method correctly tracked all of those cells . For example , while two cells in the worm #3 dataset exhibited multiple cross-layer movements ( Figure 9—figure supplement 2 ) , they were correctly tracked until the end the sequence . Furthermore , we also degenerated the zebrafish heart data to test whether our method can track dynamic 3D cell movements with unequal resolutions . Even when the zebrafish image resolution in the z-axis was reduced by sum binning with a shrink factor = 2 , our method was able to correctly track a majority of the cells ( 80% of the 30 larger cells and 53% of all 98 cells; Table 3 ) . Together , these results indicate that our method can correctly track cells in various conditions including severe noise , large movements and unequal resolution . To evaluate the tracking performance of our method on datasets with large movements , we summarized the RM values , which are listed in Table 3 and Figure 10 ( see also Figure 4 ) . Although many cell movements in the worm #3 and the zebrafish datasets had RM ≥0 . 5 , most of the worm neurons and the larger cells in the zebrafish heart were correctly tracked by the single mode . In addition , our program with the ensemble mode achieved 99 . 8% tracking of the neurons of a ‘straightened’ freely moving worm while many cell movements in this dataset had RM ≥1 . These results indicate that our method is capable of analyzing images with challenging displacements . To investigate the relationships between the cell movements and tracking error rate , we drew a scatter plot of the averaged RM and corresponding error rate for each volume ( Figure 10—figure supplement 1 ) . The results indicated by the regression lines suggest positive correlations between the RM and error rate , although the correlation trends appear to differ by dataset , implying that movement is not the only factor affecting the error rate . We also drew a scatter plot and regression line for the worm #4 dataset tracked with the ensemble mode ( Figure 10—figure supplement 2 ) . The result suggests that the error rate is not correlated with movement , perhaps because in ensemble mode the cell positions are predicted from multiple different volumes , and therefore movement from a previous volume is not closely connected with the error rate . To further assess the capabilities of our method , we compared the tracking performance of our method with that of two other state-of-the-art methods for cell tracking . The first was DeepCell 2 . 0 , a newer version of DeepCell , which is a pioneer in segmenting and tracking cells in 2D + T images using deep learning ( Bannon et al . , 2018 ) . Unlike our method , DeepCell 2 . 0 has been primarily tested on images that include cell divisions , birth and death , but not on images of deforming organs . The second was the software developed by Toyoshima et al . , which does not use a deep learning technique , but has achieved a higher accuracy in both segmenting and tracking of worm whole brain datasets than any other previous methods ( Toyoshima et al . , 2016 ) . On our worm brain datasets , DeepCell 2 . 0 tracked ~10% of cells properly ( Figure 11A , Figure 11—figure supplement 1A , Figure 11—video 1 , and Table 4 ) ; this is probably because the tracking algorithm of DeepCell 2 . 0 is optimized for the movements associated with cell divisions , but not for quickly deforming organs . Toyoshima’s method tracked ~90% of the original worm #3 neurons but only ~10% of the degenerated ‘worm #3 , 1/5 sampling’ ( Figure 11B , Figure 11—figure supplement 1B , and Table 4 ) . When tested on the zebrafish dataset , Toyoshima's method was able to detect only 76 cells , and missed some weak-intensity cells , performing worse than our method ( detected 98 cells ) . Of the detected 76 cells , 21 were incorrectly tracked from the first volume , probably because their method automatically re-fits the cell shape using a Gaussian distribution after segmentation , which can lead to a failure when fitting weak-intensity cells . Their tracking accuracy was also lower than ours ( Table 4 ) . In addition , our method was found comparable to or more efficient in terms of runtime than DeepCell 2 . 0 and Toyoshima's methods ( Table 5 ) . These results suggest that our method is more capable of tracking cells in 3D + T images than previous methods . Tracking biological objects in 3D + T images has proven to be difficult , and individual laboratories still frequently need to develop their own software to extract important features from the images obtained using different optical systems and/or imaging conditions . Moreover , even when identical optical systems are used , the optimization of many parameters is often required for different datasets . To solve these problems , we have developed a deep learning-based pipeline , 3DeeCellTracker , and demonstrated that it can be flexibly applied to divergent datasets obtained under varying conditions and/or different qualities . We analyzed multiple image series of worms , zebrafish , and tumor spheroids , which differed in terms of nuclear marker , intensity level , noise level , numbers of cells per image , image resolutions and sizes , imaging rates , and cell speed . Notably , we showed that our method successfully tracked cells in these datasets under challenging conditions , such as large movements ( see Figure 10 , Table 3 and Materials and methods ) , cross-layer movements ( Figures 6G and 7G , and Figure 9—figure supplement 2 ) , and weak intensity conditions ( Figure 7B , Figure 7—figure supplement 1 , Figure 7—video 1 ) . In addition , our method outperformed two previous state-of-the-art cell tracking methods , at least in deforming organs under challenging imaging conditions ( Figure 11 , Figure 11—figure supplement 1 , and Table 4 ) , while the other methods are likely more suited for other conditions . Furthermore , our method is comparable to or more efficient in runtime than previous methods ( Table 5 and Materials and methods ) . Running in ensemble mode on a desktop PC , our method tracked the neurons of a ‘straightened’ freely moving worm with high accuracy , which required a computing cluster with up to 200 cores in the previous study ( Nguyen et al . , 2017 ) . We consider that the high accuracy and robustness of our method is based on its use of FFN , a deep learning technique , together with its post-processing methods for tracking ( Figure 2—figure supplement 3 ) . As mentioned above , although deep learning techniques have been predicted to exhibit superior performance in 3D cell tracking , they had not been used to date because of the difficulty in manually preparing large quantities of training data , especially for 3D + T images . To solve this problem , we generated a synthetic training dataset via artificial modification of a single volume of worm 3D cell positional data , which produced excellent performance by our method . These results demonstrate that the deep network technique can be used for cell tracking by using an synthetic point set dataset , although the procedures for generating the dataset are simple . Not only is our method flexible and efficient , it can be easily used by researchers . For example , our method worked well in all the diverse conditions tested with only minor modifications . Notably , under constant imaging conditions , our method can be directly reused without modifying any parameters except for the noise level ( Tables 1 and 2 ) , making it convenient for end-users . This differs from conventional image processing methods , in which slight differences in the obtained data , such as in light intensity , resolution , the size of the target object etc , generally require the re-setting of multiple parameters through trial-and-error . Even when imaging conditions are substantially changed , our method requires only a few modifications , primarily in the segmentation process: ( 1 ) modifying the structure of the 3D U-Net ( this step can be skipped because a 3D U-Net of the same structure can be adapted to a new dataset by re-training; see Materials and methods ) ; ( 2 ) re-training the 3D U-Net; and ( 3 ) modifying parameters according to the imaging conditions ( see Tables 1 and 2 and the ‘Guide for parameters . md’ file in https://github . com/WenChentao/3DeeCellTracker ) . For re-training the 3D U-Net , manual annotation usually takes 2–3 hr for 150–200 cells , and the network can be automatically trained in 1–2 hr on our desktop PC with a single GPU . The FFN for tracking generally does not require re-training . The number of parameters to be manually determined is much smaller in our method than in conventional methods ( Table 1 ) due to its use of deep learning . The parameters can be quickly modified ( within 1 hr ) following the guide we have provided in the GitHub repository . Nevertheless , our method can be improved in two ways: more reliable tracking and a simplified procedure . Tracking reliability can be affected by large movements , weak intensities , and/or photobleaching . As revealed in our results ( Figure 6 ) , large movements such as in a freely moving worm can be resolved using the ensemble mode , which borrows the idea from ensemble learning in machine learning , that is using the average of multiple predictions to reduce the prediction error ( Polikar , 2009 ) . A similar idea , matching cells using multiple reference volumes and a clustering method instead of averaging , was applied in the previous study ( Nguyen et al . , 2017 ) , suggesting that ensemble learning is a good approach to resolving challenging movements . On the other hand , the problem of weak intensity and photobleaching remains to be solved . One possible approach would be to normalize the intensities to obtain similar images over time , although doing so might not be easy . To further simplify the entire procedure , we contemplate developing new network structures that combine additional steps . The U-Net and 3D U-Net are networks for semantic segmentation , which classify each voxel as a specific category , that is as either a cell or a non-cell region . Beyond this , networks have been designed to achieve instance segmentation by further separating objects in the same category into individual objects , eliminating the need to use a watershed for separating connected cells . Although recent advances have been made in these architectures , the focus is still on segmenting common objects in 2D images ( Liang et al . , 2016; Romera-Paredes and Torr , 2016; He et al . , 2017 ) . We suggest that instance segmentation is a possible approach for simplifying and improving cell segmentation in future studies . Another possible area for improvement is the use of FFN for tracking . By further improving the FFN structure and using more training data , the network should be able to generate more accurate matches that can directly be used for tracking cells without point set registration . We developed 3DeeCellTracker mainly using semi-immobilized worm datasets . However , it also successfully processed 3D + T images of a zebrafish dataset obtained using the SCAPE 2 . 0 system ( Voleti et al . , 2019 ) . This system is quite different from the spinning disk confocal system used for worm datasets in resolution , z-depth , and applied optical sectioning principle ( Bouchard et al . , 2015 ) . While SCAPE is an original and outstanding method for enabling ultra high-speed 3D + T image acquisition , it had been difficult to obtain or develop software that can efficiently process the 3D + T images produced by the system . In this study we tracked 3D + T images obtained from the SCAPE system by simply modifying a few parameters , which allowed us to obtained an acceptable result ( 87% of large cells correctly tracked ) . Considering that the lower performance relative to other datasets might have arisen from the difficulty in segmenting the smaller , low-intensity cells ( Figure 7B , the upper and the middle panels ) , the result may be improved by further optimization of the segmentation . We also successfully tracked a large number of cells ( ~900 ) in a 3D MTCS monitored using a two-photon microscope , a result that further supports the wide applicability of our method . Our method cannot track cells that are dividing or fusing , or many cells that enter the field of view during the recording . This is because it operates under the assumption that each cell has a unique corresponding cell in another volume in order to match cells with large movements . To handle cells with division , fusion , or entry , it will be necessary to integrate our algorithms with additional algorithms . In summary , we have demonstrated that 3DeeCellTracker can perform cell segmentation and tracking on 3D + T images acquired under different conditions . Compared with the tracking of slowly deforming cells in 2D + T images , it is a more challenging task to track cell nuclei in a semi-constrained/freely moving worm brain , beating zebrafish heart , or 3D tumor spheroid , all of which undergo considerable movements in 3D space . We consider this to be the first report on a pipeline that efficiently and flexibly tracks moving cells in 3D + T images from multiple , substantially different datasets . Our method should enable the segmentation and tracking of cells in 3D + T images acquired by various optical systems , a task that has not yet been performed . Our image processing task was performed on a personal computer with an Intel Core i7-6800K CPU @ 3 . 40 GHz x 12 processor , 16 GB of RAM , and an Ubuntu 16 . 04 LTS 64-bit operating system . We trained and implemented the neural networks with an NVIDIA GeForce GTX 1080 GPU ( 8 GB ) . The neural networks were constructed and implemented using the Keras high-level neural network API ( https://keras . io ) running on top of the TensorFlow machine-learning framework ( Google , USA ) . All programs were implemented within a Python environment except for the image alignment , which was implemented in ImageJ ( NIH; RRID:SCR_003070 ) , and the manual labeling , manual correction and manual confirmation , which were implemented in ITK-SNAP ( RRID:SCR_002010; http://www . itksnap . org ) or IMARIS ( Bitplane , UK; RRID:SCR_007370 ) . Instead of ITK-SNAP and IMARIS , one can use napari ( https://napari . org ) in the Python environment . Step 1: Because 2D images were taken successively along the z axis , rather than simultaneously , small or large displacements could exist between different layers of a 3D volume . Ideally , this should be compensated for before the segmentation procedure . Using the StackReg plugin ( Thevenaz et al . , 1998 ) in ImageJ ( NIH ) , we compensated for the displacements by using rigid-body transformations to align each layer with the center layer in worm #1 and #2 datasets . However , we did not apply this alignment in worm #3 , #4 , zebrafish , and tumor spheroid datasets but still obtained acceptable results , indicating this step may be skipped . Step 2: Cells in the same image could have very different intensities , and detecting weak cells is generally difficult . To solve this problem , we applied local contrast normalization ( Goodfellow et al . , 2017 ) through a sliding window ( 27 × 27 × 3 voxels ) so that all cells had similar intensities . This normalization was applied to the nucleus marker images only for tracking and did not affect the calculation of the signal intensities for either red fluorescent protein or GECI . We used 3D U-Net structures similar to that shown in the original study ( Çiçek et al . , 2016 ) . The network received a 3D image as input and generated a 3D image of the same size with values between 0 and 1 for each voxel , indicating the probability that the voxel belonged to a cell region ( Figure 2—figure supplement 1 ) . We used different structures of 3D U-Net for different imaging conditions , in order to capture as much information as possible of each cell within the limitations of the GPU memory . Such modification of the U-Net structures is preferred but not necessary . The same structure can be reused on different datasets because of the flexibility of deep learning methods . The 3D U-Net structure shown in Figure 2—figure supplement 1A was used on our datasets worm #1 and worm #2 , which have identical resolution , but we also successfully used the same structure in the binned zebrafish dataset ( see below ) , which had very different resolution . For dataset worm #3 , which had a lower resolution , we reduced the number of maxpooling and upsampling operations so that each voxel in the lowest layers corresponds to similar sizes as in the datasets worm #1 and worm #2 . We also reduced the sizes of the input , output , and intermediate layers because of the lower resolution . As the smaller structure occupied less GPU memory , this allowed us to increase the number of convolutional filters on each layer so that the capacity of the network was increased ( see Figure 2—figure supplement 1B ) . For the zebrafish dataset , although it has even lower resolution in the x and y axes , because the sizes of zebrafish cardiac cells are larger than worm neurons , we used the same number of maxpooling and upsampling operations as in datasets worm #1 and worm #2 . We adjusted the sizes of layers in the x , y , and z dimensions to a unified value ( =64 ) because the resolution in the three dimensions are not very different in the zebrafish dataset ( Figure 2—figure supplement 1C ) . For simplicity , we reused the structures A and B in Figure 2—figure supplement 1 for worm #4 , tumor spheroid and binned zebrafish dataset ( see Table 2 ) . The U-Net can be trained using very few annotated images ( Ronneberger et al . , 2015 ) . In this study , we trained six 3D U-Nets: ( 1 ) for datasets worm #1 and #2 , ( 2 ) for dataset worm #3 , ( 3 ) for freely moving dataset worm #4 , ( 4 ) for the zebrafish dataset , ( 5 ) for the dataset tumor spheroid , and ( 6 ) for the binned zebrafish dataset . Each 3D U-Net used one 3D image for training . Note that , although datasets worm #1 and #2 are substantially different with respect to signal intensity and cell movements , the same trained 3D U-Net was used . The image was manually annotated into cell regions and non-cell regions using the ITK-SNAP software ( http://www . itksnap . org ) . We used the binary cross-entropy as the loss function to train the 3D U-Net . Because the raw image sizes were too large ( 512 × 1024 × 28 , 256 × 512 × 20 , 180 × 260 × 165 , etc . ) for computing in the GPU , we divided the raw images into small sub-images that fit the input sizes of the three 3D U-Nets structures ( 160 × 160 × 16 , 96 × 96 × 8 , or 64 × 64 × 64 ) , and combined the cell/non-cell classifications of sub-images to form a final classification of the whole image . To improve the 3D U-Net performance , we increased the training data by data augmentation: We applied random affine transformations to the annotated 3D images by ‘ImageDataGenerator’ class in Keras . The affine transformation was restricted in the x-y plane but not in the z-direction because the resolution in the z-direction is much lower than that in the x-y plane for worm datasets and the tumor spheroid dataset ( see panel A in Figure 3 , Figure 3—figure supplements 1 , 2 and 3 , and Figure 8 ) . Although the zebrafish dataset has similar resolutions in the x , y , and z directions , we applied the same affine transformation for simplicity . We trained the U-net for datasets worm #1 and #2 using a 3D image from another dataset independent of #1 and #2 but obtained under the same optical conditions , and its classifications on datasets worm #1 and #2 were still good ( panel C in Figure 3 , Figure 3—figure supplements 1 and 2 ) , indicating a superior generalization ability of the 3D U-net . Because only one dataset is available for the specific resolutions of dataset worm #3 , #4 , the zebrafish heart , and the tumor spheroid , we trained the 3D U-Net by using the first volume of 3D + T images of each dataset and then applied the 3D U-Net to all the following 3D images of the datasets . The 3D U-Net generated probability outputs between 0 and 1 , which indicated the probability that a voxel belonged to a cell-like region . By setting the threshold to 0 . 5 , we divided the 3D image into cell-like regions ( >0 . 5 ) and non-cell regions ( ≤0 . 5 ) . The cell-like regions in the binary images were further transformed into distance maps , where each value indicated the distance from the current voxel to the nearest non-cell region voxel . We applied a Gaussian blur to the distance map to smooth it , and searched for local peaks which were assumed to be cell centers . We then applied watershed segmentation ( Beucher and Meyer , 1993 ) , using these centers as seeds . Watershed segmentation was applied twice; the first application was 2D watershed segmentation for each x-y plane , and the second application was 3D watershed segmentation for the entire 3D space . Two segmentations were required because the resolutions in the x-y plane and the z-dimension differed . An initial matching , that is , a set of correspondences between cells in two temporally adjacent volumes , is the first step for cell tracking in our pipeline and critical for the final tracking accuracy . The correspondences can be estimated based on the relative cell positions , assuming that these positions do not substantially change even during organ deformation . By comparing the similarity between the relative positions of two cells in different volumes , we can determine whether they are the same cells . One conventional method to represent relative positions is fast point feature histograms ( FPFH ) ( Rusu et al . , 2009 ) . The PR-GLS study ( Ma et al . , 2016 ) successfully used the FPFH method to match artificial point set datasets . However , we found that FPFH yielded a poor initial match for the datasets considered in this study ( Figure 2—figure supplement 3A ) , perhaps because of the sparse distribution of the cells . We thus designed a three-layer feedforward network ( FFN ) to improve the initial match ( Figure 2B ) . The three-layer structure generated good match results comparable with those of more complex structures with four or six layers . By comparing the representations between two points , the network generated a similarity score between two cells . The initial matching based on the similarity score by the FFN was more accurate than that achieved by the FPFH method ( Figure 2—figure supplement 3B ) . Our FFN + PR-GLS approach generated more accurate predictions of cell positions than FPFH + PR-GLS and the simple method of affine alignment ( using an implementation of the Coherent Point Drift Algorithm: https://github . com/siavashk/pycpd ) ( Figure 2—figure supplement 3C and D ) . In our FFN , the input of the network contained the position information of two points . Each point was represented by the normalized positions of the 20 nearest neighboring points ( Figure 2B ) . The 20 nearest neighbor positions ( d1→ , d2→ , … , d20→ ) were given by 3D vectors because they were extracted from the 3D image . To normalize the points , each of the 20 positions was divided by the mean distance d ( d=120∑k=120dk→ ) . The normalized positions were then sorted by their absolute values in ascending order . Finally , the mean distance d was included as the last value , so each point was represented by a 61D vector . We utilized the first fully connected layer to calculate the learned representation of the relative positions of each point as a 512D vector ( Figure 2B , the first hidden layer after the input ) . We then applied a second fully connected layer on these two 512D vectors to compare the representations of the two points . The resulting 512D vectors ( the second hidden layer after the input ) were processed by a third fully connected layer to obtain a single similarity score between 0 and 1 , which indicated the probability of two points originating from the same cell . We matched the two points with the highest scores in two different volumes , ignored these two points , and matched the next set of two points with the highest scores . By repeating this process , we obtained an initial match ( Figure 2C , panel 4–1 ) . In this study , we only trained one FFN based on one original image of dataset worm #3 and used the network in all the datasets including worms , zebrafish and tumor spheroid . To train the network , we first performed segmentation on a single volume of dataset worm #3 and obtained a point set for the centers of all cells . Because we required a large number of matched point sets for training , and because manually matching point sets is time-consuming and impractical , we created an synthetic training dataset by applying random affine transformations to the point set described above and adding small random movements to each point according to following equations: ( 1 ) x′→=Ax→+ε1→+ε2→ Also see Figure 2—figure supplement 2A for an illustration . Here x→ is the mean-removed 3D position {x1 , x2 , x3} of each point in the original point set , while x'→ is the transformed 3D position . A is a matrix to apply the random affine transformation . More specifically , A=I+U , where I is a 3 x 3 identity matrix , and U is a 3 x 3 random matrix with each element Uij from a uniform distribution . We used Uij~ uniform-0 . 05 , 0 . 05 in this study . ε1→ is the 3D vector for adding random movements to each point in a point set , while ε2→ is for adding even larger random movements in a subset of points ( 20 out of 175 cells ) to simulate serious errors from segmentation procedures . We used ε1 , i~uniform-2 , 2 and ε2 , i~uniform-5 , 5 in this study . By randomly generating Uij , ε1 , i and ε2 , i , we could generate an arbitrarily large number of new point sets with new positions from the original point set . After that , we chose a specific point A and another point B from each generated point set and the original point set , respectively , and we calculated their relative positions as inputs for the FFN . In half of the cases , points A and B are corresponding , that is , they come from the same cell , while in the other half , point A is from a cell adjacent to the cell of point B , thus they are not corresponding . In this study , we used 576 , 000 newly generated pairs of points A and B for training the FFN ( Figure 1C ) . We used binary cross-entropy as the loss function to train the FFN . During the training , the performance of matching by FFN was gradually improved using an independent test dataset of two point sets ( Figure 2—figure supplement 2B ) . The initial match calculated using FFN was corrected using the expectation–maximization ( EM ) algorithm in the PR-GLS method , as described in the original paper ( Ma et al . , 2016 ) . In the original study , the initial match ( by FPFH ) was recalculated during the EM iterations; however , in most datasets , we calculated the initial match only once ( by the FFN ) before the EM steps were performed , which did not cause problems . Only for the dataset worm #4 with very large movements , we recalculated initial matching by FFN after every 10 iterations , in order to improve accuracy . After the PR-GLS corrections , we obtained coherent transformations from the points of each volume to the subsequent volume ( Figure 2C , panel 4–2 ) . The FFN + PR-GLS can predict new cell positions at time t from the cell positions at t-1; this is the default mode of our pipeline , which is referred to as the single mode ( Figure 6C ) . At a sufficiently high sampling rate , the single mode is reasonable because the movement from t-1 to t is much smaller than that from t-i ( i > 1 ) to t , making the prediction from t-1 more reliable . When cell movements are very large ( such as in a freely moving worm ) , and the sampling rate is not sufficient , the prediction from t-1 to t becomes less reliable , and the average of multiple predictions from t-i to t may be more accurate . Therefore , we developed an approach using the average of multiple predictions , referred to as the ensemble mode ( Figure 6C ) . We tested this mode only on the worm #4 dataset , which had quite large movements ( Figures 6B and 10 , and Table 3 ) , because the runtime of the ensemble mode is much longer than that of the single mode , that is , the runtime of the FFN + PR-GLS component is proportional to the number of predictions used . Specifically , the runtime including segmentation and tracking for worm #4 was approximately 30 volume/h in the single mode and 6 volumes/h in the ensemble mode . In the ensemble mode , we calculated an average of up to 20 predictions from previous time points . In cases for which t ≤ 20 , the averages was calculated from time points [t-1 , t-2 , … , 1]; in cases for which t > 20 , it was calculated over [t-d , t-2d , … , t-20d] , where d is the quotient ( t-1 ) //20 . By applying the PR-GLS method to the initial match , we obtained a more reliable transformation function in which all obvious incorrect matches were corrected . However , small differences still existed in a few cells , which could accumulate over time to become large differences without correction . Thus , we included an additional automatic correction step , in which the center position of each cell was moved slightly toward the centers of each 3D U-Net detected region ( for details , see Figure 2—figure supplement 4 ) . After correction , all cells were moved to the estimated positions with their shapes unchanged from volume #1 . For most of the datasets , we applied only one correction; for the worm #4 and tumor spheroid datasets , we applied corrections up to 20 times , until achieving convergence . If multiple cells overlapped in the new positions , we applied the watershed method again to assign their boundaries . In carrying out this step , we calculated the images of the tracked cell regions based on an interpolation of the raw image from volume #1 , when the resolution along the z axis was much lower than that in the x-y plane ( i . e . all datasets except for the worm #4 and the zebrafish ) . We did this in order to obtain more accurate estimates for each cell region . We manually corrected the segmentation only in volume #1 . We superimposed the automatically segmented regions of the volume #1 3D image onto the raw 3D image in the ITK-SNAP software , and we discarded false positive regions , such as autofluorescence regions and neuronal processes . If any cells were not detected ( false negative error ) , we reduced the noise level parameter in the pre-processing ( Table 1 ) which can eliminate such errors , or we manually added these cells in ITK-SNAP . Oversegmentation and undersegmentation regions were corrected carefully by considering the sizes and shapes of the majority of cells . The overall error rates depend on the image quality , but we usually found that around 10% of cells required manual correction which usually took 2–3 hr ( for 100–200 cells ) . We counted the tracking errors of all cells by visually inspecting the tracking results in each volume ( Figure 1—figure supplement 2 ) . To confirm the tracking of worm’s neurons , we combined two 3D + T images—the raw images and the tracked labels—in a top–bottom arrangement displayed as a hyperstack by the ImageJ software to compare the cell locations in each volume . As the cells in worms’ datasets primarily moved in the x-y plane ( Figures 3F and 9L , and Figure 3—figure supplement 2F ) , we observed the correspondence between the raw images and the tracked labels in each x-y plane to identify tracking errors in the results . To confirm our tracking of the tumor spheroid dataset , we applied the same method although there were many small cross-layer movements in the images ( Figure 8F and G ) . It was more difficult to confirm the tracking results in the hyperstack images for the cells in the freely moving worm and the zebrafish heart than in the semi-immobilized worm and tumor spheroid , due to the frequent occurrence of large movements of cells across layers ( Figures 6G and 7G ) . Therefore , we superimposed images of the tracked labels onto the raw images and imported them into IMARIS ( Bitplane , UK ) , and then visually checked the tracking results of each cell individually in 3D mode . For the zebrafish dataset with repeated oscillations , we tracked and checked all 98 cells from 1000 volumes . Because the freely moving worm engaged in irregular movements in three directions , visual checking was more challenging , and thus we only tracked and checked the initial 500 volumes of 3D images out of 1519 original volumes ( Figure 6F ) . Large movements of cells is one issue that makes tracking challenging . To evaluate how challenging each cell movement is , we defined the 'relative movement' ( RM ) of cell A at time t as: RM = ( movement of cell A from t-1 to t ) / ( distance between cell A and its closest neighboring cell at t ) . Figure 4A middle and right panels illustrate two cells moving in one dimensional space with RM ≥0 . 5 . In this condition , a very simple tracking method , ‘search for the closest cell at the next time point’ will mistakenly match the cell A at t = 2 to the cell B at t = 1 . Therefore , we argue that movements with RM ≥0 . 5 are more challenging than movements with RM <0 . 5 . It should be noted that large movement is not the only challenge for tracking . For example , the zebrafish datasets have much higher error rates if we evaluate the tracking in all cells ( Table 3 ) , which is likely caused by the weak intensities in these small cells and the photobleaching ( Figure 7B and F; Figure 7—figure supplement 1 ) . After having tracked the worm’s neurons from the first volume to the last volume , we extracted activities from the regions corresponding to each neuron . By measuring the mean intensities of each neuron in both channels corresponding to the GECI and the positional markers , the activity was computed as GCaMP5G/tdTomato in dataset worm #1 . We similarly extracted calcium activities ( GCaMP ) of heart cells in the zebrafish dataset and the FRET activity in the tumor spheroid dataset . Because DeepCell 2 . 0 currently only supports tracking cells in 2D datasets , we tested it using two layers of images ( z = 9 and z = 16 ) from worm dataset #3 , which include relatively large numbers of cells . We excluded cells that disappeared or appeared due to cross-layer movements for a fair comparison . We supplied DeepCell 2 . 0 with the precise segmentations from our method in order to focus on comparing the performance of the tracking algorithms . We tested Toyoshima’s software using two worm datasets and one zebrafish dataset . For the zebrafish dataset , we only tested the initial 100 volumes because their method required a longer time for the imaging processing than ours did ( see Table 5 ) . For these two methods , we communicated with the corresponding authors and did our best to optimize the parameters based on their suggestions . Nevertheless , it is possible that those analyses could be further optimized for our datasets . We tested the runtimes of our method and the previous methods for different datasets ( see Table 5 ) . Because DeepCell 2 . 0 currently can only track 2D + T images , the runtime was estimated by using the average runtime for one layer and multiplying that by 21 layers , in order to compare it with our method . As a result , DeepCell 2 . 0 required a runtime comparable with our method . On the other hand , Toyoshima’s software took a much longer time than our method to process the worm and zebrafish datasets . In our method , the initial matching using our custom feedforward network is performed in a pairwise fashion , so the time complexity is O ( n2 ) , where n is the number of detected cells . In our tested datasets with 100 ~ 200 cells , this did not take a long time , for example , ~3 . 3 s were required for matching 98 cells between two volumes ( zebrafish ) , or ~8 . 6 s for 164 cells ( worm #1a ) using our desktop PC . In cases where the n is too large , the runtime may become much longer , for example , ~4 min for 901 cells ( tumor spheroid ) . If both n and volume number are too large , it may be required to optimize the method to reduce the runtime , for example , by restricting the calculation of matchings in a set ( 100 ~ 200 or more ) of representative cells ( e . g . large/bright cells ) , while movements of other non-representative cells can be estimated from the movements of these representative cells utilizing the coherency of these movements in a deforming organ . The techniques used to culture and handle C . elegans were essentially the same as those described previously ( Brenner , 1974 ) . Both TQ1101 lite-1 ( xu7 ) and AML14 were obtained from the Caenorhabditis Genetics Center ( University of Minnesota , USA ) . Young adult hermaphrodites were used in the imaging experiments . For pan-neuronal expression , NLS::tdTomato::NLS ( Frøkjær-Jensen et al . , 2016 ) and NLS::GCaMP5G::NLS ( in which GCaMP5G ( Akerboom et al . , 2012 ) was codon-optimized for C . elegans and attached to NLS at N- and C-termini ) were fused with the rab-3 promoter ( Stefanakis et al . , 2015 ) using a GATEWAY system ( Thermo Fisher Scientific ) . Germline transformation into lite-1 ( xu7 ) ( Liu et al . , 2010 ) was performed using microinjection ( Mello et al . , 1992 ) with a solution containing pYFU251 rab-3p::NLS::GCaMP5G::NLS ( 25 ng/µl ) , pYFU258 rab-3p::NLS::tdTomato::NLS ( 20 ng/µl ) and OP50 genome ( 55 ng/µl ) to obtain the strain KDK54165 . The strain lite-1 ( xu7 ) was used to reduce the blue light-induced activation of the worm’s sensory system ( Liu et al . , 2010 ) . Independent transgenic lines obtained from the injection produced similar results . In this study , we used four worm datasets . The 3D images in datasets worm #1 and #2 were obtained using our custom-made microscope system , OSB3D ( see below ) . The worm strains for datasets worm #1 and #2 are KDK54165 ( RRID:WB-STRAIN:KDK54165 ) and AML14 wtfEx4[rab-3p::NLS::GCaMP6s: rab-3p::NLS::tagRFP] ( RRID:WB-STRAIN:AML14 ) ( Nguyen et al . , 2016 ) , respectively . The 3D + T images of the dataset worm #3 were published previously with the worm strain JN2101 Is[H20p::NLS4::mCherry]; Ex[tax-4p::NLS-YC2 . 60 , lin-44p::GFP] ( Toyoshima et al . , 2016 ) . The 3D + T images of the freely moving worm dataset ( worm #4 ) were published previously with the worm strain AML14 ( Nguyen et al . , 2016; Nguyen et al . , 2017 ) . In three ( #79 , #135 , and #406 ) out of the 500 volumes of the worm #4 dataset , most of the cells disappeared from the volume or a large amount of noise occurred , which rendered the dataset impossible to analyze . These volumes were therefore manually skipped for tracking , that is the cells were assumed not moved from the previous volumes . We upgraded our robotic microscope system ( Tanimoto et al . , 2017 ) to a 3D version . We used a custom-made microscope system that integrated the Nikon Eclipse Ti-U inverted microscope system with an LV Focusing Module and a FN1 Epi-fl attachment ( Flovel , Japan ) . The excitation light was a 488 nm laser from OBIS 488–60 LS ( Coherent ) that was introduced into a confocal unit with a filter wheel controller ( CSU-X1 and CSU-X1CU , respectively; Yokogawa , Japan ) to increase the rotation speed to 5 , 000 rpm . The CSU-X1 was equipped with a dichroic mirror ( Di01-T-405/488/561 , Semrock ) to reflect the 488 nm light to an objective lens ( CFI S Fluor 40X Oil , Nikon , Japan ) , which transmitted the GCaMP fluorescence used for calcium imaging and the red fluorescence used for cell positional markers . The laser power was set to 60 mW ( 100% ) . The fluorescence was introduced through the CSU-X1 into an image splitting optic ( W-VIEW GEMINI , Hamamatsu , Japan ) with a dichroic mirror ( FF560-FDi01 , Opto-line , Japan ) and two bandpass filters ( BA495-540 and BA570-625HQ , Olympus , Japan ) . The two fluorescent images were captured side-by-side on an sCMOS camera ( ORCA Flash 4 . 0v3 , Hamamatsu , Japan ) , which was controlled by a Precision T5810 ( Dell ) computer with 128 GB RAM using the HCImage Live software ( Hamamatsu ) for Windows 10 Pro . A series of images for one experiment ( approximately 1–4 min ) required approximately 4–15 GB of space , and were stored in memory during the experiment , then transferred to a 1-TB USB 3 . 0 external solid-state drive ( TS1TESD400K , Transcend , Taiwan ) for further processing . For 3D imaging , the z-position of the objective lens was regulated by a piezo objective positioner ( P-721 ) with a piezo controller ( E665 ) and the PIMikroMove software ( PI , Germany ) . The timings of the piezo movement and the image capture were regulated by synchronized external edge triggers from an Arduino Uno ( Arduino , Italy ) using 35 ms intervals for each step , in which the image capture was 29 . 9 ms . For each step , the piezo moved 1 . 5 µm , and one cycle consisted of 29 steps . We discarded the top-most step because it frequently deviated from the correct position , and we used the remaining 28 steps . Note that one 3D image was 42 µm in length along the z-axis , which was determined based on the typical diameters of neuronal cell bodies ( 2–3 µm ) and of a young adult worm’s body ( 30–40 µm ) . Each cycle required 1015 ms; thus , one 3D image was obtained per second . This condition was reasonable for monitoring neuronal activities because the worm’s neurons do not generate action potentials ( Goodman et al . , 1998 ) and because many neuronal responses change on the order of seconds ( Nichols et al . , 2017 ) . We also tested a condition using 10 ms for each step and 4 . 9 ms for an exposure with the same step size and step number per cycle ( i . e . 2 . 3 volumes of 3D images per second ) , which yielded a comparable result . For cyclic regulation of the piezo position , we used a sawtooth wave instead of a triangle wave to assign positional information because the sawtooth wave produced more accurate z positions with less variance between cycles . Sample preparation and imaging have been described previously ( Voleti et al . , 2019 ) . In brief , GCaMP and dsRed were expressed in the cytosol and the nuclei of myocardial cells , respectively . The 3D + T images obtained with the SCAPE 2 . 0 system were skew corrected to account for the oblique imaging geometry and analyzed with 3DeeCellTracker . The 3D coordinates obtained from this study were used to extract calcium dynamics in the cells in the previous study ( Voleti et al . , 2019 ) . The tumor spheroid was cultured according to the previous procedures ( Vinci et al . , 2012; Yamaguchi et al . , 2021 ) . In brief , a suspension of HeLa cells ( RRID:CVCL_0030 , RCB0007 , Riken Cell Bank; certified as mycoplasma-free and authenticated by using STR profiling ) expressing the FRET-type ERK sensor EKAREV-NLS ( Komatsu et al . , 2011 ) was added to a PrimeSurface 96 well plate ( MS-9096U , Sumitomo Bakelite ) for 3D cell culture at a density of 1200 cells/well . The grown spheroids were transferred to 35 mm-dishes coated with poly-L-lysine ( Sigma-Aldrich ) and further grown with DMEM/F-12 , no phenol red ( ThermoFisher ) containing 10% FBS and 1% penicillin and streptomycin . The 3D + T images of the spheroid were recorded using a two-photon microscope equipped with a water-immersion objective lens ( N25X-APO-MP 25x CFI APO LWD objective , Nikon ) and a high-sensitivity gallium arsenide phosphide ( GaAsP ) detector ( A1R-MP+ , Nikon ) . An 820 nm optical pulse generated by a Ti:Sapphire laser was used for the excitation . The fluorescence was split using two dichroic mirrors ( FF495-Di03 and FF593-Di02 , Opto-line ) , and the split lights below 495 nm and between 495–593 nm were detected by independent channels ( CH1 and CH2 , respectively ) . The fluorescence signals were integrated four times to increase the signal-to-noise ratio . Each step size was 4 µm along the z-axis , and each volume comprised of 54 steps , requiring 5 min for recording . Cells that experienced cell death or cell division were excluded manually after the events when evaluating the tracking accuracy . The code for tracking cells and for training neural networks , the demo data , the pre-trained weights of the neural networks , and instructions for installation and use of the code are available in http://ssbd . qbic . riken . jp/set/20190602/ ( Demos190610 . zip ) . An updated version of the code can be found in https://github . com/WenChentao/3DeeCellTracker . The guides for using the codes and for setting parameters have also been included in the same GitHub repository . In the original version of the paper , we tracked the datasets by our original version of programs based on the package "3DeeCellTracker 0 . 2" ( see the version information at https://pypi . org/project/3DeeCellTracker/#history ) . The runtime for the tracking process was not optimized at that time , which sometimes could lead to a long runtime for tracking , especially in the ensemble mode and/or when the cell number is large . In addition , our previous tracking program did not hide the details irrelevant to the end-users and did not provide useful feedback on the intermediate segmentation/tracking results to users . To solve these two issues , we first improved the performance of the codes related to “FFN” and “PR-GLS” using the vectorization techniques available in the “NumPy” package in Python . We also improved the performance of the codes related to “accurate correction” by extracting and performing only once the previously repeated calculations in the loops . Note that we did not change our programs related to segmentation ( 3D U-Net + watershed ) , which may still require considerable time depending on factors such as the image size and structure of the 3D U-Net . As a result , our new program accelerated the speed by 1 . 7–6 . 8 times in the tested datasets ( Table 6 ) . Such acceleration was especially pronounced in ensemble mode ( worm #4 ) and when the cell number is large ( 3D tumor spheroid ) . Second , we simplified the program by moving the details into the package . As a result , users can track cells by simply running a few commands in Jupyter notebooks . We also added new functions to show the intermediate results of segmentation and tracking so that users can now use these results to guide their setup of the segmentation/tracking parameters . To use these new features , users can follow the updated instructions in the README file in our GitHub repository to install the latest version of 3DeeCellTracker ( currently 0 . 4 . 0 ) and use our Jupyter notebooks .	Microscopes have been used to decrypt the tiny details of life since the 17th century . Now , the advent of 3D microscopy allows scientists to build up detailed pictures of living cells and tissues . In that effort , automation is becoming increasingly important so that scientists can analyze the resulting images and understand how bodies grow , heal and respond to changes such as drug therapies . In particular , algorithms can help to spot cells in the picture ( called cell segmentation ) , and then to follow these cells over time across multiple images ( known as cell tracking ) . However , performing these analyses on 3D images over a given period has been quite challenging . In addition , the algorithms that have already been created are often not user-friendly , and they can only be applied to a specific dataset gathered through a particular scientific method . As a response , Wen et al . developed a new program called 3DeeCellTracker , which runs on a desktop computer and uses a type of artificial intelligence known as deep learning to produce consistent results . Crucially , 3DeeCellTracker can be used to analyze various types of images taken using different types of cutting-edge microscope systems . And indeed , the algorithm was then harnessed to track the activity of nerve cells in moving microscopic worms , of beating heart cells in a young small fish , and of cancer cells grown in the lab . This versatile tool can now be used across biology , medical research and drug development to help monitor cell activities .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods", "Note", "in", "proof" ]	[ "computational", "and", "systems", "biology", "tools", "and", "resources", "neuroscience" ]	2021	3DeeCellTracker, a deep learning-based pipeline for segmenting and tracking cells in 3D time lapse images
Deep understanding of tooth regeneration is hampered by the lack of lifelong replacing oral dentition in most conventional models . Here , we show that the bearded dragon , one of the rare vertebrate species with both polyphyodont and monophyodont teeth , constitutes a key model for filling this gap , allowing direct comparison of extreme dentition types . Our developmental and high-throughput transcriptomic data of microdissected dental cells unveils the critical importance of successional dental lamina patterning , in addition to maintenance , for vertebrate tooth renewal . This patterning process happens at various levels , including directional growth but also gene expression levels , dynamics , and regionalization , and involves a large number of yet uncharacterized dental genes . Furthermore , the alternative renewal mechanism of bearded dragon dentition , with dual location of slow-cycling cells , demonstrates the importance of cell migration and functional specialization of putative epithelial stem/progenitor niches in tissue regeneration , while expanding the diversity of dental replacement strategies in vertebrates . Exposed to the oral cavity of most jawed vertebrates , teeth are continually worn down through food acquisition , a process that must be compensated to maintain the oral apparatus functional ( Berkovitz and Shellis , 2016 ) . The most common , and likely ancestral , solution to this problem is polyphyodonty , or continuous replacement of the dentition , through shedding of functional teeth ( Jernvall and Thesleff , 2012; Tucker and Fraser , 2014 ) . Whereas polyphyodonty prevails in the majority of vertebrate taxa , mammals show a substantial reduction in the number of tooth generations , a phenomenon often associated with increased dental complexity in this group ( Kielan-Jaworowska et al . , 2004 ) . Particularly , the vast majority of mammals display a diphyodont dentition , producing only two sets of teeth—deciduous or ‘milk’ teeth and permanent replacement teeth . Rodents such as Mus musculus , one of the most commonly used vertebrate animal models , are even more extreme and develop only one generation of teeth , a condition known as monophyodonty ( Jernvall and Thesleff , 2012 ) . As a result , although mouse developmental studies of the past decades have revealed many of the signaling pathways controlling primary tooth development , relatively little is known about the molecular mechanisms underlying lifelong tooth regeneration in vertebrates . Recent studies in polyphyodont and diphyodont species from all major vertebrate groups have shown that , with the exception of some teleost fish such as salmonids ( Fraser et al . , 2006; Jernvall and Thesleff , 2012; Vandenplas et al . , 2016 ) , tooth replacement is generally dependent on a specialized epithelial structure , the dental lamina ( DL ) . This structure first emerges as a sheet of epithelial cells that invaginates from the oral epithelium ( OE ) , and plays a crucial role in the developmental events contributing to primary tooth formation . The DL is maintained throughout the lifetime of polyphyodont species , and continues to grow beyond the previously generated tooth to create a free margin , the successional dental lamina ( SDL ) , which constitutes the initiation site of replacement teeth . In monophyodont or diphyodont species such as mouse ( Dosedělová et al . , 2015 ) , minipig ( Buchtová et al . , 2012 ) , ferret ( Jussila et al . , 2014 ) , and chameleon ( Buchtová et al . , 2013 ) , both the DL and associated SDL degenerate relatively rapidly following formation of either one or two sets of teeth , thus highlighting the importance of maintaining these epithelial structures for tooth replacement . Coherent with this , the DL is considered to be the main source of putative odontogenic stem cells required for long-term tooth renewal ( Huysseune and Thesleff , 2004; Smith et al . , 2009; Jernvall and Thesleff , 2012; Juuri et al . , 2013; Thiery et al . , 2017 ) , and label-retaining cells ( LRCs ) have been initially mapped to this region in the polyphyodont leopard gecko ( Handrigan et al . , 2010 ) . More recently , putative stem/progenitor populations have been identified in the distal DL bulge of the American alligator ( Wu et al . , 2013 ) , in the SDL and OE of sharks ( Martin et al . , 2016 ) , as well as in the superficial DL of pufferfish ( Thiery et al . , 2017 ) , thus indicating variations in dental regenerative strategies among polyphyodont species . Molecularly , compelling evidence from a number of reptilian species , teleost fish , shark , and ferret has revealed a conserved set of transcription factors and signaling molecules exhibiting similar expression patterns in the dental apparatus . In particular , the canonical Wnt pathway has emerged as a strong candidate signaling pathway associated with tooth replacement , because of its local activation ( as revealed by the expression of Wnt/β-catenin target genes ) in most studied polyphyodont and diphyodont species ( Handrigan and Richman , 2010a; Fraser et al . , 2013; Gaete and Tucker , 2013; Wu et al . , 2013; Jussila et al . , 2014; Rasch et al . , 2016 ) , but also because of its capacity to produce a large number of supernumerary teeth when activated in snakes and transgenic mice ( Järvinen et al . , 2006; Wang et al . , 2009; Gaete and Tucker , 2013 ) . Furthermore , stabilizing Wnt/β-catenin signaling was recently shown to induce formation of replacement tooth germs in the normally degenerating mouse SDL , thus providing further support for the important role of this pathway in controlling tooth induction and regeneration ( Popa et al . , 2019 ) . However , only low-scale , single- or cross-species candidate gene approaches based on mouse expression patterns have been used so far to characterize polyphyodont dentitions , and the complete molecular signature of vertebrate tooth regeneration is still unknown . In this study , we examined one agamid lizard model , the bearded dragon ( Pogona vitticeps ) , which offers an excellent model system to study the events that define tooth development and regeneration because of its uncommon dentition with distinct modes of implantation and regenerative capacities along the same jaw—acrodont , monophyodont teeth posteriorly and pleurodont , polyphyodont teeth anteriorly ( Cooper et al . , 1970; Di-Poï and Milinkovitch , 2016 ) . This lizard is one of the rare extant vertebrate species with such marked heterodonty , which allows direct comparisons of extreme types of vertebrate dentition in a single model . Unlike the dentition of other polyphyodont reptiles investigated so far , our data further indicate that bearded dragons possess a relatively slow replacement process ( based on a ‘one-for-one’ strategy ) as well as a permanent , proliferating SDL that only initiates replacement at anterior polyphyodont locations , making this model extremely attractive for elucidating the key importance of DL and SDL epithelial structures in vertebrate tooth initiation and regeneration . Indeed , we show here that the early degradation of the SDL is not the only mechanism restricting vertebrate tooth replacement ( Buchtová et al . , 2012; Dosedělová et al . , 2015 ) . Instead , our developmental data and large-scale transcriptomic analysis of microdissected dental tissues unveil the critical importance of SDL patterning at various levels , including directional growth but also gene expression levels , dynamics , and regionalization , for tooth regeneration . Particularly , we identify a large number of yet uncharacterized developmental genes associated with dental patterning in regenerating and non-regenerating teeth , so our lizard model is also uniquely suited for revealing new signaling pathways critical for tooth replacement in vertebrates . Additionally , our bearded dragon findings reveal a novel mechanism of continuous tooth replacement in vertebrates , which shares with the leopard gecko a source of putative odontogenic progenitor/stem cells in the DL but closely parallels the shark by the presence of a separate second source of cells originating from the OE . This novel dental replacement strategy definitely demonstrates that the bearded dragon provides a rare opportunity to expand our understanding of tooth developmental dynamics and diversity , while providing a unique model system to study the function and specialization of epithelial structures and dental progenitor/stem cell niches present in distinct groups of vertebrates . Altogether , and given the well-established history in captivity and key advantages of the bearded dragon lizard for research ( Ollonen et al . , 2018 ) this emerging model organism offers a powerful system to elucidate the developmental and genetic basis of both evolutionary novelty and tooth regeneration , thus allowing to answer both developmental and evolutionary questions critical for the design of more appropriate therapies targeted at congenital dental disorders and tooth loss in humans . Agamid lizards possess an atypical heterodont dentition with both acrodont , monophyodont and pleurodont , polyphyodont tooth types , which vary widely in terms of number and shape among species and over ontogeny ( Cooper et al . , 1970 ) and personal observations ) . We first examined the gross anatomy of the dentition and associated DL in the bearded dragon , using micro-computed tomography ( CT ) imaging . Hatchling animals possess seven posterior triangular acrodont teeth attached to the jaw bones on each jaw quadrant , whereas the most anterior tooth positions on the premaxilla ( three teeth , upper jaw ) and dentary ( two teeth , lower jaw ) are occupied by pleurodont teeth ( Figure 1A , B ) . At this early stage , the central-most tooth projecting forwards in the midline of the premaxilla consists of the large egg-tooth , which is later replaced by a regular pleurodont successor ( Figure 1A ) . To visualize oral soft tissue , including the DL structure , we stained the jaws of juvenile bearded dragons with the contrast agent phosphotungstic acid ( PTA ) before CT-scanning . As expected , the DL forms a continuous sheet of epithelium spanning the whole jaws , without clear separation between the pleurodont and acrodont dentitions ( Figure 1C , D ) . However , when compared to regions occupied by pleurodont teeth , the DL is shorter both at the level of acrodont teeth and in interdental regions , a morphology that is in line with the absence of tooth replacement at these particular positions ( Figure 1D ) . We next assessed ontogenetic changes in bearded dragon tooth development , by comparing the post-ovipositional embryonic morphogenesis of the two dentition types with histological stainings ( Figure 1E–N ) . As already shown in a previous study focusing on primary tooth morphogenesis ( Handrigan and Richman , 2010b ) , odontogenesis at acrodont positions starts around 15 days post-oviposition ( dpo ) with the emergence of non-functional vestigial teeth developing directly from the OE ( Figure 1E ) . These superficial teeth abort morphogenesis prematurely following early mineralization , and functional acrodont teeth subsequently develop through classical stages of odontogenesis from a primary DL invaginating from the OE ( Figure 1E–J ) , as already described in a number of reptilian species ( Handrigan and Richman , 2010a; Buchtová et al . , 2013 ) . Around the time of tooth eruption , a small SDL emerges on the lingual side of acrodont teeth , but in contrast to previous reptile observations ( Richman and Handrigan , 2011; Buchtová et al . , 2013 ) , our data indicate that the SDL is not rudimentary but rather continues growing in size until hatching ( Figure 1I–J ) and even afterwards at postnatal stages ( see Figure 2B , C ) . Importantly , whereas the developmental stages are morphologically equivalent for the whole dentition , our direct comparisons of acrodont and pleurodont teeth clearly indicate a developmental delay in the initiation of pleurodont teeth , which only occurs around 40 dpo . Furthermore , despite the overall morphological similarities between the two types of teeth , pleurodont teeth show an elongated primary DL and thus develop deeper inside the mesenchyme than acrodont teeth ( Figure 1L ) . Finally , a different directional growth of the SDL is also evident , with pleurodont SDL growing lingually to accommodate the replacement tooth into the jaw , and acrodont SDL projecting downwards towards the jaw bone ( Figure 1J and Figure 2 ) . To investigate odontogenesis at the molecular level , we first inspected the expression pattern of conserved dental genes known to be associated with DL and SDL structures ( SHH , PITX2 , NOTCH1 , and LEF1 ) by in situ hybridization ( ISH ) , using various embryonic developmental stages in acrodont teeth: ‘vestigial tooth/DL’ stage ( 24 dpo ) , ‘bell’ stage ( 28 dpo ) , and ‘mineralization/SDL development’ stage ( 48 dpo; Figure 2A ) . The bearded dragon expression pattern of SHH and PITX2 is comparable to previous reports in polyphyodont species ( Fraser et al . , 2008; Handrigan and Richman , 2010b; Jussila and Thesleff , 2012; Wu et al . , 2013; Rasch et al . , 2016 ) . PITX2 marks the primary DL at early stages , then becomes strongly expressed throughout the enamel organ and DL in both vestigial and bell stage teeth , and finally remains in the cervical loops , odontoblasts , DL , and SDL at mineralization stage . In contrast , SHH localizes specifically to the inner enamel epithelium of vestigial teeth and functional teeth , moving downwards towards the cervical loops at later stages , but no expression is detected in the SDL ( Figure 2A ) . Notch signaling regulates stem/progenitor activity in several tissues and is particularly required for both tooth morphogenesis and maintenance of murine incisor stem cells ( Mitsiadis et al . , 2005; Felszeghy et al . , 2010 ) . Similarly to the mouse dentition , NOTCH1 is predominantly expressed in the epithelial compartment of functional teeth , including in the DL tip at 24 dpo and OE at both 24 and 28 dpo , but its expression is also noticeable in the condensed mesenchyme of vestigial teeth at early developmental stages . Finally , our analysis of the Wnt readout gene LEF1 indicates a condensed expression in the primary DL and at the tip of the SDL , a pattern previously observed in other squamate species ( Richman and Handrigan , 2011; Gaete and Tucker , 2013 ) but contrasting with previous reports in bearded dragon acrodont SDL ( Richman and Handrigan , 2011 ) . Indeed , consistent with the equivalent morphology and persistence of SDL in both acrodont and pleurodont dentitions ( Figure 1J , N ) , all tested genes were expressed in a similar pattern in both types of embryonic teeth ( Figure 2 and data not shown ) . Even further , our investigation of postnatal animals indicates the persistence of a highly proliferative SDL in both acrodont and pleurodont teeth at juvenile and adult stages ( Figure 2B , C ) , as assessed by detection of proliferating nuclear cell antigen ( PCNA ) proliferation markers and BrdU incorporation . Additionally , in contrast to juvenile chameleon lizards ( Buchtová et al . , 2013 ) , no signs of SDL deterioration or increased apoptosis were observed in the acrodont SDL ( Figure 2B ) . However , whereas PCNA , LEF1 , and PITX2 remain expressed in both acrodont and pleurodont teeth , careful examination of these markers indicates both oral-aboral and labial-lingual asymmetry in the pleurodont SDL , in contrast to the scattered pattern in acrodont SDL ( Figure 2B ) . A complementary pattern was observed by immunohistochemistry ( IHC ) for SOX2 , which is excluded from the very tip of pleurodont DL ( Figure 2B ) , as already observed in some other regenerating species ( Juuri et al . , 2013; Popa et al . , 2019 ) . Altogether , our data indicate that despite the lack of initiation of replacement teeth at acrodont positions in bearded dragon , the SDL does not degenerate and continues growing in size into adulthood . To our knowledge , this is the first example of long-term SDL maintenance in non-polyphyodont dentition , suggesting that this process is independent of and not causally related to tooth replacement . Furthermore , the observed differences in spatial organization of cell proliferation and developmental gene expression between postnatal acrodont and pleurodont dentitions likely reflect the need of directional growth and focal gene expression foreshadowing initiation of pleurodont tooth replacement . To assess the existence and location of putative slow-cycling stem/progenitor cells in bearded dragon teeth , we performed BrdU pulse-chase assays in juvenile animals . BrdU was administered for seven days ( pulse ) , and samples were obtained after 0 , 28 , and 58 days of chase time . To facilitate identification of putative stem/progenitor cells in both acrodont and pleurodont teeth , we distinguished cells that have stopped proliferating ( post-mitotic or terminally differentiated cells ) from cells that are still cycling by the use of BrdU immunostaining combined with PCNA detection at each time point ( Figure 3A ) . In addition , the relatively long half-life of PCNA increases the likelihood of correctly identifying slow-cycling BrdU/PCNA-double positive ( BrdU+PCNA+ ) cells ( Kurki et al . , 1986; Bravo and Macdonald-Bravo , 1987 ) . At day 0 , the majority of PCNA+ cells are also positive for BrdU , demonstrating the efficiency of our one-week pulse labeling in detecting most cell proliferation ( Figure 3A ) . The overall amount of BrdU+PCNA+ cells then drops dramatically by day 28 , and only clusters of cells remain throughout the dental epithelium in both types of teeth . Additionally , post-mitotic , single BrdU+ cells are present in some dental parts , including along the DL and in the more superficial regions of the OE . The latter cells likely represent the quick turnover of the epithelium itself . At day 58 , the few remaining BrdU+PCNA+ cells representing putative slow-cycling stem/progenitor cells are exclusively concentrated in two distinct regions of pleurodont teeth , the OE and the deep part of the DL ( DDL ) . Interestingly , however , double positive cells are more scarce in the DL of acrodont teeth while remaining in the OE ( Figure 3A ) . Our quantification of BrdU+PCNA+ cells in the pleurodont and acrodont epithelial regions confirms these findings , although the total number of BrdU+PCNA+ cells is relatively similar between the two dentitions ( d58; Figure 3C ) . Particularly , inspection of the distribution of BrdU+PCNA+ cells in four different epithelial subdivisions , from the OE to the SDL , highlights a high proliferative DDL in pleurodont teeth at day 0 ( region III; Figure 3B , D ) . Since the DDL represents the SDL budding site , we presume that this proliferation pattern may trigger , at least in part , the observed directional growth of the pleurodont SDL towards the lingual portion of the jaw . Consistent with this hypothesis , acrodont teeth show a relatively reduced proportion of BrdU+PCNA+ cells in the DDL region , likely reflecting the lack of directional growth of non-regenerating SDL . At day 58 , comparisons of the proportion of BrdU+PCNA+ cells confirm our previously observed variation in the location of LRCs in acrodont and pleurodont teeth . In particular , the DDL and OE regions represent together about three quarters of all LRCs in pleurodont teeth , thus supporting the existence of two distinct sites of putative stem/progenitor cells ( Figure 3D ) . In contrast , LRCs are more equally distributed between the DDL , OE , and SDL in acrodont teeth , indicating that altered distribution of LRCs is associated with lack of cyclic tooth replacement ( Figure 3C , D ) . To further assess the potential stemness of identified LRCs in bearded dragon dentition , we analyzed the expression of conserved putative stem/progenitor cell markers previously shown to be associated with dental tissues in vertebrates . Of note , tooth replacement is yet poorly studied in polyphyodont species , and only one dental progenitor marker , SOX2 , has been validated in several vertebrate species ( Juuri et al . , 2012; Juuri et al . , 2013; Martin et al . , 2016 ) . However , hair follicle stem cell markers such as LGR5 and IGFBP5 have also been shown to co-localize with LRCs at least in the leopard gecko DL ( Handrigan et al . , 2010 ) . Therefore , we compared the expression pattern of these three markers in pleurodont versus acrodont teeth of our bearded dragon model , using SOX2 IHC or LGR5/IGFBP5 ISH together with BrdU detection ( Figure 3E , F ) . Consistent with the BrdU-retaining pattern in pleurodont teeth , we first confirmed the co-expression of SOX2/LGR5/IGFBP5 or only SOX2 within the DDL and OE LRC sites , respectively ( Figure 3E ) . However , whereas acrodont teeth show similar co-expression of SOX2 in the OE LRCs , no LGR5 , IGFBP5 , or SOX2 expression could be co-detected in the scattered BrdU+ cells of acrodont DL . Altogether , these data clearly indicate the presence of separate putative dental epithelial stem/progenitor populations in bearded dragon , including one pleurodont-specific DDL population relatively similar to the leopard gecko DL ( Handrigan et al . , 2010 ) , and one shared pleurodont/acrodont OE population directly adjacent to the tooth and resembling the superficial taste/tooth junction recently identified in the spotted catshark ( Martin et al . , 2016 ) . The striking similarities between the SOX2+ stem/progenitor populations in the superficial OE of bearded dragon and spotted catshark suggest that cell migration from the OE could also contribute to tooth replacement in reptiles , as shown in sharks ( Martin et al . , 2016 ) . To test this hypothesis , we performed cell tracking experiments in bearded dragons both in vivo and ex vivo , using the lipophilic dye DiI ( Figure 4A , B ) . Labeling was carried out in the OE directly adjacent to pleurodont or acrodont teeth , and samples were collected at regular time intervals after DiI administration . Strikingly , in vivo tracking in pleurodont dentition revealed that within two weeks , cells labeled in the OE ( d0 ) shifted to the DL ( d7 ) and then SDL ( d14 ) , but also contributed to replacement teeth ( d14; Figure 4A ) . A similar ending-point was observed after four weeks of tracing ( data not shown ) , and a corresponding pattern was observed in acrodont teeth , indicating that migration from the OE still occurs despite the lack of replacement . Importantly , imaging of single ex vivo dental cultures confirmed the observed migratory pathway of DiI labeled cells along the DL , from OE to SDL ( Figure 4B ) , further demonstrating the contribution of OE progenitor/stem cells to tooth replacement . Particularly , the similar migration pattern observed in both dentition types suggests that , while not sufficient to induce replacement , the maintained growth of the DL/SDL might rely on this OE population . To test this hypothesis , we assessed the functional role of the putative OE stem/progenitor zone by manually removing the entire proximal OE in dental tissue cultures ( Figure 4C ) . As shown in Figure 4C , disruption of the OE leads to a significant decrease in proliferating DL/SDL cells ( see also quantification in Figure 4D ) as well as to some degeneration of DL/SDL structures in some samples , when compared to intact cultures , thus confirming the regulatory role of bearded dragon OE on DL/SDL growth and , eventually , tooth replacement . To obtain the first high-throughput molecular profiling of tooth replacement in vertebrates , we exploited the unique heterodont features of the bearded dragon to analyze and directly compare the transcriptome of pleurodont versus acrodont dental tissues . We particularly focused on the SDL , which shows divergent expression patterns in our candidate-based approach ( see Figure 2 ) , and the dental mesenchyme surrounding it , using staining-guided laser microdissection ( see microdissected areas in Figure 5A ) and Illumina RNA sequencing . At a cut-off of False Discovery Rate ( FDR ) -adjusted p-values<0 . 05 , we obtained 206 and 368 differentially expressed genes in the SDL and dental mesenchyme , respectively ( Figure 5A , Supplementary File 1 ) . Importantly , candidate genes already identified in the bearded dragon SDL , including PITX2 , LEF1 , and SOX2 were all retrieved in our transcriptome data , thus supporting our ISH experiments ( see Figure 2B ) . Furthermore , apart from LEF1 that was only slightly overexpressed in the pleurodont SDL ( pleurodont/acrodont fold change of 2 . 2 ) , many homeotic transcriptional regulators ( BARX1 , ALX1/ALX4 , SIX3 , DLX2 ) and members of classical signaling pathways ( BMP7 , WNT11 , FGF7 ) were highly differentially expressed between the two dentition types . A more thorough analysis of gene ontology ( GO ) category distribution among all differentially regulated genes indicates that more than 45% belong to the category ‘developmental process’ ( Figure 5B ) . Surprisingly , however , our list of developmental genes revealed only a partial overlap ( 18% ) with genes already known to be expressed in vertebrate dental tissues ( Figure 5B ) , indicating that regulation of the SDL in polyphyodont dentitions involves many genes and signaling pathways still uncharacterized and/or non-expressed in the monophyodont mouse dentition . The differential expression of a selection of newly identified genes was confirmed using quantitative PCR ( Figure 5C ) . Of the selected genes , SIX3 and ISL1 were significantly upregulated in the pleurodont epithelium , FOXI1 and BARX1 were downregulated in the pleurodont mesenchyme , and ALX1 was significantly enriched in both pleurodont epithelial and mesenchymal tissues ( Figure 5A , C ) . To further characterize their expression pattern during odontogenesis , ISH analysis of ALX1 , SIX3 and ISL1 was performed at both early postnatal ( stage equivalent to transcriptomic analysis ) and embryonic developmental stages ( Figure 6 ) . Interestingly , the relatively higher expression of ALX1 and SIX3 in pleurodont versus acrodont tissues was apparent in all stages tested by ISH , thus indicating variations in gene expression starting from early tooth developmental stages . However , similarly to the different classical dental markers tested ( see Figure 2 ) , differential ISL1 expression between the dentition types was only apparent at late tooth developmental stages , suggesting that this gene plays a role in tooth replacement . As already revealed by our transcriptomic data , the expression pattern of ISL1 and SIX3 was largely restricted to the SDL postnatally in both types of teeth , while being detected in both epithelium and dental papilla mesenchyme at embryonic dental stages . In contrast , ALX1 was confined to the developing pleurodont mesenchymes at all embryonic stages , but was later detected both in the pleurodont SDL and surrounding mesenchyme of postnatal teeth , thus supporting our RNA sequencing results and indicating dynamic expression pattern during tooth development and SDL patterning . The lack of lifelong replacement of oral dentition in many different lineages including classical model organisms has hampered the analysis of tooth regeneration in vertebrates . Furthermore , studies investigating polyphyodont species have so far largely focused on single- or cross-species comparisons of selected candidate genes based on mouse expression patterns , and the exact molecular basis of tooth renewal is still unclear . Here , we show that the bearded dragon , one of the rare vertebrate species with both polyphyodont and monophyodont teeth , constitutes a key model for filling this gap , allowing direct comparison of different dentition types . As already shown in non-teleost polyphyodont species , our findings first confirm the critical importance of both DL and SDL structures in tooth renewal . However , in contrast to a previous hypothesis ( Buchtová et al . , 2012; Dosedělová et al . , 2015 ) , the early degradation of the SDL is not the only mechanism restricting vertebrate tooth replacement . Indeed , bearded dragons maintain a proliferative epithelial SDL structure even in non-replacing teeth , indicating that the SDL maintenance is independent of and not strictly associated with tooth renewal . These data also contrast with previous observations made in acrodont teeth from chameleons , the second family of acrodont lizards , where the SDL was shown to diminish in size at postnatal stages independently of increased apoptosis ( Buchtová et al . , 2012 ) . The bearded dragon posterior dentition might thus represent an intermediary stage between polyphyodonty and complete monophyodonty , at least at the cellular level . In favor of that , and in contrast to chameleons , the implantation mode in most agamids is similar to intermediate pleurodont-acrodont dentition patterns observed in extinct acrodont iguanian lizards ( Averianov et al . , 1996; Simões et al . , 2015 ) . Haridy even recently proposed that the posterior tooth phenotype in bearded dragons might result from pleurodont to acrodont remodeling changes through ontogeny ( Haridy , 2018 ) . The latter hypothesis is coherent with our developmental data , as no clear morphological variations were observed in the two types of developing teeth , except for the length of the embryonic primary DL . The origin of such difference in the developing DL is unclear but might be linked to the delayed initiation of pleurodont teeth , when compared to acrodont teeth , which would extend the timing of DL proliferation . Importantly , differences were also noticed in the directional growth of SDL at hatchling and postnatal stages , with the pleurodont and acrodont SDL growing lingually or projecting more downwards towards the jaw bone , respectively . This differential directional SDL growth likely constitutes a key process to accommodate the replacement tooth , and was accompanied by an asymmetric expression pattern of conserved dental genes such as LEF1 and PITX2 , but also of proliferation , towards the pleurodont SDL tip . Interestingly , focal localization of LEF1 expression and/or Wnt signaling activity was already noticed in the squamate SDL ( Handrigan et al . , 2010; Gaete and Tucker , 2013 ) , alligator DL bulge ( Wu et al . , 2013 ) , shark SDL ( Rasch et al . , 2016 ) , and ferret SDL ( Jussila et al . , 2014 ) . However , our large-scale transcriptomic analysis now indicates that a large number of yet uncharacterized developmental genes and signaling pathways are , in fact , likely associated with SDL outgrowth and/or induction . Furthermore , besides differences in gene expression dynamics and regionalization between pleurodont and acrodont dentitions , our data show drastic changes in gene expression levels in both SDL and surrounding mesenchymal tissues , indicating the importance of multi-level dental organization for initiation of tooth replacement . Along this line , the overall conserved expression levels of Wnt signaling members in our comparative pleurodont-acrodont transcriptome analysis suggest that this conserved pathway might primarily regulate SDL growth , a process maintained in bearded dragon acrodont dentition , rather than directly induce tooth replacement . This hypothesis is consistent with our observations that separate processes regulate SDL growth and initiation of replacement teeth , further indicating that SDL maintenance is not causatively associated with tooth replacement . Among newly identified developmental factors differentially regulated between monophyodont and polyphyodont dentitions , several genes encoding homeobox transcription factors were highly upregulated in regenerating teeth . For example , ALX1 , a regulator of cartilage and craniofacial development in vertebrates ( Zhao et al . , 1993; Zhao et al . , 1996; Dee et al . , 2013 ) , was one of the most highly upregulated genes in both pleurodont SDL and mesenchymal tissues , and it is likely that the severe and relatively early craniofacial phenotypes of ALX1 mutations ( frontonasal dysplasia ) before initiation of odontogenesis have hampered identification of this gene as a regulator of tooth development and/or replacement in mammals . Similarly , SIX3 plays crucial roles in the development of the vertebrate sensory system ( Oliver et al . , 1995; Loosli et al . , 1999; Del Bene et al . , 2004 ) , but this gene has not yet been detected in the developing mouse dentition ( Nonomura et al . , 2010 ) . Finally , ISL1 has been reported to be selectively expressed in the dental epithelium of mouse incisor , while being absent from molar epithelium ( Mitsiadis et al . , 2003 ) , but this tooth-specific patterning has never been linked with continuous tooth growth or replacement in previous vertebrate studies . Based on the dynamic expression pattern observed in the developing epithelial and/or mesenchymal tissues , further investigation of these genes should provide key information on the signaling pathways controlling odontogenesis , including SDL outgrowth and/or initiation of replacement teeth . As already suggested by several studies in polyphyodont species ( Handrigan et al . , 2010; Wu et al . , 2013; Martin et al . , 2016; Thiery et al . , 2017 ) , our identification of bearded dragon slow-cycling LRCs co-expressing stem/progenitor markers provides further evidence of the importance of epithelial stem/progenitor cell populations in regulating tooth renewal in vertebrates ( Figure 7 ) . However , the mechanism of continuous tooth replacement identified in our model differs from the leopard gecko , the first inspected lizard species ( Handrigan et al . , 2010 ) , by the presence of two separate sources of putative stem/progenitor cells in the DL and OE ( Figure 7 ) , thus indicating alternative regenerative strategies and some specialization of stem/progenitor niches in squamates . Particularly , we show that migration from SOX2+ LRCs at the DL/OE junction contributes to the proliferative growth of DL and SDL structures , a process shared by both acrodont and pleurodont teeth ( Figure 7 ) . These findings support our hypothesis about the intermediate phenotype of bearded dragon acrodont dentition and closely parallel the regenerative dental strategy of sharks , which exhibit a similar LRC population at the taste/tooth junction ( Figure 7 ) . It is then tempting to speculate that SOX2+ OE cells of bearded dragons could also be bipotent , contributing to both taste and dental lineages ( Martin et al . , 2016 ) . In support of this , SOX2 has been recently identified as a conserved marker of dental progenitors but also of taste bud and non-gustatory epithelial fate in vertebrates ( Ohmoto et al . , 2017 ) . Furthermore , similarly to sharks , the OE of bearded dragons contains a heterogeneous population of cells , consisting of self-renewing epithelial progenitors and taste buds in close proximity to the DL ( Figure 3A and data not shown ) . However , anatomical studies have revealed strong heterogeneity in the presence and location of taste buds in lizards and snakes ( Uchida , 1980; Schwenk , 1985 ) , indicating that this potential bipotency may not be a general mechanism conserved across squamates . It is also still unclear if the newly identified OE LRC population exists and contributes to tooth replacement in other reptiles , as the OE has not been analyzed in previous lizard and snake studies despite the strong SOX2 expression observed in this region ( Juuri et al . , 2013; Gaete and Tucker , 2013 ) . The presence of one OE LRC population might also be directly linked to a specific tooth replacement rate and/or strategy in squamates , in a similar way as the heterogeneity in DL reported in teleosts with different replacement states ( Fraser et al . , 2006; Jernvall and Thesleff , 2012 ) . Indeed , replacement of pleurodont teeth in agamid lizards is relatively slow ( Cooper et al . , 1970 ) and this study ) and based on a ‘one-for-one’ strategy ( a single replacement tooth is formed at any one time for a single tooth position ) , whereas many squamates exhibit a ‘many-for-one’ replacement state with several replacement teeth formed ahead of function at a single tooth position . In this context , the conservation of only one putative stem/progenitor cell population in the DL might constitute a more efficient positioning strategy for generating multiple tooth generations , as exemplified by the leopard gecko ( Handrigan et al . , 2010 ) . Our characterization of the second source of LRCs in the bearded dragon pleurodont DL indicates striking similarities with the leopard gecko in terms of location , directly beside the lingually protruding SDL in the DDL , but also in terms of stem/progenitor marker expression profile , suggesting evolutionary conservation of this cell population at least among lizards ( Figure 7 ) . However , the exact function of putative stem/progenitor niches has remained unclear since their discovery in the leopard gecko . Our pleurodont-acrodont comparisons in bearded dragons now provide the first evidence of functional specialization of putative dental stem/progenitor populations in vertebrates . Specifically , we show that the DL population is likely strictly necessary for tooth replacement , as it is not detected in monophyodont bearded dragon teeth . The altered proliferation pattern , including lack of asymmetry and increased presence of BrdU-retaining cells , as well as differences in gene expression levels and location in the acrodont SDL suggest a key role for the DL population in the complex SDL organization required for initiation of tooth replacement ( Figure 7 ) . In contrast , the maintained OE population in the whole bearded dragon dentition contributes to SDL maintenance and growth , a process that is not causatively associated with initiation of tooth replacement . In conclusion , our results demonstrate that the various key dental features of the emerging bearded dragon model offer a powerful system to elucidate the developmental and genetic basis of evolutionary novelty and tooth regeneration in vertebrates . Particularly , the direct comparative analysis of pleurodont and acrodont dentitions indicates the importance of SDL patterning , in addition to maintenance , for vertebrate tooth replacement . This patterning process happens at various levels , including directional growth but also gene expression levels , dynamics , and regionalization , and involves a large number of yet uncharacterized developmental genes and signaling pathways . Finally , the two separate putative stem/progenitor populations identified in bearded dragon pleurodont teeth reveal a new vertebrate dental regenerative strategy associated with cell migration and functional specialization of putative stem/progenitor cells . All embryonic and postnatal stages of bearded dragons ( Pogona vitticeps ) were obtained from our animal facility at the University of Helsinki . For embryonic stages , fertilized eggs were incubated on a moistened vermiculite substrate at 29 . 5°C , and embryos were removed at regular interval after oviposition to obtain stages of interest . Embryos were staged on the basis of their external morphology according to a complete developmental table available for this species ( Ollonen et al . , 2018 ) . For general tooth morphology , heads were fixed overnight at 4°C in 4% paraformaldehyde ( PFA ) , dehydrated in an increasing alcohol series , and CT-scanned using a Bruker SkyScan 1272 instrument ( parameters: 70 kV , 142 µA , Al 0 . 5 mm filter , 13 . 2 µm resolution ) . For visualizing the dental lamina ( DL ) , dissected jaws were fixed as above , dehydrated to 70% ethanol , and stained for 2–4 weeks with 0 . 3% phosphotungstic acid ( PTA ) in 70% ethanol , as described before ( Metscher , 2009 ) . Stained jaws were then CT-scanned in 70% ethanol ( 90 kV , 111 µA , Al 0 . 5 + Cu 0 . 038 mm filters , 2 µm resolution ) . All CT-scans were reconstructed using Bruker NRecon 1 . 7 . 0 . 4 software , and 3D volume rendering and segmentation of DL and/or teeth were done manually using Advanced 3D Visualization and Volume Modeling 5 . 5 . 0 ( RRID:SCR_007353 ) . Following dissection , tissues were fixed overnight in 4% PFA at 4°C , and late embryonic stages showing mineralization were decalcified for 1–12 weeks , depending on sample size , in 20% ethylenediaminetetraacetic acid ( EDTA ) containing 0 . 5% PFA . Fixed and decalcified tissues were then progressively dehydrated into 100% methanol , embedded into paraffin blocks using a Leica ASP200 embedding machine , and sectioned at 7 μm in sagittal ( for pleurodont teeth ) or coronal ( acrodont teeth ) plane with a Microm HM355 microtome . Hematoxylin and eosin ( H and E ) staining of dental sections was done according to standard protocols , and images were acquired with an AX70 microscope , DP70 camera , and DP controller 1 . 2 . 1 . 108 software ( Olympus ) . IHC fluorescent staining was performed as described before ( Di-Poï and Milinkovitch , 2016 ) , using overnight incubation at 4°C with primary antibodies known to recognize reptile and/or chicken epitopes: proliferating cell nuclear antigen ( PCNA; 1:300 , mouse monoclonal , BioLegend , cat# 307901 , RRID:AB_314691 ) , 5-bromo-2'-deoxyuridine ( BrdU; 1:200 , rat monoclonal , Abcam , cat# ab6326 , RRID:AB_305426 ) , and SRY-box 2 ( SOX2; 1:200 , rabbit polyclonal , Abcam , cat# ab97959 , RRID:AB_2341193 ) . Last , incubation with Alexa Fluor-conjugated secondary antibodies ( Alexa Fluor-488: goat anti-rat IgG , Thermo Fisher Scientific , cat# A-11006 , RRID:AB_2354074; Alexa Fluor-568: goat anti-rabbit IgG , Thermo Fisher Scientific , cat# A-11011 , RRID:AB_143157 ) was performed for 1 hr at room temperature ( RT ) , and slides were mounted with Fluoroshield mounting medium ( Sigma-Aldrich ) containing 4′ , 6′-diamidino-2-phenylindole ( DAPI ) . Nuclear DNA fragmentation of apoptotic cells was labeled in situ using the TUNEL method ( in situ cell death detection kit , Roche ) , according to the manufacturer’s instructions . Before TUNEL staining , tissues were pre-treated with proteinase K ( 10 μg/ml , 15 min at RT ) and Triton X-100 ( 0 . 2% , 10 min at RT ) . Positive controls were performed by treating tissues with DNAse I before staining . Images were acquired either with a Nikon Eclipse 90i widefield microscope and Nikon NIS-Elements Advanced Research 4 . 30 . 01 software ( RRID:SCR_014329 ) , or with a Leica DM5000B widefield microscope and Leica Application Suite X ( LAS X ) 3 . 4 . 2 12 . 4 . 18 software ( RRID:SCR_013673 ) . Images were processed with Fiji/ImageJ ( RRID:SCR_002285; Schindelin et al . , 2012 ) and/or Adobe Photoshop CC ( RRID:SCR_014199 ) using levels adjustment . Digoxigenin ( DIG ) -labeled antisense riboprobes corresponding to Pogona vitticeps paired-like homeodomain 2 ( PITX2 , 544 bp ) , lymphoid enhancer binding factor 1 ( LEF1 , 460 bp ) , sonic hedgehog ( SHH , 931 bp ) , notch 1 ( NOTCH1 , 865 bp ) , leucine rich repeat containing G protein-coupled receptor 5 ( LGR5 , 754 bp ) , insulin-like growth factor-binding protein 5 ( IGFBP5 , 820 bp ) , ALX homeobox 1 ( ALX1 , 722 bp ) , sine oculis homeobox homolog 3 ( SIX3 , 639 bp ) , and ISL LIM homeobox 1 ( ISL1 , 675 bp ) genes were designed based on the publicly available Pogona vitticeps genome sequence ( Georges et al . , 2015 ) . ISH was performed on paraffin sections as described previously ( Eymann et al . , 2019 ) , using a hybridization temperature of 63°C . Following hybridization , sections were washed and blocked from non-specific antibody binding with blocking solution ( 2% Roche blocking reagent , 5% goat serum ) before incubating overnight at 4°C ( or 1 hr at RT ) with anti-DIG antibodies conjugated to alkaline phosphatase ( 1:2000 , sheep polyclonal , Sigma-Aldrich , cat# 11093274910 , RRID:AB_2734716 ) . A staining solution containing 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium was applied for 1–3 days at RT to visualize hybridization . Finally , slides were mounted using Dako Faramount aqueous medium ( Agilent ) and imaged with an Olympus AX70 microscope . Images were processed with Adobe Photoshop CC ( RRID:SCR_014199 ) , and ISH staining was enhanced for visualization by fake-coloring the selected blue pixels with the ‘Select color range’ and ‘Brush’ tools . 5-bromo-2'-deoxyuridine ( BrdU , Sigma-Aldrich ) was orally administered to Pogona vitticeps ( n = 10 ) twice daily ( 80 mg/kg body weight ) for a period of 7 days . Animals were euthanized 0 ( n = 4 ) , 28 ( n = 3 ) , or 58 ( n = 3 ) days after BrdU pulse labeling . A later time point ( 112 days after pulse ) was also tested , resulting in none or very few quantifiable LRCs . Chase time points were selected based on previous works on dental tissues ( Handrigan et al . , 2010; Wu et al . , 2013; Martin et al . , 2016 ) . For assessing cell proliferation , dental tissues were fixed and processed as described above , and PCNA/BrdU-double IHC was performed on sectioned acrodont and pleurodont teeth at each time point . The dental epithelium was divided into four distinct regions ( see Figure 3B ) , and BrdU/PCNA-double positive cells were quantified manually in each region using the ‘Count tool’ on Adobe Photoshop CC ( RRID:SCR_014199 ) . For each time point , cell counts were done using a minimum of three animal replicates , with 2–4 acrodont or pleurodont teeth per replicate and four sections per tooth covering similar coronal ( acrodont teeth ) or sagittal ( pleurodont ) sectioning planes . Statistical significance was evaluated with Student’s t-test in Microsoft Excel ( RRID:SCR_016137 ) . DiI ( 1 , 1′-dioctadecyl-3 , 3 , 3′ , 3′-tetramethylindocarbocyanine perchlorate; Biotium ) was first dissolved into 100% ethanol at 25 mg/ml , and subsequently diluted in 0 . 3M sucrose to achieve a final concentration of 5 mg/ml . Juvenile animals were anesthetized by injection of propofol ( 1 mg/ml; Sigma-Aldrich ) into the ventral coccygeal vein , and small amounts of DiI:sucrose solution were injected into the oral epithelium ( OE ) using a Hamilton syringe ( 702 RN , 25 µl volume ) and custom needles ( small RN , 26 s gauge ) . Animals were euthanized 1 , 7 , or 14 days after DiI administration ( n = 6 ) , and dental tissue sections were directly mounted with Fluoroshield mounting medium ( Sigma-Aldrich ) . For ex vivo DiI labeling , the skin of dissected jaws from euthanized Pogona vitticeps was removed , and about 1–2 mm thick slices were cut through the jaws with a scalpel . DiI:sucrose was administered into the OE using the tip of standard 30G needle syringe . Slices were placed on Corning Transwell-clear 3 μm nucleopore membrane inserts ( Sigma-Aldrich ) in 6-well plates filled with DMEM/F12 medium ( Gibco ) containing 10% fetal calf serum , 100 U/ml penicillin-streptomycin ( Gibco ) , 0 . 1 mg/ml ascorbic acid ( Sigma-Aldrich ) , 1:100 nystatin ( Gibco ) , and 1:100 amphotericin B ( Gibco ) . Slices were cultured at the air-liquid interface ( 37°C , 5% CO2 ) for 2 weeks , and imaged every other day with a Lumar V12 microscope and Zen Digital Imaging for Light Microscopy 1 . 1 . 2 . 0 ( RRID:SCR_013672 ) . For better visualization , DiI fluorescence images were overlaid with brightfield images . For OE removal experiments , the OE directly adjacent to the tooth in ex vivo tissue slices from one side of the jaw was carefully removed with a needle , without disturbing the mesenchyme or the DL . Slices from corresponding teeth on the other side of the jaw were used as controls ( n = 3 animals per group ) . Slices were cultured for 7 days , as described above , and then fixed overnight at 4°C with 4% PFA before processing into paraffin blocks and sectioning at 5 μm . Cell counts of PCNA-positive cells were done manually in the DL and SDL using the ‘Count’ tool on Adobe Photoshop CC ( RRID:SCR_014199 ) , and statistical significance was evaluated with Student’s t-test in Microsoft Excel ( RRID:SCR_016137 ) . Jaws of bearded dragons at early postnatal stage ( 4 days post-hatching , corresponding to well-developed SDL in both pleurodont and acrodont teeth ) were dissected in ice-cold phosphate buffered saline ( PBS ) and subsequently decalcified for 6 hr in 20% EDTA at 4°C with hourly changes of solution . Following decalcification , EDTA was washed out in PBS , and jaws were snap-frozen in liquid nitrogen and embedded in OCT compound ( Tissue-Tek ) . Thick sections of 25 µm were prepared using a Leitz 1720 cryostat and mounted on Superfrost Ultra Plus slides ( Thermo Scientific ) at −25°C . Completed slides were dipped into ice-cold 70% ethanol and stored at −25°C while sectioning . For dental tissue visualization , slides were washed in ice-cold MilliQ water for 1–2 min , stained in cold 1% ( w/v ) cresyl violet solution in 50% ethanol for 10 s , and then washed in cold 70% and 100% ethanol for 1–2 min . After air-drying , the stained sections were stored at −80°C before use . Staining-guided laser microdissection was used to separate both the SDL and surrounding mesenchymal tissue from sections of both acrodont and pleurodont teeth , using a Zeiss PALM Microbeam device equipped with a Zeiss AxioCam IC camera and PALM Robo software ( RRID:SCR_014435 ) . As cresyl violet staining allows epithelium to be well-distinguished , the SDL structure , defined as the free end of the DL epithelium that projects away from the tooth ( see Figure 5A ) , was collected first under high magnification; mesenchymal tissue surrounding the now-removed SDL was then collected ( see Figure 5B ) , thus ensuring precise tissue separation during collection . Tissue from multiple pleurodont and acrodont teeth from single individuals ( n = 3 ) was pooled together to increase RNA yield . Total RNA was extracted using the RNeasy Plus Micro Kit ( Qiagen ) , according to the manufacturer’s instructions , and RNA integrity and concentration was assessed using Bioanalyzer RNA 6000 pico assays ( Agilent ) . Ribosomal RNA ( rRNA ) depleted libraries were produced using the Nugen Ovation SoLo kit , and paired-end sequencing was done on a NextSeq 500 platform ( Illumina , U . S . A . ) . Quality control checks on raw sequence data were performed with the FastQC tool v0 . 11 . 8 ( RRID:SCR_014583 ) . Quality trimming , length filtering , and adapter sequence removal of reads were done using the Trimmomatic tool ( RRID:SCR_011848; Bolger et al . , 2014 ) , and the SortMeRNA pipeline ( RRID:SCR_014402; Kopylova et al . , 2012 ) was used for additional rRNA filtering . The STAR software ( RRID:SCR_015899; Dobin et al . , 2013 ) was used for sequence alignment to the Pogona vitticeps genome ( Georges et al . , 2015 ) and for generating count data . Count data were subsequently used for normalization ( Robinson and Oshlack , 2010 ) , and differential gene expression analysis was conducted with edgeR ( RRID:SCR_012802; Robinson et al . , 2010 ) , using three biological replicates . Genes with False Discovery Rate ( FDR ) -corrected p-values<0 . 05 were considered differentially expressed . The resulting list of genes was subjected to Gene Ontology ( GO ) -term functional annotation using the tool available on GO Consortium website ( http://geneontology . org/ ) , and genes from the GO category ‘developmental process’ were compared with vertebrate tooth gene expression ( http://bite-it . helsinki . fi/ ) and mouse gene expression ( http://www . informatics . jax . org/expression . shtml ) databases . Complementary DNA was generated by reverse transcription using the QuantiTect kit ( Qiagen ) , according to the manufacturer’s instructions , and analyzed by qPCR using iTaq Universal SYBR Green Supermix ( Bio-Rad ) and CFX96 real-time PCR detection system ( Bio-Rad ) . Amplicons were designed from 80 to 95 bp in length with a melting temperature of 61°C , based on the publicly available Pogona vitticeps genome sequence ( Georges et al . , 2015 ) . The housekeeping gene β-actin ( ACTB ) was identified as the best internal control for normalization . Each reaction was performed with four technical replicates using cDNA from three biological replicates . Results were analyzed with CFX Manager 3 . 1 . 1517 . 0823 software ( RRID:SCR_017251 ) and Microsoft Excel ( RRID:SCR_016137 ) using the double ΔCt method . The following primers were used: forward primer ( fp ) ALX1 , 5’-GCAGTTCCGTTGTGACTTC-3’; reverse primer ( rp ) ALX1 , 5’-ATCTGTCCGAGGTGAATGG-3’; fp SIX3 , 5’-CTCTACCACATCCTGGAGAAC-3’; rp SIX3 , 5’-TTCCTGGTAGTGAGCTTCG-3’; fp ISL1 , 5’-TGCGGCAATCAAATCCAC-3’; rp ISL1 , 5’-GGTTACATTCCGCACACTTC-3’; fp forkhead box I1 ( FOXI1 ) , 5’-GGCTATACTGGTTCAGTCCTC-3’; rp FOXI1 , 5’-ACTTCAGTGCCCTCTCTTG-3’; fp BarH-like homeobox 1 ( BARX1 ) , 5’-AAGGTGGAGGGCTTGAATC-3’; rp BARX1 , 5’-TGTCAACTGCTCGCTACTG-3’; fp ACTB , 5’-CCTGGAGAAGAGCTACGAAC-3’; rp ACTB , 5’-AGAAAGACGGCTGGAAGAG-3’ .	All multicellular organisms , from lizards to humans , must be able to repair and regrow damaged tissue . This includes not only healing after an injury , but also replacing parts of the body that suffer wear and tear . For example , many animals shed and replace worn out teeth throughout their life , but the number of times this occurs varies greatly between species . Much of the understanding about how teeth grow and develop has come from researching mice . However , mice only develop one set of teeth , making them a poor ‘model’ for studying how species such as fish and reptiles can re-grow and replace their teeth . Recent studies of these species has shown that regenerating teeth relies on a specialised structure known as the dental lamina . In mice , the dental lamina forms but then quickly disappears , preventing new sets of teeth from developing . In most animals that regrow their teeth , however , the dental lamina keeps growing beyond the most recently produced tooth to create an area where its replacement will emerge . Now , Salomies et al . have identified other strategies involved in tooth replacement from studying the bearded dragon lizard , a rare example of an animal that continuously regenerates some , but not all , of its teeth . Analysing the cells in different parts of the re-growing teeth from bearded dragon lizards revealed new features of the dental lamina . Specifically , Salomies et al . found that a previously uncharacterized set of genes within the dental lamina could determine whether or not a tooth will be replaced . Further experiments using microscope imaging revealed that bearded dragon lizards use two distinct groups of stem cells – specialised cells that have the potential to develop into various cell types in the body – to re-grow their teeth . These experiments demonstrate how the bearded dragon lizard uses a previously unknown mechanism to regenerate its teeth , combining elements used by gecko lizards and sharks . These findings are an important step towards understanding the different strategies animals can use to maintain and regenerate their teeth . The knowledge gained could one day help design better therapies for patients suffering from inherited dental disorders or tooth loss .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "evolutionary", "biology" ]	2019	The alternative regenerative strategy of bearded dragon unveils the key processes underlying vertebrate tooth renewal
In multicellular organisms , cells adopt various shapes , from flattened sheets of endothelium to dendritic neurons , that allow the cells to function effectively . Here , we elucidated the unique shape of cells in the cornified stratified epithelia of the mammalian epidermis that allows them to achieve homeostasis of the tight junction ( TJ ) barrier . Using intimate in vivo 3D imaging , we found that the basic shape of TJ-bearing cells is a flattened Kelvin's tetrakaidecahedron ( f-TKD ) , an optimal shape for filling space . In vivo live imaging further elucidated the dynamic replacement of TJs on the edges of f-TKD cells that enables the TJ-bearing cells to translocate across the TJ barrier . We propose a spatiotemporal orchestration model of f-TKD cell turnover , where in the classic context of 'form follows function' , cell shape provides a fundamental basis for the barrier homeostasis and physical strength of cornified stratified epithelia . The epidermis of the skin is a stratified epithelial cellular sheet that forms physical barriers on the body surface . Epidermal barrier dysfunctions cause not only lethal congenital disorders , but also a predisposition to the development of allergic diseases ( McGrath and Uitto , 2008; Akiyama , 2010; Kubo et al . , 2012; Weidinger and Novak , 2015 ) . The mammalian epidermis has two main physical barriers , an air-liquid interface barrier formed by the stratum corneum ( SC ) and a liquid-liquid interface barrier formed by tight junctions ( TJs ) ( Kubo et al . , 2012 ) . The TJ is a specialized intercellular adhesion complex that is crucial for epidermal barrier function , as it seals the paracellular space of epithelial cellular sheets ( Tsukita et al . , 2001; Van Itallie and Anderson , 2014; Krug et al . , 2014 ) . Maintenance of the TJ barrier is crucial for proper formation of the SC ( Sugawara et al . , 2013 ) , and thus for maintaining the physical barriers of the skin . The epidermis consists of keratinocytes that proliferate only in the basal layer and move upward to sequentially form stratum spinosum , stratum granulosum ( SG ) , and SC , and finally shed off from the top layer of the SC ( Figure 1A ) . During the course of its upward movement , a keratinocyte becomes flattened at the SG , forms TJs with adjacent cells in the SG2 layer ( the second cell layer of the SG ) , loses its TJs in the SG1 layer ( the top layer of the SG ) and is finally cornified to form the SC ( Figure 1A ) ( Kubo et al . , 2012; Yoshida et al . , 2013 ) . Despite its critical importance for the skin homeostasis , the TJ barrier of the epidermis consists of only a single layer of TJs forming a honeycomb-like mesh ( TJ honeycomb ) in the SG2 layer ( Figure 1A ) ( Furuse et al . , 2002; Kubo et al . , 2009; Yoshida et al . , 2013 ) . How can the TJ barrier be maintained while keratinocytes translocate across the single-layered TJ honeycomb for cell turnover ? 10 . 7554/eLife . 19593 . 003Figure 1 . Multi-dimensional visualization of epidermal TJs and TJ-bearing cells in mouse-ear skin . ( A ) 3D structure of the epidermis . ( B ) En face image of ZO-1-positive honeycomb in mouse-ear epidermis showing double-edged polygons ( * ) and single-edged polygons ( # ) . ( C ) Regularity in the size of the ZO-1-positive polygons represented in ( B ) and Figure 1—figure supplement 1 , shown by the mean ± SEM [error bars] ( one-way ANOVA multiple comparison test ) . ( D ) 3D image of a ZO-1-positive double-edged polygon in en face view ( top ) and 90°-rotated side view of the yellow-dotted rectangle ( bottom ) . Upper exterior polygon , yellow arrowheads; lower interior polygon , white arrows . See Video 1 . ( E ) Regularity of relative Z-axis position . Boxplots show the median , minimum , maximum , and interquartile range ( one-way ANOVA multiple comparison test ) for the ZO-1-positive polygons represented in Figure 1—figure supplement 2 . ( F ) In vivo live images of Venus in the ear of ZO-1-Venus mice ( left column ) and their schematics ( right column ) . Yellow arrowheads and green edges , edges of a Venus-positive polygon; white arrows and purple edges , edges of a newly appearing Venus-positive polygon; black arrows , Venus-positive edges connecting each vertex of the two polygons . See Video 4 . Scale bars , 10 µm . TJ , tight junction; SC , stratum corneum . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 00310 . 7554/eLife . 19593 . 004Figure 1—source data 1 . Percentage of double-edged polygons in ZO-1-positive honeycomb . The number of single- and double-edged polygons in 20 square en face images of 15376 µm2 from five independent assays . A represented image is shown in Figure 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 00410 . 7554/eLife . 19593 . 005Figure 1—source data 2 . Size of the ZO-1-positive polygons . The size of single- and double-edged polygons in ZO-1-positive honeycomb statistically analyzed in Figure 1C . How to define areas of polygons are shown in Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 00510 . 7554/eLife . 19593 . 006Figure 1—source data 3 . Z-axis position of the ZO-1-positive polygons . The Z-axis position of single- and double-edged polygons in ZO-1-positive honeycomb statistically analyzed in Figure 1E . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 00610 . 7554/eLife . 19593 . 007Figure 1—figure supplement 1 . Areas of exterior , interior and single-edged polygons . ( A–C ) Drawings of the ZO-1-positive edges shown in Figure 1B . The areas of polygons evaluated in Figure 1C are depicted for the exterior [pink in ( A ) ] and interior [yellow in ( B ) ] polygons of double-edged polygons ( * ) , and for the single-edged polygon [green in ( C ) ] adjacent to the double-edged polygons . The area of the single-edged polygon includes the overlapping area with the adjacent exterior polygons ( red arrows ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 00710 . 7554/eLife . 19593 . 008Figure 1—figure supplement 2 . Relative Z-axis position of TJ polygons in TJ honeycomb evaluated in vivo . A representative image of a ZO-1-positive double-edged polygon surrounded by six single-edged polygons evaluated in Figure 1E . The Z-axis position of each polygon was defined by an average of the Z-axis positions ( numbers ) of its vertices , analyzed by Imaris software ( purple arrowheads , external polygon; yellow arrowheads , internal polygon; green arrowheads , adjacent single-edged polygons ) . Scale bar , 10 µm . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 00810 . 7554/eLife . 19593 . 009Figure 1—figure supplement 3 . Epidermal TJ in ZO-1-Venus transgenic mice . ( A ) Colocalization of occludin , a transmembrane protein located at the TJs , with Venus in whole-mounted epidermal sheet from the ear skin of a ZO-1-Venus transgenic mouse . No morphological changes were observed in the TJ honeycomb in ZO-1-Venus transgenic mice . ( B ) Skin section staining of ZO-1-Venus transgenic mice after intradermal injection of Sulfo-NHS-LC-biotin as a tracer ( Furuse et al . , 2002; Yokouchi et al . , 2015 ) . Venus-positive junctions were observed to limit the inside-out permeation of the tracer ( yellow arrows ) , indicating that Venus successfully labeled the TJs without any apparent change in their occlusive function . Scale bars , 10 µm . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 009 In this study , we uncovered a mechanism of spatiotemporal coordination that replaces TJs from one cell to another to maintain TJ barrier homeostasis during cell turnover in the stratified epithelium of the epidermis . This model provides a fundamental structural basis for the integrity , physical strength , and homeostasis of the epidermis . To elucidate the mechanism of TJ barrier homeostasis in the epidermis , we investigated the three-dimensional ( 3D ) structure of the TJ honeycomb and its time-dependent changes . TJs were visualized in whole-mounted epidermis prepared from mouse-ear skin via immunostaining of zonula occludens-1 ( ZO-1 ) , an intracellular TJ scaffold protein ( Figure 1B ) . The murine ear epidermis consists of a basal layer , a spinous layer , and three layers of SG cells , and exhibits a regular structure of vertically aligned SC cells ( corneocytes ) and SG cells ( Mackenzie , 1969; 1975; Kubo et al . , 2009 ) . Two-dimensional ( 2D ) -projected images of the whole-mounted epidermis confirmed previous observations of one single-layered TJ honeycomb in the epidermis ( Figure 1B ) . We noticed that a significant number of TJ polygons in the TJ honeycomb of murine ear epidermis were double-edged ( 9 . 8 ± 0 . 6% , mean ± SEM , five independent assays; Figure 1B ) . The 3D observations revealed a striking regularity in the size and relative Z-axis position of the inner and outer polygons compared to their adjacent single-edged polygons . The outer polygons were larger ( 1044 . 3 ± 20 . 6 µm2 , mean ± SEM , n = 23 ) , and the inner ones smaller ( 814 . 7 ± 15 . 7 µm2 , n = 23 ) , than their adjacent single-edged polygons ( 905 . 8 ± 12 . 6 µm2 , n = 66; Figure 1C and Figure 1—figure supplement 1 ) . The inner ( smaller ) polygon was located lower in the epidermis than the outer ( larger ) one in each double-edged polygon ( n = 99; Figure 1D and Video 1 ) . Furthermore , comparison of the average Z-axis position of the vertices of the polygons revealed that the outer polygon was located significantly higher ( 0 . 294 ± 0 . 024 µm , n = 55 ) , and the inner one significantly lower ( −0 . 751 ± 0 . 036 µm , n = 55 ) , than the adjacent single-edged polygons ( −0 . 076 ± 0 . 009 µm , n = 55; Figure 1E and Figure 1—figure supplement 2 ) . We designated the inner ( smaller ) and outer ( larger ) double-edged polygons as the interior and exterior polygons , respectively . 10 . 7554/eLife . 19593 . 010Video 1 . En face 3D imaging of TJ honeycomb in mouse-ear skin . Representative 3D image of a ZO-1-positive double-edged polygon , shown in Figure 1D . TJ , tight jjunction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 010 We next observed dynamic changes in the TJ honeycomb structure to identify regulatory mechanisms that maintain the TJ barrier . In vivo 3D live imaging of the transgenic mouse-ear skin , in which epidermal TJs were labeled with recombinant ZO-1 fused to the fluorescent protein Venus ( Nagai et al . , 2002 ) ( Figure 1—figure supplement 3 ) , revealed sporadic appearance and disappearance of TJ polygons ( Videos 2 and 3 ) . Closer observation of a particular TJ polygon identified the systematic temporal order of events for its dynamic appearance and disappearance . Namely , a new smaller ( interior ) polygon appeared beneath a pre-existing ( exterior ) polygon , forming a double-edged polygon as observed in the fixed samples ( Figure 1D ) , followed by disappearance of the exterior polygon ( Figure 1F and Video 4 ) . Replacement of the old exterior polygon by a new interior polygon resulted in natural translocation of the cell body placed between the two polygons from the inside to the outside of the TJ barrier ( discussed below in the cell turnover model , Figure 4B ) . These findings indicate that the double-edged polygons are where cells translocate across the TJ barrier and thus are the key structures for barrier homeostasis . 10 . 7554/eLife . 19593 . 011Video 2 . In vivo 3D live imaging of epidermal TJs in the ear skin of a ZO-1 Venus mouse . Sporadic disappearance of TJ polygons ( polygons marked with an asterisk ) . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 01110 . 7554/eLife . 19593 . 012Video 3 . In vivo 3D live imaging of epidermal TJs in the ear skin of a ZO-1 Venus mouse . Sporadic appearance of new TJ polygons ( polygons marked with an asterisk ) . TJ , tight juction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 01210 . 7554/eLife . 19593 . 013Video 4 . In vivo 3D live imaging of edge-by-edge replacement of epidermal TJs in a particular polygon . Sequential appearance and disappearance of particular TJ polygons , shown in Figure 1F . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 013 Previous biotin permeation assays evaluated in vertical sections of mouse ear skin have not demonstrated barrier leakage at any TJ ( Furuse et al . , 2002; Kubo et al . , 2009 ) , suggesting that the barrier homeostasis of TJs is maintained even while the TJ polygons are being replaced . Does the interior polygon gain barrier function before its paired exterior polygon loses the barrier function during cell translocation across the TJ barrier ? We tested this hypothesis in whole mounted epidermis , using exfoliative toxin ( ETA , MW = ~31 kDa [Amagai et al . , 2000] ) as a tracer that digests the extracellular portion of desmoglein 1 ( Dsg1 ) ( Figure 2—figure supplement 1A–B ) . The interior TJ polygons exhibited barrier function that protected the extracellular portion of Dsg1 against ETA ( representative images , Figure 2A–C and Video 5; quantitative data of double-edged polygons [1 . 689 ± 0 . 020 times , mean ± SEM , n = 149] and their adjacent single-edged polygons [0 . 810 ± 0 . 006 times , n = 736] , Figure 2D ) . These observations indicate that the development of the occlusive paracellular barrier of the interior TJ polygons prevents barrier leakage during cellular translocation from inside to outside the TJ barrier ( discussed below in Figure 4B and C ) . Details of the barrier function of the interior TJ polygon , such as a dependency on molecular size or electric charge , remain to be determined , due to the limitations of in vivo permeation assays in stratified epithelia . 10 . 7554/eLife . 19593 . 014Figure 2 . Characterization of interior polygons in the epidermal TJ honeycomb . ( A and E ) En face images of double-edged [A and lower panels in E] and single-edged [upper panels in E] TJ polygons in the ETA-induced cell sheet of the epidermis ( Figure 2—figure supplement 1 ) . ( A ) Subcellular localization of Dsg1-EC and angulin-1 at double-edged polygons . ( B ) 90°-rotated image of the yellow-dotted rectangle in ( A ) ( see Videos 5 and 6 ) . ( C ) Schematic of ( B ) . N , nuclei; gray dotted lines , cell borders predicted from background cytosolic staining . ( D ) Fluorescence intensity of Dsg1-EC at the center of double- and single-edged polygons . The boxplots show the median , minimum , maximum , and interquartile range ( Student’s t-test ) . ( E ) Subcellular localization of tricellular TJ components ( angulin-1 and tricellulin ) at the single- and double-edged polygons . Yellow arrowheads , edges of the exterior polygon; white arrows , edges of the interior polygon; red arrowheads , vertical edges connecting the vertices of double-edged polygons; yellow arrows , vertices of single-edged polygons; Dsg1-EC , immunostained signals for extracellular portion of desmoglein 1 , presumably representing desmosome-accumulating Dsg1 . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 01410 . 7554/eLife . 19593 . 015Figure 2—source data 1 . Fluorescent intensity of Dsg1-EC at single- and double-edged polygons . Fluorescent intensity of Dsg1-EC at the center of ZO-1-positive polygons in ETA-induced cellular sheet statistically analyzed in Figure 2D . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 01510 . 7554/eLife . 19593 . 016Figure 2—figure supplement 1 . Preparation of the exfoliative toxin ( ETA ) -induced cell sheet and isolation of TJ-bearing keratinocytes from the skin . ( A–C ) Schematics of a vertical section of the epidermis before ETA treatment ( A ) , after intradermal injection of ETA ( B ) , and after trypsin treatment ( C ) . Green squares , TJ; red squares , desmosome; SG , stratum granulosum; SC , stratum corneum . Both SC and SG cells are drawn as f-TKD ( see Figure 4A–C ) . ( A ) TJs are located at the apical edges of the SG2–SG2 cell contact . ( B ) Desmosomal junctions are digested by ETA beneath the TJ barrier but not above it , as the permeation of ETA is limited at the TJ barrier , and a stratified cell sheet consisting of SC , SG1 , and SG2 cells is separated from the skin ( Yoshida et al . , 2013; Yokouchi et al . , 2015 ) . The en face images of the resulting ETA-induced cell sheet are shown in Figure 2A and E . ( C ) SG1 and SG2 cells are dissociated from the ETA-induced cell sheet by trypsin , as trypsin digests both TJs and desmosomal junctions . ZO-1 , the TJ intracellular scaffold protein , remains undigested [green squares in ( C ) ] . The cytosolic portion of Dsg1 also remains undigested in the cytoplasm ( not shown ) . The representative en face images of the isolated SG2 cells are shown in Figure 3F and Video 7 . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 01610 . 7554/eLife . 19593 . 017Figure 2—figure supplement 2 . Subcellular localization of claudin-1 at double-edged polygons . Subcellular localization of claudin-1 at the ZO-1-positive double-edged polygons in the ETA-induced cell sheet . Yellow arrowheads , edges of the exterior polygon; white arrows , edges of the interior polygon . Scale bar , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 01710 . 7554/eLife . 19593 . 018Video 5 . Characterization of the interior ( smaller ) polygon of double-edged TJ polygons . The newly formed polygon limits ETA permeation ( Figure 2A–C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 018 To characterize the molecular players that mediate the formation of the double-edged polygons , we further investigated the newly formed interior polygons . Claudin-1 , a major transmembrane protein of epidermal TJ ( Furuse et al . , 2002 ) , was localized to both the exterior and interior polygons ( Figure 2—figure supplement 2 ) . In simple epithelia , most TJs are formed as conventional bicellular TJs ( bTJs ) on the apical edges between two cells , and tricellular TJs ( tTJs ) are formed at tricellular contacts among three cells ( Figure 4—figure supplement 1A and B ) ( Furuse et al . , 2014 ) . In stratified epithelia of the epidermis , we observed the localization of tTJ-specific proteins ( angulin-1 and tricellulin ) on all edges of the interior polygons of the double-edged ZO-1-positive polygons , while the exterior polygons were bTJs ( Figure 2B , E and Video 6 ) . This is a striking contrast to the single-edged polygons , where tTJ-specific proteins were found as dots or short lines at the vertices as in simple epithelia ( Figure 2E ) . These findings indicate that the interior TJ polygons consist of tTJs , suggesting that the edges of the interior polygons are tricellular contacts between three TJ-forming cells ( discussed below in Figure 4B and Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 19593 . 019Video 6 . Characterization of the interior ( smaller ) polygon of double-edged TJ polygons . The newly formed polygon is a tTJ polygon ( Figure 2E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 019 Conventional schemas of skin often describe epidermal keratinocytes as rectangles in vertical 2D sections and assume their 3D cell shape to be a hexagonal prism ( Figure 3A ) ( Goldsmith et al . , 2012 ) . Indeed , corneocytes are regularly stacked in the murine ear SC ( Mackenzie , 1969; Christophers , 1972 ) . However , the regular stack of hexagonal prism cells is inconsistent with our in vivo observations of the regularity of the sizes of TJ polygons ( Figure 1C versus Figure 3—figure supplement 1 ) . Moreover , the hexagonal prism structure would require unrealistic sliding of the cell columns for cell turnover , resulting in the breakage of the TJ barrier ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 19593 . 020Figure 3 . Characterization of the shape of SG2 cells and structures of epidermal TJ honeycomb . ( A–C ) Three possible space-filling structures of the SG . ( D ) Regular interdigitated stacks of f-TKD cells . SG2 cells are displayed at the top of cell columns . ( E ) En face view of the TJ honeycomb ( green edges ) on the f-TKD cell stacks shown in ( D ) . ( F ) Six representative polyhedral shapes ( a–f ) of isolated SG2 cells , corresponding to the TJ polygons ( green edges ) in ( D ) and ( E ) . The cells were visualized by cytoplasmic staining for the intracellular portion of desmoglein 1 ( Dsg1-IC , middle column ) , with ZO-1-positive TJ at the edges ( left and middle column ) , and their 3D rendered images ( right column ) . See Video 7 for 3D rendered images . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 02010 . 7554/eLife . 19593 . 021Figure 3—figure supplement 1 . Possible cell turnover and TJ replacement in the conventionally proposed structure of SG and SC with hexagonal prism cells . ( A ) 3D schematic ( left ) and a vertical section ( right ) of the structure . ( B–E ) Temporal order of cell turnover and TJ replacement ( 3D structure , left panels; vertical-sectional views , right panels ) . TJ honeycomb is formed at the apical edges of SG2–SG2 cell junctions in ( B ) ; Two SG2 cells are vertically aligned as a result of an SG3 cell differentiated to an SG2 cell in ( C ) ; The entire cell column slides upward for cell turnover in ( D ) ; Cell turnover is completed when the upper SG2 cell differentiates to become an SG1 cell ( red ) in ( E ) . This process is inconsistent with our in vivo observations , because ( 1 ) the interior TJ polygon newly formed by the two vertically aligned SG2 cells ( purple polygon in ( D ) ) is of the same size as the pre-existing TJ polygon , and ( 2 ) the upward sliding of the cell column requires a total breakdown of their lateral desmosomal junctions in ( D ) . Green edges in left panels , pre-existing TJ honeycomb; green closed rectangles in right panels , pre-existing TJ; purple edges in left panels , newly produced TJ polygons; purple closed rectangles in right panels , newly produced TJ; red closed rectangles in right panels , desmosomes . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 021 Our in vivo observations can be coherently explained if we assume the SG cells have the shape of the Kelvin’s tetrakaidecahedron , which was originally proposed by Lord Kelvin in 1887 as the optimal natural space-filling shape with minimal surface area ( Figure 3B ) ( Thomson , 1887 ) . A flattened variation of Kelvin’s tetrakaidecahedron ( f-TKD ) shape ( Figure 3C ) was once observed in the corneocytes at the surface of murine ear skin in scanning electron microscopic studies ( Allen and Potten , 1976; Menton , 1976 ) but has been overlooked for several decades . To investigate whether the 3D shape of SG cells is f-TKD , similarly to corneocytes , we isolated SG2 cells from mouse ear epidermis via sequential treatment with ETA and trypsin ( Figure 2—figure supplement 1A–C ) . The immunofluorescent staining of isolated SG2 cells and their 3D rendered images demonstrated that the TJ-bearing SG2 cells have an f-TKD–like polyhedral shape , with TJs on their edges ( Figure 3F and Video 7 ) . Moreover , in vivo observations of the isolated SG2 cells ( Figure 3F ) confirmed the variations in the shapes of TJ polygons in the regular stacks of f-TKD cells , depending on their relative Z-axis position ( Figure 3D and E ) . We thus concluded that the basic shape of SG2 cells is f-TKD . 10 . 7554/eLife . 19593 . 022Video 7 . Three-dimensional imaging of TJ-bearing SG2 cells isolated from mouse ear epidermis . Representative polyhedral shapes of isolated SG2 cells with TJs on their edges , shown in Figure 3F . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 022 By integrating our in vivo observations on replacements of TJ polygons from edge to edge on the polyhedral cells , we propose an f-TKD turnover model , which describes a sophisticated spatiotemporal orchestration mechanism of cell-turnover to maintain the TJ barrier ( Figures 4 and 5 ) . 10 . 7554/eLife . 19593 . 023Figure 4 . Cell turnover and TJ replacement in the SG2 f-TKD cells . ( A and B ) Cell turnover and TJ replacement in the SG2 layer of regular interdigitated stacks of f-TKD cells [3D structure of the cell stack with SG2 cells displayed at the top of cell columns , left panels in ( A ) and ( B ) ; 3D structure of the TJ barrier , right panels in ( A ) ; vertical sectional views , middle panels in ( B ) ; en face view of the TJ honeycomb , right panels in B] . Four SG2 cells ( * ) are located higher than their six adjacent SG2 cells in phase 0 in ( A ) , and they differentiate to become SG1 cells in phase 2 . The differentiation of SG3 cells to SG2 cells [red arrowheads in A and B] in phase 1 results in the vertical alignment of two SG2 cells , between which tTJ polygons ( purple edges ) are formed . Differentiation of the top SG2 cells ( shown in red ) to SG1 cells is completed in phase 2 . ( C ) Schematics showing the occlusive function of single-edged TJ polygons ( phase 0 and 2 ) and an interior tTJ polygon ( phase 1 ) against the permeation of ETA . Green edges , bTJs; purple dots and edges , tTJs . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 02310 . 7554/eLife . 19593 . 024Figure 4—figure supplement 1 . Localization of TJ strands on the hexagonal prism cells in simple epithelia and on f-TKD cells in stratified epithelia . ( A ) The two meshes of TJ strands facing each other between adjacent cells bind together like Velcro tape to seal the paracellular space ( Tsukita et al . , 2001 ) . bTJs seal the paracellular space between two adjacent cells ( green mesh ) , and tTJs ( three purple meshes ) narrow the tricellular tube ( arrow ) , the paracellular space among three cells ( Furuse et al . , 2014 ) . ( B–C ) Hexagonal prism cells in simple epithelia ( B ) and f-TKD SG2 cells in stratified epithelia ( C ) . Conventional bTJs are located at the apical edges of lateral cell-cell contacts ( green edges ) ( Furuse et al . , 2014; Van Itallie and Anderson , 2014 ) , and tTJs are formed with some vertical elongation along the edges of the tricellular contact ( purple lines ) ( Furuse et al . , 2014 ) . Presumed localization of TJ strands for bTJs and tTJs is shown as green mesh and purple mesh , respectively . ( D ) Formation of a double-edged TJ polygon on f-TKD cells in stratified epithelia during cell turnover , corresponding to phase 1 in Figure 4B–C . The exterior and interior TJ polygons are bTJ polygons and tTJ polygons , respectively . The tTJ is formed on the edges at the tricellular contact among two vertically aligned SG2 cells and one laterally adjacent SG2 cell . ( E ) Formation of bTJs or tTJs at a tricellular contact between three SG cells in the epidermis . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 02410 . 7554/eLife . 19593 . 025Figure 4—figure supplement 2 . Regularity of polygon size in the TJ honeycomb on f-TKD cells in stratified epithelia . ( A ) Representative en face image of the TJ honeycomb ( green edges ) on the f-TKD cell layer . Double-edged polygons are marked by * . Any two adjacent SG2 cells ( shown in two different colors ) are located at different heights ( z-axis ) , as shown in Figure 3D . ( B ) The exterior polygon ( top ) is larger than its adjacent single-edged polygons ( middle ) , any of which is larger than the interior polygon ( bottom ) . The area of single-edged polygons is defined to include the overlapped area with the exterior polygons ( arrows ) as shown in Figure 1—figure supplement 1 . Our in vivo observation definitively confirmed that the single-edged polygons were smaller than the exterior polygons and larger than the interior polygons ( Figure 1C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 02510 . 7554/eLife . 19593 . 026Figure 5 . f-TKD turnover model . An entire cycle of the turnover of a set of f-TKD SG2 cells is shown on the outer blue spiral arrow , from the appearance of aTJ polygon ( time point #1 ) to its disappearance ( time point #11 ) on a particular SG2 f-TKD cell ( the center yellow cell of the set of f-TKD cells ) . Three-dimensional cartoons and en face views of the yellow cell at each timepoint are depicted on the spiral and straight orange arrows , respectively . The relative Z-axis position of the cell changes from the lowest ( time point #2 ) to the next highest ( time point #9 ) , while the adjacent SG2 cells ( asterisked cells ) are turned over . The TJ polygon of the cell becomes larger in a stepwise manner via edge-by-edge TJ replacement ( timepoint #2–#9 ) . The phases 0–2 correspond to those in Figure 4B . See Video 8 . Green edges , bTJs; purple edges , tTJs . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 02610 . 7554/eLife . 19593 . 027Figure 5—figure supplement 1 . Computational simulation of f-TKD model for continuous cell turnover . A computational simulation for the f-TKD model confirmed that the proposed three local rules ensure TJ homeostasis with continuous cell turnover in a stratified cell sheet ( 2D movie , Video 9; 3D movie , Video 10 ) . En face view of the time-dependent change in the TJ honeycomb reproduced the cycle of edge-by-edge TJ replacement shown in Figure 5 and Video 8 . Polygons replaced in the next panel are marked with * , # , and % in the yellow , pink , and blue layer cells , respectively . Only the TJ-bearing SG2 cells are depicted in three different colors , depending on their cell height . The tTJs at the vertices of single-edged TJ polygons are not displayed , to better illustrate the honeycomb-mesh structure of TJs . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 027 In this model , the cell turnover across the TJ barrier is completed in two phases ( Figure 4A and B and Video 8 ) . The cell turnover occurs at the cell column where there is an SG2 cell at a higher Z-axis position than any of its adjacent SG2 cells ( phase 0 ) . In the first phase ( phase 1 ) , an SG3 cell differentiates to become an SG2 cell , and then an interior TJ polygon is formed , leading to the appearance of a double-edged polygon , as observed in vivo ( Figure 1F ) . The interior TJ polygon consists of tTJs formed among three SG2 cells: two vertically aligned SG2 cells and a laterally adjacent SG2 cell ( Figure 4B ) . The upper SG2 cell naturally exits the TJ barrier , not by its own upward migration , but by the appearance of the interior TJ polygon on its basal side followed by the disappearance of the exterior TJ polygon . The precise subcellular structure of the interior TJ polygon predicted in the f-TKD model is shown in Figure 4—figure supplement 1D . The proposed model reproduces the in vivo observation of the regularity in the size of the TJ polygons ( Figure 1C vs Figure 4—figure supplement 2 ) and in the Z-axis location of the polygons ( Figure 1E vs Figure 4A ) . 10 . 7554/eLife . 19593 . 028Video 8 . Entire sequences of the edge-by-edge TJ replacement on a particular f-TKD cell ( see Figure 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 028 In the second phase ( phase 2 , Figure 4A and B ) , the upper SG2 cell differentiates to become an SG1 cell , the exterior bTJ polygon disappears , and the interior tTJ polygon becomes a bTJ polygon . The precise order for formation of the functional TJs , whereby the interior TJ polygon gains barrier function before the exterior TJ polygon disappears ( Figure 4C ) , is critical for TJ barrier homeostasis while cells are turned over . After the second phase , the surrounding six f-TKD SG2 cells go through phases 0–2 and differentiate to become SG1 cells one by one ( from #3 to #9; Figure 5 ) , accompanied with an upward movement of the entire cell sheet ( Video 8 ) . As a result , the TJ polygon on the SG2 cell at the center increases in size in a stepwise manner via translocation of the cell from the lowest to the next highest of its six adjacent SG2 cells ( from #3 to #9; Figure 5 ) until the next turnover cycle starts ( from #9 to #11; Figure 5 ) . The characteristic feature of the proposed f-TKD turnover model ( Figure 5 ) is that the spatiotemporal regulation of dynamic TJ replacement and continuous cell turnover across the TJ honeycomb can be explained by the following three simple local rules governing the f-TKD cells: ( 1 ) bTJs are formed between two SG2 cells , and tTJs are formed at the tricellular cell contact among three SG2 cells; ( 2 ) An SG2 cell located higher than its six adjacent SG2 cells differentiates to become an SG1 cell; ( 3 ) The differentiation of a SG3 cell to become a SG2 cell is synchronized with the differentiation of the pre-existing SG2 cell to become a SG1 cell in each cell column ( Figure 4B ) . To confirm that these three local rules ensure the maintenance of TJ barrier homeostasis , we developed a mathematical model implementing our f-TKD turnover model in silico . The in silico model consists of stacked layers of optimally packed SG cells , each of which is an f-TKD , similar to the geometric model of the corneocyte cornified envelope ( Feuchter et al . , 2006 ) . The nominal values for the model parameters were determined so that the simulated dynamics reproduce the experimentally observed ratio of double-edged polygons among all the polygons ( 9 . 8 ± 0 . 6% , Figure 1B ) , as well as the average time required for cell turnover from entering to exiting the SG2 layer ( 24 hr ) , estimated from the average cell turnover rate in mouse ear epidermis ( Potten , 1975 ) . Computational simulation of the in silico model ( Figure 5—figure supplement 1 , and Videos 9 and 10 ) reproduced the entire sequence of TJ replacement on f-TKD cells , including formation of TJs , transition of the f-TKD cells , and upward movement of differentiated keratinocytes as a stratified cell sheet , while maintaining the relative location of individual f-TKD cells . It therefore confirmed the maintenance of TJ barrier homeostasis , cell-cell adhesion , and the physical strength of the cell sheet during cell turnover . 10 . 7554/eLife . 19593 . 029Video 9 . 2D movie of computational simulation for the maintenance of SG2 layer integrity and continuity of TJ honeycomb during cell turnover in the f-TKD model of the epidermis . The 2D movie demonstrates the SG dynamics for 33 hr , with five frames per second and 7 . 2 min between frames . Green edges , conventional bTJ; purple edges , tTJ . bTJ ( green edges ) is formed at the apical edges of lateral SG2–SG2 cell contact faces , and tTJ ( purple edges ) is formed at the tricellular cell contact edges among three SG2 cells . tTJs at the vertices of single-edged TJ polygons were omitted from the display to better illustrate the honeycomb-mesh structure of TJ . Only the TJ-bearing SG2 cells are colored with three different colors depending on cell height . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 02910 . 7554/eLife . 19593 . 030Video 10 . 3D movie of computational simulation for the maintenance of SG2 layer integrity and continuity of TJ honeycomb during cell turnover in the f-TKD model of the epidermis . The 3D movie demonstrates the SG dynamics for 66 hr , with 20 frames per second and 7 . 2 min between frames . Green edges , conventional bTJ; purple edges , tTJ . bTJ ( green edges ) is formed at the apical edges of lateral SG2–SG2 cell contact faces , and tTJ ( purple edges ) is formed at the tricellular cell contact edges among three SG2 cells . tTJs at the vertices of single-edged TJ polygons were omitted from the display to better illustrate the honeycomb-mesh structure of TJ . Only the TJ-bearing SG2 cells are colored with three different colors depending on cell height . TJ , tight junction . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 030 In this study , we demonstrated the biological links between the shape of TJ-forming cells and a mechanism for the maintenance of TJ barrier homeostasis in the epidermis , as a representative example of how tissues adopt form to follow function . Our proposed f-TKD cell turnover model suggests that the local spatiotemporal orchestration of cell differentiation in the SG cell layer enables the constituent f-TKD cells to be renewed while maintaining TJ barrier homeostasis in cornified epidermis . The regular columnar stack of flattened corneocytes in murine ear epidermis demonstrates a regular zig-zag interdigitation pattern between two adjacent cell columns ( Figure 6—figure supplement 1 reproduced from Figure 1 of Mackenzie [1975] ) ( Mackenzie , 1969; Christophers , 1972; Menton , 1976; Ball , 2001 ) . This regular interdigitation pattern was spontaneously reproduced by our f-TKD model ( Figure 6A ) , in which cell differentiation in the SG2 cell layer occurs in turn in pairs of adjacent cell columns . The f-TKD model thus provides a coherent mechanistic explanation of how the regular stacks of corneocytes are constructed . 10 . 7554/eLife . 19593 . 031Figure 6 . Spatiotemporal orchestration of cell differentiation in the f-TKD model generates interdigitated stacks of corneocytes . ( A ) The f-TKD model provides a coherent explanation of how the regular interdigitation of corneocytes is produced in the SG2 layer . Arrows show SG3 cells that are newly aligned to the columnar stack . One cycle of cell turnover ( orange arrow ) corresponds to one cycle indicated in Figure 5 . The yellow-colored cell is differentiated from SG3 to SG1 in the time course . ( B ) Previously proposed columnar unit concept of epidermal structure and turnover . ( C ) Our proposed f-TKD model for epidermal homeostasis . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 03110 . 7554/eLife . 19593 . 032Figure 6—figure supplement 1 . Interdigitated stacks of corneocytes as a footprint of spatio-temporally regulated differentiation in SG2 cells . ( A ) A section of hamster epidermis originally shown in ( Mackenzie , 1975 , Figure 1 ) clearly demonstrates columnar stacks of corneocytes and SG cells . The red rectangular areas are magnified in ( B ) . ( B ) Regular zig-zag pattern of interdigitation between corneocytes suggests the synchronization of cell differentiation between adjacent cell columns . The numbers indicate the order of terminal differentiation suggested by the spatial relationships between corneocytes . DOI: http://dx . doi . org/10 . 7554/eLife . 19593 . 032 The f-TKD model further suggests that homeostasis of the SC–SG layers is maintained by the spatiotemporal orchestration of cell differentiation in the SG2 cell layer , rather than by the kinetics of stem cell proliferation and differentiation in the basal layer . Our model is consistent with a pioneering computational simulation suggesting that the regular 3D stacking structure of the SC can be spontaneously formed by randomly supplied cells ( Honda et al . , 1996 ) . Our model also accords with in vivo cell-tracing studies demonstrating a random supply of cells from the spinous layer to the SG layer ( Doupé et al . , 2010 ) , and in vivo live observations of the upward movement of spinous layer cells funneling into preexisting cell columns of the SG ( Rompolas et al . , 2016 ) . In the 1980s , the epidermal proliferative unit ( EPU ) concept postulated that each column of flattened corneocytes corresponds to a set of basal layer stem cells that proliferate directly under the column ( Potten and Allen , 1975; Potten , 2004 ) . In the EPU model , the regular interdigitation pattern of the SC is explained by the regular kinetics of basal layer stem cells in each EPU . However , recent in vivo cell-fate-tracing studies demonstrated more random cell fate decisions in the basal cell layer , with EPU model-like upward cell movement funneling into the cell columns of the SG , leading to a new concept: the epidermal differentiation unit ( EDU ) ( Rompolas et al . , 2016 ) . However , the underlying mechanism dictating the regular interdigitation pattern of cell columns remains enigmatic ( Figure 6B ) . In our f-TKD model , the cell columns exist only in the SG and SC layers , rather than extending from the basal layer through to the SC . The cells originate from stem cells in the basal layer and are randomly supplied to a spinous layer . Once the cells enter the SG layer , cell turnover in adjacent columns is tightly coordinated in a spatiotemporal manner , leading to the regular interdigitation pattern ( Figure 6C ) . Epidermal homeostasis is maintained by balancing cell proliferation in the basal layer , cell translocation ( differentiation ) from the basal to the spinous layer , cell integration to the SC/SG layer , and cell shedding from the top of the SC as squames . Future studies are needed to explore how this balance among critical processes in the epidermal layers is regulated to maintain a constant thickness of the epidermis . The mammalian epidermis is a representative stratified epithelium . Nutrients for stratified cells are mostly supplied from the basal connective tissue via diffusion through paracellular pathways . If all the keratinocytes in the epidermis formed TJs , cells located in the upper epidermis would likely starve due to their segregation from the nutrient supply by multi-layered TJ barriers . Therefore , it is biologically reasonable that the TJ barrier is single-layered in stratified epithelia ( Kubo et al . , 2012; Yoshida et al . , 2013 ) . Our observations revealed that TJ formation is restricted only between SG2 cells ( Figure 4—figure supplement 1E ) . The molecular mechanisms that coordinate the sequential differentiation steps from SG3 to SG1 cells and restrict TJ-forming activity to SG2 cells are currently unknown . The f-TKD model may help to reveal these mechanisms in future studies . Various mammalian epidermis shows a regular interdigitation pattern in the SC ( Christophers , 1972; Mackenzie et al . , 1981 ) , suggesting that the f-TKD cell turnover mechanism governs cell differentiation in mammals . Further investigation on whether this characteristic interdigitation pattern is observed in the SC of other vertebrates , such as amphibians , reptiles and birds , may reveal the general applicability of the f-TKD model to cornified stratified epithelia . Other mechanisms could be involved in maintaining TJ barrier homeostasis in simple epithelia and non-cornified stratified epithelia , where apoptotic cells are extruded to the outside TJ barrier by adjacent cells that migrate into the basal side of the apoptotic cells and form multi-junctional TJs ( Pentecost et al . , 2006; Eisenhoffer and Rosenblatt , 2011 ) . The actual structure of the optimal space-filling shape with minimal surface area could be more complex than Kelvin’s model ( Lewis , 1943; Williams , 1968; Weaire and Phelan , 1994 ) , and the shape and alignment of corneocytes are much more variegated in human skin compared to mouse ear skin ( Mackenzie et al . , 1981 ) . Nonetheless , the basic concept of the spatiotemporal orchestration of SG cell differentiation in the f-TKD model would be sufficient to explain the regular interdigitation pattern of corneocytes observed in various types of cornified skin ( Mackenzie et al . , 1981 ) . The f-TKD cell turnover model of the SG can be applied to stratified stacks of variously shaped polyhedral cells and provides a fundamental basis for the maintenance of barrier homeostasis during cell turnover in cornified stratified epithelia . Female 8- to 12-week-old C57B6/J mice were used as wild type in all experiments . To establish ZO-1-Venus transgenic mice , a transgenic vector was constructed with the involucrin promoter vector ( pH3700-pL2 ( Carroll and Taichman , 1992 ) , kindly provided by Dr . Lorne Taichman ) , which contained the first involucrin intron , an SV40 intron , a β-galactosidase gene , and an SV40 polyadenylation site . Mouse cDNA encoding the full length of ZO-1 and Venus ( Nagai et al . , 2002 ) cDNA ( kindly provided by Dr . Atsushi Miyawaki ) was replaced with the β-galactosidase gene in the involucrin promoter vector . The transgenic vector was injected into the fertilized egg from C57B6/J male and F1 female of C57B6/J × C3H . Transgenic mice were screened by direct observation of mouse-ear skin using fluorescence microscopy and the involucrin-promoter-driven Venus-ZO-1 mouse line was established by crossing with Balb/c mice more than eight times ( RRID:MGI:5805289 ) . All animal protocols were approved by the Animal Ethics Review Board of Keio University and conformed to the National Institutes of Health guidelines . Mouse skin samples were embedded in optimal cutting temperature compound ( Sakura Finetek , Japan ) , frozen in liquid nitrogen , and sectioned using a cryostat , as described previously ( Yoshida et al . , 2013 ) . The frozen sections were processed immediately by incubation in 95% ethanol at 4°C for 30 min , followed by 100% acetone at room temperature for 1 min and immunostained as described previously ( Yokouchi et al . , 2015 ) . Epidermal sheets were prepared from the ventral side of mouse-ear skin and immunostained , as described previously ( Yokouchi et al . , 2015 ) . For the preparation of ETA-induced SC-SG sheets ( Figure 2—figure supplement 1 ) , 100 μL of 44 µg/mL recombinant ETA ( Hanakawa et al . , 2002 ) in PBS containing 1 mM CaCl2 was injected intradermally into the ventral skin of the mouse ear and incubated for 30 min at 37°C . The ETA-induced SC-SG sheets were stripped away from the ventral skin , fixed by incubation in 95% ethanol on ice for 30 min , and immunostained , as described previously ( Yokouchi et al . , 2015 ) . Cell sheet samples were mounted in a whole-mount fashion using Mowiol ( Millipore , Germany ) . Samples were observed under a Leica TCS sp5 laser scanning confocal microscope equipped with a 63× objective using 0 . 4–0 . 5 µm optical slices . 3D reconstruction images were built using Leica sp5 software and Imaris software ( Bitplane , Switzerland ) . Images and movies were processed using Adobe Photoshop CS6 , Adobe Illustrator CS6 , and Apple QuickTime Pro . The following primary antibodies were used: polyclonal antibodies against claudin-1 ( ab15098; Invitrogen , Carlsbad , CA , RRID:AB_301644 ) , intracellular portion of desmoglein1 ( sc20114; Santa Cruz Biotechnology , Dallas , TX , RRID:AB_2293011 ) , angulin-1 ( lipolysis-stimulated lipoprotein receptor ) ( Masuda et al . , 2011 ) at a 1:200 dilution , monoclonal antibodies against ZO-1 ( T8-754; kindly provided by Dr . Masahiko Itoh ) ( Itoh , 1991 ) at a 1:10 dilution , tricellulin ( kindly provided by Dr . Sachiko Tsukita ) ( Ikenouchi et al . , 2005 ) , and occludin ( MOC37 ) ( Saitou et al . , 1997 ) . The extracellular portion of desmoglein one was detected by single-chained scFv ( 3–30/3 hr ) ( Ishii et al . , 2008; Yoshida et al . , 2013 ) . Species-specific secondary antibodies and streptavidin-labeled Alexa Fluor 488 , 568 , and 647 ( Invitrogen ) were used for detection at a 1:200 dilution . Cell nuclei were stained with Hoechst 33258 ( Invitrogen ) . A TJ permeation assay was performed according to a modified version of the procedure described previously ( Yokouchi et al . , 2015 ) . For the permeation assay with a protein biotinylation reagent ( Figure 2—figure supplement 1 ) , 30 μL of 10 μg/mL recombinant ETA ( Hanakawa et al . , 2002 ) in PBS containing 1 mM CaCl2 was injected into the dermis on the ventral side of the mouse ear , followed by injection of 50 μL of 10 mg/mL Sulfo-NHS-LC-Biotin ( 556 Da , #21335; Thermo Fisher Scientific , Waltham , MA ) 10 min later . After a 30-min incubation , skin samples were biopsied and embedded in optimal cutting temperature compound ( Sakura Finetek ) . For the ETA permeation assay , 50 μL of 10 μg/mL recombinant ETA in PBS containing 1 mM CaCl2 was injected intradermally on the ventral side of adult mouse ears ( Hanakawa et al . , 2002; Yoshida et al . , 2013 ) . After 30 min , the skin was biopsied and the ETA-induced bulla roof was stripped away and fixed by incubation in 95% ethanol on ice for 30 min and immunostained . The barrier function of the interior polygons against ETA permeation was quantitatively analyzed as follows . ETA-induced bulla roof was immunostained for ZO-1 , the extracellular portion of Dsg1 , angulin-1 and nuclei in five independent studies . Using a confocal microscope , 20 square en face images ( 15 , 376 µm2 ) were taken from each bulla roof of five mice . To quantitatively compare the single- and double-edged polygons , the relative fluorescent intensity of the extracellular portion of Dsg1 at each pixel was determined in the en face image by calculating the ratio of its fluorescence against the total average fluorescent intensity of the image . The fluorescent intensity of each polygon was determined by its average at the center area ( a circle of 320 µm2 ) . ETA-induced SC–SG cell sheets were prepared as described above . The cell sheets were floated on 500 µL of a 1:1 mixed solution of 0 . 5 g/L trypsin/0 . 53 mmol/L EDTA solution ( 32778–34 , Nacalai Tesque , Japan ) and 2 . 5 g/L trypsin solution ( 35555–54 , Nacalai Tesque ) and incubated at 37°C for 20 min . Cells were dissociated from the SC by pipetting . The trypsin solution containing SG cells was diluted with 1 mL of PBS , passed through a cell strainer ( 100 µm diameter pores , 352360 , Corning , Corning , NY ) , collected in a 15-mL conical tube ( Corning ) , and centrifuged at 180 G for 5 min at room temperature . The pellet was fixed in 10 mL of 95% ethanol on ice for 30 min . The fixed cells were pelleted by centrifugation at 720 G at room temperature for 5 min , dissociated in 1 mL of PBS , collected on a slide glass using cytospin ( Thermo Fisher Scientific ) at 450 G at room temperature for 5 min , and immunostained as described above . The shape of the isolated SG2 cells was visualized by immunostaining for the cytoplasmic portion of Dsg1 ( i . e . , staining for the precursor proteins in the endoplasmic reticulum and cytoplasmic vesicles and for the trypsin-digested proteins in the remaining desmosomes and endosomes ) . Three-dimensional rendering of the immunostained SG2 cells was performed by Imaris software ( Bitplane ) . ZO-1 Venus mice were anesthetized with isoflurane ( AbbVie , North Chicago , IL ) and oxygen and air; the isoflurane concentration was initially set at 4% and gradually lowered to 1 . 2% in a constant oxygen flow ( 0 . 15 L/min ) . To prevent movement , the dorsal side of the mouse ear was fixed to the table with double-sided adhesive tape . The two-photon microscope is a custom-made upright microscope ( BX61WI , Olympus , Japan ) attached to a mode-locked titanium-sapphire laser system ( Chameleon Vision II , Coherent , Santa Clara , CA ) that achieves a 950 nm laser with a 140-fs pulse width and an 80-MHz repetition rate ( Morikawa et al . , 2012 ) . Images ( 512 × 512 pixels ) were acquired by z-stack scanning at 0 . 6 μm intervals with a 25× objective lens ( XLPLN25 × WMP; NA 1 . 05 , Olympus ) and an Olympus FV1000 scanning unit using Fluoview software ( FV10-ASW , Olympus ) . Emitted fluorescence was detected using an external photomultiplier tube ( R3896; Hamamatsu Photonics , Japan ) after reflection via a dichroic mirror ( 580 nm cut-off ) and passing through an emission filter ( 500–550 nm ) . Acquired images were processed using Imaris software ( Bitplane ) . Epidermal sheets of mouse-ear skin were immunostained for ZO-1 in five independent assays as described above . Using the confocal microscope , 20 square en face images of 15 , 376 µm2 were taken from each sample . The number of single- and double-edged polygons was counted manually . The percentage of double-edged polygons was determined as the average of five independent assays . The size of single- and double-edged polygons ( Figure 1—figure supplement 1 ) was measured using ImageJ software ( NIH ) , for 10 square en face images ( 15 , 376 µm2 ) that included 8–10 TJ polygons with their whole edges . Comparison of the Z-axis relative position between polygons was performed on the double-edged polygons surrounded by six single-edged polygons for 55 square en face images ( 15 , 376 µm2 ) . The Z-axis position of a polygon was defined as the average of the Z-axis positions of its vertices analyzed by Imaris software ( Figure 1—figure supplement 2 ) . The relative Z-axis position of a polygon ( Figure 1E ) was calculated with respect to the average of the Z-axis position for a set of eight polygons ( an exterior and an interior polygon , and their six adjacent single-edged polygons ) . We used Prism six software ( GraphPad , La Jolla , CA ) for all statistical analyses . We developed a 3D in silicof-TKD turnover model consisting of stacked layers of optimally packed SG cells , each of which was an f-TKD . TJ-bearing SG2 cells were shown in three different colors depending on the vertical placement of the cell . A cycle of the SG2-to-SG1 transition of a cell starts with the appearance of the tTJs and is completed by their disappearance . In our model simulation , the SG2-to-SG1 transition of a cell is demonstrated by the gradual change of an SG2 cell to a transparent SG1 cell . The transition of a cell is governed by two local rules: ( 1 ) The differentiation of a SG3 cell to become a SG2 cell is synchronized with the differentiation of the pre-existing SG2 cell to become a SG1 cell in each cell column . ( 2 ) An SG2 cell located higher than its six adjacent SG2 cells differentiates to become an SG1 cell after a waiting period that is stochastically determined with a uniform probability between 0 and 9 . 6 hr . This uniform distribution was chosen to best reproduce the average turnover time ( 24 hr ) , estimated from the experimentally observed cell turnover rate in mouse ear skin ( Potten , 1975 ) and the percentage of double-edged polygons ( 9 . 8% , Figure 1B ) . The average time required to complete the TJ disappearance during SG2-to-SG1 transition was set to be 2 . 4 hr = 24 × 9 . 8/100 . Additionally , the following two rules for TJ production were implemented in the algorithm: ( 1 ) Bicellular TJs ( green ) are produced on edges shared by two SG2 cells when edges are in contact with SG1 cells; ( 2 ) tTJs ( blue ) are produced on edges shared by three SG2 cells . In our simulations , tTJs on double-edged polygons ( where TJ replacement takes place ) are displayed , but tTJs at vertices of single-edged polygons are not displayed for the sake of simplicity ( Figure 5—figure supplement 1 , Videos 9 and 10 ) . All simulations were conducted using Matlab version R2015b ( The MathWorks , Natick , MA ) .	The skin surface – known as the epidermis – is made up of sheets of cells that are stacked up in layers . One of the roles of the skin is to provide a protective barrier that limits what leaks into or out of the body . A particular layer of the epidermis – referred to as the stratum granulosum – is primarily responsible for forming this barrier . The cells in this layer are sealed together in a zipper-like fashion by structures known as tight junctions . New skin cells are continuously produced in the lowest cell layers of the epidermis , and move upwards to integrate into the stratum granulosum layer to replace old cells ( which also move upwards to leave the layer ) . How stratum granulosum cells are replaced without disrupting the tight junction barrier was not well understood . Yokouchi et al . used a technique called confocal microscopy to examine the stratum granulosum cells in the ears of mice , and found that the shape of these cells forms the basis of the barrier that they form . These cells resemble a flattened version of a shape called Kelvin’s tetrakaidecahedron: a 14-sided solid with six rectangular and eight hexagonal sides . This structure was proposed by Lord Kelvin in 1887 to be the best shape for filling space . Tight junctions are present on the edges of the flattened Kelvin’s tetrakaidecahedron . Further experiments revealed that the tight junctions move from cell to cell in a spatiotemporally-coordinated manner in order to maintain a continuous barrier throughout the stratum granulosum as cells are replaced . A newly formed stratum granulosum cell appears beneath the cell that it will replace . The shape of these cells enables a new barrier of three-way tight junction contacts to form between them and the neighboring cells in the stratum granulosum . After this barrier has formed , the upper cell leaves the stratum granulosum . Future research could address how cells adopt the flattened Kelvin’s tetrakaidecahedron shape , and discover why tight junctions only form in one layer of the epidermis .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cell", "biology" ]	2016	Epidermal cell turnover across tight junctions based on Kelvin's tetrakaidecahedron cell shape
Sleep has been conserved throughout evolution; however , the molecular and neuronal mechanisms of sleep are largely unknown . The hypothalamic hypocretin/orexin ( Hcrt ) neurons regulate sleep\wake states , feeding , stress , and reward . To elucidate the mechanism that enables these various functions and to identify sleep regulators , we combined fluorescence cell sorting and RNA-seq in hcrt:EGFP zebrafish . Dozens of Hcrt-neuron–specific transcripts were identified and comprehensive high-resolution imaging revealed gene-specific localization in all or subsets of Hcrt neurons . Clusters of Hcrt-neuron–specific genes are predicted to be regulated by shared transcription factors . These findings show that Hcrt neurons are heterogeneous and that integrative molecular mechanisms orchestrate their diverse functions . The voltage-gated potassium channel Kcnh4a , which is expressed in all Hcrt neurons , was silenced by the CRISPR-mediated gene inactivation system . The mutant kcnh4a ( kcnh4a-/- ) larvae showed reduced sleep time and consolidation , specifically during the night , suggesting that Kcnh4a regulates sleep . Sleep is a fundamental behavior that benefits the brain and sleep disorders affect a large portion of the world’s population ( Shepard et al . , 2005 ) . Thus , it is essential to identify and understand the role of the neuronal circuits and genes that regulate sleep . The hypothalamus centralizes sleep regulation and maintains essential physiological processes , including growth , reproduction , body temperature , stress , reward , feeding , and circadian rhythms ( Salin-Pascual et al . , 2001; Mignot et al . , 2002; Saper et al . , 2005; Saper , 2006; Begg and Woods , 2013; Dietrich and Horvath , 2013; Sternson , 2013; Muindi et al . , 2014 ) . These functions are mediated by several hypothalamic nuclei that interact with various neuronal networks . Some of these nuclei , such as the suprachiasmatic nucleus ( SCN ) , which is the master circadian oscillator ( Welsh et al . , 2010 ) , have been well characterized both anatomically and physiologically , while the neuronal identity and function of other nuclei is less understood ( Rolls et al . , 2010; Sakurai and Mieda , 2011; Dietrich and Horvath , 2012 ) . The hypocretin ( Hcrt , also called orexin ) neurons secrete the Hcrt neuropeptides and are located in the lateral hypothalamus ( LH ) . These hypothalamic neurons project to wide areas in the brain , including the tuberomammillary nucleus , paraventricular thalamic nucleus , arcuate nucleus , and monoaminergic nuclei ( Sakurai , 2007 ) . They were initially implicated in feeding behavior and sleep\wake cycles ( de Lecea et al . , 1998; Sakurai et al . , 1998 ) . Their role in sleep regulation was further strengthened since loss of Hcrt neurons causes the sleep disorder narcolepsy , which is characterized by sleep\wake fragmentation , increased body mass , and cataplexy ( loss of muscle tone , often triggered by emotional stimuli ) ( Chemelli et al . , 1999; Lin et al . , 1999; Peyron et al . , 2000; Schuld et al . , 2000; Adamantidis et al . , 2007; Burgess et al . , 2012 ) . However , extensive research showed that the function of Hcrt neurons is much broader and also includes regulation of energy homeostasis , pain , emotion , stress response , and reward ( Sakurai , 2007; Rolls et al . , 2010; Boutrel et al . , 2013 ) . The Hcrt neurons regulate this variety of brain functions through interactions with peptide-secreting neurons and with the monoaminergic , dopaminergic , and limbic systems , among others ( Tsujino and Sakurai , 2013 ) . How do Hcrt neurons serve as a multifunctional hypothalamic system ? Clearly , secretion of the neuropeptide Hcrt is a key pathway . A single hcrt gene encodes for the precursor polypeptide prepro-Hcrt , which is cleaved to produce two Hcrt neuropeptides . The actions of the Hcrt neuropeptides are mediated via two Hcrt G-protein–coupled receptors ( Hcrtrs ) ( Sakurai , 2007 ) . In addition , the synaptic release of glutamate from Hcrt neurons has been shown to affect the activity of post-synaptic target neurons ( Henny et al . , 2010; Schone et al . , 2012 ) . However , Hcrt neurons contain additional proteins that are likely involved in mediating their development , plasticity , and diverse functions . To date , only a few Hcrt-neuron–specific genes were substantially characterized and , except for hcrt , none of them are exclusively expressed in Hcrt neurons ( Meister and Håkansson , 1998; Broberger , 1999; Reti et al . , 2002; Blouin et al . , 2005; Crocker et al . , 2005; Belcher et al . , 2006; Florenzano et al . , 2006; Appelbaum et al . , 2007; Honda et al . , 2009; Silva et al . , 2009; Appelbaum et al . , 2010; Cvetkovic-Lopes et al . , 2010; Dalal et al . , 2013; Liu et al . , 2015 ) . A comprehensive and specific gene-expression profiling of Hcrt neurons will enhance the understanding of Hcrt neuronal networks and its diverse functions . Three studies have described the gene-expression profile of Hcrt neurons in rodents ( Honda et al . , 2009; Cvetkovic-Lopes et al . , 2010; Dalal et al . , 2013 ) . First , RNA array was used to study the effect of loss of Hcrt neurons on the expression of hypothalamic transcripts in Hcrt-neuron–ablated mice ( Honda et al . , 2009 ) . Later , using affinity purification of RNAs and transgenic mice that express FLAG-tagged poly ( A ) -binding protein , specifically in Hcrt neurons , polyadenylated mRNA was isolated and classified ( Cvetkovic-Lopes et al . , 2010 ) . Finally , the translating ribosome affinity purification technique that targets HCRT-producing neurons , was used to isolate Hcrt cell-specific RNA in mice ( Dalal et al . , 2013 ) . These extensive studies resulted in a list of genes expressed in Hcrt neurons . However , in the opaque mammalian brain , isolation of the entire Hcrt neuron population is challenging because a few thousand Hcrt cells are intermingled with other hypothalamic neurons . In addition , all studies used microarray technology , which limits gene resolution and requires a priori knowledge of transcript content . The zebrafish has become a valuable model for the study of specific neuronal populations in live animals . It is a simple and diurnal vertebrate that combines powerful genetic tools with conserved anatomy and function of the brain ( Severi et al . , 2014; Thiele et al . , 2014; Romano et al . , 2015 ) . In the last two decades , behavioral criteria have been used to characterize sleep in zebrafish ( Zhdanova et al . , 2001; Prober et al . , 2006; Yokogawa et al . , 2007; Sigurgeirsson et al . , 2011; Elbaz et al . , 2012 ) . Similar to mammals , the Hcrt neurons are located in the zebrafish hypothalamus but , in contrast to mammals , the zebrafish Hcrt system contains only a few neurons , making it a relatively simple system to study ( Kaslin et al . , 2004; Faraco et al . , 2006 ) . Functional studies using Hcrt-neuron–specific genetic ablation , as well as genetic manipulation of the hcrt ligand and receptors , showed that the Hcrt system regulates sleep and wake in zebrafish ( Prober et al . , 2006; Yokogawa et al . , 2007; Elbaz et al . , 2012 ) . In addition , the zebrafish Hcrt neurons induce feeding behavior ( Yokobori et al . , 2011 ) , as is the case in mammals . Recently , in order to study Hcrt-neuron specification , a screen for regulatory factors was conducted in the early stages of zebrafish development [26 hr post-fertilization ( hpf ) , ( Liu et al . , 2015 ) ] . Similar to mammals ( Dalal et al . , 2013 ) , microarray gene-expression analysis revealed that the LIM homeobox transcription factor Lhx9 , which is widely expressed in the brain , including in the Hcrt neurons , can induce the specification of Hcrt neurons ( Liu et al . , 2015 ) . In the present work , we used 7-days-post-fertilization ( dpf ) transgenic zebrafish larvae expressing EGFP under the promoter of hcrt ( Appelbaum et al . , 2009 ) , to identify genes that regulate Hcrt-neuron function . The hcrt:EGFP larvae were used to specifically isolate Hcrt neurons by fluorescence-activated cell sorting ( FACS ) . Using whole transcriptome RNA sequencing ( RNA-seq ) , meticulous bioinformatic analysis , and extensive anatomical validations , a novel set of Hcrt-neuron–specific genes was identified . Furthermore , the role of the voltage-gated potassium channel Kcnh4a in regulating sleep architecture was studied . In order to isolate the Hcrt neurons , the transgenic hcrt:EGFP zebrafish ( Appelbaum et al . , 2009 ) , which enables specific visualization and manipulation of the entire population of Hcrt neurons ( 16-–20 cells per larva ) , was used . At 7 dpf , the heads of hcrt:EGFP larvae ( Figure 1A , B ) were dissociated , and EGFP-positive ( EGFP+ ) cells from the cell suspension sample ( Figure 1C ) were sorted by FACS ( Figure 1D–G ) . The sorting thresholds were set to accurately detect the small amounts of cells expressing EGFP while avoiding the auto-fluorescent cells derived primarily from the eyes of the larvae ( Figure 1B ) . In order to calibrate the threshold and additional FACS parameters , α-tubulin:EGFP-injected larvae ( Figure 1E ) , which expressed EGFP in the entire central nervous system ( CNS ) , were also FAC-sorted . To avoid off-target sorting of EGFP-negative ( EGFP- ) cells and to set the threshold of EGFP+ cells , we applied the same parameters and filters to a cell suspension sample derived from wild-type ( WT ) larvae ( Figure 1F ) . As expected , the number of EGFP+ cells sorted from hcrt:EGFP larvae ( Figure 1G ) was low compared with the number of cells sorted from α-tubulin:EGFP-injected larvae . EGFP+ cells were not detected in WT larvae ( Figure 1F ) . Using this technique , we collected 300 EGFP+ and 300 EGFP- cells from hcrt:EGFP larvae in three independent experiments . To verify that the EGFP+ cells were Hcrt neurons , RNA extraction was performed , followed by reverse transcription PCR ( RT-PCR ) assays . While hcrt and egfp were detected in EGFP+ cells , they were not amplified in EGFP- cells ( Figure 1H ) . These results show that the EGFP+ cells mostly contain Hcrt neurons , while the EGFP- group contains a heterogeneous population of cells from the whole larva head . Since the amount of RNA extracted from 300 cells was extremely low ( below 1 pg/µl ) and required amplification before deep sequencing , RNA was extracted from a third control group of cells derived from a whole head of 7 dpf WT larvae . This group helped to distinguish false positive signals that might have resulted from the amplification , and covered genes that were widely expressed in the head and not restricted to 300 EGFP- cells . The RNA of the three groups: EGFP+ , EGFP- , and whole head , was subjected to RNA-Seq and bioinformatic analysis ( Figure 1I ) to obtain a list of Hcrt-neuron–enriched genes . 10 . 7554/eLife . 08638 . 003Figure 1 . Isolation of Hcrt neurons , RNA-seq , and experimental design . ( A ) Dorsal view of 6 dpf hcrt:EGFP larvae . ( B ) Dorsal view of the hypothalamus region of 6 dpf hcrt:EGFP larvae expressing EGFP in Hcrt neurons . ( C ) Cell suspension from the whole head of 6 dpf hcrt:EGFP larvae . ( D–G ) The cells were sorted based on size and fluorescence intensity . The fluorescence thresholds ( gray curve ) were set based on larvae expressing EGFP under the control of α-tubulin promoter ( positive control ) ( E ) and WT larvae ( negative control ) ( F ) . Positive EGFP cells ( EGFP+ ) sorted from hcrt:EGFP larvae are marked with gray shade ( G ) . ( H ) PCR amplification of hcrt and egfp was performed on cDNA synthesized from EGFP+ and EGFP- cells sorted from hcrt:EGFP larvae . ( I ) FAC-sorting yielded two groups of cells: Group I containing EGFP+ and Group II containing EGFP- cells . A third group contained cells from whole head of WT larvae . The cDNA of groups I and II was amplified and the three groups were then subjected to RNA-seq and bioinformatic analysis to obtain a list of Hcrt-neuron–enriched genes . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 003 We aimed to identify novel players that regulate the myriad of processes coordinated by Hcrt neurons . Thus , the RNA-seq data from EGFP+ , EGFP- , and whole-head groups ( http://www . ncbi . nlm . nih . gov/sra , PRJNA283169 ) were analyzed in silico . Initially , the raw read counts were normalized to transcripts per million ( TPM ) , and a gene was considered to be preferentially expressed in the Hcrt cells only if its normalized expression level was at least 100 TPM in EGFP+ cells . In addition , the expression levels were required to be at least 7 times more abundant in the EGFP+ than in both controls . These criteria stipulated a high level of specificity to the EGFP+ samples relative to the control samples . The bioinformatic analysis identified 20 transcripts that were found to meet these criteria ( p<0 . 01 , Figure 2A ) . Among the 20 transcripts , 12 were annotated genes and 8 were non-annotated transcripts . Notably , the hcrt gene was expressed at a level of 300 TPM in EGFP+ and below 10 TPM in both control samples . The identification of an hcrt gene confirmed the specificity of the cell sorting , the RNA-seq , and the bioinformatic analysis . 10 . 7554/eLife . 08638 . 004Figure 2 . The expression pattern of selected candidate Hcrt-neuron–specific genes . ( A ) Table presenting the top 20 Hcrt-enriched transcripts . ( B–M ) Dorsal view of whole-mount ISH-stained 2 dpf WT larvae . Based on the RNA-seq and the bioinformatic analysis , the expression pattern of selected candidate Hcrt-neuron–specific genes was determined . The expression pattern of hcrt ( B ) was used for comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 00410 . 7554/eLife . 08638 . 005Figure 2—source data 1 . Hcrt-neuron enriched transcripts . All the genes detected have p <0 . 01 with Bonferroni correction for multiple testing ( Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 005 In order to validate the bioinformatic results and to determine the spatial expression pattern of the candidate genes , whole-mount in situ hybridization ( ISH ) was performed on 2 dpf WT larvae ( Figure 2B–M ) using gene-specific probes for the enriched genes ( Figure 2A ) . Nine of them were found to be expressed in the hypothalamic area ( hcrt , star , dennd1b , kcnh4a , fam46a , si:dkey-58b18 . 8 , cuff23873 . 1 , npvf , and npffr; Figure 2B–M ) . Five transcripts ( adra , ptgs2b , grpr , cuff64723 , and cuff77494 , ) could not be amplified , and the expression of six transcripts ( elovl7b , cuff34876 , cuff70256 , cuff442204 , cuff57637 , and cuff77484 ) was not detected at the 2 dpf larval stage . To test whether these genes were expressed in later developmental stages , their expression was studied in adults . However , only elovl7b showed a detectable expression in the hypothalamus ( Figure 3J–J" ) . The hypothalamic expression pattern of the candidate genes was similar to the expression pattern of hcrt ( Figure 2B ) , suggesting that the candidate genes may be expressed in Hcrt neurons . 10 . 7554/eLife . 08638 . 006Figure 3 . Selected candidate genes are expressed in Hcrt neurons . ( A–I'' ) Double fluorescent staining of the candidate genes ( red ) and EGFP ( green ) was performed in 2 dpf hcrt:EGFP larvae using whole-mount ISH and immunofluorescence , respectively . White arrows indicate representative co-expressing cells . All confocal images show single plane view of 0 . 5 µM width . ( J–M'' ) Double fluorescent ISH and immunofluorescence experiments in brain sections of hcrt:EGFP adult zebrafish . Co-localization of candidate genes ( red ) and EGFP ( green ) in Hcrt neurons is shown . All images show single plane view of 0 . 5 µM width . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 006 The high percentage of genes that showed hypothalamic expression hints at significant efficiency of the FACS and RNA-seq experiments . Thus , in order to find more Hcrt-neuron–specific genes , we relaxed the bioinformatic parameters to 10 TPM and 3 . 6 times higher abundance in EGFP+ cells than in the control groups . This analysis revealed 212 transcripts that met the criteria ( p<0 . 01 , Figure 2—source data 1 ) , among them , 146 were non-annotated ( called cuff-serial number ) and 66 were annotated genes . The functional roles of the annotated genes are diverse and include , for example , regulation of metabolism [such as ELOVL fatty acid elongase 7b ( elovl7b ) ] , sleep ( lhx9 ) , synaptogenesis and synaptic plasticity [such as the guanine nucleotide exchange gene ( denndbl ) ] . Some of the non-annotated transcripts were likely long , non-coding RNA ( lncRNA ) since they were longer than 200 bp , did not include a coding sequence , and were located in intergenic regions ( Perkel , 2013 ) . lncRNAs regulate transcription and epigenetic processes and may be involved in the regulation of splicing and translation ( Mercer et al . , 2009 ) . Notably , some non-annotated transcripts were located in the zebrafish genome near Hcrt-enriched genes . In addition to the 8 genes tested ( Figure 2C–J ) , we attempted to examine the expression of selected candidate genes that demonstrate relatively lower enrichment in Hcrt neurons ( Figure 2—source data 1 ) . We selected zgc:171844 , H6 homeobox 3 ( hmx3 ) , and lhx9 , which were located in the bottom 100 genes in the list ( Figure 2—source data 1 ) . Previous work showed that lhx9 is expressed in Hcrt neurons in mammals and zebrafish ( Dalal et al . , 2013; Liu et al . , 2015 ) and that hmx3 is expressed in Hcrt neurons in the early stages of zebrafish development ( Liu et al . , 2015 ) . Similar to the genes that demonstrated high fold change ( Figure 2A ) , these three genes were also expressed in the hypothalamus , where hcrt is expressed ( Figure 2K–M ) , suggesting that a large portion of the 212 transcripts ( Figure 2—source data 1 ) may be expressed in the Hcrt neurons . Single-probe ISH analysis showed that selected candidate transcripts are expressed in the hypothalamus and that their spatial expression pattern is reminiscent of the expression of the hcrt gene ( Figure 2 ) . To test whether these transcripts are expressed in Hcrt neurons , we performed whole-mount fluorescent ISH using probes for the candidate genes , coupled with immunofluorescence staining using EGFP antibody , in hcrt:EGFP 2 dpf larvae and adults . To verify the efficiency and specificity of this assay , co-localization of hcrt and EGFP was initially confirmed ( Figure 3A–A’' ) . Double staining showed that among the 11 transcripts tested , 8 transcripts ( star , dennd1b , kcnh4a , fam46a , hmx3 , zgc171844 , lhx9 , and si:dkey-58b18 . 8 ) co-localized with EGFP in Hcrt neurons ( Figure 3B–I'' ) . While kcnh4a , hmx3 , lhx9 and dennd1b were expressed in most Hcrt neurons , star , fam46a , and zgc171844 , were expressed in a subset of the Hcrt neurons . In addition to their expression in Hcrt neurons , these transcripts were also expressed in other brain regions , particularly other hypothalamic areas and the forebrain . In contrast , si:dkey-58b18 . 8 demonstrated relatively weak expression that was predominantly apparent in Hcrt neurons ( Figure 3E–E'' ) . Further anatomical analysis in hcrt:EGFP adult brain sections was performed on four transcripts ( Figure 3J–M'' ) : elovl7b , which did not show expression in the earlier developmental stages ( Figure 3J–J'' ) , kcnh4a ( Figure 3K–K'' ) , dennd1b ( Figure 3L–L'' ) , and zgc171844 ( Figure 3M–M'' ) . Double staining in adults showed that kcnh4a and elovl7b are detected in all Hcrt neurons , while dennd1b is expressed in about half of the Hcrt neurons and zgc171844 in about a third of the neurons . Notably , the portion of co-localization with EGFP in larvae was similar to that in adults . Altogether , the anatomical results validated the RNA-seq and bioinformatic analysis , which provide a comprehensive list of Hcrt neuron-specific transcripts . The spatial expression of these transcripts in subpopulations of Hcrt neurons indicates that Hcrt neurons are not a uniform population but rather heterogeneous neurons . Understanding the function of these transcripts will provide the basis to elucidate the mechanism that regulates the multifunctions of Hcrt neurons . The histochemical assays revealed transcripts expressed in Hcrt neurons . However , three candidate transcripts ( cuff . 23873 , npvf , and npffr; Figure 2H–J ) labeled distinct hypothalamic populations of neurons that intermingled with Hcrt neurons , but co-localization was not detected ( Figure 4 ) . These cell populations were located in the immediate vicinity of the Hcrt neurons in the hypothalamus . While , cufff . 23873 ( Figure 4A–C'' ) and npffr ( Figure 4D–F'' ) were also expressed in the forebrain area , npvf ( Figure 4G–I'' ) showed a specific hypothalamic expression pattern . The finding of transcripts that were apparently not expressed in Hcrt neurons in the transcriptome , could be due to the adhesion of hypothalamic cells adjacent to Hcrt neurons during the FACS procedure , or because these transcripts are also expressed by Hcrt neurons but below ISH detection levels . Nonetheless , these transcripts are predominantly expressed in hypothalamic neurons and may interact with Hcrt neurons to form neuronal networks that mediate the functions of Hcrt neurons . 10 . 7554/eLife . 08638 . 007Figure 4 . Candidate genes are expressed in cell populations located adjacent to Hcrt neurons . ( A–I'' ) Fluorescent ISH and immunofluorescence experiments in 2 dpf larvae and adult hcrt:EGFP zebrafish showing three candidate genes that are expressed adjacently to Hcrt neurons within the hypothalamus . ( A–A'' , D–D'' , G–G'' ) Dorsal view of the heads showing the whole expression pattern of the genes in 40 µM z-stack . ( B–B'' , E–E'' , H–H ) Dorsal view of single 0 . 5 µM plane in 2 dpf larvae . ( C–C'' , F–F'' , I–I'' ) Dorsal view of single 0 . 5 µM plane in adult brain section . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 007 The mechanism that regulates the specific expression of transcripts in Hcrt neurons and the identity of the transcription factors ( TFs ) is unclear . To identify candidate TFs that can regulate multiple Hcrt-neuron–specific genes , a map of possible TF binding sites was generated based on the 48 most enriched transcripts . Conserved sequences in the predicted regulatory region of each Hcrt-neuron–specific gene were characterized , and the matched TFs that potentially bind to these sequences were identified . This analysis revealed 68 putative TFs ( Figure 5A and Figure 5—source data 1 ) , among them , 13 showed significant enrichment in the top 48 Hcrt-specific transcripts ( p<0 . 005 , Figure 5A ) including nr6a1 , which is a regulator of hcrt in mice ( Tanaka et al . , 2010 ) . Notably , this analysis suggests that several specific TFs regulate numerous Hcrt-neuron–specific genes ( Figure 5A ) . For example , the heat shock transcription factor 1 ( hsf1 ) is predicted to regulate 25 Hcrt-neuron–enriched genes: hcrt , ptgs2 , ttn , hspa1l , grpr , elovl7b , slc4a1 , lhx9 , c16orf45 , soat2 , tsen54 , nos1 , rfx4 , syt10 , trpc7 , ntng1 , cacng4 , myh4 , dennd1b , sgsm1 , pde2a , wscd1 , adra1a , kcnh4a , and hmx3 . To test whether this predicted TF is expressed in Hcrt neurons , fluorescent ISH , using a probe for hsf1 , and fluorescent immunostaining using an antibody against EGFP , were performed on the brain section of adult hcrt:EGFP zebrafish ( Figure 5B–C'' ) . This assay showed wide brain expression of hsf1 and confirmed that hsf1 is also expressed in Hcrt neurons . In mammals , hsf1 is a key activator of stress conditions , and the Hsf1 null mice showed major brain morphological alterations ( Santos and Saraiva , 2004 ) . In zebrafish , hsf1 is essential for recovery from ischemic injury in the brain ( Tucker et al . , 2011 ) . In addition to hsf1 , the TF binding-site analysis revealed the enrichment of TFs that regulate metabolic processes ( pax4 , hnf1 , ppara , lhx3 , creb , and foxo4 ) , such as control of the levels of glucagon , insulin , somatostatin , lipids , and glucose ( Peyron et al . , 2000; Reti et al . , 2002; Prober et al . , 2006; Plengvidhya et al . , 2007; Primo et al . , 2008; Reiner et al . , 2009 ) . In addition , ap-2 , a TF that is required for sleep-like behavior in C . elegans ( Turek et al . , 2013 ) , was predicted to regulate the transcription of 13 Hcrt neuron-specific genes ( Figure 5A ) . The identification of mutual TF binding sites in the regulatory sequences of Hcrt-neuron–specific genes suggests that several key TFs regulate the development and function of Hcrt neurons . 10 . 7554/eLife . 08638 . 008Figure 5 . Predicted TFs that regulate the expression of Hcrt-neuron–specific transcripts . ( A ) TFs with a p <0 . 005 and their target Hcrt-neuron–specific genes . For each of these transcription factors , a combined score for each gene is calculated according to all of the predicted binding sites in its promoter . Therefore , a gene with an overrepresented binding site of a TF will have a high score for that TF and a lower p value . ( B–C'' ) Double fluorescent ISH of hsf1 and immunofluorescence staining of EGFP in hcrt:EGFP adult brain section . Single plane ( 0 . 5 µM width ) view of the Hcrt-neuron region ( C–C'' ) . Arrows mark representative EGFP and hsf1 co-expressing cell . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 00810 . 7554/eLife . 08638 . 009Figure 5—source data 1 . Predicted Hcrt-neuron enriched transcription factors and their target genes . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 009 Among the candidate genes ( Figure 2—source data 1 ) , the voltage-gated potassium channel kcnh4a was of particular interest because of its genomic location , expression pattern , and predicted role . Two kcnh4 are present in zebrafish: kcnh4a ( KR733682 ) located in chromosome 3 and kcnh4b ( XM_690738 ) located in chromosome 12 . In contrast to the broad expression of kcnh4b ( data not shown ) , kcnh4a is expressed specifically in the forebrain and hypothalamus in larvae ( Figure 2E ) . Double ISH and immunofluorescence staining revealed that kcnh4a is localized in all Hcrt neurons in both larvae and adults ( Figure 3D–D'' , K–K'' ) and like Hcrt neurons , hypothalamic kcnh4a-expressing neurons are glutamatergic ( Figure 6—figure supplement 1 ) . Intriguingly , kcnh4a is localized only a few kilobase pairs ( kbp ) downstream to hcrt on the genome , and a synteny analysis showed that the genomic organization of this locus is conserved with mammals ( Figure 6A ) . In humans , kcnh4a is located only 3782 bp downstream to the hcrt gene , while in zebrafish , the distance between the genes is 5517 bp ( Figure 6A ) . Although kcnh4a expression is not restricted to Hcrt neurons , the genomic proximity of the two genes suggests a mutual transcription regulation . Indeed , a significant portion of the TFs predicted to regulate hcrt can also bind to kcnh4a regulatory sequences ( 48 out of 52 , Figure 5—source data 1 ) . 10 . 7554/eLife . 08638 . 010Figure 6 . The genomic location , phylogenetic reconstruction and structure of kcnh4a , and the generation of kcnh4a-/- zebrafish . ( A ) Synteny analysis shows similar genomic context of hcrt in zebrafish and mammals . Notably , kcnh4a is located a few kbs downstream to hcrt in zebrafish and mammals . ( B ) The 16-exon kcnh4a gene ( black box = exon , white box = UTR ) encodes for a voltage-gated potassium channel that includes an N-terminal chain ( black bar ) , pore and voltage-sensing domains ( S1-6 , grey bar ) , and the C-terminal chain ( red bar ) . ( C ) A cladogram-style phylogenetic tree depicting the evolutionary conservation of Kcnh4a protein among vertebrates . The tree shows topography as well as distance indicated by the branch support values above corresponding branches . ( D ) Generation of CRISPR-mediated kcnh4a-/- zebrafish . A 14 bp deletion was introduced in exon 5 that encodes to the S2 domain . A short mutant allele was visible on agarose gel . ( E ) Quantitative reverse transcription PCR shows reduction of 59% in the expression levels of kcnh4a mRNA in kcnh4a-/- 6 dpf larvae ( p <0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 01010 . 7554/eLife . 08638 . 011Figure 6—figure supplement 1 . Hypothalamic kcnh4a-expressing neurons are glutamatergic . Double in-situ hybridization against kcnh4a ( red , A , B ) and gad67 ( green , A' ) , or vglut2b ( green , B' ) . All pictures are on a single optical plane of 0 . 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 011 Vector cloning and sequencing of kcnh4a showed that the gene consists of 16 exons , which include the 3456 bp coding sequence ( CDS , Figure 6B ) . The start codon is located within the second exon , preceded by 1884 bp 5’ UTR . Exon 16 includes the stop codon , followed by 1286 bp 3’UTR . Structural bioinformatic analysis of the protein sequence revealed that the Kcnh4a contains the evolutionarily conserved six S domains that characterize the potassium voltage-gated ion channels ( Figure 6B ) ( Choe , 2002 ) . Domains S1–S4 constitute the voltage-gated domain that senses changes in membrane potential ( Bezanilla , 2000; Swartz , 2008 ) , whereas domains S5–S6 form the selectivity pore through which ions can flux ( Bezanilla , 2000; Choe , 2002; Swartz , 2008 ) . Next , phylogenetic analysis revealed that the zebrafish Kcnh4a protein is evolutionarily conserved with vertebrate orthologs ( Dereeper et al . , 2008 ) . As expected , the zebrafish Kcnh4a protein showed the highest homology to the Kcnh4 of another fish ( Larimichthys_crocea ) and , to a lesser extent , to the mammal Kcnh4 proteins ( Figure 6C ) . In order to test the function of Kcnh4a , we established a clustered , regularly interspaced , short palindromic-repeat ( CRISPR ) -based kcnh4a mutant zebrafish ( kcnh4a-/- ) . A 14 bp deletion mutation was generated in exon 5 , which encoded part of the pore domain . This deletion introduced a premature stop codon and is predicted to result in truncated protein ( Figure 6D ) . Furthermore , quantitative reverse transcription PCR ( qRT-PCR ) showed a reduction of 59% of kcnh4a mRNA levels in kcnh4a -/- compared to WT-sibling ( kcnh4a+/+ ) 6dpf larvae ( p<0 . 001 , Figure 6E ) . The founder ( F0 ) fish was outcrossed with WT fish , and experiments were performed on the progeny of inter-crosses between F4 heterozygous fish ( kcnh4a+/- ) . To study the rhythmic locomotor activity of kcnh4a-/- zebrafish , high-throughput video-tracking systems were used ( Elbaz et al . , 2012 ) . The locomotor activity of kcnh4a-/- ( n=85 ) , kcnh4a+/- ( n=209 ) , and kcnh4a+/+ ( n=98 ) was monitored during day and night ( 14 hr light/10 hr dark ) . As expected , larvae from all three genotypes exhibited rhythmic locomotor activity that peaked during the day ( F[2 , 180] = 14 . 98; p<0 . 0001 , mixed-effect model with repeated measures; Figure 7A-A’ ) . Notably , kcnh4a-/- larvae were slightly hyperactive ( average: 13 . 84 ± 0 . 11 ) compared to kcnh4a+/- ( average: 13 . 29 ± 0 . 07 , t = -4 . 19 , df = 180 , p<0 . 01 ) and kcnh4a+/+ sibling larvae during the day ( average: 12 . 98 ± 0 . 1 , t = 5 . 73 , df = 180 , p<0 . 0001 ) . During the night , the differences in locomotor activity were even lower , and the kcnh4a-/- larvae were slightly more active ( average: 8 . 05 ± 0 . 13 ) than the kcnh4a+/+ larvae ( average: 7 . 64 ± 0 . 12 , t = 2 . 28 , df = 180 , p<0 . 05; Figure 7A-A’ ) . These results show that the loss of Kcnh4a mildly increases larval locomotor activity , particularly during the day . 10 . 7554/eLife . 08638 . 012Figure 7 . Sleep time and quality are reduced in kcnh4a-/- larvae during the night . ( A ) The locomotor activity of kcnh4a-/- ( n=85 ) , kcnh4a+/- ( n=208 ) , and kcnh4a+/+ ( n=98 ) is shown . kcnh4a-/- larvae exhibit increased locomotor activity compared with kcnh4a+/- and kcnh4a / under LD conditions . ( B ) kcnh4a-/- larvae showed a significant reduction in sleep time compared with kcnh4a+/- and kcnh4a+/+ during the night . Bar charts represent the average total locomotor activity ( A' ) and sleep time ( B' ) for each genotype . Values are represented as means ± SEM . ( C , D ) The number of sleep\wake transitions ( C ) and the length of sleep bout ( D ) are decreased in kcnh4a-/- larvae during the night . Recording of locomotor activity and sleep was performed in 6 dpf larvae continuously during 24 hr under a 14 hr light/10 hr dark cycle ( white and black bars represent light and dark periods , respectively , p<0 . 05 , p<0 . 01 , *p<0 . 0001 , with repeated measures of ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 01210 . 7554/eLife . 08638 . 013Figure 7—figure supplement 1 . Sleep time is increased following sleep deprivation ( SD ) . ( A ) At 6dpf , larvae were sleep deprived for 6 hr during the night under constant dark conditions ( DD ) and sleep time was monitored in the following nine hours . ( B , C ) Sleep was recovered in sleep-deprived larvae . Statistical comparisons were performed using Student’s t-tests ( p<0 . 05 ) . Dark and gray horizontal bars represent night and subjective day , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 08638 . 013 Similar to humans , the zebrafish is a diurnal vertebrate that sleeps during the night ( Zhdanova , 2011; Elbaz et al . , 2013 ) . Using well-established behavioral criteria , sleep in larvae was defined as a period of one or more minutes of immobility , which is associated with an increase in arousal threshold ( Prober et al . , 2006; Elbaz et al . , 2012 ) . A previous study has shown that six hours of sleep deprivation ( SD ) during the night reduced locomotor activity in the following day ( Zhdanova et al . , 2001 ) . Similarly , six hours of SD during the night increased sleep time during the following day in 6 dpf larvae ( Figure 7—figure supplement 1 ) . Thus , similar to mammals , sleep in zebrafish larvae is regulated by circadian and homeostatic processes . Voltage-gated potassium channels are activated by membrane depolarization and contribute to neuronal repolarization and repetitive firing ( Choe , 2002 ) . Considering this role and the expression of kcnh4a in all Hcrt neurons , we tested whether it regulates sleep and wake . Similar to humans , zebrafish are diurnal animals; thus , all three genotypes ( kcnh4a-/- , kcnh4a+/- , and kcnh4a+/+ ) slept more during the night than during the day ( F[2 , 180] = 14 . 52; p<0 . 0001 , mixed-effect model with repeated measures , Figure 7B ) . Remarkably , while the amount of sleep was similar in all genotypes during the day ( average: kcnh4a-/- = 2 . 76 ± 0 . 22; kcnh4a+/- = 2 . 80 ± 0 . 14; and kcnh4a+/+ = 2 . 87 ± 0 . 21 ) , sleep time was reduced in kcnh4a-/- larvae compared with kcnh4a+/- and kcnh4a+/+ larvae during the night ( average: kcnh4a-/- = 13 . 08 ± 0 . 27; kcnh4a+/- = 14 . 78 ± 0 . 17; and kcnh4a+/+ = 15 . 46 ± 0 . 25 , t = -6 . 55 , df = 180 , p<0 . 0001 , Figure 7B , B' ) . In order to examine the sleep architecture , we quantified the number of sleep\wake transitions and the length of sleep bouts . While the number of sleep\wake transitions did not change during the day ( kcnh4a-/- = 3 . 52 ± 0 . 16; kcnh4a+/+ = 3 . 79 ± 0 . 15 ) , their number was decreased in kcnh4a-/- larvae during the night ( average: kcnh4a-/- = 12 . 68 ± 0 . 19; kcnh4a+/+ = 13 . 62 ± 0 . 18 , transitions/hr , F[1 , 80] = 13 . 16; p<0 . 0005; df = 80 , p<0 . 0005 , Figure 7C ) . In addition , the kcnh4a-/- larvae exhibit shorter sleep-bout length specifically during the night ( average: kcnh4a-/- = 2 . 21 ± 0 . 05; kcnh4a+/+ = 2 . 43 ± 0 . 05; minhour df = 81 , p <0 . 005 , Figure 7D ) . Thus , the reduction in the number and length of sleep episodes causes global reduction in sleep time during the night in kcnh4a-/- larvae . These results show that the loss of Kcnh4a affects sleep time and sleep consolidation , specifically during the night . It also suggests that Kcnh4a regulates sleep by repolarization of the membrane potential in sleep-regulating neurons . How the hypothalamic Hcrt neurons regulate diverse and fundamental physiological functions and what is the molecular mechanism that controls sleep are largely open questions . We revealed the molecular profile of the Hcrt neurons and functionally demonstrated the role of Kcnh4a in regulating sleep . Using FAC-sorting of the whole Hcrt neuronal population and RNA-seq of minute amounts of RNA , 212 Hcrt-neuron–specific transcripts were identified . Combination of fluorescent ISH and immunofluorescence assays confirmed that several transcripts are expressed in Hcrt neurons . The high efficiency and specificity of these anatomical experiments suggest that a large portion of the candidate genes ( Figure 2—source data 1 ) are expressed in Hcrt neurons . Indeed , lhx9 and hmx3 , which were ranked lower in the list of candidate genes , were previously shown to be expressed in the early stages of Hcrt-neuron development ( Dalal et al . , 2013; Liu et al . , 2015 ) , and we confirmed these observations in 7 dpf larvae . Thus , these results provide a comprehensive list of genes that are likely to mediate the multifunctions of Hcrt neurons . Comparison between the Hcrt-neuron–specific candidate genes isolated in zebrafish and mice ( Dalal et al . , 2013 ) showed that eight genes ( rfx4 , lhx9 , scg2 , vgll2 , ptprn , creb3l1 , sgsm1 , and fam46a ) are found in both vertebrates . This genetic similarity is reasonable; however , performing similar cell isolation technique and bioinformatic analysis in both species would have likely increased the list of shared Hcrt-neuron–specific genes . Accumulated data on mammals and zebrafish suggest that the Hcrt neurons are not a homogenous population ( Appelbaum et al . , 2007; Dalal et al . , 2013 ) . Indeed , our co-localization studies showed a diversity of spatial gene expression in Hcrt neurons , varying from partial to complete overlapping with hcrt . Thus , the molecular signature suggests that these neurons are divided into subpopulations that may cope with the wide variety of functions of Hcrt neurons . The development and diverse functions of Hcrt neuron subpopulations are predicted to be regulated by Hcrt-neuron–expressed TFs , which target an ensemble of Hcrt-neuron–specific genes . The role of the isolated Hcrt-neuron–specific genes is diverse . Large arrays of genes are involved in the regulation of metabolism , endocrine systems , synaptic function , neurogenesis , reward , wake , and sleep ( Figure 2—source data 1 ) . These functions are correlated with the diverse role of Hcrt neurons . A group of genes includes metabolic and endocrine genes , such as the protein tyrosine phosphatase receptor ( ptprn ) , which is implicated in insulin regulation ( Saito , 1993; Primo et al . , 2008 ) , and the steroidogenic acute regulatory protein ( star ) , which regulates the production of steroid hormones from cholesterol in the mitochondria ( Stocco et al . , 2005; Ings and van der Kraak , 2006 ) . Another gene involved in metabolism is elovl7b , which regulates fatty acid metabolism and energy homeostasis . This gene has been linked to lipodystrophy , obesity , and other metabolic disturbances ( Brolinson et al . , 2008; Chen et al . , 2013 ) . These metabolic genes are likely part of the mechanism that regulates feeding and obesity . Thus , in addition to hcrt , an imbalance of the fatty acid and glucose-regulating genes and pathways may contribute to the metabolism-related symptoms of narcoleptic patients . An array of Hcrt-neuron–specific genes are involved in neurogenesis and synaptic plasticity . For instance , the synaptic vesicle protein synaptotagmin X ( syt10 ) , which is involved in vesicular trafficking and Ca ( 2 ) -dependent exocytosis ( Zhao et al . , 2003; Craxton , 2010 ) . In addition , the voltage-dependent calcium-channel ( cacgn4 ) gene regulates the biophysical properties of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid ( AMPA ) receptors ( Sager et al . , 2009 ) and secretogranin II ( scg2 ) encodes to a secretory protein and mediates the packaging and sorting of neuropeptides into secretory vesicles ( Fischer-Colbrie et al . , 2005; Kassahn et al . , 2009 ) . Additional examples include the denndbl , which is involved in axon guidance , synaptic plasticity , and synaptic vesicle exocytosis ( Wu et al . , 2011 ) , and netrin G1 ( ntng1 ) , which is part of the mechanism that regulates axon guidance during development ( Nakashiba et al . , 2000; Seo et al . , 2009 ) . Altogether , these genes are likely to play a key role in the mechanism that regulates neuritic processes , synaptic plasticity and activity in Hcrt neurons . In addition to annotated genes , the RNA profiling also revealed a set of long , non-coding RNA ( lncRNAs ) enriched in Hcrt neurons . lncRNAs regulate gene transcription and expression via various molecular mechanisms . Several studies show that lncRNAs regulate the expression of protein-coding genes , with their genomic loci adjacent to the locus of the specific lncRNA ( Goodrich and Kugel , 2006; Mercer et al . , 2009 ) . Supporting these observations , among the 16 lncRNAs that were enriched in Hcrt neurons , several were located in the genome next to the Hcrt-neuron–specific genes . For instance , the lncRNA cuff23873 is placed between the genes hacl1 and ankrd28 while si:dkey-58b18 . 8 is located in the intergenic region between pim2 and rpp40 . Thus , these results suggest that Hcrt-neuron–specific lncRNAs regulate transcriptional and post-transcriptional processes of Hcrt-neuron–specific genes . The identification of hundreds of Hcrt-neuron–specific candidate genes enabled us to predict the TFs that regulate the expression of these genes . We found a significant enrichment of TFs , which regulate metabolism , sleep , and other physiological processes ( Figure 5 ) . For example , the hepatic nuclear factor 1 homeobox ( hnf1 ) regulates the expression of genes involved in lipid and glucose transport ( Reiner et al . , 2009 ) . In the Hcrt neurons , it was expected to regulate 35 out of the 48 Hcrt-neuron–specific genes ( Figure 5 ) , including five metabolic genes ( soat2 , f2rl1 , scg2 , grpr , and elovl7b ) . Another key TF is the peroxisome proliferator-activated receptor alpha ( ppara ) , which plays a role in lipid metabolism and satiety ( Fu et al . , 2003; Chakravarthy et al . , 2009 ) . In the Hcrt neurons , this TF is expected to regulate the transcription of 33 Hcrt-neuron–specific genes , such as ptprn , soat2 , f2rl1 , scg2 , and grpr , which are associated with the balancing of metabolism . Notably , the TF ap1 , which is associated with sleep induction ( Turek et al . , 2013 ) , is expected to be a regulator of 13 Hcrt-neuron–specific genes , including lhx9 , which regulates sleep ( Dalal et al . , 2013 ) . Intriguingly , ap1 can also regulate the expression of Hcrt-neuron–specific synaptic genes , such as cacgn4 , nos1 , and dennd1b . Since sleep regulates synaptic plasticity in Hcrt axons ( Appelbaum et al . , 2010 ) , ap1 might mediate the molecular mechanism that links sleep with synaptic plasticity in Hcrt neurons . The different players that are expressed in Hcrt neurons modulate the diverse roles of the neurons; however , these functions are also mediated by other hypothalamic neuronal networks . Aside from the Hcrt-neuron–specific genes , we identified three transcripts , cuff8731 , npvf , and npffr1 , which are expressed in cells located adjacent to Hcrt neurons . The neuropeptide VF precursor ( npvf ) and its receptor ( npffr1 ) regulate nociception , anxiety , learning , and memory ( Liu et al . , 2001 ) . The NPVF/NPFFR1 system also controls pain and analgesia through interactions with the opioid system ( Liu et al . , 2001 ) . The opioid system is formed , among others , by Nociceptin that forms the nociceptin/orphanin FQ ( N/OFQ ) system . This system makes synaptic contacts with Hcrt neurons , inhibiting their activity via pre- and post-synaptic mechanisms . The nociceptin/orphanin FQ ( N/OFQ ) system also exerts diverse actions in the hypothalamic–pituitary–adrenal ( HPA ) axis , and is implicated in the neurobiological control of stress and associated adaptive behaviors ( Delaney et al . , 2012 ) . More specifically , Hcrt neurons are essential in the generation of stress-induced analgesia ( SIA ) , and N/OFQ blocks SIA via inhibition of Hcrt neuron activity ( Xie et al . , 2008 ) . Altogether , NPVF/NPFFR and Hcrt neurons may interact in the hypothalamus to regulate morphine- and stress-induced analgesia . Among the candidate Hcrt-neuron–specific genes , we studied the role of kcnh4a , which is located adjacent to hcrt in the genome and is expressed in all Hcrt neurons . We found that sleep time , sleep\wake transitions , and sleep-bout length are decreased in kcnh4a-/- larvae during the night . Since potassiumvoltage-gated channels repolarize the cell membrane ( Choe , 2002; Bezanilla , 2000; Swartz , 2008 ) , loss of kcnh4a may reduce potassium efflux and induce repetitive hyperpolarization , and , ultimately , nighttime wakefulness . Supporting this role , the ether-a-go-go-gene–related ( ERG ) potassium channel blockers selectively increased waking activity at night in zebrafish ( Rihel et al . , 2010 ) . The importance of potassium channels for sleep regulation has also been demonstrated in flies . Genetic screen of fly mutants revealed the short sleeper shaker mutant . The shaker gene encodes a voltage-dependent potassium channel and regulates sleep need and efficiency ( Cirelli et al . , 2005 ) . Notably , loss of the shaker and kcnh4a potassium channels similarly affects nighttime sleep , while daytime sleep is unaffected in kcnh4a-/- larvae . In mice , loss of the voltage-dependent potassium channel Kcna2 decreases non-rapid-eye-movement ( NREM ) sleep and increases wakefulness ( Douglas et al . , 2007 ) . These findings suggest that sleep is regulated by neuronal-circuit–specific potassium channels in flies , zebrafish , and mammals . According to this model , in zebrafish , the presence of Kcnh4a in the excitatory Hcrt neurons suggests that Kcnh4a regulates their activity and , ultimately , sleep and wake . In kcnh4a-/- larvae , the absence of Kcnh4a may cause hyperexcitability of the Hcrt neurons that induces the activity of downstream arousal-promoting targets , such as the paraventricular thalamic nucleus ( Huang et al . , 2006 ) and the locus coeruleus ( Carter et al . , 2010 ) . However , since the expression of Kcnh4a is not restricted to Hcrt neurons and its effect on firing rates in specific neuronal population is not clear , further neurophysiological studies are required to link Kcnh4a , neuronal activity , sleep and wakefulness . The Hcrt transcriptome identified Kcnh4a as a sleep regulator and provides a platform for future studies on the molecular mechanism that regulates Hcrt-neuron–dependent physiological processes , such as feeding and sleep-wake cycles . In addition , it may also help to identify Hcrt-neuron–specific antigens that trigger the autoimmune response , leading to the specific elimination of Hcrt neurons in narcolepsy ( Mahlios et al . , 2013 ) . Since the transparent zebrafish offer a wide array of tools to manipulate genes and visualize neuronal-circuit activity in live animals , a future combination of CRISPR-mediated mutants , genetically encoded Ca2+ indicators , and optogenetic tools are expected to elucidate the functional role of the Hcrt-neuron–specific genes in a neuronal-circuit–specific manner . These experiments will facilitate our understanding of the mechanism controlling the multifunctional Hcrt neurons . The hcrt:EGFP , kcnh4a-/- , kcnh4a+/- , kcnh4a+/+ , and WT fish were kept in a fish facility under a 14 hr light/10 hr dark cycle ( LD ) at 28°C ( Elbaz et al . , 2012 ) under optimal maintenance conditions , in accordance with the animal protocol approved by the Bar-Ilan University Bioethics Committee . Larvae were generated by paired mating , and raised in incubators and larvae water systems ( Elbaz et al . , 2012 ) under LD . The heads of 7 dpf hcrt:EGFP , α-tubulin:EGFP-injected , and WT larvae ( 60 larvae per group ) were collected in a 2 ml tube . Cells were then dissociated using the Papain Dissociation System ( Worthington Biochemical Corporation , Lakewood , NJ ) according to the manufacturer’s protocol . The cells were filtered with a 70 μm cell strainer ( BD Transduction Laboratories , San Jose , CA ) and washed twice with cold phosphate-buffered saline ( PBS ) . High-speed FACS was performed using an LSRII FACS machine ( BD , Bioscience , San Jose , CA ) . A two-gate FACS technique was used to select only EGFP+ cells from non-fluorescent and auto-fluorescent cells . The EGFP+ cells were differentiated using SSC-A and FSC-A strategies . As a control , two additional groups of cells were sorted: the first group contained only EGFP+ cells derived from the heads of larvae expressing α-tubulin:EGFP , and the second group contained only EGFP- cells derived from WT larvae . Then , to separate the EFGP+ cells from the EGFP- cells , PE-Cy5-A and GFP-A filters were applied . The cells were collected into a 96-well sterile plate filled with the first RNA purification buffer from the RNeasy Mini Kit ( Qiagen , Redwood City , CA ) . The samples were stored at -80°C until the RNA was purified . Three independent FACS experiments were performed , yielding three samples of EGFP+ cells ( group 1 ) and three samples of EGFP- cells ( group 2 ) . Each sample contained 300 sorted cells . Six samples ( three EGFP+ and three EGFP- ) were used for total RNA extraction using the RNeasy Mini Kit ( Qiagen , Redwood City , CA ) according to manufacturer’s protocol . Additionally , total RNA from six samples of the whole head of WT larvae ( group 3 ) was purified using the same kit . Each sample contained 60 heads . The quality and quantity of each RNA sample were assessed by Agilent's 2100 Bioanalyzer 6000 Pico Kit ( Agilent Technologies , Santa Clara , CA ) . RNA of group 1 and 2 ( Figure 1I ) was amplified using the Ovation® RNA-Seq System V2 ( NuGEN , San Carlos , CA ) . Before amplification , all samples were lyophilized using a SpeedVac instrument and then suspended in 5 μl of nuclease-free water . The cDNAs were fragmented using a Bioruptor instrument with three 10 s ( ‘on’ ) cycles of sonication interrupted by 90 s pause periods ( ‘off’ ) . The cDNAs of group 3 ( Figure 1I ) was synthesized according to standard procedures . The cDNAs were quantified using the Nanodrop and Bioanalyzer DNA 1000 Chip . The libraries were loaded on a High Sensitivity ChIP and quantified on a Qubit instrument . Illumina TruSeq protocol was used to prepare libraries from RNA samples . Twelve libraries ( group 1 , 2 , 3 , Figure 1I ) were run on 2 lanes of an Illumina HiSeq2000 machine using the multiplexing strategy of the TruSeq protocol ( Institute of Applied Genomics , Udine , Italy ) . An average of 24 million reads were obtained from EGFP+ RNA , 22 million reads from EGFP- RNA , and 175 million reads from the whole head RNA ( http://www . ncbi . nlm . nih . gov/sra , PRJNA283169 ) . Because of the difference in the amounts of RNA and the amplification process , the reads were 2×50 base pairs for the EGFP+ and EGFP- groups , and 2×100 for whole larva head groups . The RNA-seq data from the replicates were unified , obtaining three groups for further analysis: EGFP+ , EGFP- , and whole head groups . Since the amount of cells and RNA was very low , this strategy increased the read cover for each gene and resolved potential amplification bias . Cufflinks and Cuffdiffs ( http://cufflinks . cbcb . umd . edu/ ) ( Trapnell et al . , 2010; Trapnell et al . , 2012 ) were used to calculate gene-expression levels and identify differentially expressed transcripts ( statistical analysis is described below ) . The reads were mapped to the zebrafish genome ( Zv9 ) , and raw read counts were normalized to TPM . Initially , a gene was considered to be preferentially expressed in the Hcrt cells if its normalized expression level was at least 100 TPM in EGFP+ cells and sevenfold higher than the higher of the normalized expression levels estimated in the two controls . For reference , the hcrt , which was expressed at a level above 300 TPM in the EGFP+ sample and below 10 TPM in the control samples , was used for aligning the reads against the zebrafish genome , allowing only uniquely aligned reads . In order to enlarge the list of enriched transcripts in Hcrt neurons , relaxed parameters were set and the new requirements were 10 TPM and 3 . 6 times higher abundance in EGFP+ cells compared with the control groups . Following analysis of the RNA-seq data , transcripts that were enriched in Hcrt neurons were either assigned to an annotated zebrafish gene or regarded as novel transcripts . To further characterize the annotated genes , they were assigned a human ortholog using the 'Non-Zebrafish RefSeq Genes' or the 'Human Proteins Mapped by Chained tBLASTn' tracks on the UCSC genome browser ( Zv9/danRer7 assembly; http://genome . ucsc . edu/ ) . To find over-represented molecular functions , human orthologs were used as input to DAVID annotation . We focused on over-represented TFs . The DAVID default human-gene background was used . All the significantly enriched ( Benjamini-Hochberg adjusted p<0 . 05 ) TFs are presented in Figure 5—source data 1 . The conserved location of transcription factor binding sites was identified in mammalian alignments . A binding site was considered to be conserved across the alignment if its score met the threshold score for its binding matrix . The score and threshold were calculated using the Transfac Matrix Database ( v7 . 0 ) created by Biobase ( Waltham , MA ) . The expression levels of kcnh4a mRNA were determined using quantitative real-time PCR . Total mRNA was extracted from kcnh4a-/- ( n=9 batches of 8 larvae ) and kcnh4a+/+ ( n=5 batches of 8 larvae ) 6 dpf larvae , using the RNeasy Protect Mini Kit ( Qiagen , Redwood City , CA ) and according to the manufacturer's instructions . A similar amount of mRNA ( 1µg ) was reverse-transcribed using Oligo ( dT ) oligos and SuperScript III reverse transcriptase ( Invitrogen , Carlsbad , CA ) according to the manufacturer's protocol . Transcript levels were determined by Applied Biosystems 7900HT Fast Real-Time PCR System using the Quanta SYBR FAST qPCR Kit ( Quanta Biosciences , Gaithersburg , MD ) . Ef1a was used as reference gene ( Tang et al . , 2007 ) and ΔΔCT analysis was performed ( Schmittgen and Livak , 2008 ) . To prepare probes for whole-mount ISH experiments , the full coding sequences of the following genes were amplified: hcrt ( NM_001077392 . 2 ) , star ( NM_131663 . 1 ) , dennd1b ( XM_009296374 . 1 ) , kcnh4a ( KR733682 ) , fam46a ( XM_005157860 . 2 ) , npvf ( NM_001082949 . 1 ) , npffr ( NM_001171697 . 1 ) , zgc:171844 ( NM_001127478 . 1 ) , hmx3 ( NM_131634 . 1 ) , lhx9 ( NM_001017710 . 1 ) , si:dkey-58b18 . 8 ( ENSDARG00000095761 ) , elovl7b ( NM_199778 . 1 ) , gad67 ( NM_194419 . 1 ) , vglut2b ( NM_001009982 . 1 ) , hsf1 ( NM_131600 . 1 ) and cuff23873 . All PCR products were cloned into a pCRII-TOPO vector ( Invitrogen , Carlsbad , CA ) and served as a template to transcribe digoxigenin-labeled antisense mRNA probes . Larvae and adult brains were fixed in 4% paraformaldehyde over 48 hr at 4°C . All samples were then dehydrated in 100% methanol and stored at -20°C . Before further treatment , brains and larvae were rehydrated in decreasing methanol concentrations . Adult brains were embedded in 2 . 5% agarose and sectioned with the Vibratome Series 1000 Sectioning System ( Campden Instruments , Lafayette , IN ) . Transverse sections were then processed and stained as free-floating slices . ISH was performed following standard protocols . Digoxigenin- and fluorescein-labeled antisense riboprobes were transcribed in vitro using RNA Labeling Kit SP6/T7 ( Roche Diagnostics Corporation , Indianapolis , IN ) . Single probe ISH was revealed with colorimetric BM purple ( Roche Diagnostics Corporation , Indianapolis , IN ) . Double probe fluorescent ISH was performed as described previously ( Levitas-Djerbi et al . , 2015 ) . ISH was performed as described above in hcrt:EGFP 2 dpf larvae and adults . The samples were revealed using Fast Red ( Roche Diagnostics Corporation , Indianapolis , IN ) . All the procedures were based on standard protocols ( Levitas-Djerbi et al . , 2015 ) . After blocking , larvae and adult brain slices were incubated in primary rabbit anti-EGFP ( Santa Cruz Biotechnology , Santa Cruz , CA ) , diluted 1:250 . Anti-EGFP antibodies were detected with a secondary goat anti-rabbit Alexa Fluor 488 IgG ( H L ) antibody ( 2 mg/mL , A-11034 , Invitrogen , Carlsbad , CA ) . All experiments were repeated in 3-–5 larvae and adult sections . An epifluorescence stereomicroscope ( Leica M165FC ) was used to visualize live larvae expressing EGFP and fluorescent-fixed larvae and adult brain sections . Pictures were taken using the Leica Application Suite imaging software , version 3 . 7 ( Leica , Wetzlar , Germany ) . In confocal imaging of fixed embryos , the samples were mounted on slides . In live imaging of hcrt:EGFP larvae , the larvae were mounted in low-melting–point 1% agarose . Confocal imaging was performed using a Zeiss LSM710 upright confocal microscope ( Zeiss , Oberkochen , Germany ) . All images were processed using ImageJ ( National Institutes of Health , Bethesda , MD ) . The CRISPR system ( Hwang et al . , 2013 ) was used to establish the kcnh4a-/- line . The Cas9 ( Addgene plasmid no . 42251 ) and sgRNA ( Addgene plasmid no . DR274 ) zebrafish expression plasmids were obtained from Addgene ( Cambridge , MA ) . In order to prepare the sgRNA , two kcnh4a-specific oligos were designed to match the target site ( ACAACGTCTGCTTCTCCACCC ) in exon 5 . These oligos were denatured at 95°C for 5 min , then gradually cooled down to room temperature and kept at 4o°C . Before cloning , annealing of the oligos was confirmed in 2% agarose gel . The annealed oligos were cloned into the DR274 plasmid between the BsaI restriction sites and transformed into bacteria , which was selected by standard sequencing . In order to synthesize the specific sgRNA , the DR274 plasmid containing the annealed oligos was linearized with the restriction enzyme DraI , and cleaned using the standard phenol-chloroform procedure , followed by purification by the PureLink PCR Purification Kit ( #K31000 , Life Technologies , Carlsbad , CA ) . The sgRNA was synthesized using the T7 High Yield RNA Synthesis Kit ( New England Biolabs , Hitchin , UK ) . In order to prepare Cas9 mRNA , the zebrafish Cas9 vector was linearized by AgeI , and mRNA was synthesized using the mMESSAGE mMACHINE T7 Kit ( Life Technologies , Carlsbad , CA ) . One-cell–stage WT zebrafish embryos were microinjected with mixed Cas9 mRNA ( 300 ng/µl ) and transcribed sgRNA ( 12 . 5 ng/µl ) . To test the efficiency of the CRISPR system , ten 1-dpf embryos were screened for kcnh4a-specific mutation ( as described below ) . We found that 60% of the embryos carried the mutation . The founder ( F0 ) mosaic embryos were raised to adulthood and outcrossed with WT fish in order to identify F1 mutant fish . Single F1 heterozygous fish , which carry a 14 bp deletion mutation in the kcnh4a target site ( Figure 6D ) , was selected and outcrossed with WT fish . To decrease the risk for off-target mutations , heterozygous F2 and then F3 fish were outcrossed with WT fish . In all experiments , heterozygous F4 fish were intercrossed , and the assays were performed on their progeny . Genotyping of the kcnh4a-/- zebrafish was conducted by extracting genomic DNA from embryos and larvae or by the tail clipping of adult fish using the KAPA Express Extract Kit ( Kapa Biosystems Inc . , Boston , MA ) according to the manufacturer's instructions . Genomic DNA was then amplified by PCR using the following primers: forward- 5'TTCATGTTTTCCACAGAATGTGTTTTCACA3' and reverse- 5'ACCGAGGATGAAGAGCATCTCCACAG3' . The PCR product was then run on 2% agarose gel , and heterozygous , homozygous , and WT fragments could be identified by their size ( Figure 6D ) . To confirm the gel pattern , selected PCR products were sequenced . The kcnh4a+/− adult zebrafish were intercrossed and their progeny were kept under LD cycle . At 5 dpf , the larvae were individually placed in 48-well plates . At 6 dpf , larva-containing plates were placed in the Noldus DanioVision tracking system ( Noldus Information Technology , Wageningen , Netherlands ) and acclimated for one hour prior to behavioral recording . Recording was performed using the EthoVision XT 9 software ( Noldus Information Technology , Wageningen , Netherlands ) , as previously described ( Elbaz et al . , 2012 ) . Light intensity in the tracking system was 70 LUX for all experiments . To monitor rhythmic behavior during a daily cycle , larvae were maintained under the LD cycle , which was similar to the LD cycle prior to the experiment . Data analyses of total locomotor activity , sleep time , sleep\wake transitions , and sleep-bout length were performed according to the parameters previously described ( Elbaz et al . , 2012 ) . Following each behavioral experiment , all larvae were subject to genotyping ( as described above ) . SD was performed by randomized manual tapping on a petri dish that contained 6 dpf larvae . Following the SD , sleep time was monitored in sleep-deprived and control larvae ( n = 13 for each treatment ) using behavioral systems . In the RNA-seq data , statistically significant differences between the number of reads aligned to each gene ( the expression profile ) in the different tested conditions , without unifying the replicates , were identified as previously described ( Alon et al . , 2011; Ben-Moshe et al . , 2014 ) . Briefly , the expression profiles were normalized using a variation of the trimmed mean of M-values normalization method ( Robinson and Oshlack , 2010; Alon et al . , 2011 ) . Subsequently , we searched for expression differences between the EGFP+ and the control samples that cannot be explained by Poisson noise with p <0 . 01 and Bonferroni correction for multiple testing ( Alon et al . , 2011 ) . Notably , the analysis takes into account technical biases that can cause the variance to be larger than that of naive Poisson statistics ( Alon et al . , 2011 ) . Only genes with average expression >15 ( raw reads ) in the EGFP+ samples were analyzed , and only genes with fold change higher than 3 . 6 are shown in Figure 2—source data 1 . In the behavioral experiments , statistical analysis was performed using SAS v9 . 3 software ( SAS Institute , Cary , NC ) . Locomotor activity , sleep time , sleep-bout length , and sleep\wake transitions were analyzed with repeated measures of ANOVA ( SAS PROC MIXED ) , where each was modeled as a function of genotype ( kcnh4a-/- , kcnh4a+/- , kcnh4a+/+ ) , time ( 24 hr ) , and the genotype by time interaction term . LS means ( model estimated means ) differences between the genotype groups per time point were estimated from the model interaction terms and are presented with respective levels of significance and 95% confidence intervals . These were used to compare between genotypes per time point in locomotor activity and sleep experiments .	Sleep appears to be essential for all animals . The loss of a type of brain cell called the Hypocretin/Orexin ( Hcrt ) neurons causes the sleep disorder narcolepsy , which disturbs sleep patterns . These neurons also control several other fundamental behaviors and activities , including eating and processing rewards , but it is not clear how Hcrt neurons are able to influence multiple behaviors . The development and activity of a cell depends to a large extent on the genes it expresses . Yelin-Bekerman et al . have now used genetic techniques to identify a set of genes that are specifically expressed in the Hcrt neurons of zebrafish . Some of these genes are expressed in all of the Hcrt neurons , and some are only expressed in certain subsets of them . Computational methods also revealed a set of “transcription factor” proteins that regulate the expression of clusters of these genes . Yelin-Bekerman et al . focused on a gene called kcnh4a , and found that this encodes an ion channel protein that allows potassium ions to exit the neurons and stop neuronal activity ( this activity is also known as an “action potential” ) . This gene is expressed in all Hcrt neurons . Further experiments showed that zebrafish that lack the potassium channel sleep less during the night . This therefore suggests that the potassium channel is important for regulating sleep . Future studies of the genes that are enriched in Hcrt neurons could uncover the mechanisms that enable the neurons to play a role in such a diverse range of processes , including feeding and sleep-wake cycles . These studies should enhance our understanding of the role of sleep and may help to develop treatments for metabolic and sleep disorders .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2015	Hypocretin neuron-specific transcriptome profiling identifies the sleep modulator Kcnh4a
The re-assembly of chromatin following DNA replication is a critical event in the maintenance of genome integrity . Histone H3 acetylation at K56 and phosphorylation at T45 are two important chromatin modifications that accompany chromatin assembly . Here we have identified the protein kinase Pkc1 as a key regulator that coordinates the deposition of these modifications in S . cerevisiae under conditions of replicative stress . Pkc1 phosphorylates the histone acetyl transferase Rtt109 and promotes its ability to acetylate H3K56 . Our data also reveal novel cross-talk between two different histone modifications as Pkc1 also enhances H3T45 phosphorylation and this modification is required for H3K56 acetylation . Our data therefore uncover an important role for Pkc1 in coordinating the deposition of two different histone modifications that are important for chromatin assembly . During the cell cycle one of the key processes following DNA replication is the re-assembly of chromatin on the newly replicated DNA . This involves the re-deposition of both pre-existing and newly synthesised histones into intact nucleosomes ( reviewed in Groth et al . , 2007; Ransom et al . , 2010; Gurard-Levin et al . , 2014 ) . This process must be tightly controlled and in S . cerevisiae one key event is the Rtt109-mediated modification of histone H3 by acetylation of lysine 56 ( H3K56 ) ( Han et al . , 2007; Driscoll et al . , 2007 ) . Rtt109 cooperates with two different histone chaperones Asf1 and Vps75 ( Tsubota et al . , 2007; Driscoll et al . , 2007 ) and although it is generally accepted that Asf1 is the major chaperone involved in promoting H3K56 acetylation ( Keck et al . , 2011; Tang et al . , 2011 ) , recent evidence suggests a role for Vps75 in this process ( Radovani et al . , 2013 ) . However , the signalling pathways controlling Rtt109-mediated H3K56 acetylation are unknown . Another important modification that occurs coincidentally with DNA replication is H3 phosphorylation at threonine T45 ( H3T45 ) but current evidence suggests that this modification has no relationship with the closely located H3K56 acetylation event ( Baker et al . , 2010 ) . In the context of an unperturbed cell cycle , Cdc7-Dbf4 has been shown to be a major kinase involved in controlling H3T45 phosphorylation levels ( Baker et al . , 2010 ) . S . cerevisiae possesses one protein kinase C isoform , Pkc1 , and it is known to play a major role in cell wall integrity signalling ( reviewed in Levin , 2005 ) . Pkc1 has a growing number of nuclear functions and through genetic experiments this protein kinase has previously been implicated in cell cycle events ( reviewed in Levin , 2005 ) . Part of this role can be attributed to its connection to pathways sensing cell wall integrity and its effects on the activity of the downstream MAP kinase pathway . However MAP kinase-independent roles for Pkc1 have emerged and Pkc1 was recently shown to directly affect the cell cycle by inhibiting the activity of the transcriptional coactivator protein Ndd1 ( Darieva et al . , 2012 ) . This finding suggested that Pkc1p might act as a checkpoint kinase under conditions of replicative stress where progression to G2 phase is inhibited . However , mutant strains containing ndd1 mutant alleles lacking the Pkc1 phosphorylation sites did not fully re-capitulate the pkc1 mutant phenotypes , indicating that additional defects are associated with loss of Pkc1 activity . Indeed , pkc1 mutants are sensitive to hydroxurea ( HU ) treatment ( Queralt and Igual , 2005 ) , suggesting a potential role for Pkc1 in controlling the response to replicative stress . It is unclear how Pkc1 might affect replicative stress but previous genetic studies have linked PKC1 and components of the RSC chromatin remodelling complex and Pkc1 overexpression can suppress the defects associated with RSC mutants ( Chai et al . , 2005; Hosotani et al . , 2001; reviewed in Levin et al . , 2005 ) . This is suggestive of a potential role for Pkc1 in regulating chromatin assembly during DNA replication . Here , we have investigated the potential role of Pkc1 in controlling chromatin modifications during replicative stress and found that it coordinates the deposition of two histone marks on histone H3 , T45 phosphorylation and K56 acetylation . Mechanistically , Pkc1 controls H3K56 acetylation through phosphorylation-dependent activation of Rtt109 . Moreover , Pkc1 also promotes H3T45 phosphorylation and this modification is important for the coordinated deposition of H3K56 acetylation . To examine whether Pkc1 has a role in protecting cells from replicative stress we first determined cell sensitivity to hydroxyurea ( HU ) in the presence and absence of active Pkc1 . To do this we made use of strains containing temperature sensitive pkc1 alleles which could be inactivated by incubating yeast cells at 34ºC , thereby negating the confounding effects of heat shock stress response obtained using the traditionally used pkc1ts allele ( Anastasia et al . , 2012 ) . In comparison to cells containing wild-type PKC1 ( DK186 ) , temperature sensitive pkc1-14 ( DK1690 ) cells , grown at 34ºC to inactivate Pkc1 were inviable ( Figure 1A , top panel ) consistent with previous studies using a pkc1Δ deletion strain ( Queralt and Igual , 2005 ) . Loss of PKC1 function results in cell lysis in the absence of osmotic support . The presence of sorbitol saved the pkc1-14 mutant phenotype ( Figure 1A , middle panel ) but HU sensitivity was observed under these conditions ( Figure 1A , bottom panel ) . This suggested a role for Pkc1 other than signalling through the well studied downstream MAP kinase cascade which includes the kinase Bck1 ( Lee et al . , 1993 ) . Indeed bck1Δ ( Y01328 ) cells were not sensitive to HU , further emphasising a new alternative mechanism of action for Pkc1 under replicative stress conditions ( Figure 1A ) . To determine whether the kinase activity of Pkc1 is needed to protect cells from replicative stress , we attempted to rescue pkc1-14 cells by re-expressing a catalytically dead ( KD ) version of Pkc1 in the presence of HU . However , while wild-type ( WT ) Pkc1 was able to rescue the growth defect of pkc1-14 cells , the catalytically dead version was unable to do so ( Figure 1B ) . 10 . 7554/eLife . 09886 . 003Figure 1 . Pkc1p is required for cell viability and chromatin integrity in the presence of hydroxyurea . ( A and B ) Cell growth assays . ( A ) 10 fold serial dilutions of wild-type ( WT ) ( DK186 ) , pkc1-14 ( DK1690 ) or bck1Δ ( Y01328 ) cells were plated onto YPD media in the presence or absence of 1M sorbitol or ( B ) pkc1-14 ( DK1690 ) cells containing empty vector or plasmids expressing WT or kinase dead ( KD ) Pkc1 , were plated onto SD media in the presence 1M sorbitol . Cells were grown at 34ºC and where indicated , 100 mM HU was added to the media . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 003 Together , these experiments demonstrate that Pkc1 kinase activity is required for protecting cells from replicative stress . During S phase , acetylation of histone H3 at lysine 56 ( H3K56 ) is an important modification in ensuring the correct assembly of newly synthesised nucleosomes and their incorporation into chromatin ( Li et al . , 2008; Masumoto et al . , 2005 ) . We therefore examined whether Pkc1 is required for H3K56 acetylation under replicative stress conditions . Upon release from a HU block , wild-type cells showed accumulation of acetylated H3K56 ( Figure 2A , lanes 1–6 ) . In contrast , temperature sensitive pkc1-21 cells ( DK1697 ) showed much reduced levels of H3K56 acetylation when grown at 34ºC ( Figure 2A , lanes 7–12 ) even though they had progressed through S phase and arrested at G2 ( Figure 2—figure supplement 1A ) . Importantly , this effect was not limited to strains harbouring the pkc1-21 allele but was also observed in a different temperature sensitive strain , harbouring the pkc1-14 allele either upon release from a HU block ( Figure 2B and Figure 2—figure supplement 1B ) or upon entry into S phase upon release from an α factor block ( Figure 2—figure supplement 2 ) . H3K56 acetylation is deposited by Rtt109 , and this enzyme also acetylates histone H3 at lysine 9 ( H3K9 ) ( Driscoll et al . , 2007; Han et al . , 2007; Tsubota et al . , 2007 ) . Interestingly , in addition to H3K56 acetylation defects , pkc1-14 cells showed much reduced levels of H3K9 acetylation ( Figure 2B , middle panel ) which suggests a defect in Rtt109 activity . 10 . 7554/eLife . 09886 . 004Figure 2 . Pkc1p controls H3K56 acetylation . ( A-C ) Western blot analysis of the H3K56 acetylation levels in ( A ) WT ( DK186 ) or pkc1-21 ( DK1697 ) or ( B ) WT ( DK186 ) or pkc1-14 ( DK1690 ) strains grown in YPD in the presence 1M sorbitol at 34ºC , treated with 200 mM HU for 2 . 5 hrs and released into YPD with sorbitol for the indicated times . ( C ) WT ( DK186 ) cells grown in the presence or absence of cercosporamide at 30ºC in YPD , exposed to 200 mM HU for 2 . 5 hrs and released from a HU block for the indicated times . H3K9 acetylation levels are also shown in ( B ) and Histone H3 and tubulin ( Tub1 ) are shown as loading controls . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 00410 . 7554/eLife . 09886 . 005Figure 2—figure supplement 1 . The pkc1-21 and pkc1-14 mutants exhibit a mitotic delay . Cells from strains containing wild-type ( WT ) PKC1 ( DK186 ) and the temperature sensitive pkc1-21 ( DK1697 ) and pkc1-14 ( DK1690 ) mutants were grown to early log phase at 30ºC in YPD supplemented with 1 M sorbitol , blocked with 200 mM HU for 2 . 5 hrs at 34ºC , washed and released into a fresh YPD media containing sorbitol at 34ºC . Aliquots were taken at the indicated times , analysed for propidium iodine staining by FACS , and the DNA content histograms were determined for at least 30 , 000 cells . The location of peaks corresponding to haploid ( 1n ) and diploid ( 2n ) DNA content are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 00510 . 7554/eLife . 09886 . 006Figure 2—figure supplement 2 . The role of Pkc1 in controlling H3K56 acetylation . Western blot analysis of the H3K56 acetylation levels in WT ( DK186 ) or pkc1-14 ( DK1690 ) strains . Cells were released alpha factor block for the indicated times . Histone H3 and tubulin ( Tub1 ) are shown as loading controls . All experiments were performed at 34ºC in YPD plus sorbitol . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 00610 . 7554/eLife . 09886 . 007Figure 2—figure supplement 3 . Comparison of genetic and pharmacological disruption of Pkc1 activity on H3K56 acetylation . ( A ) Western blot analysis of the H3K56 acetylation levels in WT ( DK186 ) ( top ) or pkc1-14 ( DK1690 ) ( bottom ) strains . Cells were grown in the presence or absence of cercosporamide at 34ºC in YPD plus 1M sorbitol and were released from a 200 mM HU block for the indicated times . Histone H3 and tubulin ( Tub1 ) are shown as loading controls . ( B ) Quantification of the H3K56ac levels in ( A ) normalised for tubulin and total H3 levels and shown relative to the WT cells prior to release from HU block ( taken as “1”; indicated by dashed line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 007 To further explore the role of Pkc1 in promoting H3K56 acetylation , we treated wild-type cells with the Pkc1 inhibitor cercosporamide ( Sussman et al . , 2004 ) and examined H3K56 acetylation following release from a HU block . Pharmacological inhibition of Pkc1 activity reduced H3K56 acetylation but importantly did not affect another histone H3 acetylation event at lysine 18 ( H3K18 ) ( Figure 2C ) , demonstrating the specificity of action of Pkc1 . We also examined whether combinatorial genetic inactivation and pharmacological inhibition of Pkc1 has additive effects but none were observed , indicating that the effect of cercosporamide was through Pkc1 ( Figure 2—figure supplement 3 ) . We note however that the loss of Pkc1 in the temperature sensitive strains resulted in stronger reductions in H3K56ac levels than following cercosporamide treatment , most likely due to incomplete Pkc1 inhibition in the latter case . Together these results demonstrate an important role for Pkc1 in promoting H3K56 acetylation under replicative stress conditions . The reduced levels of H3K9 and H3K56 acetylation observed upon disruption of Pkc1 activity suggested that the acetyltransferase Rtt109 which deposits these modifications might be targeted by Pkc1 . One mechanism might be through Pkc1-mediated control of Rtt109 phosphorylation . Inspection of Rtt109 sequence revealed a potential phosphorylation site at threonine 46 ( T46 ) which partially matches the PKC consensus sequence ( Figure 3A ) . To establish whether Rtt109 can be phosphorylated by Pkc1 at this site , we raised a phospho-specific antibody which specifically recognised a phosphorylated peptide surrounding T46 ( Figure 3—figure supplement 1A ) . In vitro kinase assays revealed that Pkc1 was able to phosphorylate Rtt109 in the context of a purified recombinant Rtt109-Vps75 heterodimer ( Figure 3B ) . This phosphorylation was reduced to background levels following treatment with lambda phosphatase ( Figure 3B , lane 3 ) . Immunoprecipitation assays demonstrated that Pkc1 and Rtt109 could interact in vitro ( Figure 3—figure supplement 1B ) . 10 . 7554/eLife . 09886 . 008Figure 3 . Pkc1-mediated Rtt109 phosphorylation . ( A ) Location and local sequence contexts of the potential Pkc1p target site in Rtt109 ( numbers indicate amino acid positions in the protein ) . ( B ) In vitro kinase assay using HA epitope-tagged Pkc1p immunoprecipitated from DZ2 cells and purified recombinant Rtt109-Vps75 protein complex . A HA IP from wild-type yeast cells was used as a control . Where indicated , λ phosphatase was added . Rtt109 , HA-Pkc1 and phosphorylated Rtt109 T46 ( phospho-Rtt109 ) were detected by immunoblotting ( IB ) . ( C and D ) Rtt109 T46 phosphorylation in vivo . DZ5 cells were grown in SD media at 30ºC , treated with 200 mM HU for the indicated times and either ( C ) released from the HU block for the indicated times or ( D ) treated with or without cercosporamide . Rtt109 and phosphorylated Rtt109 T46 were detected by IB . Quantification of Rtt109 phosphorylation from ( C ) relative to total Rtt109 levels is shown on the right . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 00810 . 7554/eLife . 09886 . 009Figure 3—figure supplement 1 . Pkc1 and Rtt109 phosphorylation . ( A ) A phospho-specific antibody specifically recognises a phosphorylated peptide surrounding T46 . A modified ( phosphorylation at T46; phospho-Rtt109 ) DDKRVPKST ( P ) IKTC or non-modified peptide ( Rtt109 ) DDKRVPKSTIKTC were dotted in increasing concentrations ( 2 , 5 , 10 µg ) onto Hybond-C nitrocellulose membranes and immunoblotting was carried out with 1:1000 dilutions of antibodies to the phosphorylated or non-phosphorylated form of Rtt109 . ( B ) Pkc1 interacts with Rtt109 in vitro . Immunoprecipitation assay of in vitro translated Rtt109 with anti-HA antibody bound to HA-tagged Pkc1 immunoprecipitated from DZ2 strain or with an anti-HA antibody mock immunoprecipitate from a wild-type strain lacking HA-tagged Pkc1 as a negative control . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 009 Rtt109 phosphorylation at T46 could also be identified in vivo following HU treatment ( Figure 3C , lanes 1–4 ) . This phosphorylation was subsequently lost upon release from HU arrest ( Figure 3C , lanes 5–6 ) . Importantly , treatment of cells with cercosporamide prior to HU addition blocked the appearance of Rtt109 phosphorylation at T46 ( Figure 3D ) , demonstrating a role for Pkc1 in promoting Rtt109 phosphorylation under replicative stress conditions . These data therefore indicate that in cells undergoing replicative stress , Rtt109 is phosphorylated in a Pkc1-dependent manner , most likely through direct phosphorylation by Pkc1 . To establish whether phosphorylation of Rtt109 at T46 has any functional relevance , we attempted to rescue rtt109Δ cells lacking Rtt109 protein expression with plasmids encoding mutant proteins which had either lost the ability to be phosphorylated ( T46A ) or carried a phosphomimetic residue in place of T45 ( T46D ) . In the presence of HU , rtt109Δ cells grew poorly ( Figure 4A ) . This growth defect could be rescued by re-expressing wild-type Rtt109 but this rescue was lost with the non-phosphorylatable Rtt109 ( T46A ) mutant . In contrast , Rtt109 ( T46D ) was able to suppress the growth defect ( Figure 4A ) indicating that the insertion of a phosphomimetic residue at the Pkc1 phosphorylation site reinstated the activity of Rtt109 . To understand the molecular basis to this rescue , we again examined H3K56 acetylation . As expected , this was severely depleted in rtt109Δ cells but restored upon re-expression of wild-type Rtt109 ( Figure 4B , left panels ) . Rtt109 ( T46A ) only partially re-constituted H3K56 acetylation levels and the kinetics were delayed whereas Rtt109 ( T46D ) promoted rapid and high levels of H3K56 acetylation ( Figure 4B , left panels ) . These results are fully consistent with a role for Pkc1-mediated phosphorylation of Rtt109 being required for its ability to promote H3K56 acetylation . 10 . 7554/eLife . 09886 . 010Figure 4 . Rtt109 phosphorylation is needed for efficient H3K56 acetylation . ( A ) Cell growth assays . 10 fold serial dilutions of rtt109Δ ( Y01490 ) cells containing plasmids expressing WT or the indicated Rtt109 mutants , were plated onto SD media in the presence or absence of 100 mM HU . ( B ) Western blot analysis of the H3K56 acetylation ( ac ) levels in rtt109Δ cells containing empty vector ( DZ8 ) or plasmids expressing WT or the indicated Rtt109 mutants ( DZ5-7 ) grown in SD media at 30ºC and released from a 200 mM HU block for the indicated times . Total Rtt109 and Histone H3 levels are also shown . ( C ) Western blot analysis of Rtt109 levels in ( WT ) ( DK186 ) or pkc1-14 ( DK1690 ) cells containing empty vector ( con ) or a plasmid containing Myc-tagged Rtt109 , grown in SD media in the presence of 1 M sorbitol at 34ºC and released from a 200 mM HU block for the indicated times . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 01010 . 7554/eLife . 09886 . 011Figure 4—figure supplement 1 . RTT109 expression is elevated in pkc1-14 cells . RT-PCR analysis of RTT109 and HTT2 ( control ) expression in wild-type ( WT ) ( DK186 ) ( grey lines ) or pkc1-14 ( DK1690 ) ( black lines ) cells released from 200 mM HU block for the indicated times in YPD plus sorbitol at 34ºC . Data are shown relative to 18S rRNA levels . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 011 Interestingly , we found that while wild-type Rtt109 and Rtt109 ( T46D ) were inducibly expressed to high levels upon release from HU arrest , Rtt109 ( T46A ) expression was delayed and reached a lower level ( Figure 4B , right hand panels ) . This suggested that Pkc1-mediated phosphorylation might be important for stabilising Rtt109 . Indeed , we found that Rtt109 protein was expressed to lower levels in pkc1-14 cells ( Figure 4C ) . This was not due to a transcriptional defect as RTT109 mRNA levels were elevated in pkc1-14 cells ( Figure 4—figure supplement 1 ) , suggesting that these cells are attempting to compensate for the loss in Rtt109 protein . This lower level of Rtt109 likely contributes to the decreased levels of H3K56 acetylation we observed . Together , these results indicated that Pkc1-mediated Rtt109 phosphorylation is required for maintaining Rtt109 protein levels and H3K56 acetylation levels as cells recover from HU-mediated replicative stress . One prediction from these results is that any chromatin changes should manifest themselves in similar gene expression profiles when pkc1-14 cells or cells expressing Rtt109 ( T46A ) are compared . Indeed , in cells released from HU block , a highly significant overlap was observed for pkc1-14 cells or cells expressing Rtt109 ( T46A ) in genes either downregulated or upregulated compared to wild-type cells ( Figure 5A and B ) . Importantly no such significant overlaps were observed when we considered genes that were regulated in opposite directions in the two mutant strains ( Figure 5B ) . Furthermore , when we considered the biological processes that were affected , there was a strong overlap in gene ontology ( GO ) terms , with over-representation of GO terms associated with processes such as 'cell wall' and 'nucleolus' strongly associated with both mutant strains and reciprocal effects seen with 'telomere maintenance' and 'DNA helicase activity' ( Figure 5C and D ) . These data therefore further support a strong association between Pkc1 activity , phosphorylation of Rtt109 at T46 and subsequent downstream effects on chromatin structure and gene expression . One prediction of these results is that the phosphomimetic version of Rtt109 might be able to bypass the defects seen in pkc1-14 cells . However , no rescue of the growth defects of pkc1-14 cells in the presence of HU was observed ( Figure 5—figure supplement 1 ) . Therefore a working model suggested a direct role for Pkc1 in controlling H3K56 acetylation through Rtt109 phosphorylation but also that additional activities were controlled by Pkc1 ( Figure 5E ) . 10 . 7554/eLife . 09886 . 012Figure 5 . Pkc1 and RTT109 phosphorylation mutants generate overlapping gene expression defects . ( A ) Venn diagrams showing the overlap of significantly upregulated ( top ) or downregulated ( bottom ) genes in pkc1-14 ( DK1690 ) or rtt109 ( T46A ) ( DZ6 ) cells released from a 200 mM HU block . The experiment was performed by growing all strains in SD media supplemented with 1M sorbitol at 34ºC . ( B ) Prevalence of gene expression changes showing the indicated directionality of change ( shown as a percentage of genes changed in the rtt109 ( T46A ) mutant strain ) . Hypergeometric P-values are shown ( NS = non-significant ) . ( C ) Boxplots ( left ) and heatmaps ( right ) showing the rank orders of significantly changing GO terms associated with genes whose expression is consistently changed in the pkc1-14 ( DK1690 ) ( left ) or rtt109 ( T46A ) ( DZ6 ) ( right ) cells . The GO terms are ranked according to changes in the rtt109 ( T46A ) ( DZ6 ) mutant . In the heat maps , the columns indicate the P-value ranking of GO terms associated with genes that consistently increase ( left ) or decrease ( right ) their expression in the respective mutant strain . ( D ) Heat map showing the relative expression of the genes contained in the 'Nucleolus' and 'DNA helicase' categories across all conditions analysed . DK186 and DZ5 are the WT equivalent strains for the pkc1-14 ( DK1690 ) and rtt109 ( T46A ) ( DZ6 ) mutant strains . Brackets indicate genes that are consistently up- or down-regulated in both mutant strains upon release from a HU block . ( E ) Model for Pkc1 function through phosphorylating Rtt109 and promoting its ability to acetylate ( Ac ) histone H3 K56 . Question mark indicates additional activities of Pkc1 . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 01210 . 7554/eLife . 09886 . 013Figure 5—figure supplement 1 . Rescue of the HU sensitivity of pkc1-14 by a rtt109 phosphomimetic allele . 10-fold serial dilutions of wild-type ( WT; DK186 ) , pkc1-14 ( DK1690 ) containing empty vector pCM188 or pAS4106 ( expressing Rtt109[T46D] ) were spotted onto SD plates in the presence of sorbitol with or without 100 mM HU and incubated at 34°C for 3 days . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 013 Previous results in mammalian cells had shown that PKCδ phosphorylates histone H3 at threonine 45 ( H3T45 ) ( Hurd et al . , 2009 ) . As this residue is conserved in S . cerevisiae ( Figure 6A ) , and is located in the vicinity of K56 , it is possible that Pkc1-mediated phosphorylation of T45 might influence acetylation events at K56 . We therefore examined whether Pkc1 could phosphorylate H3T45 in vitro by using purified histone H3 and immunoprecipitated Pkc1p . Efficient H3 phosphorylation was observed ( Figure 6B ) and phosphorylation at T45 was confirmed as the only major phosphorylation site by mass spectrometry ( Figure 6C ) . Next we examined H3T45 phosphorylation and its dependence on Pkc1 in vivo . As cells were released from a HU-mediated block , H3T45 became strongly phosphorylated in wild-type cells ( Figure 6D ) . However , no such rises in H3T45 phosphorylation were observed upon ablation of Pkc1 activity in pkc1-14 cells ( Figure 6D , lanes 7–12 ) . 10 . 7554/eLife . 09886 . 014Figure 6 . Pkc1 promotes histone H3 T45 phosphorylation . ( A ) Location and local sequence contexts of the Pkc1p target site in histone H3 ( numbers indicate amino acid positions in the protein and human H3 is shown below ) . ( B ) In vitro kinase assay using HA epitope-tagged Pkc1p immunoprecipitated from DZ2 cells and recombinant histone H3 or GST as substrates . Protein phosphorylation was visualised by phosphorimaging ( right panel ) and total input proteins by Coomassie blue staining ( left panel ) . ( C ) Product ion spectra for the doubly-charged precursor ion [556 . 78]2+ . The spectra is fully annotated , and includes both b- and y-ions that support phosphorylation ( p ) at the residue indicated , Thr45 . ( D ) Histone H3 T45 phosphorylation in vivo . WT ( DK186 ) or pkc1-14 ( DK1690 ) cells were grown in SD media in the presence of 1M sorbitol at 34ºC and were treated with 200 mM HU and released from the HU block for the indicated times . Histone H3 and phosphorylated H3 T45 ( H3T45-P ) were detected by IB . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 014 The results therefore indicate that Pkc1 mediates histone H3 phosphorylation under conditions of replicative stress . Given the relatively close proximity of T45 to K56 in the histone H3 tail and the link of modifications of both of these residues to Pkc1 activity , we next asked whether there is a connection between H3T45 phosphorylation and H3K56 acetylation . Importantly , H3K56 acetylation was greatly reduced in cells containing the H3 ( T45A ) allele upon release from a HU block ( Figure 7A , lanes 3 and 4; Figure 7B ) . In contrast only a small change in H3K56 acetylation was observed upon release from an alpha factor block under normal growth conditions ( Figure 7A , lanes 1 and 2 ) . However , if cells were released from an alpha factor-mediated G1 arrest into HU , and the HU subsequently removed , H3K56 acetylation again accumulated following HU removal in wild-type cells . This accumulation of H3K56 acetylation did not occur in cells harbouring the H3 ( T45A ) mutant allele ( Figure 7—figure supplement 1 ) . These results suggested functional coupling between H3T45 phosphorylation and H3K56 acetylation . To further probe this phenomenon , we asked whether the Rtt109 ( T46D ) mutant could bypass the defects seen in cells containing the H3 ( T45A ) mutant allele . However , only low levels of H3K56 acetylation were observed in the presence of Rtt109 ( T46D ) ( Figure 7C ) , demonstrating that this protein was unable to suppress the defects from H3 ( T45A ) thereby indicating a crucial role for H3T45 phosphorylation in promoting H3K56 acetylation . Indeed , the Rtt109 ( T46D ) mutant couldn’t bypass the growth defects of H3 ( T45A ) cells in HU ( Figure 7D ) . We also tested whether Rtt109 ( D ) could bypass the HU-mediated growth defects of H3 ( K56R ) cells but again this mutant was unable to rescue the growth defect ( Figure 7—figure supplement 2 ) . This further emphasises the importance of Rtt109-mediated H3K56 acetylation in protecting against the deleterious effects of HU treatment , even when H3T45 was available for phosphorylation . These results lead to a model whereby Pkc1 signals through promoting phosphorylation of both Rtt109 and histone H3 to promote H3K56 acetylation . One prediction from this model is that downstream gene expression defects should be at least partially shared by pkc1-14 , H3 ( T45A ) and Rtt109 ( T46A ) mutant strains . Indeed pairwise comparisons of the gene expression changes observed between these strains following release from a HU block , showed highly significant overlaps in genes being upregulated in all cases ( Figure 7E; Figure 5B ) . Importantly though , of the 51 genes showing significant upregulation in the pkc1 and Rtt109 ( T46A ) mutant strains , 41 ( 80% ) were also upregulated in the H3 ( T45A ) mutant strain ( Figure 7E ) . Moreover , consistent co-regulation of genes in the 'DNA helicase' GO term category could be observed across all mutant strains ( Figure 7F ) . These mutants therefore exhibit common defects in gene expression . This finding provides further supporting evidence that adds to the mechanistic links we have uncovered between Pkc1 , Rtt109 , H3T45 phosphorylation and H3K56 acetylation , and strengthens our major discovery that Pkc1 coordinates histone H3 modifications during the replicative stress response . 10 . 7554/eLife . 09886 . 015Figure 7 . Phosphorylation of histone H3 T45 is required for efficient H3 K56 acetylation . ( A and B ) Western blot analysis of K56 acetylation ( H3K56Ac ) and total H3 levels in ( MSY748 ) cells containing plasmids expressing H3 ( WT ) or the H3 ( T45A ) mutant following release from an alpha factor or HU block for the indicated times . Cells were grown in SD media at 30ºC and treated with either alpha factor for 3 hrs or 200 mM HU for 2 . 5 hrs prior to release . Tubulin ( Tub1 ) levels are shown as a loading control . Logarithmically growing ( FXY19 ) cells containing a plasmid expressing the H3 ( K56R ) mutant are shown as a control in ( A ) . ( C ) Western blot analysis of T45 phosphorylation ( H3T45-P ) , K56 acetylation ( H3K56Ac ) , K18 ( H3K18Ac ) and total H3 levels in cells containing plasmids expressing H3 ( WT ) ( MSY748 ) or the H3 ( T45A ) ( H3-T45A ) mutant following release from a HU block for the indicated times . Where indicated cells also expressed Rtt109 ( T46D ) . ( D ) Cell growth assays . 10 fold serial dilutions of cells containing plasmids expressing H3 ( WT ) ( MSY748 ) or the H3 ( T45A ) ( H3-T45A ) plus either Rtt109 ( WT ) or Rtt109 ( T46D ) , were plated onto SD media in the presence ( bottom panel ) or absence ( top panel ) of 100 mM HU . ( E ) Frequency of gene expression changes upregulated in the H3 ( T45A ) mutant strain that are also upregulated in either the rtt109 ( T46A ) or the pkc1-14 cells or in both of these mutant strains . Hypergeometric P-values are shown ( NS = non-significant ) . ( F ) Heat map showing the relative expression of the genes contained in the 'DNA helicase' category across all conditions analysed . DK186 , DZ5 and MSY748 are the WT equivalent strains for the pkc1-14 ( DK1690 ) , rtt109 ( T46A ) ( DZ6 ) and H3 ( T46A ) mutant strains respectively . The experiment was performed by growing all strains in SD media supplemented with 1M sorbitol at 34ºC . Brackets indicate genes that are generally down-regulated in all of the mutant strains upon release from a HU block . ( G ) Model for Pkc1 function through phosphorylating Rtt109 and promoting its ability to acetylate ( Ac ) histone H3 K56 and acting through mediating H3 T45 phosphorylation which in turn enhances K56 acetylation . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 01510 . 7554/eLife . 09886 . 016Figure 7—figure supplement 1 . H3K56 acetylation is lost in the H3 ( T45A ) mutant strain upon release from an alpha factor block . Western blot analysis of H3K56 and H3K18 acetylation levels in MSY748 cells expressing wild-type H3 ( WT ) or mutant H3 ( T45A ) . Cells were synchronised in G1 with a-factor and released into YPD media containing 200 mM HU for 90 min . Cells were then washed and released into fresh YPD media lacking HU for the indicated times and analysed by western blotting . Histone H3 and tubulin ( Tub1 ) are shown as loading controls . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 01610 . 7554/eLife . 09886 . 017Figure 7—figure supplement 2 . Cell growth assays of strains containing histone H3 mutants in the presence of phospho-mimetic versions of Rtt109 . 10 fold serial dilutions of cells containing plasmids expressing H3 ( WT ) ( MSY748 ) , H3 ( T45A ) ( H3-T45A ) or H3 ( T45A ) plus Rtt109 ( T46D ) ( top three rows ) , H3 ( WT ) ( RMY200 ) , H3 ( K56R ) ( FXY19 ) or H3 ( K56R ) plus Rtt109 ( T46D ) ( top three rows ) , were plated onto SD media containing 1 M sorbitol in the presence ( bottom panel ) or absence ( top panel ) of 100 mM HU . DOI: http://dx . doi . org/10 . 7554/eLife . 09886 . 017 Newly synthesised DNA must be efficiently packaged into chromatin to ensure that genome stability is maintained ( reviewed in Groth et al . , 2007; Ransom et al . , 2010; Gurard-Levin et al . , 2014 ) . This must be achieved during both an unperturbed cell cycle and under conditions of replicative stress where the cell cycle must pause to ensure DNA repair occurs before recommencing the cycle . Histone H3 T45 phosphorylation and K56 acetylation are two important modifications that occur during S phase and are required for chromatin reassembly ( Masumoto et al . , 2005; Hyland et al . , 2005; Ozdemir et al . , 2005; Li et al . , 2008; Baker et al . , 2010 ) . Here we show that in S . cerevisiae Pkc1 coordinates the deposition of these two marks under conditions of replicative stress . Previous results had implicated the Cdc7/Dbf4 kinase complex in directly controlling H3T45 phosphorylation during S phase of and unperturbed cell cycle but this kinase did not affect H3K56 acetylation ( Baker et al . , 2010 ) . However , a role under replicative stress conditions was not investigated and residual levels of H3T45 phosphorylation were observed in the cdc7Δ cells , indicating the existence of another H3T45 kinase . Indeed , Cdc7/Dbf4 activity has been shown to be downregulated under conditions of replicative stress , suggesting the involvement of a different histone H3 kinase under these conditions ( Weinreich and Stillman , 1999 ) . Here we focussed on histone H3 modification control under replicative stress conditions and in combination with previous results ( Baker et al . , 2010 ) our data indicate that at least two different kinases are involved in controlling H3T45 phosphorylation , with Cdc7/Dbf4 operating during an unperturbed cell cycle and Pkc1 operating under replicative stress conditions where it also coordinates H3K56 acetylation . In cells subjected to replicative stress it is important to delay chromatin re-assembly until the DNA damage has been resolved , and one way to do this , would be to delay H3 modifications . It is not clear how this delay is achieved but Pkc1 promotes the re-activation of H3 modifications upon removal of replicative stress . Mechanistically , Pkc1 controls H3K56 acetylation through phosphorylating the Rtt109 acetyltransferase . This modification stabilises Rtt109 and thus safeguards its availability for modifying histone H3 following removal of replicative stress and re-entry into the cell cycle . It is possible that Pkc1 might also directly target other elements of the chromatin modification machinery to coordinate efficient chromatin assembly such as occurs with Cdc7-Dbf4 which can also target CAF-1 ( chromatin assembly factor 1 ) ( Gerard et al . , 2006 ) as well as H3T45 . Under replicative stress conditions , H3T45 phosphorylation appears to be critically important for coupled H3K56 acetylation , and this phosphorylation event is promoted by Pkc1-mediated H3T45 phosphorylation . This is likely direct as histone H3 is efficiently phosphorylated at this site in vitro . This leads to a model ( Figure 7G ) whereby Pkc1 promotes the coordinate modification of these two chromatin modifications and H3T45 phosphorylation is functionally coupled to H3K56 acetylation . It is not clear how this coupling is mediated but one possibility is that H3T45 phosphorylation blocks histone binding to DNA , thus allowing H3K56 to be exposed and hence accessible for acetylation by Rtt109 . More complex scenarios are of course possible such as phosphorylated H3T45 acting as binding platform for recruiting a histone mark reader which then coordinates H3K56 acetylation . Indeed , further complexities in the regulatory links between histone H3 modifications and Pkc1-mediated signalling are suggested by the observation that Pkc1 levels appear to be reduced in H3 ( T45A ) mutant strains ( data not shown ) . Future studies will be required to distinguish between and dissect out the underlying molecular mechanisms . One outcome of perturbed Pkc1 activity and its downstream molecular events following HU treatment is likely to be defective chromatin assembly and preliminary evidence suggest that this might be the case ( data not shown ) . Our data show a consistent sensitivity of pkc1 , rtt109 ( T46A ) and H3 ( T45A ) mutants to HU treatment which complements previous data showing that H3 ( K56A ) alleles are also sensitive to HU ( Matsubara et al . , 2007 ) and supports a role in replicative stress . However , the response to other stress inducing conditions such as MMS and 6-AU treatment is less clear . For example , the H3 ( T45A ) strain is sensitive to both MMS and 6-AU whereas the rtt109 ( T46A ) strain is only sensitive to MMS ( data not shown ) . Thus different elements of the regulatory network we have uncovered may be used to respond to different stress inducers . Another open question is how Pkc1 is specifically targeted/activated during replicative stress but one potential regulator would the cell cycle-dependent changes of the cell wall and signalling through the cell wall integrity pathway . Alternatively , other upstream pathways might be involved , and one potential link might be through TOR signalling as TORC2 has been shown to phosphorylate and activate Pkc1 ( Nomura and Inoue , 2015 ) and TOR signalling has also been linked to H3K56 acetylation ( Chen et al . , 2012 ) . Whether such a route operates during replicative stress or is employed under different conditions is not clear but re-emphasises the potential widespread importance of connections between Pkc1 and histone H3 modifications we have identified . Our work may be directly relevant to mammalian systems where PKCδ has been shown to phosphorylate H3T45 under apoptotic conditions ( Hurd et al . , 2009 ) , although others have identified AKT as an additional H3T45 kinase under DNA damage conditions ( Lee et al . , 2015 ) . It is not clear whether PKC is linked to H3K56 acetylation in mammals but this modification is induced by replicative stress and its deposition is mediated by p300/CBP , the functional homologue of Rtt109 ( Das et al . , 2009; Vempati et al . , 2010 ) . Indeed , atypical protein kinase C zeta has been shown to phosphorylate and control the histone acetylation output of CBP in the context of neuronal differentiation ( Wang et al . , 2010 ) . Given these links between PKC and CBP/p300 it appears likely that our findings will be directly relevant to human cancer where oncogene-driven replicative stress is an important contributory event ( Bartek et al . , 2007 ) and therefore warrants further investigation . For bacterial expression: pET28a+His-Rtt109 and pET3a-Tr-Vps75-Flag was kindly provided by Paul Kaufman ( Tsubota et al . , 2007 ) . For yeast expression; pVD67 ( pAS1995; encoding wild-type GFP-Pkc1 ) , pVD61 ( empty vector ) and pVD124 ( kinase-dead GFP-Pkc1 ) were kindly provided by Martha Cyert ( Denis and Cyert , 2005 ) . pAS4104 encoding Myc-Rtt109 , pAS4105 ( myc-Rtt109T46A ) and pAS4106 ( myc-Rtt109T46D ) were made by inserting BamHI-cleaved PCR product generated using the templates pAS4101-4103 and primer pair ADS4723/4724 into pCM188 . pAS4101 was generated by amplifying the cDNA sequence of the RTT109 gene encoding full-length protein from yeast genomic DNA and cloning into pGBKT7 using EcoRI/SalI restriction sites . pAS4102 ( encoding Rtt109[T45A] ) and pAS4103 ( encoding Rtt109[T45D] ) were created by QuikChange mutagenesis ( Stratagene ) using the primer pairs ADS4727/ADS4728 , ADS4729/4730 and pAS4101 as a template . pGal1Vps75-flag was kindly provided by Tom Owen-Hughes . His-tagged fusion proteins were prepared using Ni-NTA agarose beads ( Qiagen , UK ) according to the manufacturer’s protocol and HA epitope-tagged Pkc1p were prepared essentially as described previously ( Darieva et al . , 2012 ) . Recombinant histone H3 was made commercially ( Active Motif ) . Pulldown assays with immunoprecipitated HA-Pkc1 from DZ2 cells and in vitro translated Rtt109 were carried out as described previously ( Pic-Taylor et al . , 2004 ) . To detect epitope-tagged derivatives by western analysis , anti-HA ( Roche ) and anti-tubulin TAT-1 ( CRUK ) , H3K56ac ( Active motif ) , H3K9ac ( Abcam ) , H3K18ac ( Abcam ) , H3 ( Active motif ) , myc epitope ( Santa Cruz ) , H3T45-P ( Active motif ) antibodies and Supersignal west dura substrate ( Pierce ) were used . Rabbit polyclonal monospecific antibody against the phosphothreonine at amino acid 46 on yeast Rtt109 ( Rtt109 T46-P ) was generated from immunizing rabbits with a KLH-conjugated peptide ( H-DDKRVPKST ( PO3H2 ) IKTC-NH2 ) , and then cross-affinity purification of polyclonal antibodies specific to modified Rtt109 T46-P or non-modified pRtt109 was performed ( Eurogentec ) . For immunoblotting analysis we used 1:1000 dilution for both antibodies . Protein kinase assays were performed as described previously ( Darieva et al . , 2012 ) using HA epitope-tagged Pkc1p immunoprecipitated form DZ2 cells and recombinant H3 or His-Rtt109/Vps75 protein complex as a substrate . Proteins were isolated from denaturing polyacrylamide gels and proteolytically digested as described previously ( Darieva et al . , 2012 ) . Digested samples were analysed by LC-MS/MS using an UltiMate® 3000 Rapid Separation LC ( RSLC , Dionex Corporation , Sunnyvale , CA ) coupled to an Orbitrap Elite ( Thermo Fisher Scientific , Waltham , MA ) mass spectrometer . Peptide mixtures were separated using a gradient from 92% A ( 0 . 1% FA in water ) and 8% B ( 0 . 1% FA in acetonitrile ) to 33% B , in 44 min at 300 nl min-1 , using a 75 mm x 250 μm i . d . 1 . 7 μM BEH C18 , analytical column ( Waters ) . Peptides were selected for fragmentation automatically by data dependant analysis . Data were analysed as described previously ( Darieva et al . , 2012 ) but phosphorylated peptide product ion spectra were also manually validated . The yeast strains used are listed in Supplementary file 1 . Yeast cells were grown , GAL1-promoter-driven constructs induced , and transformations performed , as described previously ( Darieva et al . , 2003 ) . Yeast cultures were treated with 40 nM cercosporamide ( Santa Cruz ) to inhibit PKC activity , synchronised in G1 phase by treatment with α-factor and in early S phase by hydroxyurea ( HU ) treatment as described previously ( Darieva et al . , 2012 ) . DNA content analyses were performed using propidium iodide-stained cells as described previously ( Darieva et al . , 2006 ) . Growth sensitivity tests were performed as described previously ( Darieva et al . , 2012 ) except that temperature sensitive pkc mutant strains were grown at 30°C or 34°C . The experiments were performed in duplicate and were repeated several times . RNA extraction and real-time reverse transcription-PCR ( RT-PCR ) analysis were carried out as described previously ( Darieva et al . , 2006 ) using the following primer pairs: ADS1587/1588 ( HTT2 ) , ADS4732/4733 ( RTT109 ) and ADS3978-ADS3979 ( 18S rRNA ) . All data were normalized to 18S rRNA levels in the same cells . Duplicate RNA samples were labelled and hybridised after a single round of amplification to Affymetrix Yeast Genome 2 . 0 array chips . All the cell strains used in this experiment were grown in the presence of 1M sorbitol at 30ºC , treated with 200 mM HU for 2 . 5 hr and released from HU block at 34ºC . Samples were obtained from: ( 1 ) Wild-type ( DK186 ) and pkc1-14 ( DK1690 ) mutant strains . The mRNA expression data were collected at three timepoints ( wild-type blocked in HU , wild-type and pkc1-14 released from HU block for 75 min . ( 2 ) rtt109Δ mutant cells ( Y01490 ) containing vectors expressing wild-type ( WT ) Rtt109 ( DZ5 ) or Rtt109 ( T46A ) ( DZ6 ) . The mRNA expression data were collected at three timepoints ( WT blocked in HU , and Rtt109 ( WT ) and Rtt109 ( T46A ) released from HU block for 80 min . ( 3 ) Wild-type cells ( MSY748 ) expressing wild-type histone H3 ( MSY748 ) or H3 ( T45A ) ( H3-T45A ) . The mRNA expression data were collected at three timepoints ( WT blocked in HU , and both WT and H3 ( T45A ) released from HU block for 80 min ) . The measurements were then normalised by using robust multi-array average ( RMA ) approach ( Irizarry et al . , 2003 ) . The primary data are deposited in ArrayExpress; accession number E-MTAB-3359 . Limma ( Smyth , 2005 ) was used to provide log ( foldChange ) , T-statistic and P-values . In all cases , fold changes were calculated relative to the wild-type strain released from HU block . Individual gene expression changes were considered significant if they satisfied the criteria of fold change >1 . 2 and P-value <0 . 05 . The T-statistics , P-values and Fold Changes were used by Piano ( Väremo et al . , 2013 ) to build aggregate P-values for all of the genes in each GO term , based upon the values of these statistics for each of the probe sets matching to the term . GO terms were then ranked separately by aggregate P-values , within each of the five P-value directional classes ( upregulated or downregulated in both conditions ) , and this ranking provided a consensus score . To create heatmaps of expression changes of genes within GO terms z-scores were calculated for the set of arrays testing a pair of strains across different conditions and an average z-score for each condition determined . Each paired experiment was treated independently and the z-scores were clustered and depicted in a heatmap using pheatmap ( http://cran . r-project . org/web/packages/pheatmap/index . html ) .	Prior to cell division , DNA must be copied so that each new cell gets a complete copy of the cell’s genetic instructions . But DNA is so long that it is stored in a heavily compacted form in the nucleus of the cell , with the strands of DNA coiled around several proteins called histones . Before the DNA is copied , it must be unfurled . Then each new copy of DNA must be repackaged to fit compactly inside the nucleus of each new cell . If errors occur in the process of copying DNA , it can lead to genetic mutations that may cause diseases like cancer . To prevent this , cells have mechanisms to identify errors and correct them before the DNA is repackaged . This requires a pause to allow the repairs to occur before the DNA recoils . However , it is not completely clear how this process is controlled . Now , Darieva et al . show that an enzyme called protein kinase C ( or Pkc1 for short ) is essential to repackaging DNA after the errors are corrected . Several experiments showed that Pkc1 plays an important role when cells were exposed to stressful conditions that potentially cause errors in DNA copying . Specifically , Pkc1 helps prepare the third histone protein ( histone H3 ) so that DNA can recoil around it . Pkc1 waits until the stressful conditions have passed and the DNA has been repaired to make the necessary changes . Once the stress has passed , Pkc1 adds a phosphate to another enzyme called Rtt109 that prepares the histone . The Pkc1 simultaneously contributes to another necessary change to histone H3 . These new details about DNA repackaging may help researchers understand how cells protect against DNA copying errors , and how this process goes wrong in cancer .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "genetics", "and", "genomics" ]	2015	Protein kinase C coordinates histone H3 phosphorylation and acetylation
Seizures are often followed by sensory , cognitive or motor impairments during the postictal phase that show striking similarity to transient hypoxic/ischemic attacks . Here we show that seizures result in a severe hypoxic attack confined to the postictal period . We measured brain oxygenation in localized areas from freely-moving rodents and discovered a severe hypoxic event ( pO2 < 10 mmHg ) after the termination of seizures . This event lasted over an hour , is mediated by hypoperfusion , generalizes to people with epilepsy , and is attenuated by inhibiting cyclooxygenase-2 or L-type calcium channels . Using inhibitors of these targets we separated the seizure from the resulting severe hypoxia and show that structure specific postictal memory and behavioral impairments are the consequence of this severe hypoperfusion/hypoxic event . Thus , epilepsy is much more than a disease hallmarked by seizures , since the occurrence of postictal hypoperfusion/hypoxia results in a separate set of neurological consequences that are currently not being treated and are preventable . For proper neuronal functioning , brain tissue oxygen levels are normally maintained through exquisite regulation of blood supply ( Erecińska and Silver , 2001 ) . However , during neurological events when brain tissue oxygen levels fall below the severe hypoxic threshold ( pO2 < 10 mmHg ) , neuronal dysfunction and behavioural disturbances are observed ( Farrar , 1991; van den Brink et al . , 2000; Maloney-Wilensky et al . , 2009 ) . Following termination of a seizure , behavioral impairments related to the specific brain structures that participated in the seizure are expressed ( Leung et al . , 2000; Gallmetzer et al . , 2004 ) . Todd’s Paresis , for example , specifically refers to moderate to severe motor weakness following seizures and typically subsides within a few hours ( Todd , 1849 ) . The symptoms are so similar to ischemic stroke that Todd’s paresis is often misdiagnosed ( Mathews et al . , 2008; Masterson et al . , 2009 ) . The occurrence of these behavioral impairments have been attributed to the affected cortex being ‘exhausted’ ( Franck and Pitres , 1878 ) or silenced due to increased inhibition ( Gowers , 1901 ) , but these conjectures are not supported . Despite being first described over 160 years ago , the cause of postictal behavioral disturbances has not been determined . A few case studies of persons experiencing Todd’s paresis demonstrated that the affected cortex was inadequately perfused during this period ( Mathews et al . , 2008; Yarnell and Paralysis , 1975; Rupprecht et al . , 2010 ) . Thus , we hypothesized that following termination of a seizure , a long-lasting and severe hypoxic episode would occur and manifest as behavioral dysfunction . This simple idea is intriguing , testable and provides a critical new insight into this broad reaching neurological disease by explaining some of the detrimental consequences associated with epilepsy . But to date , a continuous and detailed examination of local tissue oxygenation and blood flow following seizures has not been carried out . Most investigations of the hemodynamics associated with seizures have concentrated on changes before or during the seizure and not during the postictal period . It is well established that there is a dramatic , transient increase in blood flow during seizures ( Gibbs et al . , 1934; Penfield et al . , 1939 ) with a corresponding decrease in oxygenation that quickly recovers ( Bahar et al . , 2006; Suh et al . , 2006 ) . The few studies that investigated changes in blood flow during the postictal period have not yielded consistent observations with reports of local hypoperfusion ( Rowe et al . , 1991; Newton et al . , 1992; Leonhardt et al . , 2005 ) and hyperperfusion ( Fong et al . , 2000; Tatlidil , 2000; Hassan et al . , 2012 ) . These inconsistencies were likely due to the variable time-points after the seizure when blood flow was measured . To circumvent this problem we systematically investigated local oxygen levels and blood flow following evoked and spontaneous seizures in rats and mice . We then relied on those results to guide our clinical study and used arterial spin labeling ( ASL ) MRI to measure postictal hypoperfusion following spontaneous seizures in people with epilepsy . We further hypothesized that the enzyme cyclooxygenase-2 ( COX-2 ) , which is primarily responsible for catalyzing arachidonic acid to the PGH2-derived prostanoids ( Hla and Neilson , 1992 ) and plays a major role in neurovascular coupling ( Niwa et al . , 2000; Lecrux et al . , 2011; Lacroix et al . , 2015 ) , would mediate postictal hypoperfusion/hypoxia . We reasoned that electrographic seizures would engage this system and lead to a pronounced vasoconstriction and subsequent severe hypoxia . Downstream of this , we anticipated that elevated free calcium in vascular smooth muscle , particularly calcium conducted by L-type channels ( Putney and McKay , 1999 ) , would sustain this vasoconstriction . Here we demonstrate that blocking these pathways prevents postictal severe hypoperfusion/hypoxia and spotlights its contribution to postictal behavioral impairments . A bipolar electrode and oxygen-sensing probe ( optode ) were chronically implanted into CA1 and CA3 , respectively , of the dorsal hippocampus ( Figure 1L ) . The optode continuously recorded the partial pressure of oxygen ( pO2 in mmHg ) in awake , freely-moving rats before , during and , most importantly , after brief electrographic seizures over several weeks . During typical behavioral states , hippocampal pO2 levels vary within the normoxia range between 18 and 30 mmHg ( Figure 1A ) , as previously reported ( Erecińska and Silver , 2001 ) . 10 mmHg oxygen was chosen as the threshold for defining severe hypoxia since several independent studies have demonstrated that pO2 levels at or below 10 mmHg cause significant changes to cellular physiology and brain injury . Hypoxia-dependent gene expression via Hypoxia-Inducible Factor 1 transcriptional activation occurs as an exponential function with respect to decreasing pO2 ( Jiang et al . , 1996 ) . In a HeLa cell culture system , 10–15 mmHg was determined to be the half-maximal response and since the steep portion of the exponential curve is at lower pO2 levels , most gene expression occurs below this threshold . Restricting middle cerebral artery blood flow in the cat led to cell death at a threshold of 7–8 mmHg ( Farrar , 1991 ) . Furthermore , both the depth and duration of severe hypoxia following traumatic brain injury are good predictors for clinical outcome ( van den Brink et al . , 2000; Maloney-Wilensky et al . , 2009 ) . 10 min of pO2 below 10 mmHg was associated with increased risk of death , and further increases with lower pO2 and longer duration of severe hypoxia ( van den Brink et al . , 2000 ) . A systematic review of clinical traumatic brain injury found that 10 mmHg was a suitable threshold for defining severe hypoxia associated with worse outcomes ( Maloney-Wilensky et al . , 2009 ) . In defining tumor hypoxia , which is thought to exacerbate cancer pathophysiology , 8–10 mmHg is the suggested threshold ( Höckel and Vaupel , 2001 ) . Thus , 10 mmHg is a reasonable threshold for defining severe hypoxia and integrating the entire area below 10mmHg combines the depth and duration to reveal the total hypoxic burden . 10 . 7554/eLife . 19352 . 003Figure 1 . Seizures induce severe postictal hypoxia . ( A ) Local tissue oxygenation in the hippocampus of an awake , freely-moving rat ( blue ) . Green denotes normoxia while red denotes severe hypoxia . ( B ) Representative oxygen profile before , during , and after a spontaneous seizure . The inset expands the time-scale during the seizure with the green and red lines denoting the beginning and end of an 80 s seizure . This inset corresponds to the white vertical block near the beginning of the full oxygen recording . ( C ) Representative oxygen profile before , during , and after a 106 s electrically kindled seizure . ( D ) Scatterplot of the relationship between the duration of kindled seizures in the dorsal hippocampus ( primary afterdischarge ) and the degree of severe hypoxia expressed as the total area below the severe hypoxic threshold ( 10 . 0 mmHg ) by time ( min ) . Symbols with both the same colour and shape are from the same animal ( n = 14 ) . The line of best fit ( y = 11 . 85x+7 . 57 ) is indicated . R square = 0 . 55 , p<0 . 0001 . ( E–J ) Hypoxyprobe immunohistochemistry . ( E ) Representative image from control rat with close-ups of CA3 ( H ) and hilus ( G ) . ( F ) Representative image following a seizure with close-ups of CA3 ( J ) and hilus ( I ) . Scale bar for ( E , F ) = 400 µm . Scale bar for ( G–I ) = 150 µm . ( K ) Densely stained neurons in the hilus and CA3 were quantified . There are significantly more stained cells in the hilus and CA3 following seizures . Data are mean ± SEM . p<0 . 05 ( t-test ) . ( L–N ) Comparing hippocampus and neocortex . ( L ) Location of chronic hippocampal implants . ( M ) Location of chronic neocortical implants . Bipolar electrodes were used for stimulating and recording seizure and O2 sensors for continuous oxygen recordings . C . C . is corpus callosum , L5 is layer 5 . ( N ) Inset displays mean seizure duration ±SEM ( n = 5 ) . Hippocampus had significantly longer seizures . p<0 . 05 . The mean pO2 ( opaque ) ±SEM ( transparent ) over time recorded in motor neocortex dorsal hippocampus . ( O ) Quantification of ( N ) . Hypoxia was more severe in the hippocampus relative to neocortex as assessed by the area below 10mmHg . p<0 . 05 ( t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 00310 . 7554/eLife . 19352 . 004Figure 1—figure supplement 1 . Postictal severe hypoxia generalizes to other seizure models . Both stimulations resulted in severe hypoxia . A 40 s seizure was elicited with MES stimulation and resulted in 104 . 1 min of severe hypoxia ( red ) . Following 2 min of 3 Hz stimulation , hippocampal pO2 dropped below 10 mmHg for 90 . 4 min ( orange ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 004 Oxygen levels were then monitored in the intrahippocampal kainate model of temporal lobe epilepsy ( Rattka et al . , 2013 ) before , during , and after a spontaneous seizure . We measured a brief and small drop in hippocampal oxygenation at seizure onset and a subsequent increase in oxygenation during the electrographic seizure . However , oxygen levels dropped precipitously to below the severe hypoxic level ( pO2 < 10 mmHg ) following seizure termination ( Figure 1B ) . Even more striking was that the severe hypoxia was sustained for an hour . We then systematically examined this phenomenon in a higher throughput model of evoked focal seizures: electrical kindling . Using electrical kindling we elicited an afterdischarge ( i . e . electrographic seizure ) by applying brief stimulation ( 1 ms pulse widths at 60 Hz for 1 s ) through the chronically implanted bipolar electrode . A brief dip in oxygenation at seizure onset and subsequent increase was observed ( Figure 1C ) . Importantly , the severe hypoxic event following a kindled seizure was markedly similar in magnitude and duration compared to spontaneous seizures ( Figure 1C ) . Since no substantive differences in postictal hypoxia were observed between the two models , we chose to evoke afterdischarges with kindling stimulation as our principal model for subsequent experiments to maximize data collection efforts . Daily repeated stimulation of the dorsal hippocampus gave rise to a wide range of primary afterdischarge durations across rats ( 10–50 s ) and allowed us to determine that there is a strong positive and linear relationship between the duration of the hippocampal seizures and the severity of postictal hypoxia , as revealed by integrating the area below 10 mmHg ( Figure 1D ) . To provide convergent evidence for this postictal hypoxia phenomenon , we also used an immunohistological staining technique based on pimonidazole ( HypoxyprobeTM-1 ) , which selectively labels hypoxic cells ( Chapman et al . , 1981; Höckel and Vaupel , 2001 ) . Rats were injected with pimonidazole and afterdischarges elicited . Since postictal severe hypoxia in the rat hippocampus lasted a mean of 73 . 1 ± 7 . 8 min ( Figure 1N ) , we examined brains one hour post-seizure . We observed significantly increased numbers of pimonidazole-labeled cells in the dentate hilus and CA3 regions ( Figure 1E–K ) relative to non-seizure controls . Thus , two different and unrelated techniques identified a severe hypoxic period following brief seizures . We then asked the question whether the phenomenon could be observed in a different structure . Rats were chronically implanted with an electrode in the corpus callosum to elicit neocortical afterdischarges and an optode was implanted into layer V of motor neocortex ( Figure 1M ) . Postictal severe hypoxia was also present in the motor cortex where the typically shorter neocortical afterdischarges gave rise to higher minimum pO2 and a more rapid return to baseline oxygen levels , relative to hippocampus ( Figure 1N ) . Quantification of severe hypoxia in these two structures revealed significantly worse severe hypoxia in the hippocampus ( Figure 1O ) , likely due to longer seizure durations . To further ensure that postictal hypoxia was not specific to the method of seizure induction we also tested two other experimentally-induced seizure models . Maximal electroconvulsive shock ( MES ) which models generalized convulsions ( Young et al . , 2006 ) and 3 Hz electrical stimulation , which drives epileptiform activity and behavioral seizures ( Teskey and Racine , 1993 ) both induced a long-lasting hypoxic event in the hippocampus ( Figure 1—figure supplement 1 ) similar to those observed following spontaneous or electrical kindled seizures . We conclude that a postictal severe hypoxic event follows epileptiform seizures regardless of the eliciting technique . Severe hypoxic events are often the consequence of inadequate blood flow mediated by vasoconstriction and determining the contribution of reduced blood flow to hypoxia can identify new routes for intervention . We first determined the relationship between local tissue hypoxia and blood flow using an implantable laser Doppler flowmetry ( LDF ) probe to measure relative changes in hippocampal blood flow while simultaneously monitoring pO2 levels in the CA3 region of the rat dorsal hippocampus ( Figure 2A ) . Figure 2B displays concurrent measures of mean tissue pO2 with blood flow . Blood flow increased during afterdischarges , which was consistent with previous observations ( Metzger , 1977 ) . Importantly , the postictal severe hypoxia is accompanied by reduced blood flow to 53 . 0 ± 8 . 3% relative to baseline between 40 and 60 min post-seizure . Furthermore , blood flow recovered on a similar time-scale to tissue oxygenation indicating that the postictal severe hypoxia is , in part , related to hypoperfusion . 10 . 7554/eLife . 19352 . 005Figure 2 . Seizures cause postictal vessel constriction in an in vitro preparation and reduced blood flow ( hypoperfusion ) in vivo . ( A ) Location of implants for simultaneous blood flow and pO2 recordings . These two probes were placed at opposing angles to leave room for cable attachment . ( B ) The simultaneous measurement of mean blood flow and mean pO2 in the hippocampus following brief seizures ( n = 5 ) . ( C ) Nifedipine pre-treatment ( 15 mg/kg ) caused an elevation of baseline pO2 , inhibited severe hypoxia , and increased the rate of recovery ( n = 5 ) . Inset reveals no difference in seizure duration . ( D ) Quantification of ( C ) . Nifedipine pre-treatment reduced the amount of severe hypoxia ( area below 10mmHg ) . p<0 . 05 ( within-subject ANOVA ) . ( E ) Validation of 3 Hz stimulation in young rats ( P28–P35 ) . Mean oxygen profiles following standard kindling and 3 Hz stimulation ( n = 6 ) . ( F ) Quantification of ( E ) . No significant differences were found in the amount of severe of hypoxia ( area below 10 mmHg ) . ( G ) Acute hippocampal in vitro slice preparation for measuring postictal vessel constriction . 3 Hz stimulation was applied to the Schaeffer collaterals and imaging was captured in stratum radiatum of CA1 . ( H , I ) Representative images of CA1 arteriole pre-stimulation ( H ) and post-stimulation ( I ) . Scale bar for ( H ) and ( I ) is 15 µm . ( J ) Lumen diameter over time in stimulated and sham controls . Mean lumen diameter is reduced following 3 Hz stimulation ( n = 11 ) relative to sham ( n = 8 ) between 60 and 90 min post-stim . Data displayed as mean ± SEM . p=0 . 001 ( t-test ) . ( K ) 30 min following nifedipine ( 50 µM; n = 7 ) or vehicle ( n = 5 ) application slices were stimulated . Mean lumen diameter was significantly different between 60 and 90 min post-stim . p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 00510 . 7554/eLife . 19352 . 006Figure 2—figure supplement 1 . Post-seizure administration of nifedipine prevents severe postictal hypoxia . ( A ) Nifedipine ( 15 mg/kg ) was administered immediately after a seizure . No differences were observed in seizure duration , however nifedipine initiated a quicker return to baseline pO2 ( n = 5 ) . ( B ) Quantification of ( A ) . Nifedipine ( post-seizure administration ) caused a significant reduction in severe hypoxia . p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 006 If the postictal hypoxia is due to a vascular event then blocking calcium conductance should reduce vasoconstriction and prevent hypoxia . Pre-treatment with the L-type calcium channel antagonist nifedipine 30 min prior to seizure elicitation did not alter the afterdischarge duration ( Figure 2C ) indicating that nifedipine effects can be attributed to a neurovascular mechanism . We observed higher pO2 levels at baseline ( +9 . 0 mmHg , p<0 . 01 , paired ANOVA ) , throughout the postictal period , and an earlier return to baseline relative to vehicle ( Figure 2C ) . Moreover , nifedipine pre-treatment significantly attenuated the area below the severe hypoxic threshold ( pO2 < 10 mmHg ) ( Figure 2D ) . We also observed that nifedipine was effective at reducing postictal severe hypoxia when administered immediately following termination of the afterdischarge ( Figure 2—figure supplement 1A , B ) . This is not surprising since L-type calcium channels play a key role in the tonic phase of vascular smooth muscle contraction ( Putney and McKay , 1999 ) . These results provide key evidence that hypoxia is largely mediated by hypoperfusion and that L-type calcium channel activity plays a role in sustaining this response . While laser Doppler flowmetry provides inferential evidence that blood flow was reduced postictally , we sought a direct measure of changes in blood vessel diameter following epileptiform-like activity in the hippocampus . In young ( P25-P40 ) freely moving rats , we first determined that 2 min of 3 Hz stimulation resulted in a remarkably similar profile of severe hypoxia compared to the standard 1 s stimulation evoked afterdischarges ( Figure 2E , F ) . We then moved the 3 Hz stimulation protocol to an acute hippocampal slice preparation ( Figure 2G ) and imaged hippocampal arterioles in the synaptic layer of CA1 ( stratum radiatum ) before , during , and after Schaffer collateral stimulation ( Figure 2H , I ) . We observed sustained 13 . 8 ± 4 . 0% arteriole constriction following stimulation ( 60–90 min post-stim , Figure 2J ) . Poiseuille’s law dictates that a 13 . 8% reduction in diameter would lead to blood flow that is 55 . 2% of baseline , which correlates well with our laser Doppler flowmetry measurements ( 53 . 0 ± 8 . 3% at 40–60 min , Figure 2B ) . Pretreatment with nifedipine ( 50 µM ) caused significant vessel dilation ( 108 . 4 ± 2 . 9% relative to baseline , p<0 . 05 , paired t-test ) and prevented 3 Hz stimulation-induced vessel constriction ( Figure 2K ) . Our data obtained using laser Doppler flowmetry , nifedipine treatment , and arteriole constriction in slice all provided convergent evidence that severe hypoxia following seizures is mediated by vasoconstriction and hypoperfusion . Based on our time-course data generated from model systems , we asked whether there was postictal hypoperfusion within one hour of a spontaneous ictal event in epilepsy patients . All patients had intractable focal epilepsy and the mean age was 32 years ( range 20–42 years ) . Two patients were excluded from the study due to excessive motion during the acquisition of postictal ASL images , leaving 10 patients for analysis ( Table 1 ) . The mean age at seizure onset was 12 . 3 years ( range one month-29 years ) . Four patients had normal MRI scans , three had malformation of cortical development , one had an occipital cavernoma , one had nonspecific changes and one had postsurgical changes as a result of previous surgery . Cerebral blood flow ( CBF ) was quantified in postictal and baseline ( interictal , >24 hr without a seizure ) ASL scans . Postictal ASL images were subtracted from baseline scans to reveal areas of hypoperfusion . Maximal changes in blood flow were quantified in mL/100 g/min and the brain area where this change occurred was compared to ictal EEG to determine concordance between local hypoperfusion and seizure activity ( Table 2 ) . 10 . 7554/eLife . 19352 . 007Table 1 . Characteristics of patients recruited to ASL study . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 007Patient Age Age at Seizure Onset Structural MRI findings ASL-0013834NormalASL-0034113Lesion in the right superior temporal gyrus , resolving ? ASL-004255NormalASL-006337NormalASL-0073329Bilateral subependymal heterotopias along lateral ventricle right > leftASL-0084222Right amygdala enlargement , Right occipital cavernomaASL-009409NormalASL-010261Left frontal malformation of cortical developmentASL-0112214NormalASL-012200 . 5Postsurgical changes in the right frontal lobe , right anterior temporal lobectomy10 . 7554/eLife . 19352 . 008Table 2 . Concordance of postictal ASL hypoperfusion with brain areas involved in the seizure . EEG localization of ictal activity was determined by an Epileptologist based on scalp EEG . GTC denotes when a generalized tonic-clonic seizure occurred . Areas of maximal hypoperfusion were identified by a blinded reviewer and Regions of Interest were drawn around those areas and adjacent confluent voxels for quantification ( Figure 2—figure supplement 1A , B ) . Concordance was noted if ictal activity overlapped with the area of maximal hypoperfusion . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 008Patient EEG localization of seizure [duration ( S ) ] Area of maximal hypoperfusion for ASL quantification ROI volume ( cm3 ) Concordant ASL-001Left temporal [71]Left temporal2 . 52yesASL-003Unclear [26]No Change1 . 45n/aASL-004Left frontocentral [87]Left superior posterior temporal0 . 31yesASL-006Left fronotemporal [114]Left insula , left anterior temporal1 . 46yesASL-007GTC , Left hemisphere . Maximal posterior temporo-parietal [64]Multifocal bihemispheric1 . 35yesASL-008GTC , Right temporo-occipital [155]Multifocal bihemispheric1 . 63yesASL-009Left fronto-central [85]Multifocal bihemispheric0 . 76yesASL-010GTC , Bifrontal , maximum left [166]Left frontal3 . 82yesASL-011Left fronto-temporal [56]No Change1 . 10n/aASL-012Bitemporal right>left [30]Multiple areas over right temporal posterior and superior to resection cavity2 . 19yes Maximal postictal cerebral blood flow reductions of at least 10 mL/100 g/min were seen in 8 of 10 subjects ( Table 2 ) . Looking at the region of interest ( ROI ) for all 10 subjects , we observed a mean baseline CBF of 59 . 0 ± 4 . 4 mL/100 g/ and mean CBF of 42 . 1 ± 3 . 4 mL/100 g/min following seizures , for a difference of 16 . 9 ± 4 . 1 mL/100 g/min or 26 . 6 ± 6 . 0% from baseline ( Figure 3—figure supplement 1A ) . Furthermore , the magnitude of hypoperfusion was positively correlated with seizure duration ( Figure 2—figure supplement 1B ) , just as we saw with the magnitude of hypoxia and seizure duration in the rat ( Figure 1D ) . In the two subjects without observable hypoperfusion of >10 mL/100 g/min , the seizure durations were notably shorter and may account for the absence of hypoperfusion . Excluding these two patients , CBF of 65 . 0 ± 3 . 6% relative to baseline was observed in the postictal scan . While we observed a greater change in the rat ( 53 . 0 ± 8 . 3% relative to baseline ) , the ROI for ASL quantification also included adjacent confluent voxels of less magnitude , which would decrease the reported change , and the relative heterogeneity of subjects should be noted . This heterogeneity of subjects , however , allowed us to investigate the overlap between the brain areas of maximal hypoperfusion and areas of seizure activity to determine the specificity of this phenomenon . Two examples are shown in Figure 3 ( ASL 001 , 010 ) and provide qualitative confirmation of this specificity in frontal and temporal lobe epilepsy . In all eight cases where hypoperfusion >10 mL/100 g/min was observed , we noted that the area of maximal hypoperfusion overlapped with a brain structure subjected to seizure activity ( Table 2 ) . Thus , hypoperfusion is specifically localized to those areas involved in the seizure and is indeed a local phenomenon . 10 . 7554/eLife . 19352 . 009Figure 3 . Representative seizure-specific local postictal hypoperfusion in clinical epilepsy . ( A ) ASL-010 , 26 year old female with drug resistant focal epilepsy . Ictal EEG recording on a longitudinal bipolar montage with left frontal seizure onset and spread to the right frontal region . Green bars indicate seizure onset . indicates the electrodes to localize seizure onset . Scale bar = 1 s . and also applies to ( D ) . ( B ) Seizure description and MR . Fluid-attenuated Inversion recovery ( FLAIR ) MR images demonstrating area of poor gray-white matter differentiation with subcortical white matter hyperintensity over the left frontal region ( arrows ) . R and L indicate right and left sides and also apply to ( C ) and ( F ) . ( C ) Subtraction CBF map ( inter-ictal – post-ictal ) superimposed onto the patient’s T1-weighted anatomical image indicating areas of left frontal hypoperfusion > 20 mL/100 g/min ( > 30% reduction compared normal gray matter CBF ) . ( D ) ASL-001 , 38 year old right handed male with intractable non-lesional epilepsy . Ictal EEG recording on a longitudinal bipolar montage with left temporal seizure onset ( green bar ) and spread to the left parasagittal region ( orange bar ) . ( E ) Seizure description . Patient was non-lesional . ( F ) Subtraction CBF map shows profound hypoperfusion ( >15 mL/100 g/min ) in left temporal lobe . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 00910 . 7554/eLife . 19352 . 010Figure 3—figure supplement 1 . Clinical postictal hypoperfusion is more severe with longer seizures . ( A ) Cerebral blood flow from baseline and postictal ASL scans . Each point represents in individual . Group mean ± SEM are expressed in grey . Patient number is displayed on the left of the corresponding data point ( e . g . one is patient ASL-001 ) . A significant decrease in blood flow was measured in the postictal scan *p<0 . 01 ( within-subject t-test ) . ( B ) Percent decrease in postictal blood flow as a function of seizure duration . A linear regression revealed a significant ( p<0 . 05 ) positive correlation defined by the line y = 0 . 28x+3 . 12 ( R square = 0 . 43 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 010 COX-2 is expressed throughout the rat brain ( Figure 4A ) with particularly dense expression in neurons of the hippocampus ( Figure 4B–D ) . This expression profile and the well-established role in neurovascular coupling ( Niwa et al . , 2000; Lecrux et al . , 2011; Lacroix et al . , 2015 ) made COX-2 a likely candidate responsible for postictal hypoperfusion/hypoxia . We also hypothesized that COX-1 would not play a role since it is involved in the tonic control of brain blood flow through astrocyte signaling ( Rosenegger et al . , 2015 ) . Selective inhibitors for these two main isoforms of COX have been developed and we pre-administered celecoxib or SC-560 to inhibit COX-2 and COX-1 , respectively . Neither drug had any significant effect on afterdischarge duration and as predicted , celecoxib ( Figure 4E , F ) , but not SC-560 ( Figure 4—figure supplement 1A , B ) , prevented postictal severe hypoxia . 10 . 7554/eLife . 19352 . 011Figure 4 . COX-2 activity during a seizure is required for postictal severe hypoxia . ( A ) COX-2 expression in sagittal section . Hippocampal inset is displayed ( see B–D ) . Scale bar = 2 mm . ( B–D ) Contents of inset from ( A ) . Many NeuN-expressing neurons ( B ) in DG and hilus express COX-2 ( C ) . ( D ) displays colocalization of NeuN and COX-2 . Scale bar = 200 µm . ( E ) Celecoxib pre-treatment ( 20 mg/kg ) inhibited severe hypoxia during the postictal period ( n = 5 ) . Inset reveals no difference in seizure duration . ( F ) Celecoxib caused a significant reduction in the area below 10mmHg . p<0 . 05 ( within subject ANOVA ) . ( G ) Acetaminophen ( Ace ) dose-dependently inhibited postictal severe hypoxia ( n = 5 ) . Doses are listed on inset ( 50–250 mg/kg ) . Inset reveals no difference in seizure duration . ( H ) 150 mg/kg and 250 mg/kg of acetaminophen significantly decreased the area below 10mmHg . p<0 . 05 ( within subject ANOVA ) . ( I ) Lumen diameter over time following application of acetaminophen ( 100 µM; n = 7 or vehicle ( n = 5 ) . Acetaminophen treatment prevented post-stimulation constriction ( analyzed between 60–90 min post-stim ) . *p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 01110 . 7554/eLife . 19352 . 012Figure 4—figure supplement 1 . COX-1 or postictal COX-1/2 inhibition do not inhibit severe hypoxia . ( A ) SC-560 ( 20 mg/kg ) pre-administration did not significantly alter seizure duration or the resulting oxygen profile ( n = 5 ) . ( B ) Quantification of ( A ) . No differences in the severe hypoxia were observed with SC-560 pre-treatment . ( C ) Acetaminophen ( 250 mg/kg ) was administered immediately after a seizure . No significant differences were observed in seizure duration or the resulting oxygen profiles ( n = 5 ) . ( D ) Quantification of ( C ) . No significant differences were observed in the severe hypoxia with acetaminophen post-treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 012 We reasoned that a non-selective COX antagonist should also prevent postictal hypoxia . Indeed , pre-administration of acetaminophen dose-dependently inhibited severe hypoxia without altering afterdischarge duration ( Figure 4G , H ) . Unlike nifedipine , administering acetaminophen immediately following termination of the afterdischarge had no effect on the postictal hypoxic profile ( Figure 4—figure supplement 1C , D ) . Since acetaminophen had to be present at the time of the afterdischarge , this suggests that COX-2 activity during the afterdischarge gives rise to prolonged postictal severe hypoxia . We then returned to the in vitro preparation to directly observe the effect of COX inhibition on vasoconstriction . Slices were treated with acetaminophen ( 100 µM ) or vehicle ( 0 . 1% DMSO ) and then stimulated for 2 min at 3 Hz to drive epileptiform-like activity . Indeed , acetaminophen prevented vessel constriction following stimulation ( Figure 4I ) , but unlike nifedipine ( Figure 2K ) , had no effect on vessel diameter prior to stimulation . This mirrors our data obtained in vivo . Taken together , these data support our hypothesis that COX-2 activity during a seizure ultimately leads to sustained vessel constriction and severe hypoxia . To further test the role of COX-2 in the generation of postictal severe hypoxia , we elicited afterdischarges in mice with a genetic knockdown of COX-2 function ( Figure 5A ) . Specifically , these mice ( PTGS2Y385F ) have a single amino acid substitution in cyclooxygenase active site , have intact peroxidase function , and have blunted prostanoid production in response to LPS ( Yu et al . , 2006 ) . Both mutant and C57BL/6J controls exhibit brief afterdischarges , which did not differ in duration in response to electrical kindling stimulation . However , only C57BL/6J controls showed severe hypoxia during the postictal period , while the mutants remained within normoxia ( Figure 5B , C ) . Furthermore , severe hypoxia was blocked by acetaminophen pre-treatment in C57BL/6J controls corroborating our results in rats and no further effect was seen in COX-2 mutants ( Figure 5D ) . 10 . 7554/eLife . 19352 . 013Figure 5 . Genetic knockdown of COX-2 function prevents postictal severe hypoxia . ( A ) COX-2 proteins from wild-type and mutant mice ( PTGS2Y385F ) displayed as a schematic . Both mice have functional peroxidase ( POX ) . Mutants have a point mutation in the cyclooxygenase ( COX ) active site , which prevents the conversion of arachadonic acid ( AA ) to prostaglandin G2 ( PGG2 ) . ( B ) No differences were observed in seizure duration between wild-type and mutant mice ( n = 4 ) . Wild-type mice displayed severe postictal hypoxia , while mutant mice remained normoxic in the postictal period . ( C ) Quantification of ( D ) . COX-2 mutant mice displayed a significant reduction in severe hypoxia ( area below 10 mmHg ) . p<0 . 05 ( t-test ) . ( E ) Wild-type and mutant mice were treated with acetaminophen ( 250 mg/kg ) ( n = 3 ) . Acetaminophen prevented hypoxia in wild-type mice and had no effect in mutant mice . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 013 To better understand how seizures lead to this prolonged period of hypoperfusion/hypoxia we tested several pharmacological agents with diverse mechanisms of action with an eye to identify other potential drug candidates that might prevent hypoperfusion/hypoxia . Data are summarized in Table 3 . 10 . 7554/eLife . 19352 . 014Table 3 . Investigation of mechanisms involved in postictal severe hypoxia . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 014Drug Principle known mechanism of action Δ severity of hypoxia ( Veh−Drug ) Veh × 100%Nifedipine ( 15 mg/kg ) L-type Ca2+ Channel Blocker+78 . 82 ± 7 . 632% *Nifedipine ( 15 mg/kg postictal ) +87 . 33 ± 10 . 49% Acetaminophen ( 250 mg/kg ) COX-1/2 Inhibitor+99 . 88 ± 0 . 12% *Acetaminophen ( 150 mg/kg ) +88 . 88 ± 7 . 355% Acetaminophen ( 50 mg/kg ) +48 . 85 ± 26 . 61% Acetaminophen ( 250 mg/kg postictal ) −3 . 33 ± 17 . 29% Ibuprofen ( 20 mg/kg ) +99 . 23 ± 0 . 396% *Celecoxib ( 20 mg/kg ) COX-2 Inhibitor+67 . 26 ± 27 . 42% SC-560 ( 20 mg/kg ) COX-1 Inhibitor+19 . 42 ± 15 . 37% Celecoxib ( 20 mg/kg ) + SC-560 ( 20 mg/kg ) COX-2 and COX-1 Inhibitors+95 . 80 ± 4 . 21% *Acetaminophen ( 250 mg/kg ) + Nifedipine ( 15 mg/kg ) COX-1/2 Inhibitor + L-type Ca2+ Channel Blocker+100 ± 0 . 00% CAY-10526 ( 2 mg/kg ) Prostaglandin E2 Synthesis Inhibitor+41 . 72 ± 9 . 94% Seratrodast ( 10 mg/kg ) Thromboxane A2 Receptor Antagonist−12 . 91 ± 44 . 08% Ozagrel ( 10 mg/kg ) Thromboxane A2 Synthesis Inhibitor+21 . 24 ± 16 . 42% 2-APB ( 3 mg/kg ) IP3r Antagonist + TRP Channel Blocker+45 . 27 ± 18 . 44% Chelerythrine Chloride ( 15 mg/kg ) PKC Inhibitor+25 . 54 ± 29 . 90% Milrinone ( 3 mg/kg ) Phosphodiesterase-3 Inhibitor−10 . 74 ± 34 . 31% Sildenafil ( 15 mg/kg ) Phosphodiesterase-5 Inhibitor−19 . 34 ± 14 . 03% SKA-31 ( 10 mg/kg ) IKCa Channel Activator−1 . 50 ± 12 . 68% Paxilline ( 2 . 5 mg/kg ) BKCa Channel Blocker+19 . 55 ± 14 . 98% L-Arginine ( 500 mg/kg ) Nitric Oxide Precursor−5 . 05 ± 14 . 08% Fasudil ( 10 mg/kg ) Rho Kinase Inhibitor−53 . 07 ± 17 . 27% All drugs were delivered by intraperitoneal injection pre-seizure ( unless otherwise stated ) . Statistics reported as different from chance ( one sample T-test ) . p < 0 . 05 , p<0 . 01 , p<0 . 001 . + number indicates inhibition of hypoxia . − number indicates potentiation of hypoxia . We further interrogated the COX-2 pathway by pre-administering , ibuprofen , which non-selectively inhibits both COX-1 and COX-2 . Likewise , we also co-administered celecoxib and SC-560 to specifically and concurrently inhibit both enzymes . With both methods , inhibiting COX-1 and-2 achieved greater than 95% inhibition of severe hypoxia , similar to acetaminophen . This is of interest since COX-2 inhibition alone achieved a 67 . 26 ± 27 . 42% reduction in the severity of hypoxia and SC-560 had no significant effect . These results support a role for an interaction between COX-1 and COX-2 , which has been demonstrated through COX-1/COX-2 heterodimer signaling ( Yu et al . , 2006 ) . The conversion of arachidonic acid to PGH2 by COX-2 is followed by PGE2 , PGD2 , PGF2 , PGI2 , and TXA2 production by specific synthases ( Hla and Neilson , 1992 ) and subsequent constriction or dilation though G-protein coupled receptor signaling . Specific inhibitors for each of these pathways are not available , but we were able to target two of them: PGE2 and TXA2 . We inhibited PGE2 synthesis with CAY-10526 , which resulted in a + 41 . 72 ± 9 . 94% reduction in the area below 10 mmHg . Disruption of TXA2 signaling by inhibiting its synthesis ( ozagrel ) or receptor ( seratrodast ) resulted in no significant change from vehicle . Thus , part of the COX-2 signaling pathway can be explained by PGE2 production . Since , nifedipine inhibits calcium entry into vascular smooth muscle , we postulated that inhibiting the increase in free calcium with 2-APB ( primarily through IP3 receptors and TRP channels ) should also inhibit postictal hypoxia . Indeed , we observed a 45 . 27 ± 18 . 44% decrease in the area below 10 mmHg with 2-APB pre-treatment . We then hypothesized that the rise in free calcium could activate protein kinase C ( PKC ) , which can sustain vasoconstriction by inhibiting myosin light chain phosphatase through CPI-17 ( Kitazawa et al . , 2003 ) . However , PKC inhibition was unable to prevent severe postictal hypoxia . Furthermore , activating vasodilatory pathways by increasing cAMP ( milrinone ) , ( Farrar , 1991 ) increasing cGMP ( sildenafil ) , ( van den Brink et al . , 2000 ) hyperpolarizing smooth muscle with IKCa activator ( SKA-31 ) , ( Maloney-Wilensky et al . , 2009 ) preventing astrocytic release of potassium onto smooth muscle via BKCa ( paxilline ) , or ( Leung et al . , 2000 ) increasing production of nitric oxide ( L-arginine ) provided no inhibition of postictal hypoxia . In fact , Rho Kinase inhibition with fasudil achieved the opposite effect and augmented postictal severe hypoxia perhaps by distributing blood flow to areas outside of the afterdischarge zone . Together , these data provide support for the inflexible nature of postictal hypoperfusion/hypoxia , which is not amenable to several vasodilatory treatments , and highlights the central roles COX-2 activation and the sustained calcium elevation in vascular smooth muscle . Despite taking ASDs , many persons with epilepsy still experience breakthrough seizures and 30–40% are completely refractory to ASD treatment . Given the potential negative consequences of prolonged periods of severe local tissue hypoxia , an important question for us was whether the commonly prescribed ASDs modulate this phenomenon . Fortunately , using electrical kindling stimulation well above seizure threshold did not affect seizure duration following ASD treatment and allowed us to observe the effect of ASDs on postictal severe hypoxia without that confound . Acute dosing with commonly prescribed ASDs had no significant effect on postictal severe hypoxia when compared to vehicle treatment ( Table 4 ) . Interestingly , ethosuximide reduced the severity of hypoxia by 47 . 61 ± 11 . 01% . Ethosuximide is used to treat absence seizures by inhibiting T-type calcium channels , but partial inhibition of L-type calcium channels has been reported ( Kostyuk et al . , 1992 ) . 10 . 7554/eLife . 19352 . 015Table 4 . Effect of Anti-Seizure drugs on postictal severe hypoxia . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 015Drug Principal known mechanism of action Δ severe hypoxia ( Veh−Drug ) Veh × 100%Ethosuximide ( 300 mg/kg ) T-type Ca2+ Channel Blocker+30 . 53 ± 19 . 31% Topiramate ( 50 mg/kg ) Na+ Channel Blocker , GABA Enhancement , AMPA Inhibition+4 . 34 ± 14 . 07% Bumetanide ( 2 . 5 mg/kg ) NKCC1 Transporter Inhibitor−0 . 92 ± 14 . 72% Phenobarbital ( 30 mg/kg ) GABA Receptor Agonist−1 . 22 ± 13 . 90% Levetiracetam ( 250 mg/kg ) Glutamate Release Inhibition−14 . 85 ± 19 . 71% Phenytoin ( 75 mg/kg ) Na+ Channel Blocker−15 . 76 ± 17 . 00% Lamotrigine ( 15 mg/kg ) Na+ Channel Blocker−20 . 95 ± 17 . 13% Valproate ( 150 mg/kg ) Na+ Channel Blocker , GABA Enhancement−24 . 89 ± 22 . 65% All drugs were delivered by intraperitoneal injection pre-seizure . Statistics reported as different from chance ( one sample T-test ) . p<0 . 05 . + number indicates inhibition of hypoxia . -number indicates potentiation of hypoxia . In order to determine the role of postictal severe hypoxia in postictal behavioral disruption we first began by addressing Todd’s paresis as it was originally described: motor weakness following seizures that involve motor cortex ( Todd , 1849 ) . We hypothesized that the postictal hypoperfusion/hypoxic period would be responsible for motor impairment and that this impairment would also be confined to the hypoperfusion/hypoxic period . To test this , we used nifedipine as a tool to prevent postictal hypoperfusion/hypoxia , allowing us to dissociate the effects of an afterdischarge from the combined effects of an afterdischarge with hypoperfusion/hypoxia on behavior . We targeted the corpus callosum at the level of the forelimb area of motor cortex ( Teskey et al . , 2002 ) as the site to elicit bilateral afterdischarges . The hanging bar test was used to assess motor weakness . We first measured baseline hang time ( standardized to 100% ) and then elicited an afterdischarge . Hang time was repeatedly measured at 20 , 40 and 80 min postictal ( Figure 6A ) whilst neocortical pO2 was continuously recorded throughout . As expected , pretreatment with nifedipine prevented the expression of postictal severe hypoxia without affecting seizure duration ( Figure 6B ) . Although severely hypoxic at both the 20- and 40 min time-point , the vehicle-treated seizure group displayed a significantly shorter hang time at 40 min postictal ( Figure 6C ) , suggesting that the duration of severe hypoxia is important for the generation of behavioral deficits . At 80 min post-afterdischarge , the non-treated group's pO2 levels returned to normoxia and the behavioral performance was equivalent to nifedipine treated and non-seizure groups . These results indicate that it is the postictal severe hypoxic period that is responsible for the motor weakness , not the seizure per se . 10 . 7554/eLife . 19352 . 016Figure 6 . Preventing postictal severe hypoxia prevents behavioral impairments . ( A ) Experimental timeline for neocortex experiments . Rats were injected with vehicle or nifedipine 30 min prior to a baseline hang time measurement . Neocortical seizures ( or sham ) were then elicited and rats were retested at 20 , 40 , and 80 min post-seizure . ( B ) Nifedipine pre-treatment ( 15 mg/kg ) inhibited severe hypoxia during the postictal period ( n = 4 ) . Inset reveals no difference in seizure duration . Note that during task performance motor cortex pO2 levels temporarily increased while the rat hung from the bar , likely mediated by increased blood pressure , but returned to pre-test levels following release from the bar . ( C ) Performance on the hanging bar task was steady across the experiment , except in the vehicle+seizure group ( hypoxic ) ( n = 4 for seizures , n = 3 for no seizures ) . A significant decrease was observed at 40 min post-seizure , which recovered by 80 min . Data are mean ± SEM . p<0 . 01 ( 2 way ANOVA ) . ( D ) Experimental timeline for hippocampus experiments . Rats were injected with vehicle or acetaminophen ( 250 mg/kg ) 30 min prior to hippocampal seizure induction ( or sham ) . 45 min post-seizure , rats were place in testing environment to encode the memory of 2 identical objects . 24 hr later , memory was tested by measuring the percentage of time spent investigating the novel object . ( E ) No differences were observed in seizure duration ( n = 4 ) . Acetaminophen prevented postictal hypoxia . ( F ) Rats that had seizures and vehicle treatment ( hypoxic ) performed no different from chance and significantly worse than all other groups ( n = 4 ) . Data are mean ± SEM . p<0 . 01 ( ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 016 To determine whether postictal severe hypoxia was the cause of memory disruption we turned to a hippocampal dependent memory task: novel object recognition . Following habituation to the testing environment , hippocampal afterdischarges were elicited in the presence or absence of acetaminophen and rats were exposed to two identical objects at a time of severe hypoxia ( 45 min post-discharge ) . 24 hr later , rats were re-introduced to the testing chamber in the presence of a novel and familiar object ( Figure 6D ) . Since rats prefer to explore the novel object more frequently , a preference for the novel object indicates a memory was formed for the familiar object . Rats that were severely hypoxic at the time of memory encoding ( seizure+vehicle , Figure 6E ) explored the novel and familiar objects with no significant preference 24 hr later , which indicates that they were unable to form the memory during the postictal severe hypoxia ( Figure 6F ) . Moreover , the rats that had seizures without severe hypoxia ( acetaminophen pre-treatment ) performed at the same level as controls ( acetaminophen or vehicle without seizures ) . These results , along with the ability of nifedipine to prevent Todd’s paresis , support the hypothesis that severe hypoxia following seizures , and not the seizures themselves , are the underlying cause of these specific postictal impairments . Using a multi-level approach from animal models to clinical epilepsy , we demonstrate a profound postictal hypoperfusion/hypoxic event . We also demonstrated significant post-ictal hypoperfusion in 8 of 10 patients . The reasons why two subjects did not exhibit significant post-ictal hypoperfusion are unclear , but could relate to the brief duration of their seizures and/or patient heterogeneity . In rodents we identified COX-2 activity during afterdischarges as a central player in mediating this pathophysiological response . Inhibiting COX-2 or restoring perfusion with the potent vasodilator , nifedipine , prevented ensuing severe hypoxia and averted postictal behavioral impairments . Thus , brief seizures initiate a local and long-lasting hypoperfusion/hypoxic insult , which is responsible for behavioral dysfunction in the postictal period rather than the seizure itself . Prior to this report the direction of postictal hemodynamic changes was unclear since conflicting reports indicated either postictal hypoperfusion ( Rowe et al . , 1991; Newton et al . , 1992; Leonhardt et al . , 2005 ) or hyperperfusion ( Fong et al . , 2000; Tatlidil , 2000; Hassan et al . , 2012 ) . This discordance is most likely explained by the variable timing of the single postictal scans . In our rat models we observed the most severe hypoxia occurred between 20–60 min post-seizure thus we performed clinical postictal scans in this time window and observed dramatic local hypoperfusion . Importantly , only those previous studies that imaged during this period observed hypoperfusion , which highlights the importance of timing with this phenomenon ( Rowe et al . , 1991; Newton et al . , 1992; Leonhardt et al . , 2005 ) . Moreover , the timing of postictal scans is an important consideration as these techniques may aid the ability of clinicians to locate the seizure onset zone in potential surgical candidates ( Duncan et al . , 1993 ) . A limitation of this study is the tool used to measure hypoperfusion in the rat . Laser Doppler flowmetry was used for in vivo recordings , which is an indirect measure of blood perfusion . However , we used this system in conjunction with our highly accurate pO2 recordings in the same region of the hippocampus and observed hypoperfusion and hypoxia that display similar kinetics . For a more direct measure of vasoconstriction , we used acute hippocampal slices and observed the diameter of local arterioles . The conditions in slice differ substantially from in vivo conditions and this could potentially alter the results . However , the degree of constriction observed in slice predicts a similar change in blood perfusion to what we observed with laser Doppler flowmetry . Furthermore , the vasoconstriction following seizure stimulation in slice was sensitive to acetaminophen and nifedipine , which supports the notion that this constriction follows the same set of mechanisms that occur in vivo . In conjunction with the clinical observations of hypoperfusion , these data support the central role of hypoperfusion in the generation of severe hypoxia in the postictal period . Prior to this study , few have investigated local tissue oxygenation during the seizure event itself and they identified a transient ‘dip’ at seizure onset that quickly recovers ( Bahar et al . , 2006; Suh et al . , 2006 ) . The longest of these recordings included 30 s of postictal oxygen sampling and took place under ketamine anesthesia , which dramatically reduces neurovascular coupling by inhibiting NMDA-dependent COX-2 activation ( Tran and Gordon , 2015 ) . It is therefore unsurprising that postictal hypoperfusion/hypoxia was not identified in these studies . One group discovered hypoxyprobe-positive neurons following pilocarpine-induced SE and during the period of recurrent spontaneous seizures ( three weeks post SE ) , which they attributed to the hypoxic dip that occurs during seizures ( Gualtieri et al . , 2013 ) . We also observed a transient dip in oxygenation at seizure onset , but the duration of this dip is dwarfed by the prolonged period of severe hypoxia during the postictal period and likely accounts for hypoxyprobe labeling . We identified that COX-2 plays a central role in coordinating a cascade of events to induce severe hypoxia following electrographic seizures . Activity-dependent induction of COX-2 in neurons plays an important role in normal neurovascular coupling by producing vasoactive prostanoids that act on blood vessel receptors to cause vasodilation ( Niwa et al . , 2000; Lecrux et al . , 2011; Lacroix et al . , 2015 ) . Under pathophysiological activation during electrographic seizure activity , we observed that vasodilation is quickly switched to prolonged vasoconstriction . We also found that inhibiting PGE2 synthesis provided partial inhibition of severe hypoxia . The EP1 and EP3 receptor subtypes for PGE2 have been observed to induce vasoconstriction ( Jadhav et al . , 2004 ) and may mediate part of this response . Our study also investigated several other pathways by which postictal severe hypoxia could be mediated and also aimed to identify other drug candidates to inhibit this phenomenon . Of particular note was nifedipine’s inhibition of severe hypoxia even when administered after the seizure . Furthermore , this strategy is important clinically since administering a drug after a seizure avoids the challenges of seizure prediction . We also tested commonly prescribed anti-seizure medications since it would be informative to know which of these medications potentially inhibit , or more importantly , exacerbate postictal severe hypoxia in patient populations . Most drugs , including anti-seizure medications , failed to modulate this postictal severe hypoxia , which we believe emphasizes that this is a pathological phenomenon and that breakthrough seizures while on antiseizure medications do not directly protect against this insult . Postictal behavioral impairments last much longer than the seizures themselves and negatively impact quality of life . The mechanism underlying these disruptions were previously unknown and no treatment options existed . Our data clearly demonstrates that two common postictal impairments , Todd’s paresis and amnesia , can be prevented by interfering with the pathways that lead to hypoperfusion/hypoxia . Thus , L-type calcium antagonists and COX-2 inhibitors are strong candidates for the treatment of these currently untreated behavioral impairments . Our discovery that brief and mild seizures lead to an extended hypoperfusion/hypoxic event is critically important because it establishes that seizures could injure the brain through postictal severe hypoxia , and not necessarily by the seizure itself . Severe hypoxia may be an important component of seizure-induced brain damage ( Jackson et al . , 1999 ) and since postictal hypoperfusion/hypoxia is COX-2 dependent , this hypothesis can be assessed . Though there is some controversy ( Rojas et al . , 2014 ) , it is generally accepted that COX-2 inhibition is neuroprotective from seizures . Indeed , with genetic or pharmacological COX-2 inhibition , seizure-induced brain damage can be dramatically reduced ( Serrano et al . , 2011; Kunz and Oliw , 2001; Hewett et al . , 2006 ) . Though at the time of these studies researchers were unaware of postictal hypoperfusion/hypoxia and the requirement of COX-2 in this response , they provided support for the hypothesis that postictal hypoperfusion/hypoxia injures the brain . Given the central role of brain injury in epileptogenesis ( Sloviter and Bumanglag , 2013 ) , preventing injury from hypoperfusion/hypoxia may be a novel preventative treatment strategy in epilepsy . Young adult male Hooded Long-Evans ( LE ) rats weighing between 250–300 g at the start of experimentation were used in this study ( Charles River , Canada ) . Young adult male C57BL6/J and PTGS2Y385F mice weighing between 21–30 g were also used ( Jackson Laboratory , RRID: IMSR_JAX:008101 ) . Rats and mice were housed individually in clear plastic cages and were maintained on a 12:12 hr light/dark cycle lights on at 07:00 hr , in separate colony rooms under specified pathogen free conditions . Food and water were available ad libitum . All experimental procedures occurred during the light phase . Electrodes were constructed from Teflon-coated , stainless steel wire , 178 µm in diameter ( A-M Systems , Sequim , WA ) . Wire ends were stripped of Teflon and connected to gold-plated male amphenol pins . Rats were anaesthetized with a 5% isoflurane , and maintained between 1% and 2% . Lidocaine ( 2% ) was administered subcutaneously at the incision site . One bipolar electrode was chronically implanted under stereotaxic control ( Paxinos and Watson , 1986 ) in the site of interest with an oxygen-sensing probe positioned nearby ( Table 5 ) . The implants were adhered and anchored to the skull using dental cement and six stainless steel screws . One of the six screws served as a ground electrode . Subsequent experimental procedures commenced no earlier than five days following surgery . 10 . 7554/eLife . 19352 . 017Table 5 . Stereotaxic coordinates for surgical implantation . DOI: http://dx . doi . org/10 . 7554/eLife . 19352 . 017Experiment Anterior ( + ) / Posterior ( − ) Lateral Right ( + ) /Left ( − ) Ventral ( from brain surface ) Rat Dorsal HippocampusElectrode: −3 . 0 mm Optode: −3 . 5 mm Electrode: 0 . 5 mm Optode: 3 . 5 mm Electrode: 3 . 5 mm Optode: 3 . 5 mm Rat Dorsal Hippocampus with LDFElectrode: −3 . 0 mm Optode: −5 . 0 mm LDF probe: −3 . 0 mm Electrode: 0 . 5 mm Optode: 2 . 2 mm LDF probe: 3 . 5 mm Electrode: 3 . 5 mm Optode: 3 . 5 mm LDF probe: 3 . 5 mm Rat Ventral HippocampusElectrode: −4 . 5 mm Optode: −3 . 0 mm Electrode: 4 . 5 mm Optode: 3 . 5 mm Electrode: 6 . 5 mm Optode: 3 . 5 mm Rat Ventral Hippocampus w/ CannulaElectrode: −4 . 5 mm Optode: −3 . 0 mm Cannula: - 5 . 8 mm Electrode: 4 . 5 mm Optode: 3 . 5 mm Cannula: −4 . 5 mm Electrode: 6 . 5 mm Optode: 3 . 5 mm Cannula: 4 . 3 mm Rat NeocortexElectrode:+1 . 0 mm Optode: 0 mm Electrode: 0 . 5 mm Optode: 3 . 0 mm Electrode: 3 . 6 mm Optode: 1 . 5 mm Mouse Ventral HippocampusElectrode: −2 . 9 mm Optode: −1 . 6 mm Electrode: 3 . 0 mm Optode: 2 . 0 mm Electrode: 3 . 0 mm Optode: 1 . 8 mm Oxygen recordings were obtained using an implantable fiber-optic oxygen-sensing device . 525 nm light pulses induce fluorescence ( measured at 650 nm ) at the platinum tip that is quenched by oxygen within a local area ( ~500 µm3 ) and uses the fluorescence decay time to derive pO2 ( Ortiz-Prado et al . , 2010 ) . The technology ( Oxylite , Oxford Optronics , United Kingdom ) does not consume oxygen while measuring absolute pO2 values . The manufacturer individually calibrates each biologically inert probe , called an optode . The implantable system was designed by JFD in collaboration with Oxford Optronics . The implant is inserted under isoflurane anesthesia . We allow at least seven days between implantation and initiation of measurements to ensure that the effects of acute trauma were minimized . pO2 measurements at 1 Hz can then be made at any time by connecting the implant to the Oxylite using an extension fiber optic lead . The probe provides accurate and continuous measurements of local pO2 levels in brain tissue in awake , freely moving animals over several weeks . We used this innovative technology to precisely measure oxygen levels before , during and , most importantly , after seizures . Unlike most animal studies , we were able to perform these measurements without the confounding effect of anesthesia . On test days , rats were connected to the EEG and oxygen-sensing system and allowed 5 min to adjust before any measurements were taken . A seizure was elicited after 100 s of baseline recording using standard kindling stimulation ( 1 s train of 1 ms pulses at 60 Hz ) delivered through a Grass S88 stimulator ( Natus Neurology , Warwick , RI ) and seizure duration and stage were recorded ( Racine , 1972 ) . Once the EEG returned to baseline following a seizure , the electrodes were disconnected , but the fiber-optic cable for oxygen-sensing was left attached . The fiber-optic cable was disconnected when pO2 levels returned to baseline . Rats were chronically implanted with an electrode and oxygen probe as described above . In the contralateral hippocampus , a stainless steel guide cannula ( Plastics One ) was implanted ( see Table 5 for coordinates ) . 0 . 4 μg of kainic acid ( Sigma-Aldrich , Canada ) in 0 . 2 μL of saline was infused into the hippocampus over 4 min ( Rattka et al . , 2013 ) using a 1 . 0 μL syringe ( Hamilton Robotics , Reno , NV ) and a microsyringe pump ( Harvard Apparatus , model 55–2222 , Canada ) . Following infusion , rats displayed status epilepticus and were allowed to recover for one week . Following recovery , rats were connected to the oxygen and EEG acquisition systems to capture spontaneous seizures . A hippocampal electrode and optode were implanted into rats as described above followed by a 1-week recovery period . Rats were then connected to the oxygen detection system and saline soaked ear clips . A suprathreshold MES stimulus was delivered through the ear clips via a GSC 700 shock generator ( model E1100DA ) ( Grason-Sradler , West Concord , MA ) . Oxygen levels were recorded before and after the delivery of a 0 . 2 s train of 60 Hz biphasic sine-wave pulses ( Young et al . , 2006 ) . Seizure duration was recorded by observing seizure behavior . Rats with implanted hippocampal electrodes and optodes were re-used for this experiment . Following baseline oxygen recording , the hippocampus was stimulated with 1 ms square-wave pulses at an intensity of 1 mA and frequency of 3 Hz for 2 min . This drove continuous seizure activity for the duration of the stimulus ( Teskey and Racine , 1993 ) . pO2 was recorded until the rats recovered from hypoxia . Twelve consecutive patients with focal epilepsy admitted to the Seizure Monitoring Unit at the Foothills Medical Centre for continuous scalp video EEG were prospectively recruited . MR images were collected using a 3T scanner with an 8-channel phase-array head coil ( GE Discovery MR750 , GE Healthcare ) . MR data were collected within one hour of a habitual seizure . Baseline ( interictal ) ASL scans were obtained following a seizure-free period of >24 hr . ASL images were collected using a pseudo-continuous ASL sequence ( 1 . 88×1 . 88 mm , 28 slices; 5 mm slice-thickness , 2525 ms post-label-delay , spiral acquisition with 1024 points and eight arms ) . Quantitative CBF maps were generated by our scanner automatically from both ASL scans in ( mL/100 gr/min ) using the formula:CBF=6000 λ ( 1−e−STT1T ) ePLDT1B2T1B ( s ) ( 1−eLTT1 ) εNEX ( △MSF* PDREF ) where T1B and T1T represent blood and tissue T1 values ( 1 . 6 s at 3 T ) respectively , λ is the partial coefficient set to 0 . 9 , ε is the efficiency and is set to 0 . 80×0 . 75 , ΔM is the difference between tag and no tag images , PDREF is the reference proton density images , NEX is the number of excitations , SF is a scaling factor of 45 , and PLD is the post labeling delay . Subsequently , CBF maps were registered onto the respective patient’s T1-weighted high-resolution image using the FLIRT toolbox Version 5 . 5 from FSL ( Jenkinson et al . , 2002 ) using 12 degrees-of-freedom . Finally , a subtraction CBF map was generated from these two normalized CBF maps ( baseline minus post-ictal ) to identify areas with hypoperfusion in patients post-ictally . We looked for hypoperfusion >10 mL/100 g/min in patients ( normal grey matter CBF = 60 mL/100 g/min ) ( Chen et al . , 2012 ) . To obtain the quantitative CBF data , a region of interest ( ROI ) was manually drawn in a blinded fashion around the single brain area exhibiting the maximal post-ictal blood flow reduction as well as adjacent voxels with CBF changes >10 mL/100 g/min . If no clear hypoperfusion was seen , or if scattered areas of hypoperfusion were seen with no dominant area of hypoperfusion , then an ROI was drawn in the region where the captured seizure originated , based on EEG . Two Epileptologists ( PF/SS ) reviewed the clinical data for all patients ( e . g . , scalp EEG , PET , and SPECT ) and reached a consensus on location of the presumed seizure onset zone ( s ) for all of a given patients seizures . In addition , the same two Epileptologists reviewed the EEG recording of the seizure that was used for the ASL study and based on blinded review of the EEG alone , they identified the brain region where that seizure began . For the purpose of comparison , the brain was divided into the following regions: anterior temporal , posterior temporal , superior parietal , inferior parietal , occipital , orbitofrontal , frontopolar , mesial frontal , and lateral frontal . Next , the location of the most prominent postictal ASL changes were compared to brain region where the presumed seizure onset zone was located . Concordance occurred if the area of maximal hypoperfusion also participated in the seizure . Hypoxyprobe Green Kit was obtained from HypoxyprobeTM ( Burlington , MA ) and used to identify cerebral hypoxia ( Noto et al . , 2006 ) . This kit contains pimonidazole and a mouse anti-pimonidazole-fluorescein isothiocyanate ( FITC ) monoclonal antibody . Pimonidazole was dissolved in phosphate-buffered saline and injected into the intraperitoneal cavity at a dose of 30 mg/kg 30 min prior to stimulation . Rats were divided into two groups . Group one was implanted without stimulation and Group two received supra-threshold stimulation . Sixty minutes after stimulation , rats were deeply anaesthetized with sodium phenobarbital and perfused with ice-cold phosphate buffered saline and 4% paraformaldehyde . Brains were rapidly removed post-fixed in 4% paraformaldehyde for 24 hr and cryoprotected in 30% sucrose for at least 48 hr . Coronal sections were cut at a thickness of 25 μm and incubated overnight in 1:50 anti-pimonidazole-FITC antibody . Sections were mounted on gelatin-coated slides and cover slipped with Kystalon ( Harleco , VWR , Canada ) . Images were taken with an Olympus BX51 microscope using a GFP filter set and captured with a QImaging QICAM 1394 camera and ImagePro Plus software . Densely stained cell bodies were counted using the cell counter on the ImageJ software ( NIH ) in the hilus ( Figure 1G , I ) and CA3 ( Figure 3H , J ) . One section between 3 . 6 and 4 . 0 mm posterior from bregma as analyzed per animal and mean cell counts were determined within each group . Group means were compared with an independent samples t-test . Naïve Long-Evans rats were perfused and fixed as was performed with hypoxyprobe immunohistochemistry . 40 μm sagittal slices were incubated in rabbit Anti-COX-2 ( 1:1000 , ab15191 , Abcam , RRID:AB_2085144 ) and mouse anti-NeuN ( 1:1000 , MAB377 , Millipore , RRID:AB_2298772 ) antibodies for 24 hr at room temperature . Sections were then stained with Alexa-594 and Alexa-488-conjugated secondary antibodies ( Jackson Immuno , West Grove , PA ) to visualize COX-2 and NeuN , respectively . Sections were mounted to gelatin coated slides , cover slipped with DPX ( Sigma-Aldrich ) , and imaged with a slide scanner ( Olympus , Canada ) . A representative image including hippocampus was displayed ( Figure 3A–D ) . The stereotaxic coordinates were slightly modified to accommodate the implantable LDF probe ( Table 5 ) . The LDF probe and oxygen-sensing probe were implanted at angles to sample adjacent and overlapping region of tissue . LDF was used to assess local blood perfusion using the Oxyflo2000 ( Oxford Optronix ) . Implantable probes were cut to a length of 4 mm to match the depth of the oxygen-sensing probe . Once implanted , recordings were performed on awake , freely moving rats , as previously published using other systems ( Hu et al . , 2008 ) . Data collected from this device were assigned an arbitrary unit on a relative scale from 0–5000 Blood Perfusion Units and collected at a rate of 1 Hz . Data were standardized to baseline ( pre-seizure ) , expressed as a percentage of baseline , and plotted with a moving average of 30 data points for noise reduction . Rats implanted with ventral hippocampal electrodes and dorsal hippocampal optodes were used in these studies . Rats were reused for several drug experiments allowing at least two days between drug treatments . Celecoxib , SC-560 , ibuprofen , chelerythrine chloride , milrinone , sildenafil , SKA-31 , CAY-10526 , seratrodast , ozagrel , paxilline , 2-APB , levetiracetam , and topiramate were obtained from Cayman Chemicals ( Ann Arbor , MI ) . Acetaminophen , nifedipine , bumetanide , ethosuximide , phenytoin , and valproic acid were obtained from Sigma-Aldrich . Lamotrigine was obtained from SelleckChem ( Houston , TX ) . Fasudil was obtained from LC laboratories ( Woburn , MA ) and phenobarbital was obtained from Strathcona Prescription Center ( Canada ) . Lipophilic drugs were dissolved in 100% DMSO , while hydrophilic drugs were dissolved in saline and injected 30 min prior to seizure induction . The seizure duration and severity of hypoxia ( area below 10 mmHg ) were compared across kindle ( seizure without injection ) , vehicle- , and drug-treated groups using a within-subject ANOVA and follow-up t-test between vehicle- and drug-treated groups . All in vivo and in vitro 3 Hz experiments were performed on P25 – P40 Long-Evans rats ( male ) and obtained from Charles River Laboratories . Hippocampal slices were prepared as previously described ( Zhou et al . , 2011 ) . In brief , rats were initially anesthetized by isoflurane inhalation in air prior to decapitation . Brains were rapidly dissected from the skull and placed in ice-cooled cutting solution containing the following ( in mM ) : 210 sucrose , 10 glucose , 2 . 5 KCl , 1 . 02 NaH2PO4 , 26 . 19 NaHCO3 , 10 MgSO4 and 0 . 5 CaCl2 , pH 7 . 4 , bubbled with 95% O2/5% CO2 at 4°C . Transverse hippocampal slices ( 370 µm ) were sectioned with a vibratome ( Leica VT1200 , Germany ) in cutting solution as described above . Slices were transferred and incubated in oxygenated artificial cerebrospinal fluid ( ACSF ) containing the following ( in mM ) : 124 NaCl , 5KCl , 1 . 25 NaH2PO4 , 26 NaHCO3 , 2 CaCl2 , 1 . 3 MgCl2 and 10 glucose for 1 hr at 33° to 35°C incubator . Slices were then kept at room temperature prior to visualization of hippocampal blood vessels . Visualization of hippocampal blood vessels were performed using differential interference contrast ( DIC ) microscope . Vascular lumen diameter was measured using transverse hippocampal slices following 2 min of 3 Hz stimulation applied to Schaffer collaterals in rat hippocampus continuously perfused with ACSF . Hippocampal blood vessels were observed for 2 to 2 . 5 hr to determine changes in lumen diameter . For experiments involving acetaminophen and nifedipine , 100 μM acetaminophen or 50 μM nifedipine was added to the ACSF , respectively . Slices were pre-incubated with acetaminophen or nifedipine via continuous perfusion for 30 min prior to stimulation . Lumen diameter was measured before and after stimulation and percentage changes in diameter were calculated to determine vascular constriction . Sham stimulation and vehicle ( 0 . 1% DMSO ) were utilized as control experiments . Rats that received seizures in this study had an electrode implanted into the corpus callosum at the level of the forelimb movement representation and an oxygen-sensing probe in layer five of the motor neocortex ( Table 5 ) . Rats recovered for at least a week before experimental procedures were initiated . Rats were introduced to and allowed to explore the testing apparatus before data were obtained . They were permitted to escape to the ends of the rod during this period but the escape routes were blocked off during testing . This served to increase motivation since rats will acclimatize to the task . The test was performed as previously described ( Hunter et al . , 2000 ) , but with modifications . The two forepaws were placed on steel bar with a diameter of 5 . 0 mm that was suspended from a height of 45 cm . Rats were allowed to hang for a maximum of 45 s and then gently removed to reduce their fatigue for subsequent trials . Each test consisted of the mean of three trials . Nifedipine ( 15 mg/kg ) or vehicle was injected into rats on the first day and the other treatment on the following day . Rats were tested immediately before a seizure ( or sham ) to establish a baseline and then connected to the EEG and oxygen recording system to elicit and record a seizure , using standard kindling stimulation . In the sham condition , rats did not previously undergo surgery . We then tested at 20 , 40 , and 80 min after a seizure ( or sham ) . Each group was compared to its own baseline using a one-way ANOVA and Dunnett’s Multiple Comparison Test to identify a deficit . Rats with ventral hippocampal electrodes and dorsal optodes were used in this study ( Table 5 ) . Rats then went through the novel object recognition task paradigm ( Clark et al . , 2000 ) , with modifications . Following recovery from surgery , rats were habituated to the environment on the morning of day 1 . In the afternoon , rats were injected with vehicle or acetaminophen ( 250 mg/kg i . p . ) to permit or prevent hypoxia following a seizure , respectively . After 30 min rats were stimulated with kindling stimulation and both EEG and local tissue oxygenation were sampled . We confirmed that acetaminophen prevented hypoxia in the rats that received it and did not affect seizure duration , suggesting we have a clear dissociation . Rats were disconnected from the data acquisition system at 45 min postictal and transferred to the testing chamber where they were introduced to two identical objects and allowed to explore them for 5 min . Rats were returned to their home cage and brought back to the testing chamber 24 hr later . This time , a familiar object and a novel object ( which differed in color , size , and texture ) were introduced and the rats were allowed to explore the objects for 5 min . The position of the objects and type of objects were counter-balanced to account for rat’s preference not related to memory . If rats adequately form the memory for the two identical objects on day 1 , they should spend more time exploring the novel object on day 2 . Each session was videotaped and the duration spent investigating each object was measured . Investigation includes orienting the snout within 2 cm of the object , but not sitting on the object . All statistical analyses were performed using Prism version 5 . 01 ( GraphPad , La Jolla , CA ) . t-tests were used for experiments with only two groups . ANOVAs were used for experiments with more than two groups and a follow-up Tukey test to identify in which group ( s ) the significant differences occur . Repeated measures statistics were used for all within subject experiments . Rodents were handled and maintained according to the Canadian Council for Animal Care guidelines . These procedures were approved by the Life and Environmental Sciences Animal Care and Health Sciences Animal Care Committees at the University of Calgary ( AC11-0073 ) . Human experimentation was approved by the University of Calgary’s Conjoint Health Research Ethics Board ( REB13-0571 ) . All patients ( or guardians of patients ) provided written informed consent .	It has long been known that after an epileptic seizure , individuals often experience an extended period of impairments that affect how the brain works . Because the brain is organized so that specific tasks happen in particular areas , seizures that affect areas of the brain that control movement are often followed by muscle weakness . Likewise , amnesia may follow a seizure that affects brain areas involved in memory . While these events reduce quality of life in people with epilepsy , they have gone untreated because we did not understand what occurs in the brain after seizures . It has been observed that the impairments that follow seizures are similar to those that follow strokes , where for a period of time blood flow to certain areas of the brain is restricted and these areas are starved of oxygen . Following a seizure , is there a local stroke-like event that is responsible for the behavioural and memory impairments ? To address this question , Farrell et al . studied blood flow in the brains of mice , rats and human volunteers with epilepsy . The experiments show that after an epileptic seizure , blood vessels become narrower , which reduces blood supply to the areas of the brain involved in the seizure and dramatically reduces oxygen levels in those same areas . Using drugs to block the activity of an enzyme called cyclooxygenase-2 or other proteins called L-type calcium channels prevented both the oxygen shortage and the behavioural impairments that follow seizures . Thus , people with epilepsy are experiencing stroke-like events after seizures that they should be able to avoid with simple medical treatments . In the future , clinical research should determine how effective these treatments are in people with epilepsy . Other experiments should reveal if a shortage of oxygen after a seizure causes noticeable brain damage and the long-term behavioural problems that are often associated with epilepsy .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	Postictal behavioural impairments are due to a severe prolonged hypoperfusion/hypoxia event that is COX-2 dependent
Motor learning occurs through interactions between the cerebellar circuit and cellular plasticity at different sites . Previous work has established plasticity in brain slices and suggested plausible sites of behavioral learning . We now reveal what actually happens in the cerebellum during short-term learning . We monitor the expression of plasticity in the simple-spike firing of cerebellar Purkinje cells during trial-over-trial learning in smooth pursuit eye movements of monkeys . Our findings imply that: 1 ) a single complex-spike response driven by one instruction for learning causes short-term plasticity in a Purkinje cell’s mossy fiber/parallel-fiber input pathways; 2 ) complex-spike responses and simple-spike firing rate are correlated across the Purkinje cell population; and 3 ) simple-spike firing rate at the time of an instruction for learning modulates the probability of a complex-spike response , possibly through a disynaptic feedback pathway to the inferior olive . These mechanisms may participate in long-term motor learning . Learning is an adaptive change in behavior that results from ‘plasticity’ in the cellular mechanisms of synaptic transmission and/or spike generation in the nervous system . Our thinking about how the brain learns has been influenced heavily by the discovery of a form of synaptic plasticity called long-term potentiation ( aka ‘LTP’ ) . Through research on reduced preparations , and electrical activation of neural pathways in artificial patterns , we have a remarkable catalog of multiple forms of cellular plasticity in neurons . But a fundamental bridge is missing . We need to monitor the plasticity that occurs in a known circuit when an animal is participating in a learning task , to reveal what actually happens in the brain during behavioral learning . Our premise is that behavioral learning involves multiple forms of plasticity at multiple sites in an essential neural circuit for the behavior . Different forms of plasticity may be deployed at different sites over different time courses during learning . Circuit dynamics will play a key role in converting plasticity to learning . Thus , it is critical to understand when and where different forms of plasticity contribute to learning , and how circuits transform plasticity into changes in neural activity patterns . We start by analyzing a form of learning that occurs on the time scale of a single learning trial , with the premise that the analysis of short-term events will provide insight into circuit dynamics and plasticity that might play roles during long-term learning . The cerebellum provides an excellent brain site to ask what happens in a neural circuit during learning . Many forms of cellular plasticity are present in the cerebellum ( Hansel et al . , 2001; Zheng and Raman , 2010; Carey , 2011; Gao et al . , 2012 ) . The cerebellar circuit is understood well , and there is a prominent theory of cerebellar motor learning ( Marr , 1969; Albus , 1971; Ito , 1972 ) based on climbing-fiber induced long-term depression of the synapses from parallel fibers to Purkinje cells ( Ito and Kano , 1982 ) . Research in intact , learning animals using neural recordings , inactivation methods , and genetically-altered mice supports the cerebellar learning theory broadly ( e . g . , Gilbert and Thach , 1977; Lisberger , 1994; Medina et al . , 2000; Christian and Thompson , 2003; van Alphen and de Zeeuw , 2002; Boyden et al . , 2004; Blazquez et al . , 2006; Ke et al . , 2009; Wulff et al . , 2009 ) . As research on motor learning has advanced , we have come to understand that learning occurs at multiple neural sites and over several time scales ( Medina and Mauk , 2000; van Alphen and de Zeeuw , 2002; Jorntell and Ekerot , 2003; Boyden et al . , 2004; Yang and Lisberger , 2010 ) . In the present paper , we use learning in the smooth pursuit eye movements of awake , behaving monkeys to unravel how a neural circuit creates short-term cerebellar learning ( Kahlon and Lisberger , 1996 , 2000; Medina et al . , 2005; Medina and Lisberger , 2008; Yang and Lisberger , 2010 ) . The cerebellar circuit ( Figure 1 ) includes a mossy-fiber/granule-cell/parallel-fiber pathway to Purkinje cells , and a climbing fiber pathway that originates in the inferior olive ( IO ) . The parallel-fiber inputs cause conventional action potentials in Purkinje cells known as ‘simple-spikes’ . An action potential on the climbing-fiber input causes an unconventional , long-duration event known as a ‘complex-spike’ or ‘CS’ . Our formulation of the cerebellar circuit includes plasticity of Purkinje cell simple-spike responses driven by conjunction of parallel fiber and climbing fiber inputs ( light pink shading in Figure 1 ) , a disynaptic pathway from Purkinje cells to the inferior olive that includes two inhibitory synapses ( light gray shading in Figure 1 ) , and a disynaptic pathway from Purkinje cells to motoneurons . 10 . 7554/eLife . 01574 . 003Figure 1 . Schematic diagram of the cerebellar microcircuit . Abbreviations are: PC , Purkinje cell; GrC , granule cell; DCN , deep cerebellar nucleus; IO , Inferior olive . Pink shading highlights our finding of CS-linked plasticity in simple-spike firing . Gray shading highlights evidence that Purkinje cell output might regulate its own climbing-fiber inputs . Arrows indicate the flow of signals into the cerebellar cortex . DOI: http://dx . doi . org/10 . 7554/eLife . 01574 . 003 We show how the circuit in Figure 1 can explain a number of related observations on the spiking activity of Purkinje cells during short-term behavioral learning . First , we demonstrate that simple-spike firing rate predicts the probability of a complex-spike response to an instruction on a given trial , and we suggest that the disynaptic inhibitory pathway from Purkinje cells to the inferior olive causes this correlation ( gray shading in Figure 1 , De Zeeuw et al . , 1998; D’Angelo , 2010; Bengtsson and Hesslow , 2006 ) . Feedback from Purkinje cells might allow the cerebellum to control some of the error signals that guide synaptic plasticity in the cerebellum ( Mauk and Donegan , 1997; Kenyon et al . , 1998; Miall et al . , 1998; Bengtsson and Hesslow , 2006 ) . Second , we provide evidence that complex-spike-linked trial-over-trial depression of simple-spike responses ( Medina and Lisberger , 2008 ) is an expression of short-term plasticity ( pink shading in Figure 1 ) , presumably in the mossy-fiber/parallel-fiber input pathways . Our paper suggests how plasticity interacts with a neural circuit during a form of short-term motor learning , and outlines circuit dynamics that also may play a role in long-term learning . We use an experimental design that causes learned changes in the direction of smooth pursuit eye movements ( Medina et al . , 2005; Yang and Lisberger , 2010 ) . In other words , visual targets start moving in one direction and , 250 ms later , undergo a change in direction . Pursuit learns to anticipate the change in target direction , leading to a small deviation in the direction of eye velocity after a single learning trial , and larger learned changes over blocks of 100 identical learning trials . Figure 2A uses traces of eye position and velocity as a function of time to provide details about the structure of each learning target trajectory . Target motion began at a speed of 20 deg/s downward , in what we will call the ‘pursuit’ direction . As illustrated in the vertical position traces , the start of the downward ramp of target motion was accompanied by small upward displacement of the target to obviate the need for a catch-up saccade , thereby providing uncontaminated initiation of downward pursuit . Then , 250 ms after the onset of target motion , the target underwent an ‘instructive’ change in direction through the addition of a rightward or leftward ramp of motion at 30 deg/s for 400 ms ( horizontal position traces in Figure 2A ) . The brief addition of horizontal target motion during downward target motion created learning target trajectories in a two-dimensional position space shown by the zigzags at the top of Figure 2A . 10 . 7554/eLife . 01574 . 004Figure 2 . Background information about trial-over-trial learning and the responses of an example floccular Purkinje cell . ( A ) The zigzags at the top of the panel show a sequence of target motions in a random-order learning block . From top to bottom , the superimposed traces show vertical position , vertical velocity , horizontal position , and horizontal velocity as a function of time from the onset of target motion . Dashed and solid traces show target and eye movement . Different colored traces show responses in consecutive trials . The arrowhead on the horizontal velocity records points out trial-over-trial learning . The pink shading shows the analysis interval for trial-over-trial learning . ( B and C ) Direction-tuning of simple-spike ( B ) and complex-spike ( C ) responses of an example Purkinje cell . Histograms at eight locations show time varying firing rates for target motion in different directions . Polar plots show tuning-curves . ( D and E ) Superimposed rasters of CS responses and graphs of the probability of CS responses as a function of time from instruction onset for instructions in the off-direction ( D ) and on-direction ( E ) for simple-spike responses . The vertical line shows the time of the instruction onset and the gray shading shows the CS analysis interval . The polar plots above the CS rasters show the direction tuning of the example Purkinje cell in magenta , and use the black zigzag to indicate the trajectory of the learning target motions . DOI: http://dx . doi . org/10 . 7554/eLife . 01574 . 004 In most of the experiments reported here , we used a fixed pursuit direction , but delivered instructions that comprised changes in direction randomly in one of the two directions along the orthogonal , learning axis . This ‘random-order’ learning paradigm has the advantage that it allows us to study trial-over-trial learning while preventing any long-term learning . Trial-over-trial learning causes the eye movement on any given ‘test’ trial to possess a small , properly-timed deflection that takes eye velocity in the direction of the change in target motion on the prior ‘instruction’ trial . In the examples in Figure 2A , the small rightward deflection of horizontal eye velocity ( arrowhead ) in the red trace ( ‘n-1’ trial ) was learned as the result of a rightward instruction in the prior trial ( ‘n-2’ , not shown ) ; the leftward deflection in the blue trace ( ‘n’ trial ) was learned as the result of the leftward instruction on the ‘n-1’ trial . Because the eye velocity responses start before the instruction , they must be related to the instruction in the prior trial , rather than to that in the current trial . We realize that trial-over-trial learning is not an adaptive response during a random-order learning block , because it causes an inappropriate learned response half of the time . We also realize that the trial-over-trial effects we report in the present paper probe only the initial phases of behavioral learning . Still , we regard the random-order paradigm as a means to get ‘under-the-hood’ and determine how neural circuits transform cellular changes into behavioral learning in a limited time frame . We expect that the phenomena we are studying also play a role in more natural learning conditions , when the same instructive stimulus occurs repeatedly . We recorded from Purkinje cells in the floccular complex of the cerebellum because prior research has demonstrated that they play important roles in both normal pursuit ( Miles and Fuller , 1975; Lisberger and Fuchs , 1978 ) and pursuit learning ( Kahlon and Lisberger , 2000; Medina and Lisberger , 2008 ) . Under normal , pre-learning conditions , the simple-spike firing rate of floccular Purkinje cells is tuned for the direction of pursuit . For example , Figure 2B illustrates histograms of firing rate during pursuit in 8 directions , as well as the direction-tuning curve , for a Purkinje cell that responded best during pursuit to the right . The CS responses of the same Purkinje cell were tuned for the opposite , leftward direction ( Figure 2C ) , and had much lower firing probabilities . Opposite direction-tuning for simple-spike and CS responses is a general rule in the floccular complex . For each Purkinje cell , we contrived a learning target trajectory that matched the cell’s direction tuning . The initial target motion was orthogonal to the best direction for the simple-spike firing of the Purkinje cell under study , for example downward in Figure 2 . The instructive change in target direction was in either the on-direction or the off-direction for the simple-spikes of the Purkinje cell , that is leftward or rightward in Figure 2 . For instructions in the off-direction for simple-spike firing ( Figure 2D ) , the CS responses occurred with a high probability in response to the instruction . For instructions in the on-direction for simple-spike firing ( Figure 2E ) , the low spontaneous rate of CS responses was inhibited by the instruction . In Figure 2D , E , and all subsequent graphs , time zero on the x-axis indicates the time of the instructive change in target direction . The onset of pursuit target motion always began 250 ms before the time of the instruction and is indicated by an upward arrow below each x-axis . We have indicated the analysis interval in each figure by shading , using gray shading to indicate the interval for analysis of the probability that an instruction evokes a CS , and light pink shading to indicate the analysis window for simple-spike firing rate . We used the interval from 75 to 175 ms after the instruction ( gray shading in Figure 2D , E ) for the analysis of CS responses throughout the paper . Essentially all the CS responses to the instruction occurred in this interval . We used the interval from 50 ms before to 50 ms after the onset of the instruction ( light pink shading in bottom panel of Figure 2A ) to measure behavioral learning , and the simple-spike correlates of learning . This interval comprises the time when trial-over-trial effects were expressed in both neural responses and pursuit behavior . Importantly , the analysis interval ends before the instruction in the current trial could have any effect on eye movement or neural responses , and before there was any chance of a CS response to the instruction . Our first observation suggests a causal link between simple-spike firing rate and the probability that a CS response occurs in response to an instruction . Simple-spike firing rate at the time of an instruction was ∼5 spike/s higher on average depending on whether or not a CS occurred later in the trial in response to an off-direction instruction . The relationship between simple-spike and CS responses was restricted to the time interval surrounding the onset of the instruction ( vertical dashed line in Figure 3C , D ) , suggesting that the relationship is not caused by low frequency fluctuations of attention or other top-down influences . Note that the increased simple-spike firing rate preceded the time of the CS response and therefore cannot be attributed to a trivial explanation such as bleed-through of the CS spike into the train of spikes used to calculate simple-spike firing rate . 10 . 7554/eLife . 01574 . 005Figure 3 . Relationship between simple-spike firing rate and probability of CS responses to an instruction . ( A–D ) Average eye velocity along the learning axis ( A and B ) , CS probability ( C ) , and simple-spike firing rate ( D ) as a function of time in off-direction learning trials . Red and black traces show averages for trials with and without a CS response to the instruction . The same data appear in ( A and B ) , but with high and low gains on the eye velocity axis . Asterisks in ( C and D ) indicate 100-ms bins when the traces differed significantly ( p<0 . 01 , two-tailed paired t-test ) . Time is relative to the onset of the instruction and the upward arrow in D indicates the time of onset of target motion . ( E and F ) Neuron-by-neuron analysis for the start of an off-direction instruction ( E ) and the end of an on-direction instruction ( F ) . The continuous and dashed traces at the top show eye and target velocity along the learning axis . The pink and gray shading show analysis intervals for simple-spike and CS responses . The scatter plot contains three symbols for each Purkinje cell ( n = 90 ) ; yellow , blue , and red symbols show data for the lowest , middle , and highest third of simple-spike firing rates across trials for each Purkinje cell . Marginal histograms summarize the distributions across the population of Purkinje cells for the three different levels of simple-spike firing rate . Black arrowheads indicate the mean for each distribution . DOI: http://dx . doi . org/10 . 7554/eLife . 01574 . 005 Similar analysis of the eye velocity responses revealed a tiny difference of ∼0 . 1 deg/s that was present before the instruction and could be seen only in high-gain records ( Figure 3A ) . The difference in eye velocity probably was too small to cause the presence vs absence of a CS response to the instructive stimulus . Low-gain eye velocity records ( Figure 3B ) showed that the eye movement responses to the instruction were not related the presence vs absence of a CS response , a control that argues against the possibility that some trials lacked CS responses simply because the monkey was not attending or was being lazy . Figure 3C , D show that the simple-spike firing rate of a Purkinje cell before an instruction is related to whether or not an off-direction instruction evokes a CS on that learning trial . Because the variation in simple-spike firing rate is present before the CS response , we suspected that simple-spike firing rate could causally modulate the probability of a CS . However , the analysis described so far stops short of assessing how strongly simple-spike firing affects CS probability , and therefore also does not give insight into the possible strength of the proposed modulation . To assess the strength of the relationship , we next used simple-spike firing as the independent variable and assessed its impact on the probability of a CS response for each Purkinje cell . To do so , we created distributions of simple-spike firing rate in each trial averaged across the analysis interval ( light pink shading on traces in Figure 3E , F ) . Then , we trisected the distributions at the mean simple-spike firing rate plus/minus 0 . 44 standard deviations , grouped the trials for each third of the distribution , and computed the probability of a CS response to the instruction ( analysis interval shown by light gray shading in Figure 3E , F ) for each third of the distribution . Each neuron showed a clear relationship between the probability of an instruction-evoked CS and the average simple-spike firing rate at the time of the instruction ( Figure 3E ) . The clouds of points for different neurons move from lower left to upper right as we step from the trials with the lower 1/3 of simple-spike firing rates ( yellow symbols ) through the trials with the middle 1/3 ( blue symbols ) to the trials with the upper 1/3 ( red symbols ) . The trend appears in the scatter plot showing three connected symbols for each of the 90 Purkinje cells in our sample , and in the marginal histograms . On average , the magnitude of simple-spike firing rate averaged 66 , 84 , and 103 spikes/s for the three segments of the distributions and the probability of a CS response averaged 0 . 26 , 0 . 34 , and 0 . 44 . Correlation analysis of the cluster of points in Figure 3E yielded a statistically-significant correlation of 0 . 33 ( p<<0 . 001 ) . The slope of the means from the marginal histograms indicated that one spike/s of simple-spike firing rate is associated with a difference of 0 . 005 in the probability of a CS response to the instruction . We conclude that simple-spike firing rate may modulate the probability of CS responses , but weakly . Another fact argues that simple-spike firing rate merely modulates CS probability and does not exert obligatory control . For the interval at the end of the pulse of instructive target motion in the on-direction for simple-spike responses , the relationship is quite weak between the probability of a CS and simple-spike firing rate at the time of the instruction ( Figure 3F ) . The cluster of points in Figure 3F yielded a correlation of −0 . 028 that was not statistically significant ( p=0 . 65 ) In this interval , stopping the target during on-direction eye motion causes image motion ( relative to the retina ) in the off-direction and evokes the CS responses . But this is not a strong learning situation , and that may explain the absence of a relationship between CS probability and simple-spike firing rate . We have named the relationship between CS probability and simple-spike firing rate ‘same-trial facilitation’ , to distinguish it from CS-linked ‘trial-over-trial depression’ ( e . g . , Medina and Lisberger , 2008 ) . The existence of same-trial facilitation ( Miall et al . , 1998 ) raises the possibility we will consider in our ‘Discussion’ , that same-trial facilitation reflects an important feature of cerebellar learning ( Mauk and Donegan , 1997; Kenyon et al . , 1998 ) . Our second observation shows that plasticity of simple-spike firing is linked to the occurrence of a CS response to an instruction . We showed in a prior publication that a CS response to an instruction on one trial is linked to a depression of simple-spike firing rate on the next trial ( Medina and Lisberger , 2008 ) . We demonstrated trial-over-trial depression there ( and here ) by dividing the stream of learning trials into pairs where the first trial of the pair delivered an off-direction instruction . In each pair , we called the first trial the ‘instruction’ trial , and the second trial a ‘test’ trial . We computed the millisecond-by-millisecond difference between the simple-spike firing rates on the two consecutive trials . We tagged the trial-over-trial difference in firing rate for each pair according to whether or not the instruction trial contained a CS response to the instruction . A ‘1-0’ pair , for example , meant that the instruction trial contained a CS response to the instruction and the test trial did not . We then averaged the trial-over-trial changes in simple-spike firing rate , first for all like pairs of trials within a single Purkinje cell , and then across Purkinje cells . Trial-over-trial depression in simple-spike firing rate appeared if the instruction trial contained a CS response to the instruction and the test trial did not ( Figure 4A , red trace , ‘1-0’ pairs ) ; trial-over-trial depression did not appear when neither trial contained a CS response ( Figure 4A , blue trace , ‘0-0’ pairs ) . It was tempting to assume that the depression was caused by CS-linked plasticity of simple-spike firing . However , the existence of ‘same-trial facilitation’ means that trial-over-trial depression could appear simply because of elevated simple-spike firing rate on the instruction trial , but not on the test trial . Thus , our next step is to take advantage of a dramatically larger dataset than we had before ( Medina and Lisberger , 2008 ) to show that CS-linked short-term plasticity is one component of trial-over-trial depression . 10 . 7554/eLife . 01574 . 006Figure 4 . Mechanisms of CS-linked trial-over-trial depression of simple-spike firing rate and same-trial facilitation . ( A and D ) Trial-over-trial change in simple-spike firing rate and eye velocity . Different colors are for pairs of successive trials with different combinations of the presence or absence of a CS response to the instruction . In A , the top graph shows data for the onset of off-direction instructions and the bottom graph shows data for the offset of on-direction instructions . ( B , C , E and F ) Each graph plots the average simple-spike firing rate for trios of trials with different blends of the presence or absence of CS responses to an instructive change in the direction of target motion . Black , red , and blue traces show the first , second , and third trials of the trio . We performed statistical comparison in 25-ms bins using a paired two-sided Wilcoxon signed rank test . Red and blue asterisks indicate time bins when the firing rate in the second trial was statistically different from the first , or the third trial was statistically different from the second ( p<0 . 05 ) . Diagonal arrow in B points to a small increase in simple-spike firing rate related to the onset of pursuit target motion in a direction orthogonal to the preferred-null axis . Pink shading indicates the analysis interval . The x-axes show time with respect to the onset of the instruction . Upward arrows below the x-axes indicate the onset of pursuit target motion . DOI: http://dx . doi . org/10 . 7554/eLife . 01574 . 006 We isolated a contribution of CS-linked plasticity by comparing the trial-over-trial depression of simple-spike firing for pairs of consecutive trials that both had CS responses ( ‘1-1’ pairs ) with that for pairs of consecutive trials that both lacked CS responses ( ‘0-0’ pairs ) . Because the ‘1-1’ and ‘0-0’ pairs have the same CS response on each trial , they also should have the same amount of same-trial facilitation on each trial; same-trial facilitation would be ‘nulled’ and would not contribute to trial-over-trial depression . Even with the effect of same-trial facilitation nulled , the upper graph of Figure 4A shows a difference in trial-over-trial depression of about 5 spikes/s between ‘1-1’ pairs ( green trace ) and ‘0-0’ pairs ( blue trace ) . We attribute this component of trial-over-trial-depression to CS-linked plasticity that occurs during trial-over-trial behavioral learning , because it persists after we control for same-trial facilitation . Overall , Figure 4A shows a complicated progression of differences between the different-colored traces because it summarizes the combined effects of same-trial facilitation and CS-linked plasticity of simple-spike firing rate . We can see the effect of same-trial facilitation by comparing pairs of trials that had the same CS response on the instruction trial , to hold trial-over-trial plasticity constant , but different CS responses on the test trial . For example , the simple-spike firing rate was the same in both trials of the pair when neither included a CS response ( ‘0-0’ ) , so the trial-over-trial change in firing rate was almost flat ( Figure 4A , blue trace ) . When only the test trial included a CS response ( ‘0-1’ , yellow trace ) , simple-spike firing rate was higher in the test trial and the trial-over-trial change in simple-spike firing rate was more positive . The same effect appears when a CS response always occurred in the instruction trial . The simple-spike firing rate was higher in test trials that contained a CS response ( Figure 4A , ‘1-1’ , green trace ) , and the trial-over-trial change in firing rate was more positive , compared to pairs that did not contain a CS response in the test trial ( Figure 4A , ‘1-0’ , red trace ) . Finally , trial-over-trial learning in eye velocity ( Figure 4D ) is closely aligned to the trial-over-trial changes in simple-spike firing rate ( Figure 4A ) for the 4 pairs of trials , supporting a causal link between trial-over-trial depression in neural firing and trial-over-trial learning in behavior . As we have noted elsewhere ( Medina and Lisberger , 2008 ) , the fact that the learned eye velocities track the trial-over-trial changes in simple-spike firing rate implies that the CS responses are highly correlated across the Purkinje cell population . If the CS responses occurred with independent probabilities in different Purkinje cells , then the average across the population of Purkinje cells would be independent of the pattern of CS responses in the different pairs of trials . The learned eye velocity would be present for all pairs , and would not differ according to the pattern of CS responses . To perform statistical testing , we averaged firing rate across the analysis interval for simple-spike firing rate in individual neurons . Comparison of different pairs revealed statistically significant differences among the four types of pairs of trials ( multiple comparisons in a two-way ANOVA , n = 90 neurons , simple-spike firing , p<0 . 01; eye velocity , p<0 . 05 ) . Additional evidence for the existence of CS-linked plasticity during learning comes from analysis of trial-over-trial changes in simple-spike firing rate at the offset of on-direction instructive target motion . Same-trial facilitation is very weak at the offset of on-direction instructions ( Figure 3F ) . However , comparison of the trial-over-trial changes in firing rate for ‘1-0’ pairs and ‘0-0’ pairs revealed about 5 spikes/s of CS-linked trial-over-trial depression in simple-spike firing rate ( bottom graph in Figure 4A , also Medina and Lisberger , 2008 ) . The difference in the firing rate averaged across the simple-spike analysis interval was statistically-significant ( paired t-test , p=0 . 002 , n = 90 ) . The presence of trial-over-trial depression in conditions that lack same-trial facilitation , albeit with a slightly different time course , implies that trial-over-trial depression is caused partly by plasticity of simple-spike firing that is linked to CS responses . Our third observation elucidates the components that contribute to CS-linked same-trial facilitation . We considered trios of trials instead of pairs , and evaluated how simple-spike firing rate evolved from the first to the second to the third trial . This approach gives us better control of the history of CS responses and allows more definitive conclusions about the components of trial-over-trial changes in neural responses . Our strategy was to divide trios of consecutive trials into groups defined by binary codes that indicate whether or not a CS response occurred in response to the instruction in each trial . We then computed the time-varying average simple-spike firing rate for trios with different patterns of CS responses to off-direction instructions . We required that the first two trials of each trio deliver an off-direction instruction , and we compared the simple-spike firing rate on the third trial after two trials with identical histories of CS responses: ‘0-0-0’ vs ‘0-0-1’; and ‘1-0-0’ vs ‘1-0-1’ . As expected , simple-spike firing rate was the same in all three trials of ‘0-0-0’ trios ( Figure 4B ) , suggesting that our strategy of isolating trios was successful at controlling the effects of past history . Each trial showed the same modest increase in simple-spike firing rate ( diagonal arrow in Figure 4B ) in response to the initial eye movements along the pursuit axis chosen for the Purkinje cell under study , but the firing rate in the analysis interval ( pink shading ) was the same for all three trials . However , for ‘0-0-1’ trios , the simple-spike firing rate was higher in the third trial ( Figure 4C , blue trace ) as expected from the same-trial facilitation documented in Figure 3 , given the presence of CS response to the instruction . The same result emerges from comparison of the responses in the third trial of ‘1-0-0’ and ‘1-0-1’ trios in Figure 4E , F ( blue traces ) , with the caveat that the traces are noisier in Figure 4F because of a paucity of ‘1-0-1’ trios . For the comparisons made in the previous paragraph , the only difference in the stimulus conditions was whether or not the instruction caused a CS response in the third trial . Because the history was the same for each comparison , the most plausible explanation for the intricacies revealed by Figure 4B , C , E , F is that same-trial facilitation of simple-spike firing rate is related to spontaneous fluctuations in simple-spike firing rate . The same outside influence might drive simple-spike firing rate and the probability of a CS response in parallel . Or , higher simple-spike firing rates might operate through disynaptic disinhibition ( e . g . , Figure 1 ) to cause a higher probability of inferior olive ( and therefore CS ) responses to a given sensory event . Our data suggest that a number of features of cerebellar organization and function collaborate to create the specific fingerprint of cerebellar learning revealed in Figures 3 and 4 . However , the most stringent test for the logic of our conclusions comes from analysis of a computational model that includes the features we think are important . The data we strove to mimic with the model were: the trial-over-trial changes in simple-spike firing rate and eye velocity for all pairs of trials with and without the CS responses to an instructive stimulus ( Figure 4A , D ) ; the absolute firing patterns for the relevant trios of trials ( Figure 4B , C , E , F ) ; and the relationship between CS probability and simple-spike firing rate ( Figure 3 ) . We created a population of 1000 Purkinje cells and 100 inferior olive neurons and connected them according to the architecture outlined by the cartoon in Figure 5 . The cartoon retains the two inhibitory synapses between Purkinje cells and inferior olive neurons to emphasize the sign of the putative pathway in the brain , even though the neuron in the deep cerebellar nucleus was not actually modeled ( Equations appear in the ‘Materials and Methods’ ) . The model incorporated a number of features that were designed to allow it to reproduce our data , and that were suggested by data from our research and others’:The simple-spike responses of each Purkinje cell varied from trial-to-trial . Each individual response , meant to reflect the simple-spike firing rate averaged across the analysis interval used for our data , was drawn from a distribution with a mean of 100 spikes/s and a standard deviation of 18 spikes/s , in agreement with our recordings ( data not shown ) . The trial-to-trial variation in simple-spike firing rate was correlated at a level of ∼0 . 16 between pairs of model Purkinje cells . This level of neuron-neuron correlation enables the neuron-behavior correlations found 250 ms after the onset of pursuit by Medina and Lisberger ( 2007 ) . Each inferior olive neuron provided a climbing fiber to 10 Purkinje cells ( Sugihara et al . , 2004 visualized at least 5 ) . The probability that an off-direction change in target motion will evoke a response in a particular inferior olive neuron on a given trial is modulated by the simple-spike firing rate of floccular Purkinje cells on that trial . On-direction instructions never evoke the CS responses in model Purkinje cells . If a model Purkinje cell receives an input from its climbing fiber on a given trial , its simple-spike firing rate on the subsequent trial is reduced by 5 spikes/s through CS-linked plasticity . CS-linked plasticity relaxes back to zero over 2 trials when the instruction-test interval is 2 . 5 s ( Yang and Lisberger , unpublished data ) . The responses of model inferior olive neurons were synchronized enough so that the trial-by-trial correlation in the occurrence of a CS between model Purkinje cells was ∼0 . 25 on trials that delivered an off-direction instruction . 10 . 7554/eLife . 01574 . 007Figure 5 . A cerebellar circuit model that reproduces the features of trial-over-trial depression of simple-spike firing rate and trial-over-trial learning in eye velocity . The schematic on the left shows the cerebellar circuit on which our model was based . Abbreviations are: PC , Purkinje cell; DCN , deep cerebellar nucleus; IO , inferior olive . ( A and D ) Trial-over-trial change in simple-spike firing rate and eye velocity in the model , after Figure 4A , D . ( B , C , E and F ) Predictions for the firing rate responses in the first , second , and third trials in trios with different combinations of CS responses , after the same panels in Figure 4 . The x-axes show time with respect to the onset of the instruction . DOI: http://dx . doi . org/10 . 7554/eLife . 01574 . 007 We ran the model for 800 trials , each of which delivered a randomly selected on-direction or off-direction instruction . We analyzed the ‘data’ from the model using the same approach we had used for the real Purkinje cells . We assumed that the eye movement on each trial was defined by the average of the responses of all 1000 Purkinje cells on that trial , scaled arbitrarily so that the full model produced eye velocities of the amplitudes in our data . Because we did not simulate the responses in each trial as a function of time , the outputs from the model are scalars on each trial . For demonstration purposes , we have used those scalars as gains on the relevant simple-spike firing rate and eye velocity traces from our data in Figure 4 . Analysis of the ‘data’ produced by the model yielded results that captured most of the features of our recordings from monkeys . The trial-over-trial changes in simple-spike firing rate ( Figure 5A ) and eye velocity ( Figure 5D ) showed the same progression as the data in Figure 4A , D . Simple-spike firing rate mimicked Figure 4A almost perfectly , but there are minor differences in polarity of the eye velocities for ‘0-1’ and ‘0-0’ pairs . We suspect that these differences are related to correlations in the CS responses across the population of Purkinje cells that we have not captured fully in our model . Figure 5B , C , E , F shows excellent agreement between the performance of the model and the data in Figure 4B , C , E , F . The model also reproduced our finding that each spike/s of simple-spike firing rate is associated with an increase of ∼0 . 005 in the probability of a CS response to an off-direction instruction ( data not shown ) . Analysis of the sensitivity of the model to its parameters revealed that almost all of the features of the model were necessary to reproduce the data . In Figure 6 , we have plotted the scalar results of each of 20 runs of the model , with different configurations shown at different locations along the x-axis . The predictions for trial-over-trial changes in simple-spike firing rate broke down if we eliminated either CS-linked plasticity ( ‘NO plast’ ) or same-trial facilitation ( ‘NO SStoCS’ ) . Trial-over-trial depression of simple-spike firing rate was fairly normal for all other configurations of the model , and also did not depend strongly on whether the response of an inferior olive neuron to the instruction was controlled by 1 , 20 , or all 1000 model Purkinje cells ( ‘F1’ , ‘F20’ , ‘F1000’ ) . Trial-over-trial changes in eye velocity were quite sensitive to the presence of neuron–neuron correlations in the simple-spike firing rate and the CS responses across the population of Purkinje cells . Thus , the model supports the intuition outlined earlier about the likely high-degree of correlation in the presence and absence of CS responses across the population of Purkinje cells . 10 . 7554/eLife . 01574 . 008Figure 6 . Sensitivity of the computer model to removal of different features . Different symbols show the amplitude predicted by the 20 runs of a given architecture of the model; different colors indicate different combinations of CS responses in successive instruction and test trials . Abbreviations for different architectures are: FULL , full model; NO plast , no trial-over-trial plasticity; NO SStoCS , no connection from DCN to IO; SS/CS r = 0; no neuron–neuron correlations in simple-spike or CS responses; SS r = 0; no neuron–neuron correlations in simple-spike responses; CS r = 0 , no neuron–neuron correlations in CS responses; F1 , F20 , F1000 , IO neurons receive inputs from 1 , 20 , or 1000 model Purkinje cells . DOI: http://dx . doi . org/10 . 7554/eLife . 01574 . 008 We conclude that the features included in our model are at least sufficient to reproduce the trial-over-trial changes in simple-spike firing rate and eye velocity in our data . Even though the features of the model are supported by knowledge of the organization of the cerebellar circuit , it is possible that other , quite different , models would make similar predictions . We also note that the model does not reproduce the absence of same-trial facilitation at the end of an on-direction learning stimulus , when CS-linked plasticity persists . Clearly , we do not understand completely the basis for CS-related same-trial facilitation . The microcircuit in Figure 1 , and its success in modeling our observations on simple-spike firing rate ( Figure 5 ) , supports the hypothesis that a learned decrease in simple-spike firing rate may cause a decrease in the probability of CS responses to the instruction , and thereby reduce the magnitude or speed of learning . It could do so through the pathway from Purkinje cells through the deep cerebellar nucleus to the inferior olive . This pathway contains two inhibitory synapses so that increases or decreases in simple-spike firing rate would facilitate or inhibit responses of neurons in the inferior olive . The possibility that such a pathway exists and plays a functional role makes three testable predictions . In vitro studies have identified plasticity at multiple sites in the cerebellar cortex and the deep cerebellar nucleus , and have outlined many intricacies of the cerebellar circuit . The studies of neural responses in intact animals have measured neural representations of motor memories and have inferred what happened in the cerebellum during motor learning . There are numerous examples where genetic ablation of specific forms of cellular plasticity cause altered-learning phenotypes . Still , prior studies have stopped short of ‘monitoring’ what happens in the cerebellum during behavioral learning . We have made the next step of creating a quantitative description of how CS-linked plasticity of simple-spike firing interacts with the cerebellar circuit in a behaving monkey to create motor learning . By combining the approaches from two of our prior studies ( Medina and Lisberger , 2008; Yang and Lisberger , 2010 ) we have made four observations that relate to cerebellar operation during short-term motor learning:A single CS causes short-term CS-linked plasticity of simple-spike firing during motor learning , perhaps due to synaptic depression at the parallel fiber to Purkinje cell synapse . Higher simple-spike firing rate at the time of an instructive stimulus for learning is correlated with , and might cause , a higher probability of a CS response . Learned depression of simple-spike responses is associated with , and might cause , a reduced probability of a CS response to an instruction . Some past learning inhibits future learning . The model in Figures 5 and 6 reproduced most of the features of our data by simulating CS-linked short-term plasticity of simple-spike firing , a disynaptic , disinhibitory pathway from Purkinje cells to the inferior olive , neuron–neuron correlations in simple-spike firing rate , and correlations in the occurrence of CS responses across the population of Purkinje cells . The success of the model shows that these four features of cerebellar anatomy and physiology are sufficient to create the physiological observations we have detailed in the present paper . The necessity of including all four features shows how plasticity and circuit organization could work together to create a complex but quantifiable set of relationships among different neural and behavioral phenomena . We think our model might reflect the actual operation of the cerebellar learning system , because available knowledge supports the existence of the four critical features of the model . CS-linked trial-over-trial depression of simple-spike firing was demonstrated before ( Medina and Lisberger , 2008 ) , and we have verified here that trial-over-trial depression results partly from single-trial plasticity . The present paper demonstrates CS-related same-trial facilitation of simple-spike firing rate during learning , and Miall et al . ( 1998 ) showed the same phenomenon during quiescence . Medina and Lisberger ( 2007 ) provided evidence for moderate neuron–neuron correlations in simple-spike firing rate within the population of floccular Purkinje cells , and a number of papers have outlined the case for synchrony across the climbing-fiber inputs ( Llinas et al . , 1974; Welsh et al . , 1995 ) . Thus , our measurements have created a hypothesis for the operation of the cerebellar microcircuit during behavioral learning that is validated by experimental data and computer simulations . Still , we are aware that other models also could reproduce our data . For example , the probability of a CS and the simple-spike firing rate at the time of the instruction could be subject to top-down influences in parallel , a situation that would not require a pathway from Purkinje cells to the inferior olive , and might alter the implications for what happens in the cerebellum during motor learning . The relationship between simple-spike firing rate on a given trial and the probability of a CS response to an instructive stimulus on the same trial has multiple implications . One implication is caution in assuming that trial-over-trial depression is due to CS-linked plasticity . Indeed , our analysis indicates that CS-linked plasticity is responsible for only half of trial-over-trial depression . The other half of trial-over-trial depression on ‘1-0’ pairs of trials is related to the tendency for simple-spike firing rate to be higher on trials with a CS . A second implication concerns a possible role in learning . If simple-spike firing rate acts through inhibitory neurons in the deep cerebellar nucleus to modulate the response of inferior olive neurons to a given instruction , then a learned response in the cerebellum could modulate its own instructive inputs ( Schweighofer , et al . , 2013 ) . There are two inhibitory synapses in the pathway from Purkinje cells to the inferior olive ( De Zeeuw et al . , 1998; D’Angelo , 2010 ) , so higher ( lower ) simple-spike firing could cause a higher ( lower ) probability of CS responses , in agreement with the data of Miall et al . ( 1998 ) during quiescence and with our data during learning . Importantly , the climbing fiber axons from the inferior olive to the cerebellum and the pathway from the deep cerebellar nucleus back to the inferior olive both cross the midline , so that Purkinje cells can control their own climbing-fiber inputs ( Ruigrok , 1997 ) . Three features of our data support the idea that learned changes in the responses of Purkinje cells could modulate their own instructive climbing fiber inputs ( Mauk and Donegan , 1997; Kenyon et al . , 1998; but see Zbarska , et al . , 2008 ) . ( 1 ) There is a correlation between CS probability and simple-spike firing rate on learning trials . ( 2 ) Learned decreases in simple-spike firing are associated with a reduction in the probability of CS responses to an instruction , both during learning blocks that repeat the same instruction on many trials ( Medina and Lisberger , 2008 ) and during single-trial learning . ( 3 ) Past learning inhibits future learning , presumably by causing a learned decrease in the simple-spike responses of Purkinje cells ( Medina and Lisberger , 2008 ) that could reduce the probability of the CS responses that would instruct learning . In addition , anatomical and physiological studies have documented the required neural pathway ( Ruigrok , 1997; Bengtsson and Hesslow , 2006 ) . We suggest that the cerebellar learning circuit is set up to be conservative by allowing only a finite amount of learning in the cerebellar cortex . Learning may need to be transferred to the deep cerebellar nucleus ( Ohyama et al . , 2006 ) so that plasticity can relax back to baseline in the cerebellar cortex before further plasticity is allowed in the input pathways to Purkinje cells . Still , we are aware that there are multiple possible explanations for the observation that past learning inhibits future learning in pursuit eye movements . Thus , our data add support for , but do not obligate the conclusion that Purkinje cells modulate their own climbing-fiber inputs . We have used physiological techniques to create a hypothesis ( Figure 8 ) that starts to outline how the cerebellar circuit architecture and plasticity might interact during motor learning . Consider a sequence of trials that delivers the same instruction on 50 consecutive trials . A CS response on trial ‘1’ leads to a properly-timed depression of simple-spike firing on trial ‘2’ . Because of the decreased simple-spike firing , there is a lower probability of a CS response on trial ‘2’ . On trial ‘3’ , the random variation in simple-spike firing rate leads to some trials with the same or higher firing than trial ‘1’ and a higher probability of a CS response to the instruction; other trials have lower simple-spike firing rate and a lower probability of a CS response . Finally , our analysis predicts that by trial ‘50’ , sustained learning has depressed simple-spike firing rate substantially ( Medina and Lisberger , 2008 ) , so that the probability of a CS response is below control values and further learning is limited in the cerebellar cortex . Additional learning might become possible at a later time if some of the changes in the cerebellar cortex are transferred to the deep cerebellar nucleus , allowing a relaxation of the depression in the cerebellar cortex . 10 . 7554/eLife . 01574 . 010Figure 8 . Schematic diagrams showing mechanisms of cerebellar learning suggested by our data . Each panel shows the spiking activity of the elements in the proposed cerebellar circuit during the course of a learning experiment . Trial ‘3’ contains two schematics illustrating the effect of spontaneous trial-to-trial variation in the simple-spike firing rate , superimposed on the effects of trial-over-trial depression and longer-term learning . Abbreviations are: PC , Purkinje cell; GrC , granule cell; DCN , deep cerebellar nucleus; IO , inferior olive; CS , complex-spike . DOI: http://dx . doi . org/10 . 7554/eLife . 01574 . 010 We expect that the CS-linked plasticity of simple-spike responses in our data is a product of CS-driven depression of the synapses from parallel fibers to Purkinje cells . For example , cannabinoid-dependent synaptic depression seems to have the requisite brief time course ( Brown et al . , 2003; Brenowitz and Regehr , 2005 ) . However , other forms of cellular plasticity may contribute , either within the cerebellar cortex ( Hansel et al . , 2001; Carey , 2011 ) or in the deep cerebellar nucleus ( Miles and Lisberger , 1981; Zhang and Linden , 2006; Zheng and Raman , 2010 ) . The oculomotor vermis also could play a role in pursuit learning ( Suzuki and Keller , 1988; Takagi et al . , 2000 ) . We expect that future work will provide positive evidence of specific roles during learning for other forms of plasticity and other brain sites of learning . Indeed , we hope that we have outlined an approach that might be used to establish roles for other forms of plasticity in the cerebellar circuit . We conducted experiments on four awake , behaving adult male rhesus monkeys . Two monkeys were used for recordings from Purkinje cells in the floccular complex during pursuit learning ( at UCSF ) , and the other two were used only for behavioral studies of pursuit learning ( at Duke ) . During each experiment session , head movement was prevented with an implanted head holder , eye position was monitored with a scleral search coil , and a stainless steel recording cylinder allowed access to the floccular complex for single neuron recordings . The head holder , eye coil , and recording cylinder had been implanted in surgical procedures that used sterile technique with the monkey under isofluorane anesthesia ( Ramachandran and Lisberger , 2005 ) . Monkeys received analgesics for several days after each surgery . The Institutional Animal Care and Use Committees at UCSF and Duke had approved all the procedures in advance . The procedures were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals . Fixation and pursuit targets comprised bright 0 . 3° or 0 . 5° spots , respectively , on a dark background . Visual stimuli were presented on a CRT monitor that was placed 30 cm from the monkey’s eye and that subtended a visual field of 59° × 47° . All experiments were performed in a dimly lit room . Each single neuron recording experiment consisted of two separate blocks . The baseline block comprised approximately 80 trials . The target moved at a constant speed of 20 deg/s for 850 ms in one of eight directions . Target motion followed a standard step-ramp trajectory ( Rashbass , 1961 ) , with a 3° eccentric step to minimize the occurrence of saccades during the initiation of pursuit . A learning block comprised about 400 trials . Each trial delivered a sequence of target motions described in the ‘Results’ . Monkeys were required to keep their eyes within a reward window that was +1° during fixation , +2° during smooth target motion , and +4 or 5° for 400 ms after an instructive change in target direction . The monkey received a droplet of juice at the end of the trial if he had satisfied the fixation requirements throughout the trial . To avoid punishing the monkey for his inescapable response latencies in learning trials , fixation requirements were suspended for the 250 ms between the onset of target motion and the time of the instructive change in target motion . The magnitude of learned eye movements was small and did not affect the chances of reward in either a positive or a negative way . To estimate horizontal and vertical eye position , we scaled the eye position signals from a magnetic search coil system . The signals were passed through an analog circuit that rejected signals at frequencies above 25 Hz ( −20 dB per decade ) and differentiated signals at lower frequencies to create voltages proportional to horizontal and vertical eye velocity . Analog signals related to eye position and velocity were sampled at 1 kHz on each channel and stored for off-line analysis with single-unit recordings . To record the activity of single Purkinje cells in the floccular complex , glass-insulated platinum–iridium microelectrodes manufactured in our laboratory were introduced daily through the previously-implanted cylinder . Purkinje cells showed a high level of spontaneous simple-spike firing and occasional complex-spike ( CS ) responses . Purkinje cells in the floccular complex showed strong modulation of simple-spike firing rate when the monkey tracked sinusoidal target motion along the horizontal or vertical axis . Extracellular action potentials were amplified conventionally , filtered with a bandpass of 300 Hz to 3 kHz , digitized at 25 KHz , and stored in the computer . We used a software window discriminator to view the spike train for each trial in the dataset and identify simple-spikes and CS responses . The firing rate for simple-spikes was computed using a reciprocal interval algorithm ( Lisberger and Pavelko , 1986 ) . The CS responses were accumulated for each learning trial in bins with widths of 100 ms , and then were converted to the probability of a CS in each bin . We analyzed the responses of 108 Purkinje cells whose CS responses remained isolated through the full learning protocol . The CS responses occur in floccular Purkinje cells in response to instructive target motions that are in the ‘off’-direction for simple-spike firing during pursuit ( Stone and Lisberger , 1990; Medina and Lisberger , 2008 ) . Our analysis is based on the large majority of the Purkinje cells ( group 1 , 90/108 ) that showed CS probabilities more than three times their spontaneous level of 0 . 09 ± 0 . 02 ( mean ± SD ) . We excluded from analysis a much smaller subpopulation ( group 2 , 18/108 ) , whose CS probabilities after an instruction were less than 0 . 2 and never exceeded three times their spontaneous level . Specific data analyses are explained at the relevant locations in the ‘Results’ , and statistical tests are described either in the ‘Results’ or in the relevant figure legends . We created a cerebellar circuit model with 1000 Purkinje cells and 100 neurons in the inferior olive . In keeping with the known anatomy of the cerebellum , each neuron in the inferior olive provided a climbing fiber for 10 Purkinje cells ( Sugihara et al . , 2004 visualized at least 5 ) . The performance of the model was nearly the same if we set the climbing fiber divergence ratio to be 5 or even 1 . We computed the firing rate for each Purkinje cell as: ( 1 ) SSi , j= ( 1−Rnn ) αi , j+Rnnβjwhere SSi , j is the firing rate of the ith Purkinje cell on the jth trial , Rnn is the trial-by-trial correlation of simple-spike firing rate between pairs of Purkinje cells; αi , j is drawn for each Purkinje cell and trial from a normal distribution with a mean of 100 and a standard deviation of 18; and βj is drawn for each trial from the same random distribution but used for all Purkinje cells on that trial . We fixed the mean and standard deviation of the distribution of simple-spike firing to match the means in our data . We set Rnn to be 0 . 3 , which yielded neuron–neuron correlations that averaged 0 . 16 . For each model inferior olive neuron , we enforced the probability of a CS to be zero when the instruction was in the on-direction for simple-spike firing , and we biased the probability of a CS on the basis of the average simple-spike firing rate of a subset of the model Purkinje cells when the instruction was in the off-direction for simple-spike firing . First , we computed the ( disynaptic ) simple-spike effect on each model inferior olive neuron as: ( 2 ) INk , j=∑i=kN ( k+1 ) N−1SSi , jNwhere INk , j signifies the averaged input to the kth inferior olive neuron on the jth trial , and N is the number of Purkinje cells’ simple-spike firing rates that are averaged together to modulate the response of each model inferior olive neuron . Then , we used a sigmoid curve to relate the probability that each inferior olive neuron would respond to an off-direction instruction: ( 3 ) Pk , j ( IO ) =0 . 1+0 . 5 ( 1+e−0 . 3 ( INk , j−100 ) ) where Pk , j ( IO ) signifies the probability that the kth model inferior olive neuron will emit a response on the jth trial . We then determined whether a CS would occur on the jth trial in the ith model Purkinje cell according to the Boolean variable: ( 4 ) CSi , j= ( RCS⋅δk , j ) <Pk , j ( IO ) where RCS is a random number normally distributed around 1 ( standard deviation: 0 . 4 ) that determines the degree of synchrony across responses of model inferior olive neurons; and δk , j is a random variable with a value between 0 and 1 that is used to convert the probability of a response in the kth inferior olive neuron on the jth trial into an all or none response . To implement the 10-to-1 relationship between inferior olive neurons and Purkinje cells , CS and P ( IO ) use indexes of i vs k , according to the relationship: k=ceil ( ( i-1 ) /10 ) . We changed the synchrony across CS responses in different model neurons by changing the range used to draw values of RCS , and we eliminated synchrony across inferior olive outputs by setting the value of RCS equal to one . We ran the model as a series of trials , just as we had done with our data , with a random stream of on- and off-direction instructions . For each trial , we calculated the weakly correlated firing rates of the population of model Purkinje cells ( Equation 1 ) , and subtracted 5 spikes/s to implement plasticity if a CS had occurred for that Purkinje cell on the prior trial . Next , we computed the CS responses on the current trial for each model Purkinje cell ( Equations 2–4 ) . CS-caused plasticity decayed by 2 . 5 spikes/s on each of the next two trials . The simple-spike firing rate for a given Purkinje cell was represented as a scalar . Thus , the results of analysis of the model outputs also were scalars that were meant to represent the mean simple-spike firing rate across the analysis interval in our data . To create figures that were easy to compare with the actual data , we used those scalars as gain factors for time-varying traces from the data . There did not seem to be any reason to simulate either neural or behavioral responses on a millisecond time case within each trial , so time is represented in our model only in terms of the progression from trial-to-trial in a learning block .	Practice makes perfect in many areas of life , such as playing sport or even just drinking coffee from a cup without spilling any . Our brains can learn and improve these motor skills through trial , error and learning , with such “motor learning” depending on the cerebellum , a part of the brain that helps to coordinate all kinds of movements . Motor learning is a product of the organization of the cerebellar circuit , which is well understood , and the “plasticity” in the synapses that determine how cerebellar neurons interact with each other . The cerebellum contains cells called Purkinje cells that receive distinctive inputs from two pathways: a pathway involving inputs from many parallel fibers , which convey signals related to sensory events or motor commands; and a pathway involving input from a single climbing-fiber , which conveys signals from a part of the brain called the inferior olive nucleus . Research on slices of brain has revealed many sites and forms of cerebellar plasticity that could participate in motor learning . In one form of plasticity , the strength of the synapses between the parallel fibers and the Purkinje cell can be changed when a signal sent along the climbing fiber arrives the Purkinje cell . Yang and Lisberger have now taken the next step by studying the cerebellum of a monkey as it performs a motor learning task . Remarkably these experiments show that the climbing fiber inputs cause plasticity of Purkinje cell activity , just as happens in the experiments on brain slices . Further , some learning in the cerebellum restricts further learning , so that the cerebellum puts boundaries on its own learning . Overall the results make clear how learning is a property of groups of neurons working together in a circuit , rather than simply of changes in the strength of specific synapses . By shedding light on what happens in the cerebellum during short-term motor learning , the work of Yang and Lisberger will benefit efforts to understand how the cerebellum is involved in motor learning on all time scales .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2013	Interaction of plasticity and circuit organization during the acquisition of cerebellum-dependent motor learning
Mammals and birds have a specialized cardiac atrioventricular conduction system enabling rapid activation of both ventricles . This system may have evolved together with high heart rates to support their endothermic state ( warm-bloodedness ) and is seemingly lacking in ectothermic vertebrates from which first mammals then birds independently evolved . Here , we studied the conduction system in crocodiles ( Alligator mississippiensis ) , the only ectothermic vertebrates with a full ventricular septum . We identified homologues of mammalian conduction system markers ( Tbx3-Tbx5 , Scn5a , Gja5 , Nppa-Nppb ) and show the presence of a functional atrioventricular bundle . The ventricular Purkinje network , however , was absent and slow ventricular conduction relied on trabecular myocardium , as it does in other ectothermic vertebrates . We propose the evolution of the atrioventricular bundle followed full ventricular septum formation prior to the development of high heart rates and endothermy . In contrast , the evolution of the ventricular Purkinje network is strongly associated with high heart rates and endothermy . The muscle of the vertebrate heart initiates and conducts the electrical impulse that initiates contraction . It has regionally distinctive properties , such as conduction velocity , underlying coordinated alternating contraction of atrium and ventricle ( Chi et al . , 2008; Gillers et al . , 2015; Keith and Flack , 1907 ) . In addition , hearts of mammals and birds have specialized atrioventricular conduction system tissues composed of distinctive cardiomyocytes ( e . g . glycogen-rich , less developed contractile apparatus , highly conductive ) that can be morphologically distinguished from the working myocardium . These components , the atrioventricular bundle , bundle branches and Purkinje fiber network , rapidly conduct and distribute the impulse to the ventricular muscle , coordinating cardiac contractions ( Davies et al . , 1952; Chiodi and Bortolami , 1967; van Weerd and Christoffels , 2016; Chuck et al . , 2004; Rentschler et al . , 2001; Miquerol et al . , 2010 ) . They enable the high heart rates that set mammals and birds apart from similarly sized ectothermic vertebrates ( Davies and Francis , 1946; Lillywhite et al . , 1999 ) . The anatomical and electrophysiological analyses that unequivocally demonstrated the specialized atrioventricular conduction system in mammals and birds , yielded , in contrast , highly divergent interpretations when applied to ectotherms ( Davies et al . , 1952; Chiodi and Bortolami , 1967; Jensen et al . , 2012 ) . Nonetheless , the most prevalent view appears to be that the vertebrate taxa that independently gave rise to mammals and birds , represented by extant amphibians and reptiles , are without a specialized cardiac conduction system ( Tessadori et al . , 2012; Jensen et al . , 2017a; Burggren et al . , 2017; Jensen et al . , 2014 , Supplementary file 1 ) . The atrioventricular bundle and its branches in both mammals and birds are located in the ventricular septum ( Davies et al . , 1952 ) . Therefore , it is conceivable that the presence of a ventricular septum , rather than high heart rates , resulted in the development of a specialized atrioventricular conduction pathway . Among ectotherms , only crocodilians ( alligators , crocodiles , and gharials ) have a full ventricular septum and the electrical activation of the ventricle of the freshwater crocodile has been reported to propagate differently from that of other reptiles ( Jensen et al . , 2012; Christian and Grigg , 1999 ) . Early anatomical works suggested that atrioventricular canal myocardium of the American alligator projects onto the ventricular septum ( Greil , 1903; Swett , 1923 ) , but there was no mention of bundle branches and the later investigators that unequivocally show the specialized conduction system in mammals and birds ( Davies and Francis , 1946 ) , could not find specialized tissues in crocodilians ( Davies et al . , 1952; Davies et al . , 1951 ) . Here , we investigated the ventricles of the American alligator for the presence of atrioventricular conduction system components to address the question of their origin . To identify any specialized components in the alligator heart , we analyzed the expression of conserved gene markers for the mammalian and chicken atrioventricular bundle and its branches ( Tbx3 , Tbx5 , Scn5a , Cntn2 ) , and for the Purkinje network and its trabecular precursor ( Gja5 , Cntn2 , Nppa and Nppb ) ( Park and Fishman , 2017 ) . Furthermore , we assessed impulse conduction patterns to investigate its origin and spread in the ventricles . Because previous anatomical investigations led to contradictory interpretations ( Supplementary file 1 ) , we use complementary functional and molecular criteria to define 'specialization' . We conclude that the alligator has a functional atrioventricular bundle but lacks a specialized Purkinje network . To investigate the possible presence of a specialized atrioventricular conduction pathway in the crocodilian heart , we made an incision in the dorsal ventricular myocardium near the crux . The crux is the intersection of the atrioventricular sulcus and the dorsal descending coronary artery , and it indicates the position of the atrioventricular node and atrioventricular bundle in the mammalian heart ( Crick et al . , 1998 ) . The dorsal incision disrupted only 10–20% of the atrioventricular junctional myocardium , yet resulted in complete atrioventricular block in three out of three hearts ( Figure 1A–B , Figure 1—figure supplement 1 ) . In three other hearts , extensive ventral and lateral incisions did not result in atrioventricular block ( Figure 1C–D , Figure 1—figure supplement 1 ) . Optical mapping of isolated hearts showed the impulse appearing at the dorsal epicardial surface at the midpoint of the ventricle along the interventricular sulcus , consistent with an origin of the impulse from the atrioventricular bundle ( Figure 1E–G ) . On the ventral side of the heart , the activation wave propagated from the ventricular apex toward the major arteries ( Figure 1—figure supplements 2 ) . These activation patterns resemble those of mammals and birds ( Rentschler et al . , 2001; Reckova et al . , 2003 ) , but differ from those of the Anolis lizard , where epicardial activation starts in the ventricular base region closest to the atria ( Figure 1—figure supplement 2 , [Jensen et al . , 2012] ) . Measurements in an opened ventricle revealed the occurrence of first activation near the crest of the septum ( Figure 1H–I ) . Together , these data suggest the presence of a functional atrioventricular bundle in the alligator septum crest ( Figure 1I ) . We further characterized the atrioventricular conduction system by determining the alligator homologues of marker genes for specific heart components in mammals and birds . We performed RNA-sequencing of micro-dissected atrium , atrioventricular junction , and ventricles of hearts of embryonic American alligators of Ferguson ( Ferguson , 1985 ) stage 27 . Transcripts were readily detected for homologues of Tbx3 in the atrioventricular junction ( Hoogaars et al . , 2004 ) , for Tbx5 in all samples but lowest in the right ventricle ( Moskowitz et al . , 2004 ) , for cardiac sodium channel Nav1 . 5 ( Scn5a ) , for pan-cardiomyocyte markers cTnT ( Tnnt2 ) and cTnI ( Tnni3 ) , and for gap junction subunit Cx40 ( Gja5 ) ( Miquerol et al . , 2004 ) ( Figure 2 ) , but not for Nav1 . 8 ( Scn10a ) ( Remme et al . , 2009 ) . We used in situ hybridization to localize expression of these marker genes in the heart . In embryonic American alligator stages 13 , 16 , and 18 , the atrioventricular canal was entirely myocardial ( Tnni3-positive ) , revealing a full , undisrupted muscular continuity between the atria and the ventricle . In mammals and birds the atrioventricular bundle can be identified by the expression of Tbx3 ( Hoogaars et al . , 2004 ) . We found Tbx3 expression in a small domain from the dorsal atrioventricular canal to the crest of the forming ventricular septum ( Figure 3AB , Figure 3—figure supplement 1 ) . The ventral part of the crest showed very little Tbx3 expression ( Figure 3A ) . This pattern resembles that of Tbx3 in the atrioventricular node and bundle of mammals and birds . Ventricular expression of Tbx5 and Scn5a was highest in the septal crest ( Figure 3A ) , similar to the pattern in mice ( Remme et al . , 2009; Arnolds et al . , 2012 ) ( Figure 3A ) . The embryonic expression patterns of Tbx3 and Tbx5 were similar across crocodilian species indicating evolutionary conservation ( Figure 3—figure supplement 1 ) . In hearts of 1-year-old juvenile American alligators the myocardial continuity of the atrioventricular canal and the ventricle , including the ventricular septum , was interrupted by collagen at the ventral and lateral sides ( Figure 1—figure supplement 1 ) . Dorsally , the atrioventricular canal myocardium was nestled between collagen of the atrioventricular valves and the atrioventricular sulcus ( Figure 4 ) . The Tbx3 identified atrioventricular bundle extended from this sheet of atrioventricular canal myocardium ( Figure 4 ) . Laterally , the Tbx3 identified atrioventricular bundle was not insulated by connective tissue in contrast to the setting in mammals and birds . Tbx3 and Tbx5 were expressed in a pattern similar to that in the embryo , except that Tbx3 was also expressed deep in the septal crest myocardium ( Figures 3C and 4 ) . Tbx3 was therefore expressed where the earliest electrical activation was found ( Figures 1H–I and 3C ) . Gja5 was expressed in the Tbx3-positive myocardium deep in the ventricular septum , resembling the expression pattern of the atrioventricular bundle in mammals and birds ( Figure 3—figure supplement 2 ) . Interestingly , the dorsal cuts in alligator hearts inducing atrioventricular block ( Figure 1 ) disrupted this Tbx3-expressing myocardium ( Figure 3D ) . In mammals and birds , the atrioventricular bundle and Purkinje network are connected by the Tbx3-expressing bundle branches on the left and right flanks of the ventricular septum . We never observed similar branches in the crocodilians , neither with histology nor in situ hybridization for Tbx3 , Tbx5 , Gja5 , and Scn5a . To investigate whether a ventricular Purkinje network was also formed in the crocodilian heart , we set out to collect appropriate markers of the Purkinje network and its embryonic precursor , the trabecular myocardium . In mammals , expression of both Nppa and Nppb marks the trabecular myocardium during heart development , and Nppa expression becomes restricted to the thin subendocardial Purkinje fiber network after birth ( Miquerol et al . , 2010; Remme et al . , 2009; Houweling et al . , 2005 ) . Nppa and Nppb are paralogue genes that are part of the evolutionary ancient natriuretic peptide gene cluster ( Houweling et al . , 2005; Takei et al . , 2011 ) . Therefore , these genes may be suitable markers to characterize the ventricular conduction system in alligators . However , the archosaur branch , including birds and alligators , have been indicated to have lost Nppa during evolution ( Takei et al . , 2011; Trajanovska and Donald , 2008 ) . Unexpectedly , we detected the expression of homologues of both Nppa and Nppb in the RNA-sequencing data of the alligator heart samples ( Figure 5 ) . Cardiac Nppb was expressed strongly , Nppa was confined to the atria ( Figure 5 ) . Sequence homology analysis showed the alligator locus resembles that of turtle and frog , which retained both Nppa and CNP3 , revealing independent divergence of this trabecular myocardium-associated gene cluster in mammals and birds ( Figure 5 ) . In embryos of alligator , mouse , and chicken , Nppb ( and in mouse also Nppa ) was expressed abundantly in the trabecular myocardium ( Figure 6 , Figure 6—figure supplement 1 ) . Nppb was absent from the layer expressing Hey2 , a marker for compact ventricular myocardium in prenatal mammals ( Koibuchi and Chin , 2007 ) ( Figure 6B ) . One-year juvenile alligators maintained a trabecular layer that expressed Nppb ( Figure 6B ) . In mammals and birds , Gja5 expression becomes confined to the Purkinje myocardium ( Miquerol et al . , 2010 ) . In alligator embryos and 1-year juvenile , Gja5 was expressed not only in the trabeculated myocardium , but also the compact myocardium , which expressed Hey2 ( Figure 6—figure supplement 2 ) . Expression of the mammalian Purkinje network marker Cntn2 ( contactin-2; [Pallante et al . , 2010] ) was not detectable ( Figure 2 ) . These expression and genetic data indicate that aspects of the trabecular phenotype are maintained in the mature alligator ventricle , as they are in other ectothermic species ( Jensen et al . , 2012 ) , whereas a Gja5-positive mammalian/avian-like Purkinje network does not appear to form . Optical mapping analysis indicated the presence of two functional layers . In two out of six animals , we observed action potentials with fractionated ( biphasic ) upstrokes , in which an initial steep deflection was followed by a lessening of the inclination only to be followed by a second steep deflection ( Figure 7A–C , Figure 7—figure supplement 1 ) . At both the dorsal and ventral side , we observed the first phase of the upstroke in the vicinity of the ventricular septum . Distinct activation maps could be generated from both the first and the second part of the biphasic upstrokes showing first activation in the mid of the ventricle at the dorsal side ( Figure 7—figure supplement 1 ) . The dorsal activation patterns based on the first part of the upstroke resembled the dorsal activation patterns of the four animals that did not show fractionated upstrokes . The biphasic upstroke became monophasic at the location near the electrode during ventricular stimulation indicating that each phase of the upstroke corresponded to impulse conduction through a different myocardial layer . To validate this hypothesis , we recorded local electrograms , which showed biphasic deflections at exactly the same regions where we recorded fractionated upstrokes ( n = 2 ) ( Figure 7D ) . Together with the gene expression data , these data suggest the presence of two layers , a trabecular myocardial layer and a compact layer , which are activated in subsequent order . Consistent with the notion of the absence in alligator of a Purkinje network , which in mammals and birds exhibits rapid propagation of the electrical impulse , QRS duration was long in the alligator indicating slow ventricular impulse propagation ( 110 ± 13 ms , as in previous studies [Heaton-Jones and King , 1994] ) . Temperature affects propagation speed ( Smeets et al . , 1986 ) and we tested whether alligators at mammalian body temperatures would have a mammal-like QRS duration . Even when heated to mammalian body temperatures , the alligator QRS duration ( 90 ± 10 ms ) was much longer than QRS duration of similarly sized mammals ( approximately 30 ms ( Detweiler , 2010 ) ; Figure 7E–F ) . We show the presence of a molecularly distinctive functional atrioventricular conduction system in the alligator heart . The presence of such a system in the crocodilian heart challenges our understanding of cardiac evolution . It suggests the development of the specialized atrioventricular conduction system predates the development of endothermy and correlates instead with the formation of a full ventricular septum . From this insight , one may be able to predict a relation between the degree of development of the ventricular septum and the presence of the atrioventricular bundle . The investigation conforms with the guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health ( NIH Publication No . 8523 , revised 1996 ) and was approved by the Institutional Animal Studies Care and Use Committee of the University of North Texas ( IACUC #1403–04 ) . We used six female alligators ( 1 . 2 ± 0 . 4 kg , avg ±sd ) to investigate the electrophysiology of the heart . The hearts of the six female alligators together with a series of embryonic alligators ( n = 3 , Ferguson stages 13 , 16 , 18 ) , and embryonic Cuvier’s dwarf caiman ( Paleosuchus palpebrosus , n = 2 , Ferguson stages 13 and 16 ) were used to molecularly characterize the heart . Ferguson ( Ferguson , 1985 ) stages 13 and 18 correspond approximately to chicken Hamburger-Hamilton ( Hamburger and Hamilton , 1992 ) stages 25 and 32 , respectively , and mouse embryonic day 12 and 15 . 5 , respectively . Additional four female alligators ( 2 . 7 ± 1 . 0 kg , avg ±sd ) were used to study the in vivo effect of temperature on the QRS duration of the ECG . All four adult alligators were lightly anesthetized with Isoflurane . Electrodes were placed at the right ( R ) and left ( L ) side of the chest ( ventral ) . ECG was recorded using PowerLab 26T ( AD-Instruments , Colorado Springs , CO , sampling rate 2 kHz ) and tracings were analyzed using Labchart Pro software . We calculated Lead I by subtracting L from R . Animals were fasted for 6 to 8 days prior to experimentation . On the day of study four alligators ( mass = 2 . 7 ± 1 . 0 kg ) were lightly anaesthetized by placing them in a sealed container with gauze soaked in isoflurane ( Isoflo; Abbott laboratories , North Chicago , IL ) . The animal's trachea was then intubated and the lungs artificially ventilated using a ventilator ( Model 552; Harvard Apparatus; Holliston , MA ) downstream of a vaporizer ( Ohmed Fluotec 4 Anesthetic Vaporizer; GE Healthcare; Buckinghamshire , UK ) providing 2% isoflurane at 5–10 breaths min−1 . A modified air pump was used to continuously pull room air through the vaporizer . Once a surgical plane of anesthesia was achieved the animal was placed ventral side up on a heat pad ( Homeothermic Blanket Control Unit , Harvard Apparatus; Holliston , MA ) . In addition , a lamp with a 40-watt bulb was placed above the animal at approximately the midpoint of the thoracic cavity . Three 12 mm surgical steel needle electrodes ( 29 gauge ) were inserted subcutaneously in the animal . Two electrodes were place approximately 2 cm below the posterior end of the sternum one on left and right lateral side of the body wall . A third electrode was placed posteriorly at the intersection of the right rear limb and the body wall . Each electrode was connected to an amplifier ( Animal Bio Amp , Adinstruments Colorado Springs , CO ) connected to a PowerLab data recording system ( 4 ST Adinstruments Colorado Springs , CO , USA ) connected to a Macintosh computer running Chart software ( Chart 8 Adinstruments Colorado Springs , CO ) . Data was collected as a sampling rate of 1000 Hz with a low pass ( 1 Hz ) and a high-pass ( 50 Hz ) filter setting . Once electrodes were in place a thermocouple was advanced approximately 3 cm into the cloaca and connected to a temperature meter ( BAT-12 , Physitemp Instruments , Clifton , NJ ) . Output for the temperature meter was connected to the PowerLab data recording system . Once the instrumentation was completed the thermal blanket and lamp were turned on . Animal core body temperature was then increased gradually from ~21 . 5°C to 36 . 0°C . At the completion of the study animals were euthanized with an intravenous injection of pentobarbital ( 350 mg/kg ) . For the ex vivo experiments , we used seven juvenile alligators ( ±1 . 3 kg ) . First , we recorded and ECG from anesthetized animals ( similar to described above ) after which the animals were put in deep anesthesia by an injection of 50 mg kg−1 body mass pentobarbital in the pre-cranial venous sinus . Subsequently the hearts were excised through a ventral incision . The excised hearts were placed in a bath of a custom-made Ringer solution at room temperature ( 22° Celcius ) and perfused through PE90 catheters inserted in the ventral wall of the left and right atria ( Ringer , in mmol/l: NaCl 120 , Tris 5 , NaH2PO4 1 , KCl 2 . 5 , MgSO4 1 , CaCl2 1 . 5 , Glucose 5 , pH adjusted to 7 . 5 with HCl ) . From the excised hearts , we recorded a pseudo-ECG ( ex vivo ) from electrodes placed in the tissue bath , 5 mm from the heart ( Powerlab 26T; AD-Instruments , Colorado Springs , CO ) . For recording of optical action potentials , we first loaded the hearts by superfusion with 10–20 µM di-4-ANEPPS ( Molecular Probes , Eugene , OR ) for 10 min which gave a preferentially loading the sub-epicardial compact wall of the ventricular myocardium . After loading , the excitation-contraction uncoupler blebbistatin ( 10–100 µM , Tocris Bioscience , Ellisville , MO ) was added to the perfusate but failed to remove motion artifacts . Excitation light was delivered by a 520 ± 5 nm light emitting diode ( Prizmatix , Southfield , MI ) and emitted fluorescence was filtered >610 nm and recorded by a CMOS-sensor , ( 100 × 100 elements , 1 kHz , MICAM Ultima , SciMedia Ltd , Costa Mesa CA ) . One heart ceased spontaneous beating and was excluded from further experiments . After the first measurement , 5 out of the remaining six hearts were loaded for a second time by perfusion of 10–20 µM di-4-ANEPPS to load the trabeculations of the ventricular lumen . Local electrograms were recorded from the epicardium of the ventral and dorsal side using a Franz electrode . In the bath , a ECG was recorded between two electrodes placed at the 2 . 5 cm from the left and right side of the heart . In three hearts , we stimulated on the apex ( dorsal ) at 0 . 5 Hz . Optical signals were analyzed using custom made Matlab-based software ( Laughner et al . , 2012 ) ( http://efimovlab . org/content/rhythm ) . Principle component analysis was performed using Matlab2014 . The optical mapping of the alligator was compared to optical mappings of Xenopus and Anolis that were re-analyzed from published findings ( Jensen et al . , 2012 ) . In three hearts , we cut the atrial and ventricular tissue in the vicinity of the crux , that is , the intersection of the atrioventricular sulcus and the dorsal descending coronary artery , which reveals the approximate position of the atrioventricular node in the hearts of mammals ( Crick et al . , 1998 ) . In the other three hearts , we cut the ventral and lateral parts of the atrioventricular junction . After the experiments , the hearts were fixed for 24 hr in 4% paraformaldehyde in phosphate buffered saline ( adjusted to pH 7 . 5 with NaOH ) and then kept in 70% ethanol . The ventricular base including atrial tissue was dissected free and embedded in paraffin and sectioned in 10 or 12 µm thick sections in either the frontal , transverse , or sagittal plane . We performed in situ hybridization as previously described ( Moorman et al . , 2001 ) . Probes for American alligator mRNA transcripts were made based on the following coordinates using UCSC Genome Browser on American alligator Aug . 2012 ( allMis0 . 2/allMis1 ) Assembly: Cntn2 ( JH737142:190 , 064–195 , 365 ) , cTnI ( JH732499:9 , 192–12 , 570 ) , Gja5 ( JH733970:656 , 237–659 , 259 ) , Hey2 ( JH734274:510 , 337–510 , 783 ) , Nppa ( JH731559:424 , 483–424 , 843 ) , Nppb ( JH731559:415 , 415–417 , 373 ) , Scn5a ( JH739807:162 , 160–168 , 550 ) , Tbx3 ( JH733970:656 , 237–659 , 259 ) , Tbx5 ( JH733970:893 , 979–894 , 506 ) . Probes for Anolis mRNA transcripts were based on UCSC Genome Browser on Lizard May 2010 ( Broad AnoCar2 . 0/anoCar2 ) Assembly; Myh6 ( chrUn_GL343680:156 , 017–156 , 398 ) . To detect collagen on tissue sections , we incubated sections for more than one hour in picro-sirius red followed by 2 min differentiation in 0 . 01M HCl . Tissue samples were collected and homogenized with a homogenizer ( Ultra-Turrax , IKA , DE ) in lysis buffer RA1 of the NucleoSpin total RNA isolation kit ( RNA II , Macherey-Nagel , DE ) . RNA isolation was performed following manufacturer’s protocol . RNA purity , concentration and integrity were determined by respectively Nanodrop ( ND-1000 , Isogen Life Science , NL ) , Qubit RNA Broad-Range ( 2 . 0 , Life Technologies ) , and Bioanalyzer RNA Nano ( 2100 , Agilent Technologies ) . Amplified double stranded cDNA was generated using the Ovation RNA-Seq system ( V2 , Nugen ) following manufacturer’s protocol with 32 ng RNA input per sample . cDNA purity , concentration and size distribution were determined by Nanodrop , Qubit DNA HS and Bioanalyzer DNA 1000 . Libraries were made using the 5500 SOLiD Fragment Library Core kit ( Life technologies ) and size distribution was determined by Bioanalyzer DNA 1000 . RNA sequencing was performed on the SOLiD 5500 Wildfire ( Life technologies ) . Sequence tags were mapped with Lifescope on the AllMis1 Aug . 2012 contig assembly of the American alligator genome . All analyses were performed using IBM SPSS Statistics 24 . Statistical significance between in vivo and ex vivo derived ECG parameters was determined using a paired sample t-test . Other group comparisons were performed using one-way ANOVA with Bonferroni test for post hoc matched pairs . Sample size ( n ) is given in each figure legend . Expression patterns in histological sections were established based on section from at least two different animals . Values are given as mean ±SEM . We considered a p-value lower than 0 . 05 as statistically significant .	Mammals and birds are referred to as ‘warm-blooded’ animals , because they maintain a constant high body temperature . This requires a lot of energy , so their bodies need to be well supplied with blood at all times . The hearts of mammals and birds contain two important structures that help them do this . The first is a full wall of muscle – called the ventricular septum – that divides the heart into left and right sides . The second is an electrical circuit made of specialized muscle cells that ensures that the heart beats fast enough by sending rapid electrical signals to the rest of the heart muscle . The circuit contains one group of cells in the ventricular septum , called the bundle of His , and another group termed the Purkinje network . Reptiles , however , do not maintain high body temperatures and are instead often thought of as ‘cold-blooded’ animals . The hearts of reptiles do not need to pump blood around the body as quickly and have different structures from warm-blooded animals . For example , most reptile hearts do not have a fully developed ventricular septum . The only exceptions are crocodiles , alligators and their relatives ( the ‘crocodilians’ ) , which do . Jensen , Boukens et al . therefore wanted to determine if a crocodilian heart also contained a specialized electrical circuit like those of birds and mammals . Previous studies that attempted to answer this question using only anatomical and electrical methods had yielded ambiguous results . As such , Jensen , Boukens et al . combined these methods with genetic techniques for a more detailed study . First , the ventricular septum of American alligators , a species of crocodilian , was examined , and found to contain a narrow tissue structure that strongly resembled the bundle of His . Indeed , if this presumptive bundle of His was cut , the electrical circuit was broken . Additional genetic analysis of this structure confirmed that genes similar to those active in the mammalian bundle of His were also switched on in alligators . However , recordings of heart activity showed that heart rates and the spread of electrical signals were both slower in alligators than in warm-blooded animals . This suggests that , although alligators have evolved some specialized muscle cells ( in the form of a bundle of His ) , their electrical circuit is still ‘incomplete’ . The lack of a Purkinje network , for example , would explain why their heart rates remain slow like other reptiles’ . Together these findings add to the current understanding of how the heart works in different animals with varying requirements for energy and blood flow . Also , since crocodiles and warm-blooded birds both evolved from ancient reptiles , detailed descriptions of their heart structures could shed more light on how warm-bloodedness first developed .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2018	Specialized impulse conduction pathway in the alligator heart
Monitoring the spread of SARS-CoV-2 and reconstructing transmission chains has become a major public health focus for many governments around the world . The modest mutation rate and rapid transmission of SARS-CoV-2 prevents the reconstruction of transmission chains from consensus genome sequences , but within-host genetic diversity could theoretically help identify close contacts . Here we describe the patterns of within-host diversity in 1181 SARS-CoV-2 samples sequenced to high depth in duplicate . 95 . 1% of samples show within-host mutations at detectable allele frequencies . Analyses of the mutational spectra revealed strong strand asymmetries suggestive of damage or RNA editing of the plus strand , rather than replication errors , dominating the accumulation of mutations during the SARS-CoV-2 pandemic . Within- and between-host diversity show strong purifying selection , particularly against nonsense mutations . Recurrent within-host mutations , many of which coincide with known phylogenetic homoplasies , display a spectrum and patterns of purifying selection more suggestive of mutational hotspots than recombination or convergent evolution . While allele frequencies suggest that most samples result from infection by a single lineage , we identify multiple putative examples of co-infection . Integrating these results into an epidemiological inference framework , we find that while sharing of within-host variants between samples could help the reconstruction of transmission chains , mutational hotspots and rare cases of superinfection can confound these analyses . The SARS-CoV-2 pandemic has caused global disruption and more than one million deaths ( World Health Organization , 2020 ) . Genomic analysis has yielded important insights into the origins and spread of the pandemic and has been an integral part of efforts to monitor viral transmission in the UK , with over 100 , 000 viral genomes sequenced as of December 2020 by COVID-19 Genomics UK ( COG-UK ) , 2020 . For purposes of genomic epidemiology , a consensus genome sequence is derived from each sample , but deep sequencing data invariably reveal some level of within-host variation ( Lythgoe et al . , 2020 ) , and minor alleles are commonly filtered out prior to phylogenetic analysis ( Meredith et al . , 2020 ) . It has been suggested that SARS-CoV-2 , like SARS-CoV-1 , evolves within an infected host as a quasispecies , with many mutations ( within-host variants ) arising which may be beneficial for the virus ( Eigen , 1993; Holland et al . , 1982; Lythgoe et al . , 2020 ) . Although there have been several previous reports on the within-host diversity of SARS-CoV-2 ( Lythgoe et al . , 2020; Nicolae et al . , 2020; Popa et al . , 2020 ) , a number of key questions remain to be resolved . Examples of underexplored questions include understanding the extent of sequencing artefacts among within-host variants , the mutational processes dominating SARS-CoV-2 evolution , the action of selection on within-host variants , and the extent to which superinfection or co-transmission of multiple lineages confound within-host diversity analyses ( Cudini et al . , 2019 ) . Better understanding of these questions can shed light on the evolution of SARS-CoV-2 and could inform efforts to reconstruct transmission chains with genomic data . To address these questions , we performed Illumina deep sequencing of over a thousand SARS-CoV-2 samples collected in March and April 2020 in the East of England . Two libraries were sequenced for each sample with separate reverse transcription ( RT ) , PCR amplification , and library preparation steps in order to evaluate the quality and reproducibility of within-host variant calls . To develop reliable methods for analysing within-host variation in the context of ongoing genomic epidemiological studies , we used the ARTIC protocol that is a common method used for SARS-CoV-2 genome sequencing by many labs around the world ( DNA Pipelines R&D et al . , 2020 ) . Analyses of the data provided insights into the extent of within-host diversity , the patterns of mutagenesis , and the extent of selection and suggested that amplification biases , hypermutable sites , and co-infection complicate the use of within-host diversity for epidemiological purposes . We generated sequencing data for 1181 samples in duplicate at a median sequencing coverage ranging from 998 to 49 , 025 read depth per replicate sample ( Figure 1—figure supplement 1; Supplementary file 1 ) . To study variable sites within an individual , including errors and within-host mutations , for each sample we first identified all variable sites with variant allele frequency >0 . 5% and at least five supporting reads . Comparison of calls between replicates revealed that , while some samples showed highly concordant variants between replicates , others showed high discordance ( Figure 1—figure supplement 2 ) . The discordance between some replicates suggested that a small number of molecules may be disproportionately amplified in some samples , amplifying both RT/PCR errors and rare within-host mutations to high allele frequencies . We quantified the discordance between replicates using a beta-binomial model ( Materials and methods ) and correlated this with the diagnostic qPCR Ct values , which are inversely related to the number of viral molecules within the samples . Samples with Ct ≥24 showed considerable discordance in allele frequencies between replicates ( Figure 1—figure supplement 3 ) , but the vast majority of samples with Ct < 24 showed good concordance . Overall , as viral loads decrease , amplification biases and artefacts are more common and can impact within-host diversity analyses using RT-PCR protocols ( Zhao and Illingworth , 2019; Zanini et al . , 2017; Illingworth et al . , 2017 ) . To reliably detect within-host variants with the ARTIC protocol , we used ShearwaterML , an algorithm designed to detect variants at low allele frequencies . ShearwaterML uses a base-specific overdispersed error model and calls mutations only when read support is statistically above background noise in other genomes ( Gerstung et al . , 2014; Martincorena et al . , 2015; see Materials and methods ) . Two samples were excluded , as they had an unusually high number of low-frequency variants unlikely to be of biological origin , leaving 1179 samples for analysis , comprising 1121 infected individuals of whom 49 had multiple samples . For all analyses , we used only within-host variants that were statistically supported by both replicates ( q-value< 0 . 05 in at least one replicate and p-value < 0 . 01 in both , Materials and methods ) . Within each sample , we classified variant calls as ‘consensus’ if they were present in the majority of reads aligned to a position of the reference genome or as within-host variants otherwise . The allele frequency for each variant was taken as the frequency of the variant in the combined set of reads from both replicates . In total , we identified 18 , 888 putative variants ( Supplementary file 2 ) , including 7672 consensus variants ( affecting 1063 sites ) and 11 , 216 within-host variants ( affecting 5545 sites ) . Within-host variants included 7102 single-nucleotide substitutions , 3223 small deletions , 542 small insertions , and 349 putative multi-nucleotide variants ( Materials and methods ) . The allele frequency spectrum was dominated by fixed mutations that were common to all viral RNA molecules in a sample , and a tail of mostly low-frequency variants , as previously described ( Lythgoe et al . , 2021; Figure 1A ) . The mean standard deviation in within-host VAF at variable sites assuming a Bernoulli distribution of variants at each site was found to be 0 . 12 ( Materials and methods ) . Given the low frequency of these variants across different hosts , the majority are likely to be the result of single mutations rather than recurrent events within the same host . Overall , within-host variation was detected in the vast majority of samples ( 95 . 1% ) , with a median of 8 within-host mutations detected per sample . This is , of course , an underestimate of the full diversity within a sample due to limited sensitivity to within-host variants at lower frequencies . The use of replicates and a base-specific statistical error model for calling within-host diversity reduces the risk of erroneous calls at low allele frequencies . We noticed a slight increase in the number of within-host diversity calls for samples with high Ct values , which may be caused by a small number of errors or by the amplification of rare alleles and that could inflate within-host diversity estimates ( Figure 1—figure supplement 3; McCrone and Lauring , 2016 ) . However , the overall quality of the within-host mutation calls is supported by a number of biological signals . As described in the following sections , this includes the fact that the mutational spectrum of within-host mutations closely resembles that of consensus mutations and the observation of a clear signal of negative selection on within-host mutations , as demonstrated by dN/dS and by an enrichment of within-host mutations at third codon positions ( Dyrdak et al . , 2019; Figure 1—figure supplement 4 ) . Analysis of the mutational spectrum can yield insights into the dominant mutational mechanisms underlying the evolution of SARS-CoV-2 during the pandemic . Consistent with previous reports ( Popa et al . , 2020 ) , the mutational spectrum of within-host variant closely resembles that of consensus variants ( Figure 2 ) . The spectrum shows two striking features: a dominance of C>U and G>U changes with weak extended sequence context , and a large asymmetry between the plus and minus strands , inferred from the high C>U/G>A and G>U/C>A ratios when mutations are mapped to the reference ( plus ) strand as has previously been described in analyses of SARS-CoV-2 consensus genomes ( Kustin and Stern , 2021; Simmonds , 2020 ) . C>U mutations account for 47 . 9% of all within-host point mutations compared to 5 . 3% for G>A mutations ( C>U mutations in the minus strand ) , and G>U mutations account for 14 . 3% of mutations compared to 1 . 9% of C>A mutations ( Materials and methods ) . Normalised for the genome sequence composition , the plus/minus strand ratios of the rates of C>U and G>U are 9 . 6-fold and 7-fold , respectively . These asymmetries appear difficult to explain whether mutations were direct replication errors . Since any given viral RNA molecule has undergone the same number of plus-to-minus and minus-to-plus replication steps , replication errors are only expected to cause these asymmetries if both steps have very different error rates . For example , if the polymerase had a high C>U error rate in both strands , we would expect a symmetric number of C>U and G>A mutations in the plus strand , as C>Us would be introduced at equal rates when copying the plus or the minus strands ( Figure 2E ) . Within a cell , the fact that minus templates are copied many times could theoretically lead to a viral population with fewer G>A mutations but at higher frequencies ( Sawicki et al . , 2007; V'kovski et al . , 2021 ) . However , the fact that the same asymmetries are observed for consensus variants , which typically represent the fixation of a single genotype , makes this an unlikely factor . Thus , unless the error rates of the RNA-dependent RNA-polymerase in SARS-CoV-2 are very different on both strands , which may be possible given its multisubunit structure ( Zhao et al . , 2020 ) , replication errors seem unlikely to explain the observed asymmetries . The strong strand asymmetries observed may be more consistent with RNA damage or RNA editing of the plus strand dominating the accumulation of mutations in SARS-CoV-2 . The plus strand is the infectious genome , exported from the replication organelles , where minus molecules reside , into the cytoplasm ( Wolff et al . , 2020 ) , translated , packaged into particles , and transmitted between cells and hosts . The plus strand is also present within cells in much larger numbers than the minus strand ( Sawicki et al . , 2007 ) . Thus , the plus strand may be expected to accumulate higher rates of damage or editing , which would manifest as strand asymmetries . If we accept this hypothesis , the dominance of C>U over G>A and G>U over C>A , when mapped to the plus strand , suggests that the dominant forms of RNA editing or damage are C>U and G>U . RNA-editing enzymes in human cells are able to mutagenise single-stranded DNA and RNA molecules and are known anti-viral mechanisms ( Hoopes et al . , 2016; Di Giorgio et al . , 2020 ) . Two families of RNA-editing enzymes in particular have been speculated to contribute to the mutational spectrum of SARS-CoV-2 , APOBEC cytosine deaminase enzymes causing cytosine to uracil transitions and ADAR enzymes causing adenosine-to-inosine changes ( A>G/U>C mutations ) ( Di Giorgio et al . , 2020; Simmonds , 2020 ) . While we see a high rate of C>U changes in the SARS-CoV-2 spectra , we see much lower rates of A>G/U>C mutations than previously suggested ( Shen et al . , 2020 ) . While the mutational spectrum induced by all APOBEC enzymes is not fully understood , the activity of the better-understood APOBEC3A and 3B enzymes in human cancers is distinctly characterised by a strong sequence context , leading to C>T and C>G changes almost solely at TpC sites ( Alexandrov et al . , 2020 ) . This contrasts with the weak context dependence observed in the SARS-CoV-2 spectrum . While RNA-editing enzymes may contribute to SARS-CoV-2 mutagenesis , direct damage of cytosine and guanine bases could also be consistent with the observed spectrum . For example , spontaneous cytosine deamination would cause C>U transitions while oxidation of guanine bases could explain G>U transversions ( Krokan et al . , 2002 ) . Both forms of damage are common mutagenic processes on DNA in human cells ( Helleday et al . , 2014 ) . Having described the mutational spectrum , we can also derive approximate estimates of the average diversity within a sample in terms of the expected number of differences between any two genomes found within a sample . Since the allele frequency of a mutation represents the fraction of viral RNA molecules in a sample carrying a mutation , we can estimate the expected mean number of differences between two genomes from the same host as 2⁢∑ipi⁢ ( 1-pi ) where pi is the frequency of a variants at position i in the genome . Importantly , this is an approximation as it assumes that each RNA molecule sequenced originated from the entire viral genome , which is not always the case given the abundance of viral sgRNAs ( Wölfel et al . , 2020; Perera et al . , 2020 ) . These estimates are also likely to be a conservative lower bound as they only include mutations at detectable allele fractions in both replicates . Across samples , we found the expected number of differences between any two genomes within a sample to be 0 . 83 ( Figure 1C ) . Phylogenetic studies estimate a mutation rate in SARS-CoV-2 of 0 . 001 mutations/bp/year ( Fauver et al . , 2020 ) or 0 . 082 mutations/genome/day . Thus , the observed within-host mutation load would be consistent with the expected acquisition of mutations during a relatively short span of several days . To investigate the possible accumulation of de novo mutations during the course of an infection , we studied 43 individuals for whom we had multiple samples collected at different timepoints ( Figure 3A , Figure 3—figure supplement 1 ) . Overall , the number of within-host variants tended to increase over time and this trend was significant using a Poisson mixed model to control for host-specific effects ( p=0 . 007; Figure 3B ) . To put this finding in context , there was considerable variation in the observed number of within-host variants among samples from the same individual , even if taken on the same day . This could be due to the bottleneck caused by the different sampling methods , which included sputum , swabs , and bronchoalveolar lavage in addition to variation introduced by samples with high Ct values ( Figure 3—figure supplement 2 ) . High variability between longitudinal samples has also been observed in six out of nine hospitalised patients in Austria ( Popa et al . , 2020 ) . The authors observed multiple instances of the fixation of a variant over the course of an infection . In contrast , we observed no change in the consensus genome sequence in any of the individuals from whom multiple samples were collected . It is possible that the accumulation of within-host mutations could in future be used to estimate time since infection . Indeed , estimates of within-host diversity have previously been used successfully to estimate the time since infection in HIV ( Giorgi et al . , 2010; Swiss HIV Cohort Study et al . , 2011; Puller et al . , 2017 ) . However , determining if a consistent signal could be observed given the extensive variation observed between repeated samples on the same day would require analysis of time-series data in a larger number of individuals with high-quality epidemiological data . To study the extent of selection acting on within-host variants and consensus variants , we calculated dN/dS ratios using the dNdScv package ( Materials and methods ) . Most commonly used software to calculate dN/dS use simple substitution models ( often using a single transition–transversion ratio ) , which can lead to considerable biases and false signals of selection under neutrality ( Martincorena et al . , 2017; Van den Eynden and Larsson , 2017 ) . dNdScv uses a Poisson framework allowing for complex substitution models including context dependence , strand asymmetry , non-equilibrium sequence composition , and estimation of dN/dS ratios for missense and nonsense mutations separately . This model is particularly suitable for datasets with low mutation density in lowly recombining or non-recombining populations , as it is the case in SARS-CoV-2 genomic data Martincorena et al . , 2017 . As expected , dN/dS ratios for consensus variants are under clear purifying selection ( Figure 4A ) , with particularly strong selection against nonsense mutations . This is similar to that observed in other RNA viruses such as Influenza Xue and Bloom , 2020 . Within-host variants reaching moderately high allele frequencies ( VAF > 10% ) display similarly strong purifying selection against missense and nonsense mutations as consensus variants ( Figure 4A ) , while within-host variants at low allele frequencies appear to be under more relaxed purifying selection , as it may be expected . Purifying selection against nonsense mutations can also be observed at the level of allele frequencies ( Figure 4B ) . Overall , the similarity of dN/dS ratios for consensus and moderate-VAF within-host variants suggests that selection within hosts , rather than during transmission , may explain a considerable fraction of the extent of purifying selection observed in consensus sequences . Some sites across the SARS-CoV-2 genome appear to have mutated independently multiple times , resulting in homoplasic sites across the viral phylogeny ( Nicola et al . , 2020 ) . Some of these sites have also been reported to recurrently appear as within-host variants , although it remains unclear to what extent these events represent recurrent sequencing errors , contamination between samples , co-infection of a sample by multiple lineages , recombination , mutational hotspots or convergent positive selection ( Lythgoe et al . , 2020; Nicola et al . , 2020; Popa et al . , 2020 ) . Co-infection by multiple lineages could possibly arise from superinfection , where an individual acquires infection from multiple sources , or by co-transmission of multiple lineages from host to host following an episode of superinfection . Figure 4E represents the distribution of recurrent within-host variants including positions where the reference allele is found in the minority . Within-host variation is observed in at least 5 ( 0 . 4% ) of our samples at 215 sites within the SARS-CoV-2 genome ( Supplementary file 4 ) . Sequencing or PCR errors are unlikely to contribute substantially to the recurrent variants observed in our dataset thanks to the use of replicates and the ShearwaterML algorithm . One mechanism by which the same within-host variant can be observed across multiple closely related samples is transmission of within-host variants between contacts , when a population of virions is transmitted between hosts . Under this scenario , sharing of within-host variants will be expected between samples with identical fixed variants at sites which do not exhibiting within-host variation . The preservation of a within-host variant is incompatible with the simultaneous fixation of new variants ( the within-host allele would either be purged or hitchhike and become fixed ) . Instead , we find that most recurrent within-host variants occur across lineages , independently of the genomic background of fixed variants present in a sample ( Figure 4E ) . This pattern is suggestive of recurrent mutation , although it could also be consistent with complex histories of superinfection and recombination . If we accept the hypothesis of recurrent mutation , there remains the question of whether this is caused by mutational hotspots or convergent positive selection . We observed that 28 . 5% ( 57/200 ) of recurrent ( n≥5 ) within-host variants are predicted to cause synonymous mutations and estimates of the dN/dS ratio corrected for the trinucleotide sequence context indicate that , as a group , these recurrent variants are under some purifying selection ( Figure 4A ) . This suggests that most recurrent within-host variants are likely to represent hypermutable sites rather than convergent positive selection . Their mutational spectrum suggests an enrichment for C>U changes , particularly at sites preceded by a pyrimidine , but the mechanisms behind their apparent hypermutability remain unclear ( Figure 2—figure supplement 1 ) . Still , this analysis does not rule out the possibility that a minority of recurrent within-host variants are the result of convergent positive selection . Indeed , a number of recurrent within-host variants are also observed to frequently occur at the consensus level in the broader COG-UK dataset ( Supplementary file 4 ) . A plausible example of positive selection is the spike glycoprotein mutation D614G , which was rapidly increasing in frequency throughout the world at the time these samples were collected: within-host variation at this position appears in 27 of our samples , with the D614G allele often present at high allele frequencies . In order to investigate the genetic background of our samples , we generated a maximum-likelihood phylogeny of all consensus genomes produced by the COG-UK consortium by the end of May 2020 , including those for the samples on our dataset ( Figure 5A ) . This showed that our samples represent a broad range of the SARS-CoV-2 genetic diversity found in the UK at that time and that the diversity observed among our samples was not primarily driven by geographical location . To explore the relationship between within-host variants and the consensus phylogeny , we identified within-host variants that are shared between samples , as illustrated by links between the tips of the phylogenetic tree in Figure 5D . This confirmed that within-host variants are often shared between samples that are distant on the consensus phylogeny . Both a high level of recombination and a large transmission bottleneck would be required for within-host variants to be maintained across long transmission chains in order to explain the simultaneous preservation of some within-host variants with the fixation of others . While recent studies have suggested that the bottleneck in SARS-CoV-2 transmission can be on the order of 102–103 virions , they also identified substantial overlaps in the fraction of shared minority variants between samples unrelated by close transmission ( Popa et al . , 2020; Mara and Chu , 2020 ) . The sharing of within-host variants between consensus genomes as divergent as 10 SNPs , indicating multiple months of separation between the samples suggests that a more likely explanation is that many of these shared within-host variants are the result of recurrent mutation or co-infection ( Figure 6D ) . Indeed , a re-analysis of the Austrian data estimated much smaller bottleneck on the order of 1–3 virions ( Martin Michael and Katia , 2021 ) . To further investigate the correspondence between shared within-host variants and transmission , we used the transcluster algorithm to infer the probability distribution of the number of intermediate hosts that separate each pair of samples ( Stimson et al . , 2019 ) . This pairwise approach accounts for the serial interval of the virus as well as its evolutionary rate using the difference in the number of SNPs between each pair of consensus genomes . Figure 6B illustrates the relationship between the mean number of shared within-host variants and the probability that the respective infections were a result of direct transmission . A weak correlation with transmission probability was found for shared within-host variants of high allele frequency , but not for those of low allele frequency . It had been proposed that co-infection by different lineages was a significant source of within-host variation in SARS-CoV-2 although this was later amended ( Lythgoe et al . , 2020 ) . Instances of co-infection are common in other RNA viruses including enterovirus and HIV ( Dyrdak et al . , 2019; Liu et al . , 2002 ) . Co-infection by multiple lineages would result in the allele frequencies of divergent bases to appear as within-host variants and reflect the proportions of either haplotype . Detecting instances of co-infection is important in the context of both transmission and to accurately characterise the within-host mutation rate ( Chris et al . , 2017; Cudini et al . , 2019 ) . To identify possible co-infected samples , we compared the site frequency spectrum for each of our samples with all of the consensus genomes in the COG-UK dataset at the end of May 2020 . A linear model was used to identify mixtures of two consensus genomes that could explain the allele frequencies of variants within a sample better than a single consensus genome . Mixtures were considered if two consensus genomes could explain at least two additional within-host variants over that of a single consensus genome without including more than one variant not found within the sample . Samples identified by the model as being a possible mixture of consensus genomes were then visually inspected to determine whether the putative co-infections were convincing . This resulted in 36 putative co-infected samples , with a representative example shown in Figure 6A . The frequencies of within-host variants of all 36 samples , along with consensus genomes that comprise the putative mixture , are shown in Figure 6—figure supplement 1 . To determine the potential impact of co-infections on transmission inference , we excluded all 36 samples and re-ran the pairwise transmission analysis using the transcluster algorithm on the remaining samples . Figure 6C indicates that the remaining within-host variants no longer correlate as well with the probability of direct transmission , suggesting that much of the correlation observed previously was potentially driven by co-infections . It had been suggested that correlations between co-infections and the consensus phylogeny could be driven by the co-transmission of multiple strains ( Lythgoe et al . , 2021 ) . However , as the transmission signal in the minority variants in the remaining samples has reduced , it is likely that in some instances co-infection of certain strains is driven by multiple episodes of superinfection ( co-infection from two different infection sources ) . As these samples and those from Lythgoe et al . , 2021 were acquired from hospitalised patients , the correspondence with the consensus phylogeny could be driven by structured superinfection within hospital COVID-19 wards . The prevalence of infection within hospitals , particularly during the first ‘wave’ of the COVID-19 pandemic was often considerably higher than in the community with up to 50% of the consultant A and E workforce at one Welsh hospital testing positive in April 2020 ( Black et al . , 2020 ) . This contrasts with the finding that approximately 6% of the UK population had been infected with SARS-CoV-2 by the end of June ( Ward et al . , 2020 ) . Within-hospital transmission was also common in the early stages of the outbreak in Wuhan , with 41% of 138 patients found to have contracted SARS-CoV-2 in hospital ( Wang et al . , 2020 ) . While it is not possible to rule out cross-contamination between samples , this contamination would have to be structured ( e . g . occurring within the hospital rather than at a sequencing centre ) to explain the correlations between the inferred co-infections and the consensus phylogeny . Batch effects were carefully controlled for by Lythgoe et al . , and as we identified a similar signal using an independent dataset with samples sequenced in replicate , structured superinfection provides a more likely explanation . In cases of higher prevalence , repeated episodes of superinfection rather than co-transmission could complicate the use of within-host minority variants in transmission inference ( Figure 6D ) . We find a considerable amount of genetic diversity within individual SARS-CoV-2 infections that cannot be explained by technical artefacts . This is consistent with the quasispecies population structure typically found in RNA viruses ( Vignuzzi et al . , 2006; Eigen , 1993 ) , including related betacoranaviruses SARS-CoV-1 and MERS ( Lu et al . , 2017; Xu et al . , 2004 ) . By analysing the frequency of variants within individual samples , we estimate that two randomly chosen genomes within a sample differ by 0 . 83 variants on average . Since most of these samples were probably collected more than a week after the time of infection , and since the SARS-CoV-2 genome is known to acquire approximately one mutation every 2 weeks , these findings are broadly consistent with the hypothesis that within-host variation is largely due to the accumulation of de novo mutations within the host or , at least , in the span of a few days . Further support for this hypothesis comes from our analysis of hosts sampled at multiple timepoints , showing that the number of within-host variants tends to increase during the course of an infection . Increased numbers of within-host variants have also been observed in immunocompromised patients: consistent with the acquisition of de novo mutations within the host ( Siqueira Juliana et al . , 2020 ) . These data show that within-host variants have a similar mutational spectrum to the consensus variants that define between-host variation . Both are characterised by clear purifying selection , as would be expected if virions with disadvantageous mutations failed to survive or propagate within the host . Strikingly , the mutational spectrum of SARS-CoV-2 exhibits a near complete asymmetry between the plus and minus strands and is dominated by C>U and G>U mutations . This seems consistent with RNA damage or RNA editing of the plus strand . Whilst current knowledge of the mutational spectrum of APOBEC enzymes on RNA molecules is insufficient to assess the likelihood that they actively contribute to SARS-CoV-2 evolution , we note that the spectrum observed is very different to that of APOBEC mutagenesis in DNA from human cells . Instead , direct damage of cytosine and guanine bases could also be consistent with the observed spectrum of C>U and G>U changes in SARS-CoV-2 . Many of the within-host variants that we have identified are shared between infected individuals located on distant branches of the consensus phylogeny , and this is congruent with previous reports that the SARS-CoV-2 consensus phylogeny has many homoplasies that also appear as within-host variants ( Lythgoe et al . , 2021 ) . It appears likely that this is largely due to recurrent mutation although co-infection between lineages could also be a relevant factor in high transmission settings . The vast majority of our samples appear to comprise a single lineage , but we find evidence of putative co-infection by multiple lineages in approximately 3–4% of samples , which is likely an underestimate of the extent of co-infection in our dataset . In other viral and bacterial diseases , within-host variants provide a valuable source of information for the inference of transmission chains ( De Maio et al . , 2018; Chris et al . , 2017; Worby et al . , 2017; Arias et al . , 2016 ) and have been shown to improve the accuracy of inferences based on consensus genomes ( Worby et al . , 2014; Hall et al . , 2015; De Maio et al . , 2016; Campbell et al . , 2018 ) . In the case of SARS-CoV-2 , initial estimates suggested that the transmission bottleneck may be large ( Lythgoe et al . , 2020; Popa et al . , 2020 ) . However , revised estimates suggest that a much smaller bottleneck is likely similar to that seen in Influenza ( Lythgoe et al . , 2021 ) . We find some evidence of a correlation between sharing of within-host variants and the probability of direct transmission , a signal that we find is partially driven by co-infection , but the current dataset lacks the epidemiological resolution to evaluate the utility of this in reconstructing transmission chains , which might depend on the prevalence of infection and other local circumstances such as superinfection within COVID-19 hospital wards . In summary , these data show that there is considerable genetic diversity in the SARS-CoV-2 viral populations within individual hosts . Within-host diversity seems consistent with the accumulation of de novo mutations during the course of infection but can also result from co-infection with different lineages . Analysis of within-host variation could potentially be used to improve the inference of transmission chains , but this approach requires caution because of the confounding effects of recurrent mutation and unrelated episodes of superinfection . More detailed studies are required to evaluate the transmission bottlenecks that govern the propagation of within-host variants from host to host and to examine the patterns of within-host variation associated with different epidemiological circumstances . The 1181 samples were taken as a random subset from a larger prospective study into SARS-CoV-2 infections at Cambridge University Hospitals National Health Service Foundation Trust ( CUH; Cambridge , UK ) , a secondary care provider and tertiary referral centre in the East of England ( Meredith et al . , 2020; Hamilton et al . , 2020 ) . This study was done as part of surveillance for COVID-19 under the auspices of Section 251 of the National Health Service Act 2006 . It therefore did not require individual patient consent or ethical approval . The COVID-19 Genomics UK ( COG-UK ) study protocol was approved by the Public Health England Research Ethics Governance Group . A single swab was taken for each sample . Two libraries were then generated from two aliquots of each sample with separate RT , PCR amplification , and library preparation steps in order to evaluate the quality and reproducibility of within-host variant calls . The ARTIC protocol v3 was used for library preparation ( a full description of the protocol used is available at http://dx . doi . org/10 . 17504/protocols . io . be3wjgpe ) . Alignment was performed using the ARTIC Illumina nextflow pipeline available from https://github . com/connor-lab/ncov2019-artic-nf ( Bull , 2020; copy archived at swh:1:rev:8af5152cf7107c3a369b996f5bad3473d329050c ) . The pipeline trims primer and adapter sequences prior to alignment to the reference genome using bwa ( MN908947 . 3 ) ( Li and Durbin , 2009 ) . Variant calling was performed using ShearwaterML , and bam2R was used to count the number of reads supporting each base at each site . Both these methods are available as part of the deepSNV R package ( Gerstung et al . , 2014; Martincorena et al . , 2015 ) . In order to create a base-specific error model for the SARS-CoV-2 genome , we randomly selected sequencing data from 100 samples ( 50 from each set of duplicates ) from the 468 samples that met the following criteria: ( 1 ) ρ value from beta-binomial model ≤0 . 02; ( 2 ) proportion of genome with at least 500× coverage ≥0 . 9 for both duplicates; ( 3 ) absolute difference in median coverage between replicates ≤20 , 000; and ( 4 ) only one sample sequenced from a given donor . For each sample , non-reference sites with VAF ≥0 . 01 were set as uninformative ( i . e . depth for all base types were set to 0 ) in order to enable calling of recurrent , high VAF mutations . However , two sites ( 11074 and 25202 ) were effectively excluded from variant calling as they were set as uninformative for all samples in the normal panel . Variants were called separately for each set of duplicates . The initial ShearwaterML calls were filtered using the following criteria: ( 1 ) total depth at variant site ≥100× and ( 2 ) Benjamini–Hochberg false discovery rate q-value ≤0 . 05 in one duplicate and unadjusted p-value ≤0 . 01 in the other duplicate . The adjustment of p-values was performed by considering the five mutation types ( three non-reference bases , insertions , and deletions ) at all sites with ≥100× coverage in the 1181 samples from a given duplicate set . Variants at consecutive sites were merged into single events if the difference in VAFs ≤0 . 05 . The VAF for variants that were called using the ShearwaterML algorithm in both replicates was calculated as x1 , j+x2 , jn1 , j+n2 , j where x1 , j and x2 , j are the alternative ( non-reference ) allele counts of variable site j in replicates 1 and 2 of a given sample , and n1 , j and n2 , j are the local read counts at the respective sites . All scripts used to perform the variant calling are available in the accompanying GitHub repository . The mean standard deviation in VAF was calculated assuming a Bernoulli distribution at each site , such that given a variant at frequency pi , the standard deviation was calculated as pi⁢ ( 1-pi ) . The average over all variable sites was then taken . The expected number of differences between two randomly chosen genomes within a sample was calculated as 2⁢∑ipi⁢ ( 1-pi ) . To quantify the level of discordance in the allele frequencies between technical replicates , for each pair replicates we first identified a set of variable sites , which are expected to contain both artefacts and within-host mutations . All non-reference variable sites with coverage in both replicates >100x and mean VAF from both replicates higher than 1% and lower than 90% were considered for the beta-binomial modelling . In Illumina sequencing protocols where adapters are ligated before amplification and PCR duplicates can be removed computationally , VAFs may be expected to show binomial variation between technical replicates . However , amplicon sequencing can lead to preferential amplification of some molecules leading to higher variation in VAFs . We quantified the extent of variation above binomial sampling in the VAFs of variable sites between replicates by fitting a beta-binomial model to the allele counts of variable sites , obtaining a maximum-likelihood estimate of the ρ ( overdispersion ) parameter for each pair of replicate samples . Let x1 , j and x2 , j be the alternative ( non-reference ) allele counts of variable site j in replicates 1 and 2 of a given sample , and n1 , j and n2 , j be the local coverages at the site . We can calculate an approximate likelihood for the beta-binomial model using the following equation , where P depicts the beta-binomial density function:L=∏i∈{1 , 2}∏jP ( xi , j , ni , j , p=x1 , j+x2 , jn1 , j+n2 , j , ρ ) A maximum-likelihood estimate for the overdispersion parameter was obtained by grid search . The overdispersion parameter , measuring the level of discordance in allele frequencies between the replicates , was found to correlate strongly with the Ct value of the sample . The mutational spectra shown in Figure 2 were generated assuming that the allele in the reference sequence ( SARS-CoV-2 isolate Wuhan-Hu-1 , MN908947 . 3 ) represents the ancestral allele . The normalised mutation rates ( r ) for each of the 192 possible changes ( j ) in a trinucleotide context were calculated as:rj=njLj⁢∑jnjLjwhere nj is the total number of mutations observed for a trinucleotide change j , and Lj is the total number of times that the corresponding trinucleotide is present in the reference genome ( MN908947 . 3 ) . When observing the same mutation in multiple samples , it is not always straightforward to determine the number of independent mutational events that this represents . For consensus variants , this can be better estimated using a phylogeny , although recurrent hotspots can cause errors in the phylogenetic reconstruction . For simplicity , mutations observed across samples as consensus variants were counted only once for the estimation of nj . For within-host variants , a more relaxed approach was used since , based on our results , multiple instances of the same within-host variant across samples are more likely to represent independent events . Within-host variants observed in more than five samples were only counted five times for the calculation of nj , to avoid a small number of highly recurrent hotspots from dominating the trinucleotide spectrum . Analyses of selection were carried out using the dNdScv package ( Martincorena et al . , 2017 ) . In contrast to traditional implementations of dN/dS developed for divergent sequences , which rely on Markov-chain codon substitution models ( Goldman and Yang , 1994 ) , dNdScv was developed for comparisons of closely related genomes . Under low divergence rates , such as those in the densely sampled SARS-CoV-2 phylogeny , observed changes typically represent individual mutational events , which can be modelled using a Poisson framework instead of Markov-chain substitution models . This enables the use of more complex and realistic substitution models , including context dependence , strand asymmetry , non-equilibrium sequence composition , and separate estimation of dN/dS ratios for missense and nonsense mutations . These changes can be important to avoid false signals of negative or positive selection under neutrality when using simplistic substitution models ( Martincorena et al . , 2017 ) . Here we used the default substitution model in dNdScv , which uses 192 rate parameters to model all possible mutations in both strands separately in a trinucleotide context , as well as two ω parameters to estimate dN/dS ratios for missense and nonsense mutations separately . dNdScv files and code needed to generate dN/dS ratios from SARS-CoV-2 data are available at https://github . com/gtonkinhill/SC2_withinhost , DOI: 10 . 5281/zenodo . 5115287 . dN/dS ratios on polymorphism data need to be interpreted with caution . This is both because dN/dS ratios can be time dependent ( Rocha et al . , 2006 ) , providing weaker signals of selection for more recent changes , and because dN/dS ratios can also behave non-monotonically with respect to selection coefficients under idealised conditions of free recombination when using nucleotide diversity within a population ( Kryazhimskiy and Plotkin , 2008 ) . The former can also result in an excess of non-synonymous changes at deeper branches in the phylogeny due to incomplete purifying selection as has been observed previously in RNA viruses ( Kustin and Stern , 2021 ) . The latter can result from strong positive selection causing a reduction in the effective population size ( and in nucleotide diversity ) of non-synonymous sites under free recombination . However , the potential loss of monotonicity should not be a concern in our analyses of SARS-CoV-2 data . This is both because free recombination between adjacent sites is unrealistic for SARS-CoV-2 and because collections of mutations were identified by comparison of derived sequences to an ancestral reference genome , instead of by comparison of two derived sequences at any one time . Consensus genomes for each replicate were generated using the ARCTIC SARS-CoV-2 bioinformatics pipeline . A multiple sequence alignment was generated using MAFFTv7 . 464 , and any sites that were discordant between replicates were set to be ambiguous ( Katoh et al . , 2002 ) . Sites that have previously been identified to create difficulties in generating phylogenies were masked using bedtools v2 . 29 . 2 using the VCF described in Nicola et al . , 2020; Quinlan and Hall , 2010 . Fasttree v2 . 1 . 11 was used to generate a maximum-likelihood phylogeny ( Price et al . , 2010 ) . To investigate transmission , samples were only considered if both replicates produced high-quality consensus genomes . When multiple samples from the same host were available , the earliest sample was used . Pairwise SNP distances were generated between the consensus genomes using pairsnp v0 . 2 . 0 ( Tonkin-Hill , 2018 ) . The distribution of the underlying number of intermediate transmission events between each pair of samples was then inferred using an implementation of the transcluster algorithm ( Stimson et al . , 2019; Tonkin-Hill , 2020 ) . The serial interval and evolutionary rate were set to 5 days and 1e-3 substitutions/site/year ( Fauver et al . , 2020; Zhang et al . , 2020 ) . Within-host variants observed in more than 2% of samples were excluded from the histograms in Figure 6B , C . Potential mixed infections were identified using a linear model by testing whether the allele frequencies in a sample could be better explained by the inclusion of an additional consensus genome from the COG-UK dataset of 29 May 2020 . Additional sample mixtures were considered if the addition of a COG-UK consensus genome could explain at least 2 iSNVs and have at most one variant that was not found in the alleles of the sample . This identified 54 putative mixtures , which were then screened manually to obtain 36 potentially mixed samples . The code used to run this analysis is available in the supplementary materials . Analysis data and code available from: https://github . com/gtonkinhill/SC2_withinhost , DOI: 10 . 5281/zenodo . 5115287 ( Tonkin-Hill et al . , 2021; copy archived at swh:1:rev:5a20ba005ddcbe9b81f701a248a2ae348d2d9bd8 ) . Dehumanised sequence data is available from the ENA under accession number ERP126512 .	The COVID-19 pandemic has had major health impacts across the globe . The scientific community has focused much attention on finding ways to monitor how the virus responsible for the pandemic , SARS-CoV-2 , spreads . One option is to perform genetic tests , known as sequencing , on SARS-CoV-2 samples to determine the genetic code of the virus and to find any differences or mutations in the genes between the viral samples . Viruses mutate within their hosts and can develop into variants that are able to more easily transmit between hosts . Genetic sequencing can reveal how genetically similar two SARS-CoV-2 samples are . But tracking how SARS-CoV-2 moves from one person to the next through sequencing can be tricky . Even a sample of SARS-CoV-2 viruses from the same individual can display differences in their genetic material or within-host variants . Could genetic testing of within-host variants shed light on factors driving SARS-CoV-2 to evolve in humans ? To get to the bottom of this , Tonkin-Hill , Martincorena et al . probed the genetics of SARS-CoV-2 within-host variants using 1 , 181 samples . The analyses revealed that 95 . 1% of samples contained within-host variants . A number of variants occurred frequently in many samples , which were consistent with mutational hotspots in the SARS-CoV-2 genome . In addition , within-host variants displayed mutation patterns that were similar to patterns found between infected individuals . The shared within-host variants between samples can help to reconstruct transmission chains . However , the observed mutational hotspots and the detection of multiple strains within an individual can make this challenging . These findings could be used to help predict how SARS-CoV-2 evolves in response to interventions such as vaccines . They also suggest that caution is needed when using information on within-host variants to determine transmission between individuals .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "epidemiology", "and", "global", "health", "genetics", "and", "genomics" ]	2021	Patterns of within-host genetic diversity in SARS-CoV-2
Rho-associated kinases 1 and 2 ( ROCK1/2 ) are Rho-GTPase effectors that control key aspects of the actin cytoskeleton , but their role in proliferation and cancer initiation or progression is not known . Here , we provide evidence that ROCK1 and ROCK2 act redundantly to maintain actomyosin contractility and cell proliferation and that their loss leads to cell-cycle arrest and cellular senescence . This phenotype arises from down-regulation of the essential cell-cycle proteins CyclinA , CKS1 and CDK1 . Accordingly , while the loss of either Rock1 or Rock2 had no negative impact on tumorigenesis in mouse models of non-small cell lung cancer and melanoma , loss of both blocked tumor formation , as no tumors arise in which both Rock1 and Rock2 have been genetically deleted . Our results reveal an indispensable role for ROCK , yet redundant role for isoforms 1 and 2 , in cell cycle progression and tumorigenesis , possibly through the maintenance of cellular contractility . The Rho-associated coiled-coil-containing protein serine/threonine kinases ROCK1 and ROCK2 are downstream effectors of the Rho subfamily of small GTPases . They are activated by interaction with Rho GTPases and act through a number of pathways to regulate the actin cytoskeleton and thus cell migration , cell-cell adhesion and cancer cell invasion ( Itoh et al . , 1999; Olson and Sahai , 2009; Thumkeo et al . , 2013 ) . Substrates for ROCKs include the myosin-binding subunit of myosin phosphatase ( MYPT1 ) and the myosin regulatory light chain ( MLC ) , which regulate actomyosin contractile forces ( Kawano et al . , 1999; Kureishi et al . , 1997 ) . Actomyosin contractility is generated when myosin regulatory light chain ( MLC ) is phosphorylated , allowing myosin II interaction with actin filaments to generate mechanical force ( Aksoy et al . , 1983; Kureishi et al . , 1997 ) . ROCKs may either directly phosphorylate MLC ( Amano et al . , 1996 ) or indirectly regulate MLC phosphorylation by phosphorylating the myosin-binding subunit of myosin phosphatase ( MYPT1 ) and inactivating it ( Kimura et al . , 1996 ) . Myosin phosphatase inactivation results from phosphorylation of two inhibitory sites , Thr696 and Thr850 , on MYPT1 ( Feng et al . , 1999; Kawano et al . , 1999; Muranyi et al . , 2005 ) . Furthermore , ROCKs phosphorylate and activate LIM kinases 1 and 2 ( LIMK1 , 2 ) , which inhibit the actin-depolymerizing protein Cofilin ( Yang et al . , 1998 ) . ROCK1 and 2 exhibit 65% overall identity and 87% within the kinase domain , and some studies suggest differential roles for the isoforms . Using RNA interference , ROCK1 was reported to be important for stress fiber formation in fibroblasts , whereas ROCK2 controls cortical contractility and phagocytosis ( Yoneda et al . , 2005; 2007 ) . Distinct roles for ROCK1 and 2 have also been described in the regulation of keratinocyte differentiation ( Lock and Hotchin , 2009 ) and cell detachment ( Shi et al . , 2013 ) . Mice in which Rock1 has been genetically deleted are born , but have defects in eyelid as well as ventral body wall closure ( Shimizu et al . , 2005 ) . Ninety percent of Rock2 null mice die in utero due to defects in the placental labyrinth layer . This indicates that ROCK1 cannot compensate for a loss of ROCK2 . However , the few Rock2 null mice , that are born , display defects similar to those described in Rock1 null mice ( Thumkeo et al . , 2003 ) . This indicates some level of functional redundancy ( Thumkeo et al . , 2005 ) . In addition to their role in cell migration , ROCKs have been reported to modulate apoptosis ( Coleman et al . , 2001; Sebbagh et al . , 2005 ) and cell proliferation ( Croft and Olson , 2006; Samuel et al . , 2011; Zhang et al . , 2009 ) . The precise role of ROCKs in cell proliferation is not clear: some reports suggest ROCK function is required for G1/S progression ( Croft and Olson , 2006; Zhang et al . , 2009 ) , but others suggest ROCK is only required for anchorage-independent growth of transformed cells ( Sahai et al . , 1999; Vigil et al . , 2012 ) . One in vivo study reported that over-activation of ROCK , by expressing the kinase domain of ROCK2 in mouse skin , led to hyperproliferation and epidermal thickening ( Samuel et al . , 2011 ) . In order to investigate the roles of ROCK1 and 2 in tumorigenesis , we have generated conditional Rock1 and 2 knockout mice and studied these in vivo , using genetically engineered mouse models of non-small cell lung cancer ( NSCLC ) and BrafV600E mutant models of melanoma , as well as in vitro , using cells isolated from these mice . It is well understood that ROCKs control key aspects of the actin cytoskeleton such as actomyosin contractility , but their role in cell proliferation is not known . Because Rock null mice die early due to developmental defects , we generated Rock1 and 2 conditional alleles ( Rock1f and Rock2f ) by inserting LoxP sites flanking exon 6 in the Rock1 locus and exons 5 and 6 in the Rock2 locus ( Figure 1—figure supplement 1A ) . These exons are located within the kinase domain and their deletion results , through frameshifts , in the absence of ROCK proteins . We first generated cells lacking ROCK1 ( Rock1∆/∆ ) , ROCK2 ( Rock2∆/∆ ) or ROCK1 and 2 ( Rock1∆/∆;Rock2∆/∆ ) by isolating mouse embryo fibroblasts ( MEFs ) from embryos with the following genotypes: Rock1f/f;Rock2wt/wt , Rock1wt/wt;Rock2f/f or Rock1f/f;Rock2f/f and infecting them with Adenovirus-expressing Cre recombinase ( Ad-Cre ) or GFP ( Ad-GFP ) . Depletion of ROCK1 and ROCK2 ( Rock1∆/∆;Rock2∆/∆ ) resulted in a strong impairment of cell proliferation in vitro , which was not observed in cells lacking either ROCK1 ( Rock1∆/∆ ) or ROCK2 ( Rock2∆/∆ ) ( Figure 1A ) , demonstrating that ROCK function is required for cell proliferation and that ROCK1 and ROCK2 act redundantly . Infection of cells with adenovirus is not 100% efficient , and when the growth of Rock1∆/∆;Rock2∆/∆ cells was monitored over a longer period of time , these cells eventually recovered their ability to proliferate ( Figure 1—figure supplement 1B ) , but western blot analysis revealed that these cells express ROCK1 and 2 in equivalent levels to wild type cells ( data not shown ) and thus likely originated from uninfected cells . 10 . 7554/eLife . 12203 . 003Figure 1 . Depletion of ROCK1 and 2 leads to defects in cell proliferation in vitro and in vivo . ( A ) Proliferation curves of MEFs with different genotypes over 6 days . The cells were seeded 3 d after adenovirus infection . Graphs show total number of cells and SD from 5 independent experiments each carried out in triplicates . p-values were calculated using Student’s t-test: p<0 . 005; * p<0 . 001 . ( B ) Rock1f/f;Rock2f/f control and Rock1∆/∆;Rock2∆/∆ MEFs were cultured for 3 days and wild-type cells were treated with H1152 , inactive blebbistatin ( + ) or active blebbistatin ( +/- ) for 48 hr . Cells from all conditions were then subjected to a colony formation assay and grown for a further 7 days . ( C–F ) Rock1f/f;Rock2f/f MEFs transformed with Trp53 DD and HRas V12 were treated with Ad Cre to generate ∆ . Cells were injected subcutaneously into CD1 nude mice and growth analyzed . The graph shows average tumor volume in mm3 and SEM for Rock1∆/∆ and control ( C ) , Rock2∆/∆ and control ( D ) , Rock1∆/∆;Rock2∆/∆ and control ( E ) . p values were calculated by ANOVA and are as indicated . ( F ) Tumors with stated genotypes were immunoblotted with indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 00310 . 7554/eLife . 12203 . 004Figure 1—figure supplement 1 . Generation of conditional Rock alleles and cell proliferation analyses . ( A ) Schematic representation of mouse ROCK1 and 2 proteins , Rock1 and 2 loci , targeted and deleted alleles . ( B ) Proliferation curves of Rock1f/f;Rock2f/f control , Rock1∆/∆;Rock2∆/∆ MEFs 11 to 16 days after seeding . Cells were seeded 3 days after adenovirus infection . Graph shows total number of cells and SD from three independent experiments , each carried out in triplicates . ( C ) Cells were treated with Y-27632 and H1152 . One day later , equal numbers of cells were plated and subjected to growth analysis . The graph shows average number of cells and SD from at least 3 independent experiments , each carried out in triplicates . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 004 As genetic depletion abrogates ROCK function long term , we investigated whether long-term treatment of cells with ROCK inhibitors caused proliferation defects . Cells treated for 48 hr with the ROCK inhibitor H1152 ( Sasaki et al . , 2002 ) had reduced proliferation ( Figure 1B ) . Similar results were observed with other ROCK inhibitors , such as GSK269962A , AT13148 , GSK429286A and chroman1 ( data not shown ) . However , the much-used ROCK inhibitor Y-27632 ( Narumiya et al . , 2000 ) had a much weaker effect on cell proliferation than H1152 ( Figure 1—figure supplement 1C ) . This is consistent with previous studies , where we have shown that Y-27632 is a less effective ROCK inhibitor than H1152 ( Sadok et al . , 2015 ) . To determine whether defects in proliferation were due to ROCK-mediated effects on actomyosin contractility , cells were treated with blebbistatin , an inhibitor of myosin II ATPase , for 48 hr . Treated cells displayed a similar proliferation defect to that observed in Rock1∆/∆;Rock2∆/∆ cells ( Figure 1B ) , arguing that the requirement for ROCK in cell proliferation is mediated through its role in maintaining actomyosin contractility . To test whether Rock1∆/∆ , Rock2∆/∆ orRock1∆/∆;Rock2∆/∆ MEFs can proliferate in vivo , MEFs isolated from Rock1f/f , Rock2f/f or Rock1f/f;Rock2f/fmice were immortalized with retrovirus expressing a dominant-negative form of Trp53 ( Trp53 DD ) ( Hahn et al . , 2002 ) , and subsequently transformed with a retrovirus expressing mutant H-Ras ( H-RasV12 ) . The cells were then treated with Ad-Cre or control virus and , 5 days later , injected subcutaneously and tumor growth monitored . No differences in tumor growth were observed when transformed MEFs , derived from Rock1f/for Rock2f/f cells , infected with Ad-Cre were compared to control virus ( Figure 1C , D ) . In contrast , transformed Rock1∆/∆;Rock2∆/∆ MEFs grew significantly slower in vivo compared to controls ( Figure 1E ) . Western Blot analysis revealed the retention of ROCK1 and 2 protein in tumors derived from Rock1∆/∆;Rock2∆/∆ ( Trp53 DD; H-RasV12 ) MEFs but complete deletion of ROCK1 in Rock1∆/∆ and ROCK2 in Rock2∆/∆ tumor samples ( Figure 1F , data not shown ) , in agreement with the observed results on proliferation in vitro . This result shows that injected Rock1∆/∆;Rock2∆/∆ cells are unable to proliferate and form tumors but the un-infected cells can proliferate and form tumors thereby accounting for the delay in tumorigenesis and implying that ROCK function is essential for proliferation and hence tumor initiation and formation . We here show that concomitant depletion of ROCK1 and ROCK2 leads to defects in cell proliferation in vitro as well as in vivo . Depletion of a single ROCK isoform however has no effect , suggesting they can act redundantly . In order to identify causes for the defects in cell proliferation seen in cells lacking ROCK1 and ROCK2 ( Rock1∆/∆;Rock2∆/∆ ) , we characterized their cell morphology . One of the most prominent responses to activated ROCK is the generation of actin stress fibers , which regulate cell shape ( Amano et al . , 1997; Ridley , 1999 ) . A previous report using RNA interference suggested that ROCK1 , rather than ROCK2 , is important for stress fiber formation ( Yoneda et al . , 2005 ) . After thorough analysis of ROCK depletion upon infection with Ad-Cre or Ad-GFP , a slight decrease in protein levels was observed after three days ( Figure 2—figure supplement 1A ) . After four days a further reduction was seen and after five days both ROCK1 and 2 were depleted ( Figure 2B–C ) . Three days after infection with Ad-Cre , Rock1∆/∆;Rock2∆/∆ cells lacked stress fibers and displayed a ‘tail retraction’ phenotype , a well-known characteristic of ROCK inhibition ( Somlyo et al . , 2000 ) . Cells had long processes and defects in detaching their tails , in contrast to their controls ( Figure 2A and Videos 1 , 2 ) . A similar phenotype was observed in cells treated with the ROCK inhibitor H1152 or the myosin II inhibitor blebbistatin but not in cells lacking either ROCK1 ( Rock1∆/∆ ) or ROCK2 ( Rock2∆/∆ ) ( Figure 2A ) . After five days , cells that lack both ROCK1 and ROCK2 had few apparent central stress fibers , unlike control cells , and surprisingly adopted a flat morphology ( Figure 2A and Videos 1 , 2 ) . This phenotype seemed to appear gradually after prolonged loss of ROCK which was seen three days after adenovirus infection . Interestingly , when random motility was analyzed between Days 3 and 4 after adenovirus infection , using ImageJ to track cells , migration speed and directionality were significantly increased in cells lacking ROCK1 and 2 ( Figure 2—figure supplement 1B ) . This was previously described by Lomakin et al . ( Lomakin et al . , 2015 ) . 10 . 7554/eLife . 12203 . 005Figure 2 . Only deletion of both ROCK1 and 2 leads to a loss of central stress fibers and a decrease in actomyosin contractility . ( A ) Images of wild-type , Rock1∆/∆ , Rock2∆/∆or Rock1∆/∆;Rock2∆/∆ MEFs , 3 days after Ad-GFP and Ad-Cre-GFP infection , stained for pMLC and phalloidin . Wild-type and Rock1∆/∆;Rock2∆/∆ cells are also shown 5 days after Ad-GFP and Ad-Cre-GFP infection . Scale bars are 50 µm . ( B , C ) Western blot analyses of ROCK targets in lysates of MEFs with indicated genotypes . Representative blots are shown , quantification of multiple biological replicates can be found in Figure 2—figure supplement 1C . ( D ) Images of MEFs with indicated genotypes plated on a thick layer of collagen and stained for phalloidin . Scale bars are 50 µm . ( E ) Images and quantification of collagen gel contraction assay using MEFs of indicated genotypes . Graph shows average data and SD from 5 independent experiments , each carried out in triplicate . p-values were calculated using Student’s t-test: *p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 00510 . 7554/eLife . 12203 . 006Figure 2—figure supplement 1 . Depletion of ROCK leads to an increase in migration speed and persistent down-regulation of Myosin Regulatory Light Chain ( Mrlc ) phosphorylation . ( A ) Western blot analysis of ROCK1 and 2 in Rock1f/f;Rock2f/f cells after indicated days of Ad-GFP and Ad-Cre-GFP infection . ( B ) Cells from time-lapse movies were tracked and their migration speed and directionality determined using using ImageJ analysis software . Data are from five independent experiments and p-values were calculated using Student’s t-test: p<0 . 01 . ( C ) Quantification of western blot analyses . Graphs show protein levels of ROCK1 , ROCK2 divided by total ERK protein levels and pMLC , pMYPT850 , pMYPT696 and pCofilin divided by their total protein levels and SEM . The data are from five independent experiments and p-values were calculated using Student’s t-test: * p<0 . 001 . ( D ) MEFs were treated with H1152 for 20 min and the phospho-proteome was analyzed by quantitative mass spectrometry . The graph shows log ratios of identified phospho-peptides from two reciprocally heavy vs . light SILAC-labelled control and H1152-treated cell populations . The ‘Significant-B’ outlier test was used to determine significantly regulated peptides in both replicates , using a Benjamini-Hochberg FDR rate of 5% . ( E ) MEFs were treated with H1152 overnight and the phospho-proteome was analyzed by quantitative mass spectrometry . The graph shows log ratios of identified phospho-peptides from two reciprocally heavy vs . light SILAC-labelled control and H1152-treated cell populations . The ‘Significant-B’ outlier test was used to determine significantly regulated peptides in both replicates , using a Benjamini-Hochberg FDR rate of 5% . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 00610 . 7554/eLife . 12203 . 007Video 1 . Rock1f/f;Rock2f/f control , 3 days after infection with Ad-GFP , 10X magnification . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 00710 . 7554/eLife . 12203 . 008Video 2 . Rock1∆/∆;Rock2∆/∆ , 3 days after infection with Ad-Cre , 10X magnification . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 008 Consistent with the levels of stress fibers , phosphorylation of MYPT and MLC were only affected in cells where both ROCKs had been removed ( Figure 2B–C , quantification in Figure 2—figure supplement 1C ) , showing that ROCK1 and ROCK2 act redundantly in regulating MYPT and MLC phosphorylation . While MYPT Thr850 phosphorylation was markedly reduced in Rock1∆/∆;Rock2∆/∆ cells , MYPT Thr696 phosphorylation was not ( Figure 2B , quantification in Figure 2—figure supplement 1C ) , suggesting that , while Thr850 is principally regulated by ROCKs , Thr696 is regulated by other kinases such as MRCK ( Wilkinson et al . , 2005 ) . ROCK1 levels were slightly but not significantly increased in Rock2∆/∆ cells and vice versa ( Figure 2—figure supplement 1C ) . Rock1∆/∆;Rock2∆/∆ MEFs spread when plated on a thick layer of collagen 5 days after adenovirus infection , unlike Rock1∆/∆ or Rock2∆/∆ cells which had a more spindle-like morphology similar to wild-type controls ( Figure 2D ) . This indicates reduced cellular contractility upon loss of ROCK1 and 2 . As a functional measure of actomyosin contractility , we used a gel contraction assay ( Hooper et al . , 2010 ) , demonstrating that cells containing only ROCK1 or ROCK2 were able to contract a collagen gel , but cells lacking both ROCK1 and ROCK2 were not ( Figure 2E ) , supporting the notion ROCK1 and ROCK2 act redundantly to maintain cellular contractility . To further investigate ROCK signaling , we analyzed the temporal profile of changes in the phospho-proteome upon ROCK inhibition using mass spectrometry . Short-term inhibition of ROCK , using H1152 , resulted in a significant decrease in phosphorylation of many phospho-sites , including several known targets of ROCK such as Cofilin-1 ( Ser-3 ) , Cofilin-2 ( Ser-3 ) , Destrin ( Ser-3 ) , and Ser-19/Thr-20 of MLC ( Supplementary file 1 and Figure 2—figure supplement 1D ) . However , the majority of these phosphorylation sites , with the exception of MLC Ser-19/Thr-20 , were only transiently affected , as their phosphorylation was fully recovered in long-term ROCK inhibited cells ( Supplementary file 1 and Figure 2—figure supplement 1E ) or Rock1∆/∆;Rock2∆/∆ cells ( Figure 2B–C ) . These results indicate that alternative phosphorylation mechanisms exist for some ROCK substrates , but that no other kinase previously implicated in MLC phosphorylation can substitute for ROCK1/2 in the generation of actomyosin contractility . In summary , we showed that short-term ( three days after Ad-Cre infection ) depletion of ROCKs induces a previously described ‘tail retraction’ phenotype but surprisingly long-term ( from five days after Ad-Cre infection ) depletion induces a further change in morphology resulting in a flattened morphology and decrease in central stress fibers . The flattened and enlarged morphology observed upon long-term depletion of ROCK1 and ROCK2 is typical of cellular senescence . Indeed , Rock1∆/∆;Rock2∆/∆ MEFs , or MEFs treated with H1152 or blebbistatin for prolonged periods exhibited a significant increase in staining for senescence-associated β-galactosidase ( SA-βGal ) ( Debacq-Chainiaux et al . , 2009 ) than control cells ( Figure 3A ) . Senescence was also observed after prolonged treatment with other ROCK inhibitors such as GSK269962A , AT13148 , GSK429286A and chroman1 ( Figure 3—figure supplement 1A ) . To determine whether abrogation of ROCK function causes senescence in vivo , we used the recently described inhibitor AT13148 that has been shown to potently inhibit ROCK and has been previously used in vivo ( Sadok et al . , 2015; Yap et al . , 2012a ) . 690cl2 mouse melanoma cells were injected intra-dermally into CD1 athymic mice , and the mice treated with AT13148 at 40 mg/kg . Senescence was detected by SA-βGal staining of tumors . We observed a decrease in tumor growth when mice were treated with AT13148 compared to controls ( Figure 3—figure supplement 1B ) . Additionally , we observed an increase in the number of SA-βGal positive cells within tumors ( Figure 3—figure supplement 1C ) . 10 . 7554/eLife . 12203 . 009Figure 3 . Rock1∆/∆;Rock2∆/∆ cells undergo senescence and show a cell cycle block . ( A ) Images of MEFs treated with H1152 , blebbistatin ( + ) , blebbistatin ( +/- ) for 5 days and Rock1f/f;Rock2f/f control , Rock1∆/∆;Rock2∆/∆ MEFs , 5 days after Ad-GFP and Ad-Cre-GFP infection and followed by SA-βgal staining . Scale bars are 50 µm . Graph shows number of SA-βgal expressing cells divided by total number of cells and SD of >100 cells from three independent experiments and p-values were calculated using Student’s t-test: * p<0 . 001 . ( B ) Images of Rock1f/f;Rock2f/f and Rock1∆/∆;Rock2∆/∆ MEFs stained with phalloidin and DAPI . Overlay images shown . Arrows in image indicate binucleate cells . Scale bars are 50 µm . Bar chart shows average data and SD of nuclei in > 150 cells from 5 independent experiments . Values were calculated using Student’s t-test: * p<0 . 05 . ( C , D ) Analysis of cell division of Rock1f/f;Rock2f/f and Rock1∆/∆;Rock2∆/∆ MEFs in time-lapse movies . ( C ) Quantification of failed cell divisions resulting in binucleate cells . ( D ) Quantification of average number of cells dividing . Graphs show average data and SD of > 300 cells from at least five independent experiments . ( E ) Cell cycle profiles of Rock1f/f;Rock2f/f and Rock1∆/∆;Rock2∆/∆ MEFs . The graph shows the percentage of cells in G2/M ( top ) , S ( middle ) and G1 ( bottom ) phase of the cell cycle . Error bars represent SD . Data are from 5 independent experiments and p-values were calculated using Student’s t-test: * p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 00910 . 7554/eLife . 12203 . 010Figure 3—figure supplement 1 . Analysis of senescence in ROCK-inhibited cells and induction of a cell cycle block in ROCK-depleted cells . ( A ) Melanoma lines were treated with the indicated inhibitors for 7 days and subjected to SA-βgal staining . Scale bars are 50 µm . ( B , C ) Intra-dermal injection of 690cl2 melanoma cells into CD1 athymic mice followed by treatment with AT13148 ( 40 mg/kg ) or vehicle . The graph shows average tumor volume in mm3 and SEM ( B ) . Values calculated by Student’s t-test are as indicated . The lungs were analyzed by SA-βGal staining . The graph shows the number of SA-βGal-positive cells per 10 fields of view ( C ) . p-values calculated by ANOVA are as indicated . ( D ) Rock1f/f;Rock2f/f and Rock1∆/∆;Rock2∆/∆ cells treated with nocodazole ( 4 µg/ml ) or vehicle for 48 hrs were subjected to propidium iodide staining and flow cytometry analysis . The graph shows the percentage of cells in G2/M ( top ) , S ( middle ) and G1 ( bottom ) phase of the cell cycle . Data are from three independent experiments and error bars represent SD . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 010 Analysis of Rock1∆/∆;Rock2∆/∆ cells by microscopy revealed a significant increase in the number of cells with two nuclei ( Figure 3B ) , indicating a failure of cytokinesis . During cytokinesis an actomyosin-based contractile ring is formed to drive cleavage furrow ingression , which subsequently results in two daughter cells . ROCK and the related Citron kinase have been shown to localize to the cleavage furrow , yet it is not entirely clear which kinase regulates MLC phosphorylation at the furrow to generate actomyosin contractility ( Kosako et al . , 1999; 2000; Madaule et al . , 1998 ) . Analysis of time-lapse movies further confirmed a role for ROCK1 and 2 in cytokinesis as a greater percentage of Rock1∆/∆;Rock2∆/∆ cells failed to divide compare to wild-type cells ( Figure 3C and Videos 3 , 4 ) . In addition to an increased number of binucleate cells , analysis of time-lapse movies of Rock1∆/∆;Rock2∆/∆ cells revealed an overall lower number of dividing cells ( Figure 3D and Videos 3 , 4 ) . These data suggest that cells depleted of ROCK1 and 2 do not initiate division further , thus indicating a cell cycle block . 10 . 7554/eLife . 12203 . 011Video 3 . Rock1f/f;Rock2f/f control , 5 days after infection with Ad-GFP , 20X magnification . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 01110 . 7554/eLife . 12203 . 012Video 4 . Rock1∆/∆;Rock2∆/∆ , 5 days after infection with Ad-Cre , 20X magnification . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 012 To analyze the cell cycle profile of Rock1∆/∆;Rock2∆/∆ cells , cells were stained with propidium iodide ( PI ) along with control cells for analysis by flow cytometry . Fewer double knockout than wild-type MEFs were in G1 or S phase , and a higher percentage of cells in G2/M phase of the cell cycle ( Figure 3E ) . The increased percentage may be due to the fact that binucleate cells in G1 will have the same DNA content as cells in G2/M . To investigate a potential cell cycle block , we treated Rock1∆/∆;Rock2∆/∆ cells and their controls with nocodazole , commonly used to arrest cells with a G2/M phase DNA content ( Blajeski et al . , 2002 ) . In controls , nocodazole led to an increased proportion of cells with G2/M DNA content as expected ( Figure 3—figure supplement 1D ) . In Rock1∆/∆;Rock2∆/∆ cells , however , distribution of cells in the three cell cycle phases is very similar between control treated and nocodazole treated cells ( Figure 3—figure supplement 1D ) . These data suggest that ROCK depletion causes a cell cycle block prior to G2/M . As there is an overall decrease in S phase cells , the block is likely to be in the G1 phase of the cell cycle . We have shown that ROCK depletion or inhibition induces cellular senescence; therefore , we next investigated whether they affected key cell cycle proteins involved in senescence . Two major pathways are activated by senescence signals . Activation of the transcription factor Trp53 induces expression of p21 , a cyclin-dependent kinase inhibitor ( CDKI ) ( Brown et al . , 1997 ) . The other pathway is induction of the CDKI p16 , which prevents CDK4/6-mediated phosphorylation of retinoblastoma ( Rb ) proteins , thereby blocking E2F-dependent transcription ( Brenner et al . , 1998; Nevins , 2001 ) . To determine whether blockade of Rb or Trp53 function would overcome senescence resulting from abrogation of ROCK function , we stably infected Rock1f/f;Rock2f/fMEFs with simian virus 40 large T-antigen ( SV40 Large T ) , which is known to inactivate the tumor suppressors Trp53 and Rb ( DeCaprio et al . , 1988; Kierstead and Tevethia , 1993 ) . The cells were then infected with Ad-Cre , to generate Rock1∆/∆;Rock2∆/∆ cells ( Figure 4—figure supplement 1A ) . Inactivation of Trp53 and Rb did not prevent the senescence phenotype caused by loss of ROCK , and cells displayed a similar growth defect to the one observed in Rock1∆/∆;Rock2∆/∆ MEFs without SV40 Large T ( Figure 4—figure supplement 1A ) . To further confirm that the senescent phenotype did not involve activation of Trp53 , MEFs were isolated from Rock1f/f;Rock2f/f;Trp53f/f mice and ROCK and Trp53 depleted in vitro by Ad-Cre infection . Rock1∆/∆;Rock2∆/∆;Trp53∆/∆ MEFs were found to undergo senescence , and were defective in their proliferation ( Figure 4—figure supplement 1B ) . Western blot analysis confirmed depletion of both ROCKs and Trp53 ( Figure 4—figure supplement 1B ) . We then treated Trp53 wild type and null cells with the ROCK inhibitors H1152 and GSK269962A and found an increase in senescence compared to control cells ( Figure 4—figure supplement 1C ) . Therefore , the classical senescence pathways did not seem to be affected in Rock1∆/∆;Rock2∆/∆ cells or cells treated for prolonged periods with H1152 . To identify protein changes associated with senescence , we used immunoblotting and quantitative mass spectrometry ( Ong et al . , 2002 ) . The protein levels of some cell cycle regulators appeared to change upon treatment with Cre recombinase alone , so for proteomics analysis we focused on long-term H1152 treatment . Annotation enrichment analysis ( Cox and Mann , 2008 ) of the proteome-wide changes after long-term ROCK inhibition revealed many cell cycle-related protein categories that were significantly modulated in ROCK-inhibited cells ( Supplementary file 2 ) , consistent with a prominent cell-cycle phenotype . However , when we analyzed proteins that are known to control G1 cell-cycle arrest and senescence , we did not detect significant changes in the levels of these proteins . These included known regulators , such as p16 , p19 , p27 , p21 , CyclinD1 , CyclinE , or phosphorylation of Rb ( Supplementary file 2 and Figure 4—figure supplement 2A , B ) . However , CDK1 , CyclinA and CKS1 protein levels were significantly reduced upon loss of and long-term inhibition of ROCK1 and 2 ( Supplementary file 2 and Figure 4A–C ) . Long-term treatment of cells with blebbistatin ( blebbistatin +/- ) , but not the inactive enantiomer ( blebbistatin + ) , also decreased the levels of CDK1 , CyclinA and CKS1 ( Figure 4B , C ) , suggesting that it is loss of contractility downstream of ROCK loss that results in down-regulation of these proteins . However , it cannot be ruled out that both ROCK inhibition/depletion as well as myosin II inhibition independently lead to a similar phenotype . 10 . 7554/eLife . 12203 . 013Figure 4 . Downregulation of CKS1 , CyclinA and CDK1 on abrogation of ROCK function . ( A ) MEFs were treated with H1152 or vehicle for 3 days and the proteome analyzed by quantitative mass spectrometry . Graph shows log ratios of identified proteins . Data are from duplicate experiments of reciprocal SILAC labelling . The ‘Significant-B’ outlier test was used to determine significantly regulated peptides or proteins , using a Benjamini-Hochberg FDR rate of 5% . ( B , C ) MEFs of different genotypes or treated with indicated inhibitors were immunoblotted for pCDK1 , total CDK1 , Cyclin A , CKS1 , ROCK1 , ROCK2 and total ERK as loading control ( B ) . ( C ) Quantification of western blot analyses . Graphs show protein levels of CDK1 , CyclinA and CKS1 divided by total ERK protein levels and SEM in indicated samples . The data are from 5 independent experiments and p-values were calculated using Student’s t-test: * p<0 . 05 , p<0 . 005 , * p<0 . 001 . ( D ) qPCR analysis of mRNA levels of Cks1b , Cdk1 and Ccna2 in indicated samples . Graphs show average normalized mRNA levels and SD from at least five independent experiments , each carried out in triplicates . p-values were calculated using Student’s t-test: * p<0 . 05 , *** p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 01310 . 7554/eLife . 12203 . 014Figure 4—figure supplement 1 . Analysis of cellular senescence in Rock1∆/∆;Rock2∆/∆;Trp53∆/∆ cells and in ROCK-inhibited Trp53 wild type and Trp53 null cells . ( A ) Rock1f/f;Rock2f/f cells , immortalized with SV40 Large T , were infected with Ad-GFP or Ad-Cre-GFP to generate Rock1f/f;Rock2f/f control and Rock1∆/∆;Rock2∆/∆ cells . These were subjected to growth analysis . The graph shows average data and SD from 3 independent experiments , each carried out in triplicates . ( B ) Rock1f/f;Rock2f/f;Trp53f/f and Rock1∆/∆;Rock2∆/∆;Trp53∆/∆ cells were seeded at equal numbers and kept for ~7 days allowing colonies to form . In parallel , cells were subjected to SA-βgal staining and western blot analysis using indicated antibodies . Scale bars in images are 50 µm . ( C ) Trp53 wild-type and Trp53 null lines were treated with the indicated inhibitors for 7 days and subjected to SA-βgal staining . Scale bars are 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 01410 . 7554/eLife . 12203 . 015Figure 4—figure supplement 2 . Protein levels of cell cycle regulators in ROCK-inhibited and Rock1∆/∆;Rock2∆/∆ cells . ( A ) MEFs were treated with H1152 or vehicle for 3 days and the proteome analyzed by quantitative mass spectrometry . Graph shows log ratios of identified proteins . The data are from duplicate experiments of reciprocal SILAC labelling . Proteins indicated are Cdkn2a/Arf ( p19 ) , Cdkn2a/Ink4a ( p16 ) , Cdkn2b ( p15 ) , Ccnd2 ( CyclinD2 ) , Ccnd1 ( CyclinD1 ) , Ccnd3 ( CyclinD3 ) , Cdk2 , Cdk4 , Cdkn1b ( p27 ) , Ccnb1 ( CyclinB1 ) . ( B ) Rock1f/f , Rock2f/f , Rock1f/f;Rock2f/f and where indicated wild type cells infected with Ad-Cre-GFP were subjected to immunoblotting with antibodies against proteins indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 015 Cyclin-dependent kinase regulatory subunits 1 and 2 ( Cks1b and Cks2 ) form a complex with a subset of cyclin-dependent kinases ( Cdks ) that are the essential regulators of cell cycle progression . Mammalian cells in which Cks is silenced exhibit a slowed G1 progression and arrest in G2/M phase of the cell cycle , and are therefore unable to enter mitosis ( Martinsson-Ahlzen et al . , 2008; Westbrook et al . , 2007 ) . It was previously reported that siRNA knockdown of CKS1 protein in Cks2-/-Cks1b+/- MEFs leads to a decrease in proliferation through defects in Cdk1 , Ccna2 and Ccnb1 transcription and expression of the equivalent proteins CDK1 , CyclinA and CyclinB ( Martinsson-Ahlzen et al . , 2008 ) . Analysis of mRNA levels revealed that , in Rock1∆/∆/Rock2∆/∆ cells , Cdk1 and Ccna2 mRNA levels were significantly reduced , whereas Cks1b mRNA remained unchanged ( Figure 4D ) . APC/CCdh1 ( anaphase-promoting complex and its activator Cdh1 ) have been previously implicated in regulating CKS1 as well as SKP2 stability ( Bashir et al . , 2004 ) . We were unable to detect SKP2 protein in the cells , but CKS1 protein was still downregulated when Cdh1 null MEFs were treated with H1152 or blebbistatin ( data not shown ) . Furthermore , MG132 treatment of cells did not prevent downregulation of CKS1 suggesting the proteasome is not responsible for this decrease ( data not shown ) . To determine whether the down-regulated cell cycle regulators are responsible for the phenotype induced by the loss of ROCK 1 and 2 , we employed shRNA knockdown of Cks1b , Cdk1 and Ccna2 . Cells depleted of CKS1 , CDK1 or CyclinA were unable to proliferate ( Figure 5A ) and had increased staining for SA-βGal ( Figure 5B ) . Interestingly , it has been shown previously that the knockout of Cks1b ( Hoellein et al . , 2012 ) or Cdk1 ( Diril et al . , 2012 ) leads to cellular senescence . We found that knockdown of Cks1b affected protein levels of CDK1 and CyclinA , while knockdown of Cdk1 affected protein levels of CKS1 and CyclinA and knockdown of Ccna2 affected levels of CKS1 and CDK1 ( Figure 5C ) , confirming a clear link between the expression of these proteins . Overall , these results indicate that cell cycle proteins CKS1 , CDK1 and CyclinA are downregulated upon loss of actomyosin contractility , and their depletion causes defects in cell proliferation and increased cellular senescence . 10 . 7554/eLife . 12203 . 016Figure 5 . Knockdown of Cdk1 and Cks1b leads to cellular senescence . ( A–C ) MEFs were infected with either control shRNA or shRNAs targeting Cks1b , Cdk1 and Ccna2 . Four days after infection , equal numbers of cells were plated and kept for seven days , allowing colonies to form ( A ) . Five days after infection with shRNAs , cells were also subjected to SA-βGal staining ( B ) and immunoblotting with antibodies against pCDK1 , total CDK1 , Cyclin A , CKS1 . ROCK2 and total ERK were used as loading control ( C ) . Scale bars are 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 016 After thorough in vitro characterization of cells lacking ROCK1 and ROCK2 , we investigated the roles of ROCK1 and ROCK2 in tumorigenesis in vivo using genetically engineered mouse tumor models . To this end , we interbred mice carrying the conditional Rock alleles with models of lung tumorigenesis and melanoma . Activation of the LSL-oncogenes KrasG12D and BrafV600E in these models by Cre recombinase also leads to deletion of conditional Rock alleles . For lung tumorigenesis , we used a NSCLC model , exploiting activation of LSL-KrasG12Dand LSL-Trp53R270H in lung epithelial cells ( DuPage et al . , 2009; Jackson et al . , 2005; 2001 ) . Mice were treated with adenovirus expressing Cre ( Ad-Cre ) and sacrificed 24 weeks after treatment . Histological analysis revealed that the total tumor area was increased in Rock1f/f mice ( Figure 6A ) . This was seen for all tumor grades ( Figure 6A ) that have been classified as reported previously ( DuPage et al . , 2009; Jackson et al . , 2005; 2001 ) . This result suggests that ROCK1 may have a tumor suppressor function in this specific mouse model , of which the mechanism requires further investigation and will be the subject of another manuscript . However , no difference was seen comparing tumor area of Rock1f/f;Rock2f/f or Rock2f/f with wild type mice ( Figure 6A ) . Western blot analysis of cell lines generated from Rock1f/f;Rock2f/f;KrasG12D; Trp53R270H-driven tumors showed that they always retained ROCK1 or ROCK2 protein ( Figure 6B ) and contained un-recombined floxed alleles of Rock1 and 2 . To investigate the consequences of complete loss of ROCK function , we infected these ‘retainer’ cell lines isolated from lung tumors arising in Rock1f/f;Rock2f/f mice , which retained ROCK1 protein ( Rock1f/f;Rock2∆/∆ ) , with Ad-Cre-GFP to excise the retained Rock1 locus . The same procedure was applied to cells isolated from tumors arising in Rock1f/wt;Rock2f/f mice ( Rock1∆/wt;Rock2∆/∆ ) . Rock1∆/∆;Rock2∆/∆ cells , but not Rock1∆/wt;Rock2∆/∆ cells , failed to proliferate when compared to the control cells in a colony formation assay ( Figure 6C ) . This further highlights the fact that one wild-type allele of Rock is sufficient to allow cell proliferation . Importantly , we find that just one allele of either Rock is sufficient to rescue the proliferation defect which was also observed using MEFs ( data not shown ) and this Rock1∆/wt;Rock2∆/∆ cell line ( Figure 6C ) . 10 . 7554/eLife . 12203 . 017Figure 6 . ROCK is essential for tumorigenesis . Cohorts of mice carrying LSL-KrasG12D; LSL-Trp53R270H alleles and the indicated Rock alleles were treated once with Ad-Cre viruses by intranasal inhalation . 24 weeks later , the mice were sacrificed and the lungs were analyzed by histopathology . ( A ) Graphs show tumor area divided by lung area for indicated genotypes and tumor grades . The data are also displayed as a sum of tumor area divided by lung area for the indicated genotypes . Graphs show average data and SEM . The total number of mice analyzed is indicated in Scatter Plots . ( B ) Cell lines derived from KrasG12D;Trp53R270H tumors of either Rock1f/f , Rock2f/f , Rock1f/f;Rock2f/f or wild-type genotype were subjected to immunoblotting with indicated antibodies . ( C ) Isolated lung tumor cell lines of Rock1∆/wt;Rock2∆/∆ genotype as well as a cell line which retained both Rock1 alleles ( Rock1f/f;Rock2∆/∆ ) were treated with Ad-GFP and Ad-Cre-GFP . 3 days after infection , cells were plated and kept for ~5 days allowing colonies to form . ( D ) Absolute quantification of ROCK1 and 2 , using selective reaction monitoring ( SRM ) . Bar graphs show the mean concentration in nM of ROCK1 ( black ) and ROCK2 ( grey ) across at least 3 biological replicates from each genetic background . Line 1–3 indicate different cell lines isolated from NSCLCs and Tumors 1–4 represent different micro-dissected tumors . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 01710 . 7554/eLife . 12203 . 018Figure 6—figure supplement 1 . Analysis of melanomagenesis in wt , Rock1f/f , Rock2f/f and Rock1f/f;Rock2f/f mice . ( A , B ) Scatter plots showing time until melanomas developed in mice carrying Tyr-CreERT2; BrafV600E alleles and where indicated Ptenf/wt , Ptenf/f and Rock alleles ( A ) as well as Kaplan-Meier survival curves ( B ) . The same number of mice indicated in scatter plots was analyzed in survival curve . ( C ) Absolute quantification of ROCK1 and 2 , using selective reaction monitoring ( SRM ) . Bar graphs show the mean concentration in nM of ROCK1 ( black ) and ROCK2 ( grey ) across at least 3 biological replicates of different cell lines isolated from mice with a Tyr-CreERT2; BrafV600E; Ptenf/wt genetic background with Rock alleles as indicated . Error bars represent SEM . ( D ) Immunoblotting of lysates from melanoma cell lines derived from tumors with ROCK1 and 2 antibodies . GAPDH or ERK were used as loading control . ( E ) Melanoma cell lines of Rock2∆/∆ genotype as well as a cell line which retained both Rock2 alleles ( Rock1∆/∆;Rock2f/f ) were treated with Ad-GFP and Ad-Cre-GFP to generate controls , Rock2∆/∆ and Rock1∆/∆;Rock2∆/∆ cells . 3 days after infection , the cells were plated and kept for ~5 days allowing colonies to form . DOI: http://dx . doi . org/10 . 7554/eLife . 12203 . 018 In order to quantify ROCK1 and ROCK2 protein abundance from each knockout , we established a custom-selected reaction monitoring ( SRM ) targeted mass-spectrometry assay ( Worboys et al . , 2014 ) . This enables the analysis of absolute concentrations of each isoform , allowing us to identify whether one isoform is more expressed than the other and thus could play a more important role . It will also allow analysis of whether loss of one isoform would lead to a compensatory increase in the levels of the other isoform . Isotopically ‘heavy’ synthetic quantotypic ROCK1 and ROCK peptides were correlated with ROCK1 and ROCK2 to calculate absolute protein abundance from each sample . SRM analysis showed that , in cell lines derived from tumors in Rock1f/f;Rock2f/fmice , either ROCK1 or ROCK2 protein was always retained ( Figure 6D ) . ROCK1 and ROCK2 protein was absent in cell lines derived from Rock1f/f or Rock2f/ftumor bearing mice , respectively ( Figure 6D ) . Micro-dissected tumors from Rock1f/f mice showed a reduction of ROCK1 protein levels but not complete depletion as seen in cell lines . This is likely due to contamination from the wild type stromal cells infiltrating the tumor . The same was also observed in tumors from Rock2f/f mice ( Figure 6D ) . In tumors analyzed from Rock1f/f;Rock2f/fmice , ROCK2 protein was retained but low levels of ROCK1 were also present ( Figure 6D ) , most probably from stromal cells . Importantly , ROCK1 protein was not up-regulated in cell lines or tumors from mice with Rock2f/f genotype in a compensatory manner and vice versa . The overall concentration of ROCK1 and 2 proteins was found to be similar , excluding potential effects driven by differential isoform expression . Similar results were also observed in mouse models of melanoma ( Dhomen et al . , 2009 ) . In the melanoma model , expression of BrafV600E from an LSL-BrafV600E allele is driven in melanocytes by a tamoxifen-activated version of Cre recombinase under the control of the melanocyte-specific tyrosinase promoter ( Tyr-CreERT2 ) ( Dhomen et al . , 2009 ) . We also used conditional melanocyte-specific expression of oncogenic Braf in combination with a conditional allele of the Pten tumor suppressor to decrease tumor latency ( Dankort et al . , 2009; Dhomen et al . , 2009 ) ( Figure 6—figure supplement 1A–B ) . No differences were found in onset of tumorigenesis ( Figure 6—figure supplement 1A ) or survival ( Figure 6—figure supplement 1B ) when comparing BrafV600E with BrafV600E;Rock1f/f or BrafV600E;Rock1f/f;Rock2f/f mice . Tumor latency was decreased in BrafV600E;Rock2f/fmice only but not in combination with Ptenf/f or Ptenf/wt ( Figure 6—figure supplement 1A ) . Survival was also found to be decreased in BrafV600E:Rock2f/fmice as well as in combination with Ptenf/f ( Figure 6—figure supplement 1B ) , suggesting that unlike in the NSCLC model , ROCK2 may be having a tumor suppressor role in this model . Nevertheless , similar to the NSCLC model , analysis of cell lines generated from these melanomas , by SRM and immunoblotting , revealed retention of either ROCK protein in mice with Rock1f/f;Rock2f/fgenotype , but efficient excision of the floxed locus in mice with Rock1f/f and Rock2f/f genotypes ( Figure 6—figure supplement 1C–D ) . These results further support our observation that ROCK protein is required for tumorigenesis , since retention of at least one Rock allele is selected for in both Ras-driven lung tumors and Raf-driven melanomas . Furthermore , treatment of melanoma cell lines with Ad-Cre-GFP , to excise the retained Rock locus , revealed that Rock1∆/∆;Rock2∆/∆ cells but not Rock2∆/∆ cells fail to proliferate compared to control cells in a colony formation assay ( Figure 6—figure supplement 1E ) . This was also seen when compared to Rock1∆/∆ cells ( data not shown ) . Collectively , analysis of ROCK1 and ROCK2 protein expression in tumors and tumor derived cell lines suggests that the absence of ROCK1 and ROCK2 is incompatible with tumorigenesis . However , either ROCK1 or ROCK2 protein is sufficient to promote tumorigenesis . Rho-kinase generates actomyosin contractility and thus has been described as an important regulator of cell motility and plasticity ( Sanz-Moreno et al . , 2008 ) . Several reports using siRNA to knockdown ROCK1 or 2 have been published suggesting that ROCK1 and 2 proteins are not redundant . These include different effects on actomyosin contractility , stress fibers and focal adhesions in fibroblasts ( Yoneda et al . , 2005 ) , opposing effects on cell proliferation and changes in cell migration in glioblastoma cell lines ( Mertsch and Thanos , 2014 ) . Using ROCK1 and 2 single knockout MEFs , differences in stress fiber formation and Cofilin phosphorylation have been described ( Shi et al . , 2013 ) . In a further study , knockdown of either ROCK1 or ROCK2 alone was sufficient to reduce anchorage-independent growth of NSCLC cell lines - also suggesting their functions are not redundant ( Vigil et al . , 2012 ) . All of these studies , however , describe very different effects of ROCK1 or ROCK2 , and at least partially contradict each other . The use of siRNA knockdown , ROCK inhibitors or germ-line knockout MEF lines has several limitations , such as off-target effects , differences in growth properties or cell morphology . To establish an isogenic system and to circumvent the severity of the embryonic phenotypes in germ-line knockouts of Rock1 and 2 ( Shimizu et al . , 2005; Thumkeo et al . , 2003; Thumkeo et al . , 2005 ) , we generated conditional targeted alleles of Rock1 and Rock2 , allowing us to analyze of the role of ROCK1 and ROCK2 , and permitting the generation of cells that lack both ROCK1 and 2 . Our data with genetic deletion in an isogenic fashion show that either ROCK1 or ROCK2 is sufficient to regulate MLC and MYPT phosphorylation and thus actomyosin contractility . Known targets such as the phosphorylation of cofilin are not affected in single or double knockout cells . However , using quantitative mass spectrometry treatment of cells with the ROCK inhibitor H1152 , we show that Cofilin phosphorylation is affected after short-term ROCK inhibition but recovers upon long-term treatment . This indicates that other kinases are able to take over and phosphorylate Cofilin . MLC phosphorylation , however , is not recovered indicating that ROCK is the major kinase regulating MLC phosphorylation . We demonstrate that the depletion of ROCK1 and 2 leads to severe defects in cell proliferation , which is not observed in cells lacking either ROCK1 or 2 . Defects in cell proliferation are also observed in cells treated long-term with ROCK inhibitors or the MyosinII inhibitor blebbistatin . Previous work with the ROCK inhibitor Y-27632 has led to contradictory results on the requirement for ROCK in cell proliferation . We believe that this may reflect the relatively low potency of Y-27632 as a ROCK inhibitor , since we found that treatment with several other ROCK inhibitors ( H1152 , GSK269962A , AT13148 , GSK429286A and chroman1 ) , which have been shown to be more potent than Y-27632 ( Sadok et al . , 2015 ) , could mimic the effects of genetic deletion of Rock1 and Rock2 . Cell proliferation defects upon deletion of ROCK1 and 2 are due to a G1 arrest and a block in cytokinesis . Interestingly , the cell proliferation defects arising from abrogation of ROCK function result in cellular senescence , both in vitro and in vivo . This cellular senescence does not appear to involve changes in levels of the cell cycle inhibitors p16 , p21 , p27 , but the down-regulation of cell-cycle proteins CDK1 , Cyclin A and the cyclin-dependent kinase inhibitor CKS1 . CKS1 is known to regulate levels of CDK1 and Cyclin A ( Martinsson-Ahlzen et al . , 2008; Westbrook et al . , 2007 ) , and can function in G1 and G2 phases of the cell cycle ( Tang and Reed , 1993 ) . Interestingly , it has been shown that Cks1b-/- cells are defective in proliferation , have reduced proportions of cells in S- and G1-phase , accumulation of cells with G2/M DNA content ( Hoellein et al . , 2012; Spruck et al . , 2001 ) and exhibit senescence . We therefore propose that an essential function of ROCK in cell proliferation is to regulate the expression of CKS1 . In addition , this ROCK-driven CKS1 down-regulation is a novel marker of senescence . CKS1 does not appear to be regulated at the mRNA level , thus further studies are required to understand how ROCK function could regulate CKS1 protein . Protein degradation was not mediated by the proteasome or APC/CCdh1 complex ( data not shown ) . APC/CCdh1 was previously shown to regulate the degradation of CKS1 and Skp2 ( Bashir et al . , 2004 ) . CKS1 and Skp2 are the targeting subunits of the SCFSkp2–Cks1 ubiquitin ligase regulating p21 and p27 degradation ( Sutterluty et al . , 1999 ) . Even though we were unable to detect Skp2 in our system , we do show that p21 and p27 levels are not affected upon loss of ROCK , suggesting a novel regulation of CKS1 and function in mediating senescence . More work is required to unravel the complete mechanism and determine how downregulation of these proteins is mediated . We observed that treatment with blebbistatin a direct inhibitor of myosin II led to cell cycle arrest , senescence and reduced levels of CKS1 , CyclinA and CDK1 , potentially suggesting that ROCK-dependent regulation of cell proliferation is through the inhibition of actomyosin contractility rather than another target of ROCK signaling . Using tumor samples as well as tumor derived cell lines , from two different genetically engineered mouse models of tumorigenesis , we find that tumor formation is incompatible with homozygous deletion of both Rock1 and Rock2 , since tumors arising in Rock1f/f;Rock2f/f mice always retained either un-recombined Rock1f/for Rock2f/f alleles and showed expression of either ROCK1 or ROCK2 , respectively . Targeting either Rock1 or Rock2 appears to have mild and divergent consequences in the mouse models . Deletion of Rock1 appears to increase tumor burden in the NSCLC model , whereas deletion of Rock2 leads to decreased tumor onset and survival in the melanoma model . These putative tumor suppressor functions of Rock1 and Rock2 do not appear to be a consequence of differential expression as the absolute concentration of the two isoforms was found to be similar by SRM . These results require further investigation and will be the basis of a separate manuscript . The retention of ROCK protein in Rock1f/f;Rock2f/f mice , however , demonstrates that ROCK function is essential for tumorigenesis . Incomplete excision of proteins that affect cell proliferation has been reported in several studies ( Kissil et al . , 2007; Peschard et al . , 2012 ) . In the tumor models used in this study , excision of Rock alleles and tumor initiation are achieved simultaneously in an inducible fashion; therefore we propose that the activation of Cre recombinase leads to the formation of a mixed population of cells in which either Rock , or both Rock1 and 2 , have been excised . The ROCK1 and 2 double-null cells , however , are unable to form a tumor as they are defective in cell proliferation and senesce . The remaining cells , that is those that have retained either Rock allele form tumors but these are the genetic equivalent of a single knockout . In summation , our data reveal that ROCK proteins act redundantly to regulate actomyosin contractility , and that this function of ROCK is necessary , by supporting levels of the cell cycle regulators Cyclin A , CDK1 and CKS1 , for senescence , cell cycle progression and proliferation both in vitro and in vivo . Recent studies have identified a role for ROCK inhibition in combinatorial treatments in BRAF-mutant melanoma ( Smit et al . , 2014 ) , as well as in Kras-driven lung cancers ( Kumar et al . , 2012 ) . Our data thus provide a molecular basis , and highlight the need , for exploiting the use of effective ROCK inhibitors as anti-cancer agents either alone or in combination . However , incomplete or specific inhibition of only one isoform of ROCK alone may not be an effective treatment , and may even lead to adverse results as our results also highlight isoform specific tumor suppressive functions for ROCK1 in lung and ROCK2 in melanoma models . All animal procedures were approved by the Animal Ethics Committee of the Institute of Cancer Research in accordance with National Home Office regulations under the Animals ( Scientific Procedures ) Act 1986 . Rock1 and 2 gene-targeting vectors were made by inserting arms of homology into the PGKneoF2L2DTA vector ( Phillipe Soriano , Addgene plasmid #13445 [Hoch and Soriano , 2006] ) , where loxP sites flank exon 6 of Rock1 ( Rock1f/f ) and exons 5–6 of Rock2 ( Rock2f/f ) . Homologous recombination was performed in Bruce4 ES cells of C57BL/6J origin ( Ozgene ) . Positive clones were identified by PCR and confirmed by Southern blotting before injection into albino C57BL/6J blastocysts to generate chimeras in a pure C57BL/6J background . Rock1- and 2-targeted mice were subsequently crossed to PGK-Flp transgenic mice ( provided by the Cancer Research UK transgenic facility ) to generate conditional knockout mice . For the lung tumor model Rock1 and 2 targeted mice were crossed to LSL-KrasG12D ( the mice were a generous gift from David Tuveson [Tuveson et al . , 2004] ) and Trp53R270H . Mice were anesthetized and treated once by intranasal inhalation using 2×106 pfu/mouse of Adeno-Cre viruses as described previously ( DuPage et al . , 2009 ) . Tumor grading is according to the Jacks’ lab classification; AAD: Atypical Adenomatous Dysplasia not forming any solid elements; G1: Uniform small rounded nuclei; G2: Enlarged nuclei with some variation and hyperchromatism , small nucleoli; G3: Markedly enlarged and variable nuclei , prominent nucleoli , occasional mitotic figures; G4: Pleomorphic nuclei , prominent nucleoli , mitotic figures easily found , focal necrosis . For the melanoma model , Rock1 and 2 targeted mice were crossed to mice carrying the transgenes LSL-BrafV600E , Ptenf/f ( Ptenf/wt , Ptenwt/wt ) and Tyr-CreERT2 ( the mice were a generous gift from Richard Marais ) . Briefly , two treatment regimens were used; mice carrying Ptenf/wt and Ptenwt/wt targeted alleles were treated with 100 µl of amoxifen ( 100 mg/ml in ethanol ) , tamoxifen was applied evenly , to the shaven skin on the dorsal surface of 8-week-old mice . Mice carrying Ptenf/f targeted alleles were treated with 0 . 5 µl of 4-Hydroxytamoxifen ( 4-OHT ) ( 10 mM in DMSO ) ; 4-OHT was applied to shaven skin on the dorsal surface of 6-week-old mice three times on three consecutive days . MEFs were isolated from E13 . 5 embryos and cultured in a low-oxygen ( 3% ) incubator ( Parrinello et al . , 2003 ) in Dulbecco’s modified Eagle’s medium containing 10% fetal bovine serum . Tumors from mouse models were incubated in trypsin and put into culture to allow cells to grow out . Melanomas were incubated at 20% oxygen for fibroblasts present in the culture to undergo senescence and melanoma cells to grow out . All cell lines were subsequently analyzed by PCR and immunoblotting and their genotypes confirmed . 690cl2 cells were generated in-house by N . Dhomen and R . Marais from tumors arising in the BrafV600E , Ptenf/fmouse model ( Dhomen et al . , 2009 ) . Trp53 MEFs were generated in-house from the Trp53 knockout mouse ( Harvey et al . , 1993 ) and their genotype confirmed by PCR . All cell lines used for this study were generated in house and tested free of mycoplasma . Adenoviruses were obtained from the Gene Transfer Vector Core at the University of Iowa . SMARTchoice lentiviral shRNA vectors were obtained from Dharmacon and produced in 293T cells co-transfected with pMD2 . G and psPAX2 packaging vectors ( http://tronolab . epfl . ch/lentivectors ) . Where indicated , cells were infected using amphotropic retrovirus produced by transfection of pBABE vectors into Phoenix packaging cells ( https://web . stanford . edu/group/nolan/protocols/pro_optimiz . html ) . pBabe-puro-RasV12 and pBABE-hygro-Trp53DD were a kind gift from Bob Weinberg ( Addgene plasmids # 1768 , # 9058 ) ( Hahn et al . , 2002 ) , pBABE-puro-SV40-LT was a kind gift from Thomas Roberts ( Addgene plasmid # 13970 ) ( Zhao et al . , 2003 ) . Primary antibodies were used , unless indicated otherwise , at a dilution of 1:1000 for immunoblotting and 1:200 for immunofluorescence . Secondary antibodies were used at a dilution of 1:10 , 000 for immunoblotting ( LI-COR Biosciences ) and 1:500 for immunofluorescence ( Alexa Fluor , Life Technologies ) The following antibodies were used: p44/42 MAPK ( ERK1/2 ) ( 1:3000 ) ( Cell Signaling ) , phospho-p44/42 ( ERK1/2 ) ( Thr202/Tyr204 ) ( Cell Signaling ) , GAPDH ( 1D4 ) ( Novus Biologicals ) , Rock1 ( H-85 ) ( Santa Cruz Biotechnology ) , Rock2 ( ROKα ) ( BD Biosciences ) , phospho-MLC ( Thr18/Ser19 ) ( Cell Signaling ) , phospho-MLC ( Ser19 ) for immunofluorescence ( Cell Signaling ) , MRCL3/MRLC2/MYL9 clone ( E-4 ) ( Santa Cruz Biotechnology ) , p21/CIP1 ( M-19 ) ( Santa Cruz Biotechnology ) , Cks1 ( Invitrogen ) , p34cdc2 ( Invitrogen ) , Cdc2 ( Y15 ) ( Cell Signaling ) , CyclinA ( CY-A1 ) ( Sigma Aldrich ) Inhibitors used were H1152 ( Calbiochem/Merck Millipore ) , GSK429286A ( Selleckchem ) , GSK269962A ( Axon Medchem ) , chroman1 ( ApexBio ) , Blebbistatin + and Blebbistatin +/- ( Calbiochem/Merck Millipore ) , nocodazole ( Sigma Aldrich ) , AT13148 was synthesized in-house ( Yap et al . , 2012b ) . Cells were seeded on glass coverslips , or inside a 1 . 8 mg/ml collagen gel layer prepared according to manufacturer's instructions ( PureCol , Advanced Biomatrix ) , and left to set in 1 µ-slide chambers ( ibidi ) . The cells were then fixed with 4% paraformaldehyde , permeabilized with 0 . 1% Triton X-100 and blocked with 2% BSA before staining with phospho-MLC ( Ser19 ) ( Cell Signaling ) and F-actin ( Phalloidin; Life Technologies ) and mounting in Vectashield Hardset mounting medium containing DAPI . Confocal images were acquired using a Zeiss LSM 710 microscope ( 40x oil immersion objective ) . Time-lapse phase-contrast microscopy was performed in a humidified CO2 chamber under a Diaphot inverted microscope ( Nikon , UK ) with a motorized stage ( Prior Scientific , UK ) controlled by Simple PCI software ( Compix ) . Cells were monitored by taking an image every 10 min over a period of 16–24 hr and their division quantified . For analysis of cell motility , cells were tracked using ImageJ analysis software ( http://rsb . info . nih . gov/ij ) and the ibidi chemotaxis and migration tool ( www . ibidi . com ) used to determine migration speed ( in µm/min ) and directionality ( displacement/track length , where displacement is the distance from the start to end point for each cell ) were determined using ImageJ analysis software ( http://rsb . info . nih . gov/ij ) and ibidi chemotaxis and migration tool ( www . ibidi . com ) . Cells were grown to approximately 80% confluence , washed once in PBS and lysed in SDS loading buffer ( 100 mM Tris pH 6 . 8 , 10% Glycerol , 2% SDS , [100 mM DTT] ) , homogenized by sonication for 5 s and boiled for 10 min . DTT , to a final concentration of 100 mM , was added after measuring protein concentration by Pierce BCA Assay ( Thermo Scientific ) and before boiling . Cell lysates containing equal amounts of protein were resolved by SDS-PAGE on precast NuPAGE Novex 10% Bis-Tris Midi Protein Gels ( Life Technologies ) and transferred onto PVDF-FL membranes ( Millipore ) . Membranes were incubated with indicated antibodies and visualized by fluorescently conjugated secondary antibodies using the Odyssey Infrared Imaging System ( LI-COR Biotechnology , Cambridge , UK ) . A total of 1x106 cells were embedded in 500 µl of a bovine Collagen I mix prepared according to the manufacturer's protocol ( PureCol , Advanced Biomatrix ) with a final concentration of approximately 2 . 1 mg/ml in a 24-well plate format . The gel was incubated at 37°C for 1 hr for it to polymerize and 500 µl of medium , containing inhibitors where indicated , was added . A pipette tip was then used to loosen the gel from the tissue culture vessel and incubated for 16 hr . The plates were scanned and percentage contraction was calculated by using the formula 100- ( [area of the gel/area of an empty well]100 ) of which areas were determined using ImageJ software ( http://rsbweb . nih . gov/ij/ ) ( Hooper et al . , 2010 ) . Cells were fixed in ice-cold 70% ethanol for at least 1 hr at 4°C and washed twice with PBS . DNA was stained with propidium iodide ( PI ) ( Sigma Aldrich ) at a concentration of 20 µg/ml and in the presence of 100 µg/ml RNaseA ( Qiagen ) . The samples were analyzed on a BD LSR II flow cytometer using BD FACSDiva software ( BD Biosciences ) . Where indicated , cells were treated with Nocodazole . Staining to detect senescence-associated beta-galactosidase ( SA-βgal ) activity was carried out as described by Debacq-Chainiaux et al . ( Debacq-Chainiaux et al . , 2009 ) . For in vivo staining , the mouse melanoma cell line ( 690cl2 ) used was generated from tumors arising in mice carrying BrafV600E , Ptenf/f , Tyr-CreERT2after induction by Tamoxifen treatment ( Dhomen et al . , 2009 ) . 2x105 melanoma cells were injected intra-dermally into the lateral flanks of 6–8 week-old CD1 athymic mice; two days after injection the mice were randomly divided into two groups ( three mice in each group ) , one group was treated with vehicle only and the other with the drug . AT13148 was administered at 40 mg/kg in 10% DMSO , 1% Tween-20 in saline , p . o . , the dosing schedule used was 2/5 days . At the end of the experiment , the tumors were removed , snap-frozen and senescence-associated beta-galactosidase activity assessed on frozen sections using the protocol described above . SILAC heavy and light labelled MEFs were treated for indicated times with H1152 , or vehicle , in duplicate ( heavy vs . light as well as light vs . heavy ) . Cells were then lysed by 2% SDS , Tris-HCl pH 7 . 5 , and protein concentrations were measured and balanced by BCA assay . The reciprocal labels of drug and vehicle treated lysates were then mixed 1:1 , resulting in two reciprocally labelled SILAC replicates ( Heavy drug + light vehicle and vice versa ) . The mixed lysates were then reduced by addition of 100 mM DTT and boiling for 5 min , followed by trypsin digestion using a Filter Aided Sample Preparation ( FASP ) protocol ( Wisniewski et al . , 2009 ) . For phospho-peptide analysis , the digests were subjected to phospho-peptide enrichment on TiO2 tips ( GL sciences ) . For total proteome analysis , the digests were initially fractionated by isoelectric focusing , using an OFF-gel fractionator ( Agilent , pH 3–10 ) . The fractions were then cleaned up using C18 ziptip columns ( Millipore ) . All samples were analyzed by LC-MS/MS using a Thermo Orbitrap-Velos mass spectrometer ( ICR proteomics core facility ) . The raw data wereas then searched and quantified using Maxquant ( Cox and Mann , 2008 ) . All subsequent statistical analysis was performed using Perseus software from the Maxquant package . Briefly , the ratios were converted to Log-2 scale , and reverse and contaminants were filtered out . Proteins only identified in one replicate were also filtered out . The ‘Significant-B’ outlier test was used to determine the significantly regulated peptides or proteins , using a Benjamini-Hochberg FDR rate of 5% . A hit was considered significant if it was found to be significantly changing in opposite directions in both reciprocal SILAC replicates . 2-Dimensional Annotation enrichment analysis was also performed in Perseus , using Broad Institute’s Gene Set Enrichment Analysis ( GSEA ) annotations , using a Benjamini-Hochberg FDR rate of 2% . All proteomics data graphs were generated in Perseus . The targeted proteomics dataset has been deposited to PRIDE: Figure 4A accession PXD002521 , Figure 2—figure supplement 1B–C accession PXD002515 . RNA was isolated using the RNeasy Mini Kit ( Qiagen ) , according to manufaturer’s instructions , and QuantiTect Primer Assays , purchased from Qiagen , were used . The Brilliant II SYBR Green QRT-PCR 1-Step system was used according to the manufacturer’s instructions ( Agilent Technologies ) . The reactions were analysed on an ABI PRISM 7900HT Sequence Detection System ( Applied Biosystems ) . A previously reported approach for absolute quantitation of proteins ( AQUA ) based on standard stable isotope labelled peptides was used ( Gerber et al . , 2003 ) . ‘Heavy’ isotopic peptides were purchased from Thermo Scientific ( USA ) . SRM was performed , as described elsewhere ( Worboys et al . , 2014 ) , for the measurement of endogenous and synthetic peptides . To select optimal proteotypic peptides to monitor by SRM , all possible proteotypic peptides of ROCK1 and 2 were empirically measured within cell lines that were wt , Rock1-/- and Rock2-/- , permitting a confirmation of peptide specificity . For each SRM analysis bands from an SDS-PAGE gel were excised between 150–250 kDa ( Precision Plus Protein Prestained Standard , Bio-Rad ) , and proteins trypsin-digested in-gel . NIDNFLSR , LLLQNELK and ESDIEQLR were chosen for ROCK1 and VYYDISSAK , QLDEANALLR and ENLLLSDSPPCR for ROCK2 . For quantitation , the two most intense transitions from each peptide were summed for each isotopic variant . A cycle time of 1 . 8 s was used , and transitions monitored in an unscheduled manner giving a dwell time of 37 . 5 ms . Standard curves were generated within sample matrices , and plotted as a function of their SILAC ratios . Subsequent regression lines were used per peptide to calculate absolute concentrations . In tumor samples , 5 fmol of each peptide was spiked into each sample . Three technical replicates were performed and averaged per peptide and each average peptide measurement was averaged per protein . Thus , for each protein a total of 9 measurements are made . The targeted proteomics dataset has been deposited to the PeptideAtlas SRM Experiment Library ( PASSEL ) with the identifier PASS00721 . Statistical tests were conducted with GraphPad Prism version 6 or Microsoft Excel 2010 . Unless stated otherwise , p-values were generated using two-tailed , paired Student’s t-test and considered significant if p<0 . 05 . p<0 . 05; p<0 . 01 * p<0 . 001 . Error bars represent standard deviation ( SD ) or standard error of the mean ( SEM ) as indicated in figure legends .	Animal cells contain a structure called the cytoskeleton , which helps give the cells their shape . This structure can rapidly disassemble and reassemble , which enables cells to change their shape , move and divide into two . Many proteins are involved in controlling these processes . In particular , two proteins called ROCK1 and ROCK2 are known to be important for helping cancer cells move . However , investigations into the exact roles of these proteins have so far produced contradictory results . Kümper et al . have now developed a more refined method of studying what ROCK1 and ROCK2 do in cells . This involves genetically engineering mice in a way that makes it possible to control whether ROCK1 and ROCK2 are produced in specific cell types and tissues . Studying cells that had been taken from these mice revealed that cells that lacked both proteins could not contract . Moreover , these cells became bigger and flattened out . This change in appearance went hand in hand with the cells becoming unable to divide and form new cells . However , cells that lacked just one type of ROCK protein were still able to divide and proliferate . As tumors form as a result of cells dividing and proliferating uncontrollably , Kümper et al . then studied how the ROCK proteins affect tumor development in mice that are susceptible to lung or skin cancer . Although cancerous cells were found that contained just one type of ROCK protein , no tumor cells were found that lacked both ROCK1 and ROCK2 , further confirming that having one ROCK protein is essential for tumor formation . Overall , it appears that within the systems studied ROCK1 and ROCK2 perform the same roles , and that ROCK proteins are indispensable for cell proliferation and hence tumor development . The next challenge will be to identify the tumor types that are highly dependent on processes driven by the ROCK proteins . Further work could then investigate whether drugs that inhibit the activity of the ROCK proteins block the growth and spread of these tumors .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods", "Statistical", "analysis" ]	[ "cell", "biology", "cancer", "biology" ]	2016	Rho-associated kinase (ROCK) function is essential for cell cycle progression, senescence and tumorigenesis
Pancreatic adenosquamous carcinoma ( PASC ) is an aggressive cancer whose mutational origins are poorly understood . An early study reported high-frequency somatic mutations affecting UPF1 , a nonsense-mediated mRNA decay ( NMD ) factor , in PASC , but subsequent studies did not observe these lesions . The corresponding controversy about whether UPF1 mutations are important contributors to PASC has been exacerbated by a paucity of functional studies . Here , we modeled two UPF1 mutations in human and mouse cells to find no significant effects on pancreatic cancer growth , acquisition of adenosquamous features , UPF1 splicing , UPF1 protein , or NMD efficiency . We subsequently discovered that 45% of UPF1 mutations reportedly present in PASCs are identical to standing genetic variants in the human population , suggesting that they may be non-pathogenic inherited variants rather than pathogenic mutations . Our data suggest that UPF1 is not a common functional driver of PASC and motivate further attempts to understand the genetic origins of these malignancies . Pancreatic adenosquamous carcinoma ( PASC ) is a rare and aggressive disease that constitutes 1–4% of pancreatic exocrine tumors ( Madura et al . , 1999 ) . Patient prognosis is extremely poor , with a median survival of 8 months ( Simone et al . , 2013 ) . Although PASC is clinically and histologically distinct from the more common disease pancreatic adenocarcinoma , the genetic and molecular origins of PASC’s unique features are unknown . A recent study reported a potential breakthrough in our understanding of PASC etiology . Liu et al . , 2014 reported high-frequency mutations affecting UPF1 , which encodes a core component of the nonsense-mediated mRNA decay ( NMD ) pathway , in 78% ( 18 of 23 ) of PASC patients . These mutations were absent from patient-matched normal pancreatic tissue ( 0 of 18 ) and from non-PASC tumors ( 0 of 29 non-adenosquamous pancreatic carcinomas and 0 of 21 lung squamous cell carcinomas ) . The authors used a combination of molecular and histological assays to find that the UPF1 mutations caused UPF1 mis-splicing , loss of UPF1 protein , and impaired NMD , resulting in stable expression of aberrant mRNAs containing premature termination codons that would normally be degraded by NMD . The recurrent , PASC-specific , and focal nature of the reported UPF1 mutations , together with their dramatic effects on NMD activity , suggested that UPF1 mutations are a key feature of PASC biology . Three subsequent studies of distinct PASC cohorts , however , did not report somatic mutations in UPF1 ( Fang et al . , 2017; Hayashi et al . , 2020; Witkiewicz et al . , 2015 ) . This absence of UPF1 mutations is significantly different from the high rate reported by Liu et al . ( 0 of 34 total PASC samples from three cohorts [Fang et al . , 2017; Hayashi et al . , 2020; Witkiewicz et al . , 2015] vs . 18 of 23 PASC samples from Liu et al; p<10−8 by the two-sided binomial proportion test ) . Although these other studies relied on whole-exome and/or genome sequencing instead of targeted UPF1 gene sequencing , those technologies yield good coverage of the relevant UPF1 gene regions because the affected introns are very short . Given this discrepancy , we sought to directly assess the functional contribution of UPF1 mutations to PASC using a combination of biological and molecular assays . We first tested the role of the reported UPF1 mutations during tumorigenesis in vivo . Liu et al . reported that the majority of UPF1 mutations caused skipping of UPF1 exons 10 and 11 , disrupting UPF1’s RNA helicase domain that is essential for its NMD activity ( Lee et al . , 2015 ) . We therefore modeled UPF1 mutation-induced exon skipping by designing paired guide RNAs flanking Upf1 exons 10 and 11 , such that these exons would be deleted upon Cas9 expression ( Figure 1A ) . We chose mouse pancreatic cancer cells ( KPC cells: KrasG12D; Trp53R172H/null; Pdx1-Cre ) as a model system . KPC cells are defined by mutations affecting KRAS and p53 ( encoded by Kras and Trp53 in mouse ) that also occur in the vast majority of PASC cases ( Borazanci et al . , 2015; Fang et al . , 2017; Hayashi et al . , 2020 ) , making them a genetically appropriate system . We delivered Upf1-targeting paired guide RNAs to KPC cells using recombinant adenoviral vectors and confirmed that guide delivery resulted in the production of UPF1 mRNA lacking exons 10 and 11 and a corresponding reduction in full-length UPF1 protein levels ( Figure 1—figure supplement 1A–G ) . We injected subcloned control and Upf1-targeted KPC cells into the tails of the pancreata of B6 albino mice ( n = 10 mice per treatment ) and monitored tumor growth and animal survival . We detected no significant differences in tumor volume or survival in mice implanted with control or Upf1-targeted KPC cells ( Figure 1B–D and Figure 1—source datas 1 and 2 ) . Tumors derived from control as well as Upf1-targeted cells displayed similar histopathological features characteristic of moderately to poorly differentiated pancreatic ductal adenocarcinomas ( Figure 1E–G ) . Moderately differentiated areas were composed of medium to small duct-like structures or tubules with lower mucin production , while poorly differentiated components were characterized by solid sheets or nests of tumor cells with large eosinophilic cytoplasms and large pleomorphic nuclei . No squamous differentiation was identified by histomorphologic evaluation and no expression of the squamous marker p40 ( ΔNp63 ) was detected ( Figure 1H and Supplementary file 1a ) . We concluded that inducing the reported Upf1 exon skipping in vivo had no detectable effects on pancreatic cancer growth or acquisition of adenosquamous features in the KPC model . However , there are several important caveats to our data . First , we cannot rule out the possibility that inducing Upf1 exon skipping in a different model system or cell type could influence tumorigenesis . Second , as our assays were performed in the complex setting of in vivo tumorigenesis , we cannot infer how inducing the reported Upf1 exon skipping might affect tumor cell proliferation in the controlled setting of in vitro growth . Third , as accurate measurement of Upf1 spicing and UPF1 protein isoforms was only possible for KPC cells prior to orthotopic injection , we cannot infer how the relative frequencies of mis-spliced UPF1 mRNA and the resulting truncated proteins may have changed during tumorigenesis . We next assessed molecular phenotypes induced with UPF1 mutations . Liu et al . measured the effects of each mutation on UPF1 splicing using a minigene assay , in which each mutation was introduced into a plasmid containing a small fragment of the UPF1 gene that was subsequently transfected into 293 T cells . Liu et al . concluded that all reported UPF1 mutations caused dramatic UPF1 mis-splicing that disrupted key protein domains that are essential for UPF1 function in NMD . Minigenes are common tools for studying splicing , but they are frequently spliced less efficiently than endogenous genes , presumably because they are gene fragments that lack potentially important sequence features that promote splicing and incompletely capture the close relationship between chromatin and splicing ( Luco et al . , 2011; Naftelberg et al . , 2015 ) . We modeled UPF1 mutations in 293 T cells in order to mimic Liu et al . ’s experimental strategy , but introduced mutations into their endogenous genomic contexts rather than using minigenes . We selected two distinct UPF1 mutations in intron 10 for these studies . We selected IVS10+31G>A ( patient 1; P1 ) because it was reportedly recurrent across three different patients ( making it equally or more common than any other mutation ) and induced strong mis-splicing on its own ( 36% mis-spliced mRNA , versus 0% for wild-type UPF1 ) ; we selected IVS10-17G>A ( patient 9; P9 ) because it had one of the strongest effects on splicing ( 90% mis-spliced mRNA ) . IVS10+31G>A was present in a homozygous state in two of the three patients carrying it , while IVS10-17G>A was present in a heterozygous state . We introduced each mutation into its endogenous context by transiently transfecting a plasmid expressing Cas9 and a single guide RNA ( sgRNA ) targeting UPF1 intron 10 as well as appropriate donor DNA for homology-directed repair , screened the resulting cells for the desired genotypes , and established clonal lines . The resulting cell lines contained the desired mutations in the correct copy numbers as well as a point mutation disrupting the protospacer adjacent motif ( PAM ) site ( Figure 2—figure supplement 1A–C ) . As neither the PAM site itself nor nearby positions were reported as mutated in Liu et al . , we additionally established a cell line in which only the PAM site was mutated as a wild-type control . We systematically tested the functional consequences of UPF1 mutations for NMD efficiency , UPF1 protein levels , and UPF1 splicing . We measured NMD efficiency in our engineered cells using the well-established beta-globin reporter system , which permits controlled measurement of the relative levels of mRNAs that do or do not contain an NMD-inducing premature termination codon , but which are otherwise identical ( Zhang et al . , 1998 ) . We did not observe decreased NMD efficiency in UPF1-mutant versus wild-type cells; instead , UPF1-mutant cells exhibited evidence of modestly more efficient NMD , although these differences were not statistically significant ( Figure 2A and Figure 2—source data 1 ) . To confirm these results from reporter experiments , we then queried levels of endogenous NMD substrates across the transcriptome . We performed high-coverage RNA-seq on each of the three 293 T cell lines that we engineered to lack or contain defined UPF1 mutations in biological triplicate , quantified transcript expression , and identified differentially expressed transcripts . We focused on NMD substrates arising from alternative splicing , as these are abundant and sensitive biomarkers of NMD efficiency that are internally controlled for gene expression variation ( Feng et al . , 2015 ) . These analyses revealed that neither UPF1-mutant cell line exhibited global increases in the expression of NMD substrates relative to wild-type cells . Instead , both UPF1-mutant cell lines exhibited modestly lower global levels of endogenous NMD substrates than did wild-type cells , mimicking the trend observed with our NMD reporter experiments . Together , these data confirm that the tested mutations in UPF1 intron 10 do not affect NMD activity ( Figure 2B–D and Supplementary file 1b and c ) . Consistent with similar NMD activity independent of UPF1 mutational status , UPF1 mutations did not cause loss of full-length UPF1 protein ( Figure 2E , Figure 2—figure supplement 1D , and Figure 2—source data 2 ) . Although UPF1 protein levels varied between the individual cell lines , this variation in UPF1 protein levels was not associated with variation in NMD efficiency and did not segregate with UPF1 mutational status . We therefore measured the levels of normally spliced and mis-spliced UPF1 mRNA . We readily detected normally spliced UPF1 mRNA in all samples by RT-PCR , but found no evidence of mis-spliced UPF1 mRNA , except in positive control samples in which we spiked in synthesized DNA corresponding to the exon skipping isoform reported in Liu et al . ( Figure 2F , Figure 2—figure supplement 1E , and Figure 2—source data 3 ) . We confirmed these results with our RNA-seq data by mapping all reads against all possible splice junctions connecting exons 9 , 10 , 11 , and 12 . These analyses revealed no evidence of splice junctions consistent with the reported exon 10 and 11 skipping or other abnormal exon skipping isoforms ( Figure 2G ) . Given the differences between Liu et al . ’s findings of common UPF1 mutations and their absence from subsequent studies of PASC , we wondered whether some of the UPF1 mutations reported by Liu et al . might correspond to inherited genetic variation rather than somatically acquired mutations . We searched for each mutation reported by Liu et al . within databases compiled by the 1000 Genomes Project , NHLBI Exome Sequencing Project , Exome Aggregation Consortium ( ExAC ) , and the genome aggregation database ( gnomAD ) ( Auton et al . , 2015; Karczewski et al . , 2020; Exome Aggregation Consortium et al . , 2016; Server EV , 2016 ) . These databases were constructed from a mix of whole-genome and whole-exome sequencing , both of which are effective for discovering variants within the relevant regions of UPF1 ( because UPF1 introns 10 , 21 , and 22 are very short , they are well covered by exon-capture technologies ) . We found genetic variants identical to 45% ( 18 of 40 ) of the reported UPF1 mutations , one of which is present in the reference human genome . Eighty-nine percent ( 16 of 18 ) of UPF1-mutant patients had one or more reported mutations that corresponded to standing genetic variation ( Figure 3A–F and Supplementary file 1d ) . The distribution of overlaps between reported UPF1 mutations and standing genetic variation depended strongly upon genic context . A large fraction of reported intronic UPF1 mutations were identical to standing genetic variation , while the majority of reported exonic UPF1 mutations were not ( Supplementary file 1d ) . Our discovery that a large fraction of the reported UPF1 mutations are present in databases of germline genetic variation was surprising for two reasons . First , when strongly cancer-linked mutations occur as germline variants , they frequently manifest as cancer predisposition syndromes . However , no such relationship is known for UPF1 genetic variants , despite their reportedly high prevalence as identical somatic mutations in PASC . Second , UPF1 is essential for embryonic viability and development in mammals ( Medghalchi et al . , 2001 ) , zebrafish ( Wittkopp et al . , 2009 ) , and Drosophila ( Avery et al . , 2011 ) . As Liu et al . reported that all UPF1 mutations caused mis-splicing that is expected to disable UPF1 protein function ( Liu et al . , 2014 ) , then those mutations should be incompatible with life when present as inherited genetic variants . Our finding that two reported mutations had no effect on UPF1 splicing when introduced into their endogenous genomic contexts offers a way to explain this incongruity , at least for the two reported lesions that we studied . Given these discrepancies , we next sought to verify the somatic nature of the UPF1 mutations described in Liu et al . , which was reportedly determined by sequencing both tumors and patient-matched controls . The GenBank accession codes reported in Liu et al . corresponded to short nucleotide sequences containing UPF1 mutations , without corresponding data for patient-matched controls . We contacted the senior author ( Dr . YanJun Lu ) to request primary sequencing data from patient-matched tumor and normal samples , but neither primary sequencing data from matched samples nor the samples themselves were available . To further explore whether UPF1 is recurrently mutated in PASC , we reanalyzed sequencing data from Fang et al . , 2017 to manually search for UPF1 mutations ( Supplementary file 1e-f ) . We focused on the two loci that contained all mutations reported by Liu et al . ( UPF1 exons 10-11 and exons 21–23 ) . Because the relevant introns are very short , they were well covered by both the whole-exome and whole-genome sequencing used by Fang et al . Using relaxed mutation-calling criteria to maximize sensitivity ( details in Materials and methods ) , we identified somatic UPF1 mutations in samples from 6 of 17 PASC patients . However , those mutations exhibited genetic characteristics expected of passenger , not driver , mutations . None of those UPF1 mutations matched the UPF1 mutations reported by Liu et al . , and only one was present at an allelic frequency equal to the allelic frequency of mutant KRAS , which is a known driver and which we detected in samples from all PASC patients ( median allelic frequencies of 12% versus 34% for UPF1 versus KRAS mutations ) . Furthermore , we also identified UPF1 mutations in samples from patients with non-adenosquamous tumors ( 3 of 34 pancreatic ductal adenocarcinomas ) , whereas Liu et al . reported finding no UPF1 mutations in non-adenosquamous pancreatic cancers ( 0 of 29 ) . In concert with the reports of Witkiewicz et al . , 2015 and Hayashi et al . , 2020 of finding no UPF1 mutations in their PASC samples , these analyses suggest that UPF1 is not a frequent or adenosquamous-specific mutational target in most PASC cohorts . UPF1’s role in the pathogenesis of PASC has been unclear and controversial given the seeming discrepancies between its mutational spectrum in different PASC cohorts . Although it is difficult to conclusively prove that a specific genetic change does not promote cancer , we were unable to detect biological or molecular changes arising from two mutations reported by Liu et al . UPF1’s status as an essential gene and our discovery that many reported UPF1 mutations occur as germline genetic variants of no known pathogenicity together suggest that other UPF1 mutations reported by Liu et al . could similarly represent genetic differences that do not functionally contribute to PASC . Our study highlights the need for continued study of the PASC mutational spectrum in order to understand the molecular basis of this disease . Mouse KPC cells ( KrasG12D; Trp53R172H/null; Pdx1-Cre ) were obtained from Dr . Robert Vonderheide and were cultured in DMEM ( GIBCO ) supplemented with 10% fetal bovine serum ( FBS ) and 1% Penicillin/Streptomycin ( GIBCO ) . All cell lines were incubated at 37°C and 5% CO2 . Guide RNAs targeting mouse Upf1 introns 9 and 11 were cloned into a paired guide expression vector ( px333 ) as previously described ( Maddalo et al . , 2014 ) . An EcoRI-XhoI fragment containing the double U6-sgRNA cassette and Flag-tagged Cas9 was then ligated into the EcoRI-XhoI-digested pacAd5 shuttle vector . Recombinant adenoviruses were generated by Viraquest ( Ad-Upf1 and Ad-Cas9 ) or purchased from the University of Iowa ( Ad-Cre ) . KPC cells were infected with ( 5 × 106 PFU ) of Ad-Cas9 or Ad-Upf1 in each well of a 6-well plate . Genomic DNA was extracted 48 hr post infection to confirm excision of Upf1 exons 10 and 11 . For PCR analysis of genomic DNA , cells were collected in lysis buffer ( 100 nM Tris-HCl at pH 8 . 5 , 5 mM EDTA , 0 . 2% SDS , 200 mM NaCl supplemented with fresh proteinase K at a final concentration of 100 ng/mL ) . Genomic DNA was extracted with phenol–chloroform–isoamylic alcohol and precipitated in ethanol , and the DNA pellet was dried and resuspended in double-distilled water . For RT-PCR , total RNA was extracted with TRIzol ( Life Technologies ) following the manufacturer’s instructions . cDNA was synthesized using SuperScript III ( ThermoFisher ) following the manufacturer’s instructions . Cells were lysed in 1X RIPA buffer with protease and phosphatase inhibitors . Fifteen micrograms of protein was separated on 4–10% acrylamide/bisacrylamide gels , transferred onto PVDF membranes , and blocked for 1 hr in 5% milk in 1× TBST . The membranes were incubated with rabbit UPF1 antibody ( CST #9435 ) used at 1:1000 dilution in 5% BSA 1× TBST overnight at 4°C or mouse Tubulin ( Sigma T9206 ) used at 1:2000 dilution in 5% milk 1× TBST for 1 hr at room temperature . Following primary antibody incubation , the membranes were washed three times with 1× TBST buffer at room temperature and probed with rabbit or mouse horseradish peroxidase-linked secondary antibody ( 1:5000; ECL NA931 mouse , NA934V Rabbit ) . The western blot signal was detected using the ECL Prime ( RPN322 ) kit and the blot was exposed to an X-Ray film , which was developed using the Konica Minolta SRX 101A film processor . KPC cells carrying Upf1 ΔE10-11 or an empty Cas9 control were mixed in 1:1 Matrigel ( BD Biosciences ) and simple media to a final concentration of 100 , 000 cells in 30 μL of total volume . Cells were orthotopically implanted into the tails of the pancreata of B6 albino mice ( Charles River ) . Ten mice were implanted with each stable , genetically engineered cell line . Tumor growth was measured weekly via 3D-ultrasound starting at 10 days post-implantation . For survival assessment , animals were sacrificed following the endpoints approved by IACUC: ( i ) animals showing signs of significant discomfort , ( ii ) ascites or overt signs of tumor metastasis or gastrointestinal bleeding ( blood in stool ) , ( iii ) animals losing >15% of their body weight , and ( iv ) animals with tumors >2 cm in diameter . Investigators responsible for monitoring and measuring the xenografts of individual tumors were not blinded . All animal studies were performed in accordance with institutional and national animal regulations . Animal protocols were approved by the Memorial Sloan-Kettering Cancer Center Institutional Animal Care and Use Committee ( 14-08-009 and 11-12-029 ) . Power analysis was used to determine appropriate sample size to detect significant changes in animal median survival , which was based on previous survival analyses ( Escobar-Hoyos et al . , 2020 ) . Survival curves and statistics were performed using PRISM . Paraffin sections were dewaxed in xylene and hydrated in graded alcohols . Endogenous peroxidase activity was blocked by immersing the slides in 1% hydrogen peroxide in PBS for 15 min . Pretreatment was performed in a steamer using 10 mM citrate buffer ( pH 6 . 0 ) for 30 min . Sections were incubated overnight with a primary rabbit polyclonal antibody against p40-DeltaNp63 ( Abcam , ab166857 ) diluted at a ratio of 1:100 . Sections were washed with PBS and incubated with an appropriate secondary antibody followed by avidin–biotin complexes ( Vector Laboratories , Burlingame , CA , PK-6100 ) . The antibody reaction was visualized with 3–3' diaminobenzidine ( Sigma , D8001 ) followed by counterstaining with hematoxylin . Tissue sections were dehydrated in graded alcohols , cleared in xylene , and mounted . For p40 ( ΔNp63 ) IHC , expression was defined based on nuclear labeling . HEK 293 T cells were cultured in DMEM media ( GIBCO ) supplemented with 10% FBS ( GIBCO ) , 100 IU penicillin , and 100 mg/mL streptomycin ( PenStrep , GIBCO ) . Cells were cultured at 37°C and 5% CO2 . Cells were split at a ratio of 1:10 once they reached 90–100% confluency as needed . Cell lines were authenticated using ATCC fingerprinting and tested regularly for mycoplasma contamination . Guide RNAs for all cell lines were designed using the GuideScan 1 . 0 software package ( Perez et al . , 2017 ) and sequences were chosen among those predicted to have the highest cutting efficiency and specificity scores . DNA oligos for all guide RNAs were synthesized by IDT and amplified using primers that appended homology arms to facilitate ligation into the pX459/Cas9 expression plasmid ( Ran et al . , 2013 ) by Gibson assembly ( Gibson et al . , 2009 ) . Gibson assembly reactions were transformed into NEB Stable Competent E . coli and resulting sgRNA expression plasmids were amplified and purified using standard protocols . Ultramers for homology directed repair ( HDR ) were designed using previously described strategies ( Richardson et al . , 2016 ) and synthesized by IDT , Inc . HEK 293 T cells were transiently transfected with 1 μg/mL sgRNA/pX459 Cas9 expression plasmid and 20 nM HDR Ultramer using Lipofectamine 2000 that was diluted in Opti-MEM Reduced Serum Medium ( ThermoFisher ) . Cells were incubated at 37°C for 24 hr , at which time the transfection medium was replaced with DMEM media ( GIBCO ) supplemented with 10% FBS ( GIBCO ) , 100 IU penicillin , 100 mg/mL streptomycin ( PenStrep , GIBCO ) , and 2 mg/mL puromycin . Cells were then incubated for 48–72 hr in DMEM media ( GIBCO ) supplemented with 10% FBS ( GIBCO ) , 100 IU penicillin , 100 mg/mL streptomycin ( PenStrep , GIBCO ) , and 2 mg/mL puromycin , at which time genomic DNA was extracted and regions of interest were amplified using the appropriate oligos . Genome engineering was validated using genomic DNA as follows . Amplicons from genomic DNA PCR were ligated into vectors using the Zero Blunt TOPO PCR cloning system ( ThermoFisher ) and the presence of the desired mutations was validated using Sanger sequencing ( GENEWIZ ) . Polyclonal cell populations were then diluted and sorted into 96-well plates using a BD FACS Aria II flow cytometer ( BD Biosciences ) , such that each well contained on average one cell , which were grown in DMEM media ( GIBCO ) supplemented with 20% FBS ( GIBCO ) , 100 IU penicillin , and 100 mg/mL streptomycin ( PenStrep , GIBCO ) . Once cells in 96-well plates reached confluency , they were transferred to 24-well plates and allowed to grow to confluency for genomic DNA extraction and Sanger sequencing . NMD efficiency was estimated using the beta-globin reporter system ( Zhang et al . , 1998 ) . HEK 293 T cells engineered with the reported mutations were plated at 10–15% confluency on pol-L-lysine coated 12-well plates . Cells were co-transfected 24 hr later with 1 µg of phCMV-MUP plasmid ( transfection control ) and 1 µg of either pmCMV-Gl-Norm ( normal termination codon ) or pmCMV-Gl-39Ter ( premature termination codon ) using Lipofectamine 3000 ( Invitrogen ) according to the manufacturer’s protocol . After 48 hr the cells were close to confluency , at which time they were lysed using 1 mL of Trizol Reagent ( Invitrogen ) per well . The lysate was collected , and total RNA was extracted according to the manufacturer’s protocol . The RNA was further purified and DNase treated using the Direct-zol RNA MiniPrep Kit ( Zymo Research ) according to the manufacturer’s protocol . Residual plasmid DNA was removed using DNaseI ( Amplification Grade , Invitrogen ) according to the manufacturer’s protocol from 600 ng of the extracted RNA . cDNA synthesis was then performed using SuperScript IV Reverse Transcriptase ( Invitrogen ) with oligo dT primers according to the manufacturer's protocol . The cDNA synthesis reaction was diluted 1:50 and 4 µL was used for a 10 µL qPCR reaction with PowerUp SYBR Green Master Mix ( ThermoFisher ) and primers specific for the reporter mRNA diluted to a working concentration of 100 nM for each primer . qPCR reactions were performed in technical triplicate for three different biological replicates in 384-well plates ( ThermoFisher ) using an ABI QuantStudio 5 Real-Time PCR System ( ThermoFisher ) . The levels of pmCMV-Gl-Norm and pmCMV-Gl-39Ter cDNA were normalized to phCMV-MUP mRNA abundance for each sample , and levels of pmCMV-Gl-39Ter cDNA were plotted relative to levels of pmCMV-Gl-Norm for each replicate in each cell type . Statistical analysis was performed using PRISM . Cells were lysed using a buffer containing 150 mM NaCl , 1% NP-40 , and 50 mM Tris pH 8 . 0 supplemented with a phosphatase inhibitor ( ThermoFisher ) and protease inhibitor ( ThermoFisher ) . After the lysis buffer was added , cells were frozen at −80°C and thawed for three cycles , and then incubated on ice for 15 min . The lysed cells were centrifuged at 10 , 000 × g for 15 min , and the supernatant was collected to determine total protein concentration . Total protein concentration was determined using a Qubit ( ThermoFisher ) and 20 μg of total protein was used for electrophoresis . Following electrophoresis , protein was transferred to nitrocellulose membrane ( Novex ) in transfer buffer containing 10% methanol overnight at 4°C . The blot was blocked with Odyssey Blocking Buffer ( LI-COR Biosciences ) for 1 hr at room temperature and probed using a 1:1000 dilution of 0 . 514 mg/mL UPF1 antibody ( Abcam , 109363 ) overnight with shaking at 4°C . Following overnight incubation , the blot was washed three times with 1× TBST buffer at room temperature and probed with rabbit secondary antibody ( IRDye 680RD goat anti-rabbit ) for 1 hr at room temperature . The blot was then imaged and UPF1 abundance was quantified using band intensity in Fiji ( v2 . 0 . 0 ) . Statistical analysis was performed using PRISM . Total RNA was extracted using the RNeasy Plus Mini Kit ( Qiagen ) . cDNA was synthesized using oligo dT primers and SuperScript IV Reverse Transcriptase ( ThermoFisher ) following the manufacturer's protocol . PCR was carried out with primers targeting UPF1 exons 9 and 12 using Q5 High-Fidelity DNA Polymerase ( NEB ) . PCR products were run in a 2% agarose slab gel and stained with ethidium bromide for visualization by UV shadowing ( Bio-Rad Molecular Imager Gel Doc XR+ ) . A positive control for the amplification of the truncated UPF1 variant missing exons 10 and 11 was synthesized as a double-stranded DNA gBlock ( IDT ) . Two different amounts ( 10 fg and 1 fg ) of this ‘spike in’ control was added to separate PCRs containing cDNA from wild-type HEK 293 T cells and amplified in the same manner as described above . To quantify the degree of exon skipping ( percent ΔE10-11 ) , a background subtraction across the entire gel was first performed in Fiji using a rolling ball radius of 100 pixels . Next , the integrated density of each band was determined , and the density of the lower band in each lane was divided by the total density in that same lane by summing the integrated densities of the upper and lower bands . Whole-genome and whole-exome sequencing data from Fang et al . , 2017 were downloaded from the Sequence Read Archive ( accession number SRP107982 ) and mapped to the UPF1 ( chr19:18940305–18979266; hg19/GRCh37 assembly ) and KRAS ( chr12:25356390–25405419; hg19/GRCh37 assembly ) gene loci . The first 10 nt of all reads were trimmed off due to low sequencing quality and the trimmed reads were mapped with Bowtie v1 . 0 . 0 ( Langmead et al . , 2009 ) with the arguments '-v 3 k 1 m 1 --best --strata --minins 0 --maxins 1000 --fr' . Mapped reads were visualized in IGV ( Robinson et al . , 2011 ) . Mutation/genetic variant calling thresholds ( read coverage depth ≥9 reads and ≥2 reads supporting the mutation/variant ) were chosen in order to allow detection of a hotspot KRAS mutation ( G12 or G13 ) in every PASC sample in the cohort . That criteria ensured that our thresholds were appropriate for discovering known cancer driver mutations in all samples . A genetic difference from the reference genome was defined as a somatic mutation if it was called in a tumor sample but not in the corresponding patient-matched normal control sample . Genome and transcriptome annotations for mapping RNA-seq data to the human ( NCBI GRCh37/UCSC hg19 ) genome were generated as described previously ( Dvinge et al . , 2014 ) . Briefly , transcriptome annotations from Ensembl ( Flicek et al . , 2013 ) were merged with isoform annotations from the MISO v2 . 0 database ( Katz et al . , 2010 ) and the UCSC knownGene track ( Meyer et al . , 2013 ) . NMD substrates were defined as those isoforms containing a premature termination codon >50 nt upstream of the last exon–exon junction . RNA-seq reads were mapped to the transcriptome with RSEM v1 . 2 . 4 ( Li and Dewey , 2011 ) and Bowtie v1 . 0 . 0 ( Langmead et al . , 2009 ) , where RSEM was modified to invoke Bowtie with the ‘-v2’ option . Reads that are unaligned after this transcriptome mapping were then aligned to the genome with TopHat v2 . 0 . 8 ( Trapnell et al . , 2009 ) , as well as mapped to a database of splice junctions that was defined by creating all possible co-linear combinations of 5' and 3' splice sites within every gene . A final file of aligned reads was created by merging the read alignments from TopHat with the read alignments from RSEM . MISO v2 . 0 ( Katz et al . , 2010 ) was used to quantify isoform expression for alternative splicing events . Differentially spliced events were defined as those that met the following criteria: ( 1 ) had at least 20 informative reads ( reads that uniquely distinguish between isoforms of a given splicing event ) , ( 2 ) exhibited either an absolute change in isoform ratio ≥10% or an absolute fold-change ≥2 , and ( 3 ) had an associated p≤0 . 05 ( computed using a two-sided t-test ) . Differential splicing analyses were restricted to splicing events arising from U2-type ( major ) introns , which constitute >99% of all introns , in order to ensure that no potential confounding effects arose from intron type . Splicing events were defined as NMD relevant if at least one , but not all , of the child isoforms was predicted NMD substrates .	Cancer is a group of complex diseases in which cells grow uncontrollably and spread into surrounding tissues and other parts of the body . All types of cancers develop from changes – or mutations – in the genes that affect the pathways involved in controlling the growth of cells . Different cancers possess unique sets of mutations that affect specific genes , and often , it is difficult to determine which of them play the most important role in a particular type of cancer . For example , pancreatic adenosquamous carcinoma , a rare and aggressive form of pancreatic cancer , is a devastating disease with a poor chance of survival – patients rarely live longer than one year after diagnosis . While the cells of this particular cancer display distinct features that separate them from other forms of pancreatic cancer , the genetic causes of these features are unclear . Using new technologies , some researchers have reported mutations in a ‘quality control’ gene called ‘UPF1’ , which is responsible for destroying faulty forms of genetic material . However , subsequent studies did not find such mutations . To clarify the role of UPF1 in pancreatic adenosquamous carcinoma , Polaski et al . used mouse and human cancer cells with UPF1 mutations and monitored their effects on tumour growth and the development of features unique to this disease . Polaski et al . first injected mice with mouse pancreatic cancer cells containing mutations in UPF1 ( mutated cells ) and cancer cells without . Both groups of mice developed pancreatic tumours but there was no difference in tumour growth between the mutated and non-mutated cells , and neither cell type displayed distinct features . The researchers then generated human mutated cells , which were also found to lack any specific characteristics . Further analysis showed that the mutations did not stop UPF1 from working , in fact , over 40% of these mutations occurred naturally in humans without causing cancer . This suggests that UPF1 does not seem to be involved in pancreatic adenosquamous carcinoma . Further investigation is needed to illuminate key genetic players in the development of this type of cancer , which will be vital for improving treatments and outcomes for patients suffering from this disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "short", "report", "cancer", "biology" ]	2021	The origins and consequences of UPF1 variants in pancreatic adenosquamous carcinoma
Expanded GGGGCC repeats in the first intron of the C9orf72 gene represent the most common cause of familial amyotrophic lateral sclerosis ( ALS ) , but the mechanisms underlying repeat-induced disease remain incompletely resolved . One proposed gain-of-function mechanism is that repeat-containing RNA forms aggregates that sequester RNA binding proteins , leading to altered RNA metabolism in motor neurons . Here , we identify the zinc finger protein Zfp106 as a specific GGGGCC RNA repeat-binding protein , and using affinity purification-mass spectrometry , we show that Zfp106 interacts with multiple other RNA binding proteins , including the ALS-associated factors TDP-43 and FUS . We also show that Zfp106 knockout mice develop severe motor neuron degeneration , which can be suppressed by transgenic restoration of Zfp106 specifically in motor neurons . Finally , we show that Zfp106 potently suppresses neurotoxicity in a Drosophila model of C9orf72 ALS . Thus , these studies identify Zfp106 as an RNA binding protein with important implications for ALS . Expanded GGGGCC repeats in the first intron of the C9orf72 gene represent the most common cause of familial amyotrophic lateral sclerosis ( ALS ) and frontotemporal dementia ( FTD ) ( DeJesus-Hernandez et al . , 2011; Renton et al . , 2011 ) . The repeat expansion has been identified in more than 40% of familial ALS and in at least 7% of sporadic cases ( Majounie et al . , 2012 ) . Several gain-of-function mechanisms for how the expanded hexanucleotide repeats cause disease have been proposed . One idea suggests that toxic dipeptide repeat ( DPR ) proteins generated by repeat associated non-ATG ( RAN ) translation of the RNA repeats cause neurodegeneration ( Kwon et al . , 2014; Mizielinska et al . , 2014; Mori et al . , 2013 ) . Another major proposed gain-of-function mechanism for C9orf72 ALS is via the formation of pathogenic repeat RNA aggregates that sequester one or more RNA binding proteins , leading to altered RNA processing and metabolism in motor neurons ( Cooper-Knock et al . , 2015; Haeusler et al . , 2014; Lee et al . , 2013; Prudencio et al . , 2015; Sareen et al . , 2013 ) . Zfp106 is a C2H2 zinc finger protein with four predicted zinc fingers and seven WD40 domains ( Grasberger and Bell , 2005 ) . The mouse Zfp106 gene is located on chromosome 2 at 60 . 37 cM ( Chr2:120 , 506 , 820–120 , 563 , 843 bp; Genome Reference Consortium , Mouse Build 38 ) . In addition , the human ortholog , ZNF106 , is located at human chromosome 15q15 . 1 ( Chr15: 42 , 412 , 437–42 , 491 , 197; Genome Reference Consortium Human Build 38 ) , a region with strong linkage to a rare recessive familial form of ALS ( Hentati et al . , 1998 ) , suggesting a possible role in human ALS . Consistent with this notion , mice lacking Zfp106 function show evidence of nondevelopmental neuromuscular degeneration , also suggestive of a possible role for Zfp106 in ALS ( Anderson et al . , 2016; Joyce et al . , 2016; van der Weyden et al . , 2011 ) . Here , we identify a previously unknown role for Zfp106 as an RNA binding protein that specifically interacts with GGGGCC repeats and with a network of ALS-associated RNA-binding proteins , including TDP-43 . In addition , we establish that knockout of Zfp106 in mice results in an ALS-like motor phenotype and that transgenic restoration of Zfp106 expression specifically in motor neurons suppresses the neurodegenerative phenotype in knockout mice . Finally , we show that Zfp106 is a potent suppressor of neurotoxicity caused by expression of GGGGCC repeats in a Drosophila model of C9orf72 ALS . Thus , these studies have important implications for the most common form of human ALS . Sequestration of RNA binding proteins by rGGGGCC repeats has been implicated in the pathology of C9orf72 ALS ( Mizielinska and Isaacs , 2014 ) . Therefore , to identify rGGGGCC-binding proteins that may be sequestered by repeat-containing RNA and involved in the pathology of C9orf72 ALS , we performed a pull-down assay using biotinylated r ( GGGGCC ) 8 RNA and identified the zinc finger protein Zfp106 as a previously unknown rGGGGCC-binding protein . Zfp106 bound specifically to r ( GGGGCC ) 8 but not the control sequence r ( AAAACC ) 8 ( Figure 1a ) . RNA electrophoretic mobility shift assay ( EMSA ) using purified Zfp106 protein demonstrated that Zfp106 bound directly to the rGGGGCC repeats and that this binding was specific since it was efficiently competed by unlabeled self oligonucleotide but not a 30× molar excess of the control oligonucleotide ( Figure 1b ) . Importantly , Zfp106 binding to r ( GGGGCC ) 4 was not competed by 30× molar excess of a mutated RNA probe of equal length , nor was it competed by 30× molar excess of rCUG repeats comprising a MBNL1 binding site ( Figure 1—figure supplement 1a ) . Likewise , MBNL1 , an unrelated RNA-binding protein involved in myotonic dystrophy ( Miller et al . , 2000 ) , bound efficiently and specifically to rCUG repeats under conditions in which Zfp106 showed no binding to the same rCUG sequence ( Figure 1—figure supplement 1b ) and , as previously described ( Reddy et al . , 2013 ) , was not able to bind GGGGCC repeats ( Figure 1—figure supplement 1a ) . Together , these data demonstrate that Zfp106 binds efficiently and specifically to GGGGCC RNA repeats . 10 . 7554/eLife . 19032 . 003Figure 1 . Zfp106 is an rGGGGCC binding protein . ( a ) Western blot analysis of Zfp106 eluted from a biotinylated-RNA pulldown of Neuro-2a nuclear proteins . Endogenous Zfp106 was specifically pulled down by ( GGGGCC ) 8 biotinylated RNA and not by the unrelated ( AAAACC ) 8 biotinylated RNA oligonucleotide . ( b ) RNA EMSA performed with purified Zfp106 protein demonstrates that Zfp106 directly binds ( GGGGCC ) 4 RNA in vitro . 30× molar excess of unlabeled self competitor , but not 30× r ( AAAACC ) 4 , competes with Zfp106 binding to r ( GGGGCC ) 4 , establishing specificity of the interaction . ( c ) Immunofluorescence with an anti-Zfp106 antibody detects Zfp106 expression in the nucleolus ( arrowheads ) and in other discrete nuclear puncta ( arrows ) in cultured Neuro-2a cells . The three images are the same section showing the red channel in ( c ) , the blue channel in ( c′ ) , and both channels in ( c′′ ) . Scale bar , 20 μm . ( d , d′ ) Immunohistochemical detection of human ZNF106 shows strong nucleolar expression in human motor neurons in the anterior horn of spinal cord . The two images are from different post-mortem individuals . Arrowheads mark nucleolar ZNF106 expression; arrows mark other nuclear and perinucleolar foci of ZNF106 expression . Scale bar , 25 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 00310 . 7554/eLife . 19032 . 004Figure 1—figure supplement 1 . Zfp106 binds specifically to rGGGGCC repeats . ( a ) RNA EMSA performed with purified Zfp106 protein demonstrates that Zfp106 binds specifically to r ( GGGGCC ) 4in vitro since binding is efficiently competed by 30× molar excess of unlabeled self competitor but not 30× molar excess of r ( AAAACC ) 4 , r ( CUG ) 8 or a mutated r ( GGGGCC ) 4 sequence [mut , r ( GAGGCCGGGACCGAGACCGAGGCC ) ] . Purified MBNL1 protein was used as a negative control for r ( GGGGCC ) 4 binding . MBNL1 does not bind to r ( GGGGCC ) 4 under the same conditions in which Zfp106 shows efficient binding . ( b ) RNA EMSA using purified Zfp106 protein demonstrates that Zfp106 does not bind r ( CUG ) 8 repeats under conditions in which the RNA binding protein MBNL1 efficiently binds r ( CUG ) 8 . MBNL1 binding to r ( CUG ) 8 is specific since binding is competed by 30× molar excess of self competitor but not 30× molar excess of r ( GGGGCC ) 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 004 Patients with C9orf72 ALS show evidence of post-mortem nucleolar stress ( Haeusler et al . , 2014 ) . Interestingly , in cultured Neuro-2a neuronal cells , Zfp106 was primarily present in the nucleolus and in other discrete nuclear puncta ( Figure 1c ) . Likewise , in lower motor neurons from post-mortem human spinal cord samples , ZNF106 was primarily observed in the nucleolus ( Figure 1d , d′ ) and also showed occasional localization in other puncta in the nucleus outside of the nucleolus ( Figure 1d′ ) . The localization of Zfp106 to the nucleolus is in accordance with previous studies showing that Zfp106 is predominantly located in the nucleolus of C2C12 myoblast cells ( Grasberger and Bell , 2005 ) , and taken together , the results presented in Figure 1 support a possible role for Zfp106 in the biology of C9orf72 neurodegeneration . As noted above , previous studies have demonstrated a requirement for Zfp106 in neuromuscular function in mice . Abnormal skeletal morphology in Zfp106tm1a ( KOMP ) Wtsihomozygous mice was briefly reported as part of a large-scale phenotypic screen of genetically modified mice by the Sanger Mouse Genetics Project ( van der Weyden et al . , 2011 ) . More recently , more extensive phenotypic studies demonstrated progressive neuromuscular disease in the absence of Zfp106 gene function ( Anderson et al . , 2016; Joyce et al . , 2016 ) . These studies , combined with the interaction of Zfp106 with the C9orf72 hexanucleotide repeat sequence , prompted us to examine Zfp106 knockout mice in additional detail and to determine whether the observed defects were autonomous to motor neurons . Therefore , we obtained the previously described Zfp106tm1a ( KOMP ) Wtsimice from the KOMP knockout consortium ( van der Weyden et al . , 2011 ) and generated the presumptive null allele ( Zfp106lacz; also referred to herein as Zfp106-null or Zfp106−/− ) using a universal germline deleter transgenic mouse line ( Ehlers et al . , 2014 ) . Consistent with published studies , we found that loss of Zfp106 function caused a progressive neuromuscular phenotype . Zfp106-null mice were born at normal Mendelian frequency and were indistinguishable from control mice until approximately 4 weeks of age , when 100% of knockout mice began to exhibit evidence of a wasting disease and loss of muscle strength ( Figure 2—figure supplement 1; Video 1 ) . Transverse sections of the quadriceps from 4-week-old Zfp106-null mice showed increased variation in fiber size and evidence of grouped atrophy with very few centrally located nuclei ( Figure 2a , b ) , consistent with neurogenic rather than myopathic disease ( Baloh et al . , 2007 ) . By three months of age , near end stage disease , Zfp106-null mice displayed extensive muscular atrophy and severe degeneration of muscle fibers ( Figure 2c , d ) . The number of choline acetyltransferase ( ChAT ) -positive motor neurons was also profoundly and significantly reduced in Zfp106 knockout spinal cords compared to littermate control mice ( Figure 2e–i ) , providing further evidence that loss of Zfp106 in mice causes a neurodegenerative phenotype . Independently , we confirmed the knockout phenotype observed with the KOMP allele ( van der Weyden et al . , 2011 ) using CRISPR-mediated gene disruption by non-homologous end joining ( NHEJ ) and creation of a frameshift mutation leading to a premature stop codon . The CRISPR-generated mutation resulted in an essentially identical wasting and histological phenotype ( data not shown ) . 10 . 7554/eLife . 19032 . 005Figure 2 . Zfp106 loss-of-function causes a severe neuromuscular phenotype in mice . ( a–h ) Evidence of profound neuromuscular pathology in Zfp106-null mice . H&E stained sections of quadriceps muscles ( a–d ) from 4-week old and 12-week old wild type ( a , c ) and Zfp106 knockout ( b , d ) mice are shown . Note the presence of grouped small angular fibers in knockout mice ( dashed circle ) , already clearly evident by 4-weeks of age ( b , arrowheads ) compared to the rounder , more evenly-shaped fibers in controls ( a ) . By 12-weeks ( end stage disease ) , Zfp106 knockout mice have profound pathology and the presence of atrophic and degenerated muscle fibers with numerous inclusions ( d , asterisks ) . Scale bar , 100 μm . ChAT immunohistochemical staining ( e–h ) of lumbar spinal cord sections from 8-week old wild type ( e , g ) and Zfp106 knockout ( f , h ) mice show a drastic reduction in the number of ChAT+ motor neurons . Arrowheads mark ChAT-expressing motor neurons in the ventral horn . Scale bar , 200 μm in all panels . ( i ) Quantification of ChAT-positive neurons in the lumbar spinal cord of wild type and Zfp106 knockout mice at 8 weeks of age shows a ~ 47% loss of motor neurons in Zfp106 knockout mice . n = 5 mice for each genotype . Data are expressed as percent of control ± SD and were analyzed by t-test . **p<0 . 001 . ( j ) Schematic of the HB9-3×FLAG-2×STREP-Zfp106-IRES-gfp ( HB9-Zfp106 ) transgene . ( k , l , m ) Transgenic expression of Zfp106 in motor neurons suppresses the wasting phenotype ( k ) , the reduction in grip strength ( l ) and the loss of lumbar spinal cord motor neurons ( m ) in Zfp106 knockout mice . Wasting data are expressed as the mean weight ± SEM . Grip strength data are expressed as the mean grip score ± SD; grip score: 1 , 1–10 s; 2 , 11–25 s; 3 , 26–60 s; 4 , 61–90 s; 5 , > 90 s . ChAT-positive neuron counts are expressed as percent of control ± SD . n = 5 female mice at 8 weeks of age for each genotype . Data were analyzed by one-way ANOVA and Bonferroni's Multiple Comparison Test . p<0 . 05; *p<0 . 001 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 00510 . 7554/eLife . 19032 . 006Figure 2—figure supplement 1 . Zfp106 knockout mice exhibit progressive and severe weight and grip strength loss over time . ( a ) Dorsal view of skinned wild type and Zfp106 knockout mice at 12-weeks of age . Zfp106 knockout mice show severe muscle wasting and kyphosis compared to wild type mice . ( b ) Female mice were weighed at the indicated ages , n = 5 for each group . Body weight of Zfp106 knockout mice was significantly reduced at 2 months of age compared to control and further reduced at 3 months of age . ( c ) Zfp106-null mice exhibit profound loss of grip strength; grip score: 1 , 1–10 s; 2 , 11–25 s; 3 , 26–60 s; 4 , 61–90 s; 5 , > 90 s . n = 5 age-matched female mice for each group tested at the indicated ages . Graphical data are presented as the mean ± SD . Data were analyzed by one-way ANOVA and Bonferroni's Multiple Comparison Test . ns , not significant; p<0 . 001 , **p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 00610 . 7554/eLife . 19032 . 007Figure 2—figure supplement 2 . Zfp106 is expressed in skeletal muscle and motor neurons . Expression of Zfp106 as detected by β-galactosidase activity in quadriceps ( a ) and spinal cord ( b ) of 8-week-old Zfp106laczflox/+ mice . β-galactosidase is expressed from a lacZ knock-in at the Zfp106 locus . A section of the quadriceps muscle from an 8-week old Zfp106laczflox/+ mouse stained with X-gal is shown in ( a ) . Arrowheads show β-galactosidase-expressing muscle fibers , weakly detectable by X-gal staining; scale bar , 100 μm . A section through the ventral horn of the spinal cord ( lumbar region ) of a Zfp106laczflox/+ mouse is shown in ( b ) . The same section is shown in all three panels in ( b ) ; the green channel shows expression of the neuronal marker NeuN; the red channel shows expression of β-galactosidase expressed from the Zfp106 locus . Arrows denote co-expression of β-galactosidase and NeuN in large motor neurons in the ventral horn of the spinal cord; scale bar , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 00710 . 7554/eLife . 19032 . 008Figure 2—figure supplement 3 . Expression of Zfp106 in wild type , knockout and HB9-Zfp106 transgenic mice . ( a ) Evaluation of Zfp106 expression by RT-qPCR in lumbar spinal cord of wild type , KOMP Zfp106 knockout and CRISPR Zfp106 knockout mice . Expression of Zfp106 mRNA is drastically and significantly reduced in lumbar spinal cord of both KOMP and CRISPR knockout mice . ( b ) Schematic of the transgenic construct used for generation of the motor neuron specific Zfp106 mice . The HB9 promoter drives expression of a 3xFLAG-2xSTREP N-terminally tagged Zfp106 and IRES-EGFP . ( c ) Evaluation of transgene expression by RT-qPCR in lumbar spinal cord at P10 . Expression of Zfp106 mRNA ranged from ~1 . 5 to 2 fold of wild type expression in two independent transgenic mouse lines , as assessed by qPCR with primers amplifying both endogenous and transgenic Zfp106 . Data in a , c were analyzed by one-way ANOVA and Bonferroni's Multiple Comparison Test . p<0 . 05; *p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 00810 . 7554/eLife . 19032 . 009Figure 2—figure supplement 4 . Rescue of ChAT+ motor neuron loss by restoration of Zfp106 to motor neurons in Zfp106 knockout mice . Representative images of transverse sections of lumbar spinal cord from wild type ( a ) , motor neuron-specific Zfp106 transgenic ( b ) , Zfp106 knockout ( c ) , and Zfp106 knockout/motor neuron-specific Zfp106 transgenic mice ( d ) , stained by anti-ChAT antibody . Arrowheads mark ChAT+ motor neurons . Scale bar , 200 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 00910 . 7554/eLife . 19032 . 010Video 1 . Zfp106-null mice exhibit a severe progressive neurodegenerative phenotype . The video shows a Zfp106 knockout mouse and a littermate control at 10 weeks of age . The Zfp106 knockout displays abnormal gait due to hind limb paralysis , severe kyphosis , wasting , and persistent hind limb clasping at tail suspension . The wild type mouse displays normal gait and normal extension of the hind limbs away from the abdomen during the tail suspension test . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 010 To gain insight into the cell autonomous requirements for Zfp106 , we used the KOMP lacZ knock-in allele ( van der Weyden et al . , 2011 ) to examine Zfp106 expression using β-galactosidase activity as a sensitive method for detecting expression . This approach showed Zfp106 expression in both skeletal muscle and motor neurons ( Figure 2—figure supplement 2a , b ) . Given the multiple lines of evidence for a role of Zfp106 in neurodegeneration , and the evident denervation observed in the knockout mice , we hypothesized that the primary requirement for Zfp106 was likely to be in motor neurons . We confirmed a highly significant and dramatic reduction in Zfp106 expression in spinal cord ( Figure 2—figure supplement 3a ) . The ~80% reduction in Zfp106 expression is consistent with previous studies ( Anderson et al . , 2016; Joyce et al . , 2016 ) . Therefore , we used a transgenic approach in mice to restore Zfp106 expression only in motor neurons under the control of the Mnx1 promoter ( herein called HB9 , HB9-Zfp106 ) on a Zfp106 germline knockout background ( Figure 2j; and Figure 2—figure supplement 3b , c ) . Importantly , restoration of Zfp106 specifically in motor neurons significantly suppressed the Zfp106 knockout wasting and muscle grip phenotypes and resulted in restoration of the number of ChAT-expressing motor neurons ( Figure 2j–m; Figure 2—figure supplement 4 ) , indicating a cell autonomous requirement for Zfp106 in motor neurons . To gain additional insights into the molecular function of Zfp106 , we used mass spectrometry to identify all proteins co-immunoprecipitated by Zfp106 from 293T cells ( Figure 3 ) . Several controls were included in the mass spectrometry experiments , including the DNA-binding factors Etv2 and MEF2C and the RNA-binding proteins Vif and Tat . The interactomes for each of the controls were compared to the Zfp106 interactome using the previously described MiST algorithm , which determines specific interactions based on enrichment , reproducibility , and specificity compared to interaction with the negative controls ( Jager et al . , 2012 ) . These experiments identified thirty-nine proteins that specifically interacted with Zfp106 ( Figure 3a; Supplementary file 1 ) . Gene ontology ( GO ) and protein complex analyses revealed a significant enrichment in interactors with roles in RNA processing , metabolism , and splicing ( Figure 3a , b; Supplementary file 2 ) , suggesting that Zfp106 may also play roles in these processes . 10 . 7554/eLife . 19032 . 011Figure 3 . Zfp106 interacts with a network of other ALS-associated RNA-binding proteins . ( a ) Network representation of the Zfp106 interactome generated by Cytoscape . Zfp106 ( yellow node ) interacts with 39 proteins with top 5% MiST score ( blue nodes ) . Curated protein-protein interactions from the CORUM database are indicated as nodes connected by green lines . ( b ) Gene Ontology ( GO ) Biological Process analysis of the Zfp106 interactome ranked by p value . The number of interacting proteins ( top 5% by MiST score ) in each category is indicated inside each bar . ( c ) Selected ALS-associated proteins within the top 5% MiST-scoring hits from the Zfp106 interactome are listed with putative functions associated with the pathobiology of ALS indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 01110 . 7554/eLife . 19032 . 012Figure 3—figure supplement 1 . Zfp106 interacts with TDP43 and Nup107 in Neuro-2a cells . Endogenous TDP43 and Nup107 coimmunoprecipitated with Zfp106 in Neuro-2a cells . Cells were transfected with 3×FLAG-2×STREP-Zfp106 or 3×FLAG-2×STREP parental vector and lysates were subjected to immunoprecipitation with anti-FLAG beads , followed by anti-FLAG ( top ) , anti-Nup107 ( middle ) , and anti-TDP43 ( bottom ) western blotting . Nup107 and TDP43 were coimmunoprecipitated by anti-FLAG IP only in lysates from cells expressing FLAG-Zfp106 . The immunoglobulin heavy chain ( IgG HC ) of the anti-FLAG antibody is indicated . Molecular weight markers in kD are indicated on the left . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 012 There is growing and compelling genetic and pathological evidence that dysregulated RNA processing and metabolism contribute to the development of ALS and C9orf72 ALS in particular ( Ling et al . , 2013; Prudencio et al . , 2015 ) . Interestingly , several of the Zfp106-interacting proteins identified in our affinity purification-mass spectrometry experiment are proteins directly associated with ALS and/or C9orf72 ALS , including splicing factors TDP-43 and FUS , the nucleolar protein Nucleolin , and the nuclear pore component Nup107 ( Freibaum et al . , 2015; Haeusler et al . , 2014; Lagier-Tourenne et al . , 2010; Ling et al . , 2013 ) ; Figure 3c; Supplementary file 1 ) . We further examined the interaction of Zfp106 with TDP-43 and Nup107 , as examples of interacting proteins , by anti-Flag-Zfp106 immunoprecipitation followed by western blot for TDP43 and Nup107 in Neuro-2a cells ( Figure 3—figure supplement 1 ) . These experiments confirmed and validated the interaction with these other ALS-associated proteins and support the notion that Zfp106 is part of an RNA metabolic pathway ( or pathways ) that when dysregulated leads to neurodegeneration . Prior studies have used Drosophila as an in vivo gain-of-function model of C9orf72 neurodegeneration ( Freibaum et al . , 2015; Tran et al . , 2015; Xu et al . , 2013; Zhang et al . , 2015 ) . Here , we used the GAL4-UAS system to coexpress mouse Zfp106 and GGGGCC repeats in the fly in a gain-of-function approach to determine if the interaction between Zfp106 and GGGGCC repeats had a functional consequence in this model ( Figure 4; Figure 4—figure supplement 1a ) . Expression of 30 tandem copies of the GGGGCC hexanucleotide in the 5′-untranslated region of a UAS-dependent gfp transcript ( 30R-GFP ) in glutamatergic neurons using OK371-GAL4 ( Mahr and Aberle , 2006 ) caused ~50% lethality due to an apparent failure of flies to eclose from the pupal case ( Figure 4a ) . Furthermore , 28 day-old OK371; 30R-GFP-expressing flies that did manage to successfully eclose exhibited profound defects in locomotor activity compared to control flies ( Figure 4b ) . No overt eclosion or locomotor phenotype was observed in flies expressing GFP only , 3R-GFP , or Zfp106 alone in glutamatergic neurons ( Figure 4a , b ) . Previous studies have also shown that expression of 30R-GFP causes a reduction in the number of active zones in Drosophila larval neuromuscular junctions ( Zhang et al . , 2015 ) . Consistent with those studies , we found that expression of 30R-GFP reduced active zones as measured by anti-Bruchpilot ( Brp ) immunofluorescence and quantification ( Figure 4c–f ) . Remarkably , co-expression of Zfp106 with 30R-GFP restored Brp expression at larval active zones ( Figure 4c–f ) and completely suppressed the ~50% lethality caused by expression of 30R-GFP ( Figure 4a ) . Moreover , co-expression of Zfp106 partially suppressed the adult locomotor defect caused by expression of 30R-GFP ( Figure 4b ) . Importantly , co-expression of Zfp106 did not result in a reduction in the expression of 30R-GFP mRNA Figure 4—figure supplement 1b ) , indicating that the suppression of the neurotoxic phenotype in Drosophila was not due to dilution of GAL4 protein . Taken together , these data show that Zfp106 is a potent suppressor of neurodegeneration in a C9orf72 ALS model , and the results from this gain-of-function model support the idea that Zfp106 functionally interacts with C9orf72 GGGGCC RNA repeats . 10 . 7554/eLife . 19032 . 013Figure 4 . Zfp106 suppresses GGGGCC-mediated neurotoxicity in Drosophila . ( a ) Schematics and Punnett squares from crosses of UAS-Zfp106; UAS-30R-GFP ( left ) and UAS-Zfp106; UAS-3R-GFP ( right ) transgenic flies to OK371-GAL4 flies . The Punnett squares show the number of flies that eclosed from the indicated crosses and show that expression of 30R-GFP resulted in failure of ~50% of flies to eclose ( 25 actual versus 54 . 5 predicted ) ; this defect was completely suppressed by co-expression of Zfp106 . The Chi-square test ( χ2 ) was used to determine the significance of the differences between the observed and expected frequencies of the expected genotypes . ( b ) Locomotor activity of 28-day post-eclosion flies of the indicated genotypes assessed by rapid iterative negative geotaxis ( RING ) assay . Data are presented as the mean height climbed ± SEM . A two-way ANOVA analysis of the locomotor data revealed a significant effect of repeat expression ( p<0 . 0001 ) and Zfp106 expression ( p<0 . 01 ) on locomotor activity , with a significant interaction between the two factors ( p<0 . 01 ) ; i . e . Zfp106 coexpression significantly suppresses neurotoxicity induced by expression of 30R ( p<0 . 01 ) by Bonferroni's posthoc Multiple Comparison Test . n = 25 male flies for each genotype; p<0 . 0001 , p<0 . 01 . ( c , d , e ) Representative NMJs of muscle 6/7 in abdominal segment A3 of third instar larvae of the indicated genotypes co-stained with anti-HRP ( neuronal marker , red channel ) and anti-Bruchpilot ( Brp ) ( active zone marker , green channel ) . Staining of the active zone component Brp shows that expression of 30R causes a reduction in Brp-positive active zones [arrowheads in ( d ) compared to control in c] . Coexpression of Zfp106 suppressed the reduction in Brp-positive active zones caused by 30R [arrowheads in e compared to d] . ( f ) Quantification of active zones as total voxel occupancy by Brp signal at each terminal shows that Zfp106 significantly suppresses the reduction in active zone caused by expression of 30R . Data were analyzed by one-way ANOVA and Bonferroni's Multiple Comparison Test . *p<0 . 001 , **p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 01310 . 7554/eLife . 19032 . 014Figure 4—figure supplement 1 . GAL4-dependent expression of Zfp106 in Drosophila does not affect 30R-gfp mRNA levels . ( a ) Anti-FLAG western blot conducted on lysates from heads dissected from transgenic flies of the indicated genotypes . The band in lane 3 migrating below the 225 kD marker indicates that UAS-3×FLAG-2×STREP-Zfp106 ( UAS-Zfp106 ) transgenic flies express Zfp106 in a GAL4-dependent manner . ( b ) Evaluation of 30R-gfp expression by RT-qPCR in flies expressing 30R alone or coexpressing 30R and Zfp106 . Zfp106 coexpression does not cause alterations in 30R-gfp mRNA levels . DOI: http://dx . doi . org/10 . 7554/eLife . 19032 . 014 The ability of Zfp106 to suppress the neurotoxic gain-of-function effect of rGGGGCC repeats has important implications for understanding and possibly ameliorating the pathology of C9orf72 ALS . Previous studies have suggested roles for Zfp106 in rDNA transcription and in pre-rRNA processing ( Ide and Dejardin , 2015; Tafforeau et al . , 2013 ) . Zfp106 was also identified as a polyadenylated mRNA binding protein in an interactome capture study aimed at defining an atlas of mammalian mRNA binding proteins ( Castello et al . , 2012 ) . Interestingly , our interactome data also suggest possible roles for Zfp106 in multiple steps of RNA metabolism and processing and interaction with key RNA binding proteins involved in various forms of neurodegeneration . Precisely how disrupting these processes leads to neurodegeneration and exactly how this relates to C9orf72 ALS/FTD remains to be determined . A simple model is that sequestration of Zfp106 by the expanded repeats disrupts normal Zfp106 function in patients with C9orf72 ALS in a sponge-like mechanism ( Cooper-Knock et al . , 2014; Mizielinska and Isaacs , 2014 ) . Alternatively , one might speculate that Zfp106 plays other functions important in the biology of C9orf72 , such as influencing dipeptide repeat protein ( DPR ) production or affecting nucleolar function or nucleocytoplasmic transport via interactions with Nucleolin or Nup107 , which have also been implicated in the pathobiology of C9orf72 ALS ( Boeynaems et al . , 2016; Freibaum et al . , 2015; Haeusler et al . , 2014; Jovičić et al . , 2015; Kwon et al . , 2014; Zhang et al . , 2015 ) . Future studies will address the molecular targets and mechanisms of action of Zfp106 in order to gain further insights into the pathogenesis of C9orf72 ALS and to identify new possible therapeutic targets for neuroprotection . A murine Zfp106 cDNA corresponding to NCBI accession number XM_006499044 . 2 and a murine Mbnl1 cDNA corresponding to NCBI accession number NM_020007 . 4 were cloned by PCR from an E13 . 5 embryonic heart cDNA library; the amplified cDNAs were subcloned into mammalian expression plasmid pcDNA4/TO ( Invitrogen , Carlsbad , CA ) modified to include N-terminal 3×FLAG-2×STREP tags . The 3×FLAG-2×STREP-Zfp106 cassette was subsequently subcloned into the EcoRI/XhoI sites of pUAST ( Brand and Perrimon , 1993; Flybase ID FBmc0000383 ) for subsequent generation of transgenic flies , and into the AgeI/SpeI sites of HB9-MCS-IRES-EGFP ( Addgene plasmid #16283 ) for generation of motor neuron-specific HB9-Zfp106 transgenic mice . The Zfp106Δ allele was generated by CRISPR-mediated genome editing ( Wang et al . , 2013 ) . 5′-atattgtgccacatcgtgtatgg-3’ and 5′-tttttgagccatacacgatgtgg-3’ sgRNAs , which target exon 1 ( ENSMUSE00001311370 ) of mouse Zfp106 , were transcribed in vitro using the MEGAshortscript T7 kit ( Life Technologies , AM1354 ) and were purified using the MEGAclear kit ( Life Technologies , AM1908 ) . Purified sgRNAs and in vitro-transcribed Cas9 mRNA were co-injected into the cytoplasm of fertilized mouse oocytes using standard transgenic technology as described previously ( De Val et al . , 2004 ) . Two independent F0 founders were independently outcrossed to wild-type mice , and F1 offspring were used for subsequent Zfp106Δ intercrosses . The mouse strain Zfp106tm1a ( KOMP ) Wtsi ( RRID:IMSR_KOMP:CSD26348-1a-Wtsi ) was generated from ES cell clone EPD0033_4_C03 , obtained from the NCRR-NIH supported KOMP Repository ( www . komp . org ) and generated by the CSD consortium for the NIH funded Knockout Mouse Project ( KOMP ) . The targeted allele contains a promoterless LacZ-Neo cassette , which also contains an En2 splicing acceptor ( En2 SA ) and polyadenylation signal ( pA ) , flanked by FRT sites; the targeted allele also has loxP sites flanking exon 2 ( Testa et al . , 2004; ENSMUSE00000454975 ) . To generate Zfp106laczΔ/+ mice , male Zfp106laczflox/+ mice were crossed to female Mef2c-AHF-Cre mice ( Verzi et al . , 2005 ) ; RRID:MMRRC_030262-UNC ) . When crossed from the female , Mef2c-AHF-Cre functions as a universal deleter line due to Cre expression in the female germline , resulting in Cre-dependent recombination in all cells in all offspring ( Ehlers et al . , 2014 ) . Transgenic HB9-3×FLAG-2×STREP-Zfp106-IRES-egfp ( HB9-Zfp106 ) mice expressing Zfp106 in motor neurons under the control of the promoter of the Mnx1 gene ( herein called HB9 ) , were generated by standard pronuclear injection of the HB9-3×FLAG-2×STREP-Zfp106-IRES-EGFP plasmid linearized by XhoI digestion . Injected embryos were implanted into pseudo-pregnant females , and F0 founders were screened by PCR . Grip strength was measured by Kondziella inverted screen test ( Deacon , 2013; Kondziella , 1964 ) . Genotyping of mutant and transgenic mice was performed by PCR on genomic DNA isolated from tail biopsies using the following primers: Zfp106Δ , 5′-caggaggctgtgctgtgata-3’ and 5′-attctaggactgggaggcaga-3′; Zfp106laczΔ and Zfp106laczflox 5′-ccccctgaacctgaaacata-3’ and 5′-gaaagtcccagctgcaagac-3’ for the mutant allele and 5′-tgaatcagctcgtaaacggac-3’ and 5′-atgaccgctcaaatcgatcaa-3’ for the wild type allele; HB9-3×FLAG-2×STREP-Zfp106-IRES-egfp , 5′-aagttcatctgcaccaccg-3’ and 5′-tccttgaagaagatggtgcg-3′ . UAS-GFP , UAS-3R-GFP and UAS-30R-GFP transgenic Drosophila melanogaster lines were a generous gift of Peng Jin ( Emory University ) and have been described previously ( Xu et al . , 2013 ) . GMR-GAL4 ( FBst0008605 ) and OK371-GAL4 ( FBst0026160 ) lines were obtained from the Bloomington Stock Center . UAS-3×FLAG-2×STREP-Zfp106 ( UAS-Zfp106 ) transgenic D . melanogaster were generated by standard w+ P-element-mediated transgenesis ( Brand and Perrimon , 1993 ) on a w1118 background ( BestGene , Inc . ) . Flies were maintained at 23°C , unless otherwise indicated . For western blot detection of the Zfp106 transgene , flies were shifted to 30°C for 3 days prior to removing heads for protein extraction to induce higher expression of GAL4 from GMR-GAL4 . Mouse Neuro-2a ( ATCC , CCL-131; RRID:CVCL_0470 ) and human HEK293T cells ( ATCC , CRL-3216; RRID:CVCL_0063 ) were obtained directly from ATCC and maintained in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum ( FBS , Gibco , Waltham , MA ) and penicillin-streptomycin ( Gibco ) . All cell lines were routinely tested for Mycoplasma contamination with the MycoAlert PLUS Mycoplasma Detection Kit ( Lonza , Switzerland , LT07-705 ) and found to be negative . Biotinylated-RNA pull-down of nuclear proteins was performed as previously described with modifications ( Marin-Bejar and Huarte , 2015 ) . 100 pmol of ssRNA oligomer [either r ( GGGGCC ) 8 or r ( AAAACC ) 8] biotinylated at the 3’ end ( Integrated DNA Technologies , Inc . ) were incubated with 50 µl of pre-washed Nucleic Acid Compatible Streptavidin Magnetic Beads ( Pierce Magnetic RNA-Protein Pull-Down kit , cat . n . 20164 ) in RNA capture buffer ( 20 mM Tris , pH 7 . 5; 1 M NaCl; 1 mM EDTA ) for 30 min on a rotary shaker . The beads were washed twice in 20 mM Tris at 4°C for 5 min each and then incubated with nuclear extract for 4 hr at 4°C . Approximately 1 × 107 Neuro-2a cells were detached from cell culture dishes when they were approximately 80% confluent with 10 mM EDTA in PBS , incubated in Harvest Buffer ( 10 mM HEPES , pH 7 . 9; 50 mM NaCl; 0 . 5 M sucrose; 0 . 1 mM EDTA; 0 . 5% NP-40 ) for 10 min on ice . Nuclei were pelleted by centrifugation at 4500 × g for 10 min at 4°C , washed for 5 min in buffer A ( 10 mM HEPES , pH 7 . 9; 10 mM KCl; 0 . 1 mM EDTA; 0 . 1 mM EGTA ) , centrifuged again at 4500 × g for 5 min at 4°C and lysed in IP buffer ( 50 mM Tris , pH7 . 4; 150 mM NaCl; 1 mM EDTA ) plus 0 . 5% NP-40 . Complete protease inhibitor and PhosSTOP phosphatase inhibitor ( Roche ) were added to all buffers . After incubation with the nuclear extracts , beads were washed 3× in IP buffer plus 0 . 05% NP-40 , and then 1× in IP buffer without NP-40 . Co-precipitated proteins were eluted from bound biotinylated RNAs in 45 µl 50 mM Tris-Cl , pH7 . 4; 150 mM NaCl; 1 mM EDTA plus 0 . 05% RapiGest ( Waters Corp . , Milford , MA ) . Precipitated proteins were then detected by mass spectrometry or western blot . 100 pmol of r ( GGGGCC ) 4 or r ( CUG ) 8 single-stranded RNA ( Integrated DNA Technologies , Inc . ) were 5′-end labeled with [γ-32P]ATP and T4 Polynucleotide Kinase ( New England Biolabs ) as previously described ( Rio , 2014 ) . Unincorporated radioactivity was removed using Illustra MicroSpin G-25 Columns ( GE Healthcare Life Sciences , UK ) . RNA EMSA was then performed as previously described ( Rio , 2014 ) by incubating the labeled RNA ( 100 , 000 cpm ) with purified Zfp106 or MBNL1 protein for 30 min at room temperature in 1× binding buffer ( 40 mM Tris-Cl , pH 8 . 0; 30 mM KCl; 1 mM MgCl2; 0 . 01% NP-40; 1 mM DTT; 5% glycerol; 10 µg/ml bovine serum albumin ) . The RNA-protein mixtures were then electrophoresed in 6% acrylamide-TBE gels . Gels were dried and exposed to X-ray film for analysis . For the competition experiments , the purified proteins were preincubated for 10 min at room temperature with 30-fold molar excess of the following unlabeled single-stranded RNAs: r ( GGGGCC ) 4 , r ( AAAACC ) 4 , r ( CUG ) 8 , or a mutated ( mut ) GGGGCC repeat sequence , r ( GAGGCCGGGACCGAGACCGAGGCC ) . Skeletal muscles were isolated from sex- and age-matched mice , mounted on cork discs with 10% tragacanth gum ( Sigma ) and immediately frozen in liquid nitrogen-cooled isopentane . Frozen muscles were then cryosectioned at a thickness of 10 μm and stained with hematoxylin and eosin as previously described ( Verzi et al . , 2005 ) . For X-gal staining , cryosections were fixed in ice cold 4% formaldehyde , 0 . 5% glutaraldehyde in PBS for 2 min and then stained as previously described ( Verzi et al . , 2005 ) . Spinal cords were dissected from mice perfused with 4% PFA , additionally post-fixed with 4% PFA overnight , equilibrated with 30% sucrose in PBS and embedded in OCT . Immunostaining for choline acetyltransferase ( ChAT ) was performed on 40 μm free-floating cryostat sections as described previously ( Qiu et al . , 2014 ) with goat anti-ChAT antibody ( 1:300; Millipore , AB144P; RRID:AB_2079751 ) and biotinylated rabbit anti-goat ( 1:200; Vector Laboratories , BA-5000; RRID:AB_2336126 ) as secondary antibody , followed by diaminobenzidine immunoperoxidase development using the Vectastain Elite ABC kit ( Vector Laboratories , PK-6105 ) and the DAB substrate kit for peroxidase ( Vector Laboratories , SK-4100 ) . Post-mortem human spinal cord tissue was obtained from the UCSF Neurodegenerative Disease Brain Bank ( NDBB ) . Formalin fixed paraffin-embedded tissue blocks were cut into 10 µm-thick sections . Tissue sections were subjected to antigen retrieval by autoclaving at 120°C for 5 min in 0 . 1M citrate buffer , pH 6 . 0 . Immunohistochemistry was performed with rabbit anti-Zfp106 ( 1:500; Bethyl , A301-526A; RRID:AB_999690 ) . Following immunostaining , digital images were captured using a high resolution digital camera ( Nikon DS-U2/L2 ) mounted on a Nikon Eclipse80i microscope using NIS-Elements ( Version 3 . 22 , Nikon ) imaging software . Mouse spinal cords were dissected and then fixed in 2% paraformaldehyde overnight at 4°C , embedded in paraffin and then sectioned at 20 μm for subsequent immunofluorescence analyses . Neuro-2a cells were grown on coverslips and fixed with 4% PFA for 10 min . Immunofluorescence was performed as described previously ( Anderson et al . , 2004 ) with the following antibodies: mouse anti-NeuN ( 1:500; Millipore , MAB377; RRID:AB_177621 ) , chicken anti-β-gal ( 1:500; Aves , BGL1040; RRID:AB_2313507 ) , rabbit anti-Zfp106 ( 1:100; Bethyl , A301-527A; RRID:AB_999691 ) . Alexa Fluor 594 goat anti-chicken ( 1:500; Molecular Probes A11042; RRID:AB_142803 ) , Alexa Fluor 488 donkey anti-mouse ( 1:500; Molecular Probes A21202; RRID:AB_141607 ) , and Alexa Fluor 594 donkey anti-rabbit ( 1:500; Molecular Probes A21207 RRID:AB_141637 ) were used as secondary antibodies . The slides were counterstained and mounted with SlowFade Gold antifade reagent with DAPI ( Life Technologies , Waltham , MA ) . Confocal images were acquired with a Leica TCS SPE laser scanning confocal system . For coimmunoprecipitation , Neuro-2a cells were transfected with 7 µg of 3×FLAG-2×STREP-Zfp106 or 3×FLAG-2×STREP parental vector per 55 cm2 plate by using GenJet in vitro DNA Transfection Reagent for Neuro-2A Cells , according to the manufacturer’s recommendations ( Signagen , Rockville , MD ) . Cells were lysed in Benzonase Lysis Buffer [20 mM Tris-HCl ( pH 7 . 5 ) , 40 mM NaCl , 2 mM MgCl2 , 0 . 5% NP-40 and Complete EDTA-free protease inhibitor ( Roche , Switzerland ) ] with the addition of 100 U/ml benzonase ( Novagen , Germany ) to digest nucleic acids . After incubation for 30 min on ice , NaCl concentration was increased to 150 mM and the lysates were further incubated with agitation for 30 min at 4°C . After clearing , the lysates were incubated with pre-washed anti-FLAG ( M2 ) -conjugated magnetic beads ( Sigma , St . Louis , MO , M8823 ) for 4 hr at 4°C . The beads were then washed five times with IP Buffer with 0 . 05% NP40 , before incubation at 70°C for 10 min in 2× NuPAGE Sample Buffer ( ThermoFisher Scientific , Waltham , MA , NP0007 ) supplemented with NuPAGE Sample Reducing Agent ( ThermoFisher Scientific , NP0004 ) to release coimmunprecipitated proteins . Samples were then subjected to SDS-PAGE and western blot using standard procedures . For western blot to detect Zfp106 in transgenic D . melanogaster , heads from 40 Gmr-Gal4 , UAS-3×FLAG-2×STREP-Zfp106 , and Gmr-Gal4;UAS-3×FLAG-2×STREP-Zfp106 flies were removed and homogenized in RIPA buffer ( 50 mM Tris-Cl , pH 7 . 4; 150 mM NaCl; 2 mM EDTA; 1% NP-40; 0 . 2% SDS ) with a Dounce homogenizer ( Wheaton , Millville , NJ ) . Homogenates were cleared by centrifugation , and total protein content was quantified with the Bio-Rad Protein Assay kit ( Bio-Rad , 5000002 ) . An equivalent amount of total protein for each sample was subjected to SDS-PAGE and western blot using standard procedures . The following antibodies were used for western blot: rabbit anti-Zfp106 ( 1:1000; Bethyl , A301-527A; RRID:AB_999691 ) ; mouse anti-FLAG ( 1:1 , 000; Sigma , F3165 ( clone M2 ) ; RRID:AB_439685 ) , mouse anti-actin ( 1:500; Developmental Studies Hybridoma Bank , JLA20; RRID:AB_528068 ) , rabbit anti-TDP-43 ( 1:1000; Proteintech , 12892–1-AP; RRID:AB_2200505 ) and rabbit anti-Nup107 ( 1:1000; Proteintech , 19217–1-AP; RRID:AB_10597702 ) . For RT-qPCR , total RNA was extracted from lumbar spinal cord from mice or from 20 flies per genotype using Trizol ( Invitrogen ) . RNA was treated for 1 hr at 37°C with DNaseI , followed by cDNA synthesis using the Omniscript RT kit ( Qiagen ) . qPCR was performed using the MAXIMA SYBR Green kit ( Thermo Scientific ) and a 7900HT Fast Real Time PCR System ( Applied Biosystems ) . The SDS 2 . 4 software package was used to extract raw qPCR data . Data were normalized to housekeeping genes by the 2−ΔΔCt method . The following primers were used for qPCR in mice: primers 5′-atttggcgttgtggatcact-3’ and 5′-gcgttcaatatgtcctgcaa-3’ for amplification of Zfp106 exon 21–22 and 5′-accacagtccatgccatcac-3’ and 5′-tccaccaccctgttgctgta-3’ for amplification of gapdh . The following primers were used for qPCR in flies: 5′-gcgcggttactctttcacca-3’ and 5′- atgtcacggacgatttcacg-3’ for actin , and 5′-gcacgacttcttcaagtccgccatgcc-3’ and 5′-gcggatcttgaagttcaccttgatgcc-3’ for egfp . HEK293T cells were transfected with 20 µg of 3×FLAG-2×STREP-Zfp106 or 5 µg of 3×FLAG-2×STREP-MBNL1 plasmid per 150 cm2 plate by using PolyJet DNA In Vitro Transfection Reagent , according to the manufacturer’s recommendations ( Signagen ) . For RNA EMSA , Zfp106 and MBNL1 were purified from HEK293T total cellular lysates using the Strep-Tag system ( Schmidt and Skerra , 2007 ) , with Strep-Tactin Sepharose ( IBA , catalog #2-1201-010 ) , according to the manufacturer’s protocol . 0 . 1% Triton X-100 was added to all buffers for cell lysis and chromatography to enrich for purified proteins . Eluted fractions were analyzed by SDS-PAGE stained with Coomassie brilliant blue . Comparable amounts of Zfp106 and MBNL1 were used in RNA EMSA . A parallel purification from HEK293T cells transfected with the 3×FLAG-2×STREP parental vector was performed and the corresponding fractions in which Zfp106 and MBNL1 were eluted were used in RNA EMSA as a negative control . Affinity purification of Zfp106 for identification of interacting proteins by mass spectrometry was performed as previously described ( Jager et al . , 2012 ) . Briefly , cleared nuclear extracts from HEK293T cells were prepared as described above ( Biotinylated-RNA pull-down section ) with the introduction of a benzonase digestion step to digest nucleic acids ( 100 U/ml for 30 min in the presence of 1 mM MgCl2 ) and incubated with pre-washed anti-FLAG ( M2 ) -conjugated magnetic beads ( Sigma , M8823 ) for 2 hr at 4°C . Following purification , complexes bound to beads were washed and then eluted in IP buffer containing 100 µg/ml 3×FLAG peptide ( Elim Biopharm ) and 0 . 05% RapiGest ( Waters Corp . ) . To analyze proteins by liquid chromatography-mass spectrometry ( LC-MS/MS ) , the eluates were digested with trypsin , and the peptides were analyzed on a Velos Pro mass spectrometry system ( Thermo Scientific ) . Peptides were matched to protein sequences by the Protein Prospector Algorithm ( RRID:SCR_014558 ) , and data were searched against a database containing SwissProt human protein sequences ( UniProt Consortium , 2015; UniProtKB , RRID:SCR_004426; http://www . uniprot . org ) . Parallel FLAG-affinity purifications were performed on untransfected HEK293T cells and HEK293T cells transfected with 3xFLAG-2xSTREP tag only , GFP-3xFLAG-2xSTREP or the following 3xFLAG-2xSTREP-tagged proteins: Vif , Tat , MEF2C , Myocardin , and Etv2 . The bait-prey datasets were analyzed with the APMS scoring algorithm MiST ( Jager et al . , 2012 ) , and an in-house curated APMS dataset used for determining machine background noise . The MiST algorithm compares peak intensities from the mass spectrum to determine abundance , reproducibility across multiple replicated experiments ( n = 5 for each condition ) , and specificity based on the uniqueness of each interaction , and then these three metrics are used to determine a composite score by the algorithm ( Jager et al . , 2012 ) . The Zfp106 interactome was compiled by selecting bait-prey pairs with a top 5% MiST score . Enrichment analysis for GO biological process terms overrepresented in the interactome was performed on the Gene Ontology Consortium website ( Gene Ontology Consortium , 2015; RRID:SCR_002811; http://www . geneontology . org ) . The top 5% MiST hits were visualized as network representation using Cytoscape ( Cline et al . , 2007; RRID:SCR_003032 ) . Protein complex analysis was performed using the manually curated CORUM protein complex database ( Ruepp et al . , 2008; RRID:SCR_002254; http://mips . helmholtz-muenchen . de/corum/ ) . For analysis of neuromuscular junction ( NMJ ) phenotypes in Drosophila , wandering third instar larvae were dissected as previously described ( Smith and Taylor , 2011 ) . Dissected larvae were fixed in 3 . 7% formaldehyde in PBS for 20 min , washed in PBT ( PBS with 0 . 4% Triton X-100 ) and blocked with 10% normal goat serum for 1 hr at room temperature . The larvae were then incubated overnight at 4°C with mouse anti-Brp antibody ( 1:50; Developmental Studies Hybridoma Bank , nc82; RRID:AB_2314865 ) to label the active zones and Cy3 goat anti-HRP ( 1:200; Jackson ImmunoResearch , 123-165-021; RRID:AB_2338959 ) to label the NMJ membrane and axons . Next , samples were washed in PBT and incubated with Alexa Fluor 488 goat anti-mouse ( 1:200; Molecular Probes , A11001; RRID:AB_2534069 ) in PBT + 10% normal goat serum at room temperature for 2 hr . Multiple confocal stacks were acquired using a point scanning confocal microscope ( Leica TCS SPE ) and then processed using the DeadEasy Synapse ImageJ plugin ( Sutcliffe et al . , 2013 ) , which provides total voxel occupancy by Brp signal at each nerve terminal . Brp signal was quantified at muscle 6/7 NMJs from abdominal segment A3 . Locomotor activity in 28-day post-eclosion flies was assayed by the Rapid Iterative Negative Geotaxis ( RING ) assay , as previously described ( Gargano et al . , 2005; Nichols et al . , 2012 ) . Briefly , 25 flies for each genotype were transferred without anesthetizing in polystyrene vials assembled in a RING apparatus . The RING apparatus was sharply tapped down three times and negative geotaxis was recorded on video for six trials for each genotype . The average height climbed at 3 s after completion of the third tap was scored for each genotype . Statistical analyses were performed using GraphPad Prism 5 . 0 ( GraphPad Software; RRID:SCR_002798 ) . Data were analyzed by one-way or two-way ANOVA followed by Bonferroni’s Multiple Comparison Test or by t-test .	Molecules of ribonucleic acid ( or RNA for short ) have many roles in cells , including acting as templates to make proteins . RNA is made of building blocks called nucleotides that are assembled to form strands . The precise order of the nucleotides in an RNA molecule can have a dramatic effect on the role that RNA plays in the body . For example , amyotrophic lateral sclerosis ( ALS ) is a deadly disease caused by the gradual loss of the nerve cells that control muscle ( known as motor neurons ) . The most common cause of inherited ALS is a genetic mutation that results in some RNA molecules having many more copies of a simple six nucleotide sequence known as GGGGCC than normal cells . RNA molecules with these “GGGGCC repeats” form clumps in motor neurons . The clumps of RNA molecules also contain proteins , but the identities of these RNA-binding proteins and the roles they play in ALS remain largely unknown . Celona et al . have now identified a new RNA-binding protein called Zfp106 , which binds specifically to GGGGCC repeats in mice and fruit flies . Removing the gene that encodes Zfp106 from mice causes the mice to develop ALS . On the other hand , restoring Zfp106 only to the motor neurons of these mutant mice prevents the mice from developing disease . This suggests that Zfp106’s role is specific to motor neurons . Indeed , fruit flies that have too many copies of GGGGCC develop severe symptoms reminiscent of ALS . Introducing a mammalian version of Zfp106 into these flies prevents them from developing the disease . The findings of Celona et al . suggest that Zfp106 might be a potential new drug target for treating ALS in humans . The next step following this work will be to find out exactly how Zfp106 regulates normal cellular processes by binding to RNA and how it suppresses ALS-like disease by binding to GGGGCC RNA-repeats .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "short", "report", "neuroscience" ]	2017	Suppression of C9orf72 RNA repeat-induced neurotoxicity by the ALS-associated RNA-binding protein Zfp106
Protein abundance differs from a few to millions of copies per cell . Trypanosoma brucei presents an excellent model for studies on codon bias and differential gene expression because transcription is broadly unregulated and uniform across the genome . T . brucei is also a major human and animal protozoal pathogen . Here , an experimental assessment , using synthetic reporter genes , revealed that GC3 codons have a major positive impact on both mRNA and protein abundance . Our estimates of relative expression , based on coding sequences alone ( codon usage and sequence length ) , are within 2-fold of the observed values for the majority of measured cellular mRNAs ( n > 7000 ) and proteins ( n > 2000 ) . Our estimates also correspond with expression measures from published transcriptome and proteome datasets from other trypanosomatids . We conclude that codon usage is a key factor affecting global relative mRNA and protein expression in trypanosomatids and that relative abundance can be effectively estimated using only protein coding sequences . Cellular growth and function depend upon the efficient expression of a large number of proteins that differ in abundance over a range from only a few molecules to millions of molecules per cell . Gene expression can be controlled at many levels , including transcription , mRNA stability , translation and protein stability . Trypanosoma brucei are parasitic protozoa that present a unique model for studies on gene expression because a relatively small number of ‘dispersed’ GT-rich RNA polymerase II ( pol-II ) promoters drive constitutive polycistronic transcription ( Wedel et al . , 2017 ) . Indeed , T . brucei have no known regulated RNA polymerase II promoters for protein-coding genes . Some stress response genes and cell cycle regulated genes are located towards the ends of the polycistonic units ( Kelly et al . , 2012 ) but there is otherwise limited clustering of functionally related genes . In addition , trypanosomes have only two genes containing known introns ( Mair et al . , 2000 ) , and every mRNA has an identical sequence trans-spliced onto the 5'-end ( Clayton , 2016 ) . These features are conserved in the related parasitic trypanosomatids , Trypanosoma cruzi and Leishmania spp . and in other trypanosomatids . This remarkable level of uniformity in terms of transcription and mRNA processing indicates that gene expression control primarily operates post-transcription . Sixty-one alternative base-triplets ( codons ) in DNA and mRNA encode for twenty different amino acids , such that many amino acids are encoded by two or up to six distinct but 'synonymous' codons . These codons can vary in their first position for three amino acids and in their third position for eighteen . Although recognised several decades ago , our understanding of the impact of inherently redundant codon usage and codon usage bias remains incomplete . One mRNA can yield 4000 molecules of protein , as measured in yeast ( Futcher et al . , 1999 ) , while the average protein:mRNA ratio in insect-form T . brucei is estimated to be 550:1; median values of three and 1650 molecules per gene , per cell , respectively ( Kolev et al . , 2010 ) . Thus , there is substantial capacity for expression control at the level of translation . Indeed , individual codons can control translation-rate in yeast ( Gardin et al . , 2014 ) , codon-dependent local translation slowdown can facilitate nascent peptide processing in eukaryotes ( Mahlab and Linial , 2014; Pechmann et al . , 2014 ) and changes in relative codon usage and cognate tRNA abundance can control differentiation-related translation programmes in metazoa ( Gingold et al . , 2014 ) . Thus , different codons are decoded at different rates by native ribosomes ( Hanson and Coller , 2018; Novoa and Ribas de Pouplana , 2012; Quax et al . , 2015 ) . Codon usage can also impact mRNA decay ( Hanson and Coller , 2018; Presnyak et al . , 2015 ) . In trypanosomatids , highly expressed genes are amplified in tandem and are enriched in GC3 codons , those codons that have a G or a C at the third position; cognate tRNA genes for these codons also display increased copy number ( Horn , 2008 ) . Although pol-II promoters appear to underpin much of the regulation of gene expression in metazoa and other eukaryotes , GC3 codons are also favoured in highly expressed genes in mammals and the evidence suggests that protein abundance is in fact predominantly controlled at the level of translation ( Schwanhäusser et al . , 2011 ) . In trypanosomatids , control of translation and mRNA stability are typically ascribed to 3'-untranslated sequences and their interactions with RNA-binding proteins and indeed , such regulation does operate ( Clayton , 2013 ) , but informatics analysis also supports a role for translational selection , the increased translation of GC3 codons ( Chanda et al . , 2007; Horn , 2008; Subramanian and Sarkar , 2015 ) . Since transcription for all but a few genes is polycistronic and constitutive , trypanosomatids present excellent model eukaryotic systems in which to investigate the impact of codon bias on gene expression control . Here , we combine experimental and bioinformatics analyses to explore the contribution that codon usage makes to differential gene expression in T . brucei and in other trypanosomatids . Our results indicate a capacity for codon bias based control of relative protein abundance over several orders of magnitude . We find that mRNA abundance is also increased by favoured codons . Remarkably , our predictions , based on trypanosomatid protein-coding sequences alone , correlate well with observed measures of steady-state mRNA and protein abundance . Prior analysis of codon usage in trypanosomatids indicated a correlation between GC3-bias and protein expression ( Horn , 2008 ) . However , the hypothesis that GC3 codons increase expression was not addressed experimentally . To test this hypothesis , we cloned wild-type and ( human ) codon-optimised ( GC3-bias is also observed in human sequences ) Gaussia luciferase ( gLUC ) genes ( Shao and Bock , 2008 ) in a T . brucei inducible expression construct ( Figure 1A ) . Importantly , we used a construct that integrates reproducibly as a single copy at a single site in the genome , eliminating position effects that arise due to integration at different genomic sites ( Alsford et al . , 2005 ) . All other sequences , except for the protein coding sequence , remained constant for each experiment performed with these constructs . The relative abundance of GC3 codons , typically found in highly expressed T . brucei genes ( Horn , 2008 ) , is illustrated in heat-map format ( Figure 1A ) . We selected gLUC because the expression of the gLUC protein can be assessed over a wide dynamic range using an activity assay , and inducible expression was employed so that we could determine whether gLUC expression was toxic to T . brucei . We assembled bloodstream form T . brucei strains with each construct integrated into the genome and first assessed gLUC expression on protein blots . We observed robust inducible expression and relative induced expression levels that were as predicted by our hypothesis , revealing 6-fold higher luciferase expression from the GC3-enriched gene as determined by densitometry ( Figure 1B ) ; three independent clones using each construct revealed similar results and we saw no evidence of toxicity . Encouraged by these initial results , we assembled constitutive expression constructs containing three new synthetic gLUC genes ( Supplementary File 1 , sheet 1 ) with the maximal or minimal number of GC3 codons or with alternating GC3 and AT3 codons; again illustrated in heat-map format ( Figure 1C ) . T . brucei strains expressing these constructs were assessed by luciferase activity assay , which again revealed relative expression levels that were significantly different and as predicted by our hypothesis ( Figure 1C ) . We observed 4-fold higher expression from the high-GC3 gene relative to the medium-GC3 gene and strikingly undetectable expression from the low-GC3 gene; in each case , a pair of independent clones derived using each construct revealed similar expression levels . These results indicate that codon usage has the potential to account for the full dynamic range of protein expression levels observed in trypanosomes . The Green Fluorescent Protein is often used as a reporter or as a protein tag for gene expression and subcellular localisation studies , and this is also the case in trypanosomatids . We , therefore , used GFP to test the impact of codon usage . As above , we assembled expression constructs containing three synthetic GFP genes ( Supplementary File 1 , sheet 1 ) with the maximal or minimal number of GC3 codons or with alternating GC3 and AT3 codons; see heat-maps ( Figure 1D ) . T . brucei strains expressing these constructs were assessed by quantitative protein blotting , which once again revealed the expected relative expression levels ( Figure 1D ) , further validating our hypothesis . Expression from the high-GC3 GFP gene was higher than from the medium-GC3 gene and , as in the case of gLUC above , expression of the low-GC3 gene was undetectable; again , pairs of independent clones using each construct revealed similar expression levels . We next assessed mRNA abundance in the strains detailed above . Initially , we used a tubulin ( TUB ) fragment as a probe for the wild-type and human codon-optimised gLUC mRNAs , since the untranslated regions in these reporter constructs are from a TUB gene . This allowed us to use native TUB transcripts as a loading control and , since the gLUC gene is short , to identify gLUC transcripts migrating below native TUB transcripts on the blot . Analysis of gLUC RNA extracted from the T . brucei strains shown in Figure 1A revealed increased expression from the GC3-enriched gene ( Figure 2A ) . These results indicated that mRNA expression was also increased by GC3 codons . To further test the link between GC3 codons and mRNA abundance , we assessed expression of the synthetic gLUC and GFP genes described above ( Figure 1C , D ) . In this case , we engineered a unique λ-phage DNA sequence immediately downstream of each protein-coding sequence such that it would be included in the 3’-untranslated region ( Figure 2B ) . T . brucei strains expressing these reporter constructs were assessed on RNA blots using a unique ‘λ−probe’ ( Figure 2B-C ) and an equivalent blot was probed with TUB as a loading control . We observed higher mRNA expression from the high-GC3 relative to the medium-GC3 or low-GC3 gLUC and GFP genes; the difference was <2 fold for gLUC but approximately 2-fold for GFP and achieved statistical significance in the latter case . Notably , low-GC3 gLUC and GFP genes yielded slower-migrating transcripts ( Figure 2B ) , which may reflect longer poly ( A ) tails , as also observed for poorly translated mRNAs with a low proportion of optimal codons in other eukaryotes ( Lima et al . , 2017 ) . These results indicate that codon usage does indeed impact mRNA expression in T . brucei . While the range of values reaches almost 2-fold differential for mRNA ( Figure 2C ) , the range is substantially greater for protein ( Figure 1C ) pointing to an impact on translation that quantitatively exceeds the impact on mRNA levels . Having established that GC3-codons increase mRNA and protein expression , we next assessed codon usage at the genomic scale ( Figure 3A ) . The average Codon Adaptation Index ( CAI ) ( Sharp and Li , 1987 ) for the full set of >7 , 000 T . brucei genes is 0 . 712 ( Figure 3A ) , with a maximal CAI of 0 . 883 for the α-tubulin genes . The reference set of highly expressed genes ( see Materials and methods ) has an average CAI of 0 . 782 ( Figure 3A ) and , consistent with the view that tandem amplification is another mechanism used by trypanosomatids to increase gene expression ( Horn , 2008 ) , there are many ‘tandem genes’ in this reference set . Importantly , the set of genes encoding detected proteins in a published proteome dataset ( Urbaniak et al . , 2012 ) , had a substantially higher average CAI ( 0 . 724 ) , relative to the proteins that were not detected ( 0 . 701 , Figure 3A ) . Just over 500 genes encoding abundant proteins detected in a Leishmania mexicana proteome study were also enriched for high CAI values ( Paape et al . , 2008 ) . When we assessed CAI in relation to location on trypanosomatid chromosomes; we found that many genes immediately adjacent to divergent or convergent polycistronic transcription initiation or termination regions displayed relatively low CAI values ( Figure 3A , B ) ; the full cohort of 343 of these ‘strand-switch’ genes yielded an average CAI of 0 . 685 ( Figure 3A ) and T . brucei chromosome 3 is shown to illustrate ( Figure 3B ) . This reflects the previous observation of base-skew asymmetry flanking these regions ( Nilsson and Andersson , 2005 ) . We also assessed codon usage at the genomic scale in Leishmania major and similarly found that tandem paralogous genes and genes at transcription switch regions have inflated and deflated CAI values , respectively; L . major chromosome 1 is shown to illustrate this ( Figure 3B ) . An analysis of codon representation in the reference set of highly expressed genes revealed the expected over-representation of GC3 codons , which is particularly pronounced when there are only two alternative codons for an amino acid ( Figure 3C ) . Indeed , in the case of xxC/T wobble codons the xxC rather than xxT version of the tRNA is always present for these amino acids ( Asn , Asp , Cys , His , Phe , Tyr ) ( Horn , 2008 ) . In contrast , when there are 3–6 alternative codons , the xxT version of the wobble tRNA is typically present ( Ile , Ala , Pro , Thr , Val , Leu , Arg , SerTCT ) and both alternative wobble codons are often over-represented in these cases ( Ile , Ala , Leu , Arg , SerTCT ) . Variant Surface Glycoprotein ( VSG ) genes in T . brucei , which are involved in immune evasion and are subject to allelic exclusion , display unusual codon usage bias ( Alvarez et al . , 1994; Chanda et al . , 2007; Cross et al . , 2014 ) . Distinct from highly expressed and indeed other gene families , A3 codons are over-represented , while G3 codons , and notably almost all U3 codons , are under-represented in the VSG genes ( Figure 3C ) . Consistent with this , we find that the cohort of 188 intact VSGs display low CAI values ( Figure 3A , B ) , with an average of 0 . 672 . The single expressed VSG produces the most abundant mRNA and protein in bloodstream form T . brucei but a high rate of translation is probably not required due to a high rate of transcription by RNA pol-I , combined with remarkably slow turnover of both the VSG mRNA ( half-life ~4 . 5 hr ) ( Ehlers et al . , 1987 ) and protein ( half-life ~33 + /- 9 hr ) ( Seyfang et al . , 1990 ) ; the mean mRNA half-life in bloodstream-form T . brucei is 16 min ( Fadda et al . , 2014 ) . Indeed , sub-optimal translation , due to low G3 and high A3-usage in VSGs , may be necessary to prevent these multi-copy genes , distributed throughout the genome , from compromising VSG allelic exclusion and the immune evasion strategy of bloodstream form T . brucei . This particular virulence strategy could fail if constitutive pol-II transcription yielded sufficient VSG at the cell surface to simultaneously present multiple variants to the host . Since codon-pair bias is observed in other genomes ( Quax et al . , 2015 ) , and some pair-bias combinations have the potential to be in conflict with increased codon optimality , we also examined these features of T . brucei protein-coding sequences . This revealed enrichment of duplicated codons ( Figure 4A ) and other examples of pair-bias ( Figure 4B , C ) . Codon co-occurrence is thought to facilitate tRNA recharging , while the over-represented ( examples include xxC/Axx , xxU/Gxx , xxC/xxC and xxG/xxG ) and under-represented ( examples include xxA/Gxx , xxU/Axx , xxC/xxG , xxG/xxC ) pairs ( Figure 4B , C ) are thought to reflect more optimal tRNA interactions at the A and P sites that impact translation fidelity ( Quax et al . , 2015 ) . We are not aware of evidence supporting a direct impact of codon pair-bias on mRNA or protein expression , so these differences were not considered further here . We do note , however , that an over-representation of xxU/Gxx pairs , and under-representation of xxC/xxG and xxG/xxC pairs , limits the accumulation of ‘optimal’ GC3 codons . The results above suggested that it might be possible to predict relative mRNA and protein expression in trypanosomatids , based on protein coding-sequences alone . To address this question , we generated a new T . brucei transcriptome dataset . This dataset comprised three replicates , was of sufficient depth to provide >114 million mapped reads , and yielded pair-wise Pearson correlation coefficients of >0 . 999 ( Figure 5—figure supplement 1 ) . This depth of coverage and correspondence provides an excellent level of resolution for the T . brucei transcriptome . We analysed this dataset alongside a published T . brucei proteome dataset ( Urbaniak et al . , 2012 ) . An initial comparison of mRNA and peptide expression yielded a Pearson’s correlation coefficient of 0 . 57 ( Figure 5A , Supplementary file 1 , sheet 2 ) . Thus , as in other cell types ( Ly et al . , 2014 ) , mRNA abundance is predictive of protein abundance . This relationship displays the best fit with a logarithmic trend-line , whereas the best fit is with a linear trend-line when only 10% of the most highly expressed mRNAs are excluded from the analysis . This may reflect poor translation of highly abundant mRNA’s , as noted previously for transcripts encoding ribosomal proteins ( Antwi et al . , 2016 ) , but might equally reflect experimental bias due to limitations in the effective dynamic range of proteome analysis ( Urbaniak et al . , 2012 ) . Short genes produce relatively more mRNA in T . brucei ( Fadda et al . , 2014; Jha et al . , 2014 ) , so we also considered gene length for our analysis . Indeed , we found that gene length is inversely correlated with both mRNA expression in our dataset ( Figure 5B , Supplementary file 1 , sheet 2 ) and with protein expression ( Figure 5C , Supplementary file 1 , sheet 2 ) and that long genes tend to make less mRNA and protein . An analysis of the correspondence between codon usage and observed T . brucei mRNA expression ( n = 7225 ) yielded a Pearson correlation coefficient of 0 . 48 ( Figure 6A , Supplementary file 1 , sheet 2 ) . We found this quite remarkable , that a measure of mRNA abundance can be predicted based on the coding-sequence alone . Since shorter coding-sequences yield more mRNA ( Figure 5B ) , we derived a simple formula including a penalty for coding-sequence length [CAI - ( 0 . 03 x √coding sequence length in kbp ) ] . This penalty improved the correspondence between CAI-values and observed mRNA expression and yielded a correlation coefficient of 0 . 52 ( Figure 6B , Supplementary file 1 , sheet 2 ) ; using this approach , >70% of our estimates of relative expression are within 2-fold of the observed values ( Figure 6C ) . Our RNA-seq data above are derived from the bloodstream-form life-cycle stage of T . brucei . The expression of some mRNAs and proteins is developmentally regulated , however . We , therefore , analysed our previously published RNA-seq data from both bloodstream-form and insect-form cells ( Hutchinson et al . , 2016 ) . Using these data , minus those 2 . 8% of genes that differed >3 fold between stages ( 201 of 7191 genes ) , we once again observed correspondence between length-adjusted CAI values and RNA-seq expression data ( Figure 6—figure supplement 1A , B ) . Pearson correlation coefficients were 0 . 54 and 0 . 53 for the bloodstream-form ( Figure 6—figure supplement 1A ) and insect-form data ( Figure 6—figure supplement 1B ) , respectively ( improved by 5 . 5% and 4 . 6% by the length-adjustment ) . RNA-seq data from insect-form cells generated by another research group ( Christiano et al . , 2017 ) also revealed correspondence with length-adjusted CAI values ( Figure 6—figure supplement 1C ) . The Pearson correlation coefficient was 0 . 5 in this case ( improved by 3% by the length-adjustment ) . Thus , our findings apply to multiple T . brucei life cycle stages and to independently generated datasets . A similar analysis of the impact of codon usage on the proteome of T . brucei ( n = 2315 ) also revealed a substantially improved correlation when we applied the same penalty for coding-sequence length . In this case , the correlation between CAI values and observed protein expression yielded a correlation coefficient of 0 . 55 ( Figure 6D , Supplementary file 1 , sheet 2 ) , which increased to 0 . 65 following the length-adjustment ( Figure 6E , Supplementary file 1 , sheet 2 ) ; again , >70% of our estimates of relative expression are within 2-fold of the observed values ( Figure 6F ) . Predictions of 'steady-state' abundance for individual proteins are prone to 'under-sampling' errors in proteome data . Since many proteins function in multi-component complexes or share similar functions with related proteins , we analysed cohorts of genes encoding components of protein complexes or related functions ( Supplementary file 1 , sheet 3 ) . For these cohorts ( n = 23 ) , codon usage-based predictions , when compared to observed measures , yielded a striking Pearson correlation coefficient of 0 . 84 ( Figure 7 ) . This further reinforces the view that codon usage bias plays a major role in controlling protein expression . Using our relative expression estimates , we derived values for numbers of both mRNA and protein molecules per bloodstream form cell ( Supplementary File 1 , sheet 2 ) . These are based on 19 , 000 non-VSG mRNAs ( Haanstra et al . , 2008 ) and 100 million non-VSG proteins per cell , similar to estimates for yeast ( Milo , 2013 ) . We also derived adjusted protein estimates based on the mRNA levels we observe , which provides a proxy adjustment for gene copy-number . There are limitations of course; including expected inflated values for cell cycle specific proteins or developmentally regulated proteins . Indeed , we see examples of such complexes in Figure 7; the abundance of the cell cycle specific kinetochore complex ( Akiyoshi and Gull , 2014 ) and the bloodstream down-regulated editosome complex ( Stuart et al . , 2005 ) are 3 . 7-fold and 2 . 7-fold below the predictions based on the trend-line , respectively ( Figure 7 ) . Notably , the analyses shown in Figure 7 also appears to reflect a particular technical limitation typically seen when using peptide-based proteomic strategies , the under-representation of membrane–associated proteins; the abundance of the o-sector of the V-ATPase ( Baker et al . , 2015 ) is 2 . 5-fold below the prediction on this plot . Despite the limitations , we believe our estimates of protein expression levels and stoichiometry will be of value for expression of recombinant proteins in trypanosomatids ( Fritsche et al . , 2007 ) and for understanding trypanosomatid biology . Favoured or optimal codons may increase protein expression by increasing the translation rate . To explore this possibility , we compared our predicted expression values with published ribosome profiling data ( Vasquez et al . , 2014 ) and observed correspondence with measures of translation efficiency ( Figure 8A ) ; the correlation coefficient was 0 . 35 ( improved by 6 . 8% by the length-adjustment ) . Optimal codons may similarly increase mRNA abundance , if ribosome interactions stabilise mRNAs . Indeed , a comparison between CAI values and published T . brucei mRNA half-life data ( Fadda et al . , 2014 ) revealed a correspondence ( Figure 8B ) ; the correlation coefficient was 0 . 37 . Notably , correspondence decreased ( 3 . 2% in bloodstream-form cells; or improved only 0 . 3% in insect-stage cells ) when the length-adjustment was applied in this case . This supports the view that codon usage bias controls mRNA stability in a length-independent manner . We suggest that optimal codons increase ribosome occupancy and translation , which in turn protects mRNA from ( length-independent ) decapping/deadenylation , thereby also increasing mRNA stability . We suspected that codon usage bias also has a substantial impact on both mRNA and protein expression in other trypanosomatids . We , therefore , analysed previously published RNA-seq and proteomic data from Trypanosoma vivax , another African trypanosome that causes disease in cattle and other livestock ( Jackson et al . , 2015 ) . Using these data , we observed correspondence between length-adjusted CAI values and both RNA-seq ( Figure 9A ) and proteomic ( Figure 9B ) expression data . The Pearson correlation coefficients were 0 . 47 and 0 . 63 , respectively ( improved by 2 . 1% and 10% by the length-adjustment ) . This was also the case ( Figure 9C ) when we analysed RNA-seq data from Leishmania mexicana ( Fiebig et al . , 2015 ) , a trypanosomatid parasite thought to have diverged from the African trypanosome lineage >120 million years ago ( Harkins et al . , 2016 ) ; the correlation coefficient was 0 . 43 ( improved by 2 . 4% by the length-adjustment ) . These findings support the view that codon usage bias , and gene length , control mRNA and protein expression in many , if not all , trypanosomatids . Taken together , our results illustrate a remarkable ability to predict mRNA and protein expression and abundance at a transcriptomic and proteomic scale based on protein-coding sequences alone . Indeed , relative expression can now be predicted for the full complement of proteins , including thousands of proteins that fail to register in current proteome datasets ( Supplementary file 1 , sheet 2 ) . To facilitate further predictions we provide extended sets of CAI values for T . brucei ( Supplementary file 1 , sheet 4 ) , T . vivax ( Supplementary file 1 , sheet 5 ) and L . mexicana ( Supplementary file 1 , sheet 6 ) . Cells have evolved mechanisms to coordinate the expression of a large number of proteins of widely differing abundance . Codon usage contributes to gene expression control but it can be challenging to investigate the impact of codon usage bias at a genomic and proteomic scale in most eukaryotes because gene expression control operates at many levels , through transcription control in particular . Trypanosomatids display constitutive RNA pol-II transcription and , therefore , rely heavily upon post-transcriptional control , consequently presenting excellent models for the study of mRNA stability and translation control . It was previously suggested that codon usage and corresponding tRNA gene dosage plays an important role in gene expression control in trypanosomatids ( Horn , 2008 ) but experimental tests of this translational selection hypothesis were lacking . We now show that codon usage has a major impact on gene expression in T . brucei and that both relative mRNA and protein expression can be predicted based on coding sequence alone . Using two distinct reporter genes in bloodstream form T . brucei , we show that codon usage has the capacity to control relative protein abundance over several orders of magnitude . Although to a lesser extent , mRNA abundance was also dependent upon codon usage . Indeed , codon optimality increases mRNA stability in yeast ( Presnyak et al . , 2015 ) , by reducing Dhh1p-dependent decay ( Radhakrishnan et al . , 2016 ) . Anticipating a major impact of codon bias on gene expression in T . brucei , we predicted relative mRNA and protein expression at a global scale using coding sequences alone . We found that codon usage was indeed predictive of observed mRNA and protein expression and report correspondence between the relative abundance of thousands of native proteins and thousands of native mRNA transcripts . We find it quite remarkable that our predictions of steady-state mRNA and protein expression , based only on coding sequences , correlate with observed abundance data; even taking no account of controls mediated by untranslated sequences or differences in protein turnover . Under-sampling is one of the major current challenges in terms of determining relative protein abundance in cells and tissues ( Ly et al . , 2014 ) . With this in mind , we predicted protein expression for cohorts of genes that encode components of protein complexes or related functions . We found a remarkable correlation between our sequence-based predictions and observed protein levels . Notably , for proteins that form multi-component complexes , protein turnover is likely dependent upon complex assembly , with destabilisation of unassembled subunits . In addition to codon usage , gene length also has an impact on mRNA and protein expression , and this is proposed to be due to increased turnover of longer transcripts ( Fadda et al . , 2014; Jha et al . , 2014 ) . This is entirely feasible because individual lesions in mRNA will reduce abundance and lesions may be proportional to length . When we applied a gene-length adjustment , this failed to improve the correspondence between codon usage and mRNA half-life , consistent with the idea suggested previously ( Fadda et al . , 2014 ) that transcript length has a greater impact on co-transcriptional degradation rather than on co-translational degradation . Perhaps the former is dominated by ( length-dependent ) endonuclease activity while the latter is dominated by ( length-independent ) exonuclease activity , following decapping and deadenylation . Ribosome profiling , nascent chain profiling ( Chadani et al . , 2016 ) and analysis of newly synthesised proteins ( Zur et al . , 2016 ) are powerful methods for probing translational efficiency and translational pausing . The former approach has been applied to T . brucei ( Vasquez et al . , 2014 ) . Our analysis of codon usage in relation to these translation efficiency data , as well as mRNA half-life data ( Fadda et al . , 2014 ) , also from T . brucei , suggested that optimal codons increase ribosome occupancy and mRNA stability . Indeed , possible crosstalk between translation and mRNA decay in T . brucei was noted previously ( Antwi et al . , 2016 ) . We propose that optimal-codons reduce mRNA turnover in trypanosomes by increasing the rate of translation , increasing the association with actively translating ribosomes , and thereby protecting mRNA from degradation . In this scenario , slow translation or stalling would promote mRNA degradation ( Deana and Belasco , 2005 ) . Indeed , mRNA decay is known to be co-translational in yeast ( Hu et al . , 2009; Pelechano et al . , 2015 ) . Mechanisms by which optimal codons impact ribosome translation rates remain a subject of intense debate ( Hanson and Coller , 2018 ) . Our findings in T . brucei suggest a more pronounced impact on translation that also impacts mRNA stability . For reference , estimates for highly expressed genes in yeast yield initiation rates of once every two seconds and elongation rates up to twenty amino acids per second ( Futcher et al . , 1999 ) . In T . brucei , it is estimated that there are 125 , 000 ribosomes and 19 , 000 non-VSG mRNA molecules per bloodstream form cell ( Haanstra et al . , 2008 ) . Notably , a number of studies point to translation initiation as the rate-limiting step in protein synthesis ( Jackson et al . , 2010 ) and ribosome occupancy was found to be similar on optimised and non-optimised constructs in yeast ( Presnyak et al . , 2015 ) . In simple terms , a constant translation initiation rate combined with variable elongation rates would be expected to yield rapidly translated mRNAs loaded with relatively few ribosomes and slowly translated mRNAs loaded with many , potentially stalled , ribosomes . In contrast to this scenario , however , our analysis indicates correspondence between optimal codons and ribosome occupancy . This suggests a coupling between elongation and initiation in trypanosomes which could be explained by slow ribosomes affecting trailing ribosomes and decreasing the initiation rate ( Chu et al . , 2014; Hanson and Coller , 2018 ) or by ribosome recycling on the same mRNA [see Supplementary information S5 in ( Jackson et al . , 2010 ) . It should be possible in the future to further refine mathematical models of gene expression and also to better understand the relative impact and evolution of codon usage in trypanosomatids . Improvements in quantitative proteomics , measurements of protein synthesis and turnover rates , a better understanding of the impact of untranslated sequences on translation and mRNA turnover and an improved understanding of the relative contributions of individual codons to translation elongation and initiation will all help make this possible . We conclude that protein-coding sequence composition contributes to differences in steady-state mRNA and protein abundance in trypanosomatids . The ability to predict relative expression , based on protein-coding sequences alone , indicates that codon usage makes a major contribution to the transcriptomes and proteomes of these cells . Our results support a model whereby translation rate is increased by optimal codons , resulting in reduced mRNA turnover . Bloodstream form T . brucei , Lister 427 ( MITat 1 . 2 ) , clone 221a cells and 2T1 cells ( Alsford et al . , 2005 ) have been confirmed mycoplasma free and their identity was confirmed by RNA-seq ( see below ) . These cells were grown in HMI-11 medium and genetically manipulated using electroporation as described ( Alsford et al . , 2005 ) . Initially , humanopt and WT Gaussia princeps gLUC genes were introduced into a T . brucei inducible expression vector . Humanopt gLUC ( EU372000 ) was amplified using the EUluc5 and EUluc3 primers ( Key Resources Table; PstI/EcoRI restriction sites are underlined ) and cloned between T . brucei TUB mRNA processing signal sequences . The entire cassette was then cloned ( BamHI/Bsp120I ) in pRPaiSL ( Alsford and Horn , 2008 ) . WT gLUC ( AY015993 . 1 ) was amplified from pUC19GLuc ( Prolume Ltd/NanoLight ) using the WTluc5 and WTluc3 primers ( Key Resources Table; XhoI/HindIII restriction sites are underlined ) and used to replace the humanopt gLUC gene in the pRPaiSL backbone . This also allowed us to remove an N-terminal signal-sequence and to add a C-terminal peroxisome-targeting signal ( SKL ) to both proteins ( see italics in the Key Resources Table in EUluc3 and WTluc3 ) . A 443 bp λ-DNA fragment was then amplified using the Lambda5 and Lambda3 primers ( Key Resources Table; HindIII/EcoRI restriction sites are underlined ) and cloned immediately downstream of the gLUC coding sequence . Once we had established that gLUC expression was not toxic , we moved the expression cassette ( BamHI/Bsp120I ) to pRPa ( Acc65I/Bsp120I ) , a vector for constitutive expression ( Alsford and Horn , 2008 ) . New gLUC and GFP genes were designed using a previously published codon usage table ( Horn , 2008 ) . These genes were synthesised ( Genscript ) and cloned ( XhoI/HindIII ) in the constitutive pRPaλ expression vector . All of these constructs were linearized with AscI prior to electroporation and insertion at the ribosomal DNA spacer locus on chromosome 2 . Tetracycline was applied at 1 µg . ml−1 for 24 hr to induce expression in pRPai-based strains . Total cell extracts equivalent to 1 × 106 cells were separated on SDS-polyacrylamide gels and subject to either standard or LICOR western blotting analysis according to the manufacturers’ instructions . For standard western blots , duplicate gels were generated and one was stained with Coomassie and the other was used to produce the nitrocellulose blot . Blots were blocked in 5% milk in TBST and washes were performed in TBST ( 0 . 05% Tween ) . Blots were then probed with 1/1000 α-gLUC primary antibody ( New England Biolabs ) and 1/2000 α-rabbit secondary antibody ( Bio-Rad ) . For LICOR blots , blocking was performed in 50 mM Tris , pH 7 . 4 , 0 . 15 M NaCl , 0 . 25% BSA , 0 . 05% Tween , 2% fish skin gelatine ( Sigma , UK ) and washes were performed in TBST ( α-gLUC ) or PBST ( α-GFP ) . Nitrocellulose blots were incubated with 1/1000 α-gLUC or 1/5000 α-GFP ( Life Technologies ) and 1/20 α-tubulin ( kind gift from Keith Gull ) , followed by 1/10 , 000 α-mouse and 1/10 , 000 α-rabbit IR Dye antibodies ( LICOR ) . Lysates for gLUC assays were prepared by adding 20 µl of 1 × luciferase cell lysis buffer ( New England Biolabs ) to 2 × 106 pelleted cells and these were directly used to perform BioLux Gaussia luciferase assays ( New England Biolabs ) according to the manufacturers’ instructions and using a TopCount plate-reader with white-walled plates . One-way ANOVA tests were carried out in GraphPad Prism ( version 7 ) . For RNA extraction , 5 × 107 bloodstream form T . brucei were collected and RNA prepared using the Qaigen RNeasy kit , according to the manufacturer’s instructions . The TUB and λ probes were generated by PCR , using the TUBF and TUBR and λF and λR primers , respectively ( see Key Resources Table ) . RNA blotting was carried out according to standard protocols and signal quantification was carried out using a phosphorimager . One-way ANOVA tests were carried out in GraphPad Prism ( version 7 ) . For RNA-seq , RNA samples were prepared from three independent bloodstream form T . brucei cultures; batches of 5 × 107 cells at a density of 1–2 × 106/ ml . RNA was prepared using the Qiagen RNeasy kit , according to the manufacturer’s instructions . RNA samples were assessed using the TapeStation Platform and the QuBit platform and all samples were normalised for an input of 1 μg . Polyadenylated transcripts were enriched using oligo d ( T ) 25 magnetic beads ( NEB ) . We used the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina platforms which is optimised for ~200 bp inserts . Following PCR enrichment and purification each sample library was again QC checked using TapeStation and Qubit . 300 ng of each sample were pooled to generate a pooled library of 233 nM . The pooled library was diluted to 4 nM before preparation for running on the NextSeq platform . Samples were loaded at 1 . 6 pM on a 2 × 75 bp High Throughput Flowcell achieving a %>=Q30 , 81 . 9G , 93 . 0% . Reads were mapped to the T . brucei 927 reference genome ( TriTrypDB , RRID:SCR_007043 ) . Bowtie 2-mapping ( Langmead and Salzberg , 2012 ) was as previously described ( Glover et al . , 2016 ) with the parameters --very-sensitive --no-discordant --phred33 ( no mismatches allowed ) . Approximately 115 million reads were aligned . Alignment files were manipulated using SAMtools ( Li et al . , 2009 ) . Read counts were normalised using edgeR ( Robinson et al . , 2010 ) . The highly expressed T . brucei reference set comprised those genes ( n = 50 ) encoding proteins of >250 amino acids that registered >25 unique peptides and >75% coverage in the proteome dataset ( Urbaniak et al . , 2012 ) . We used an online CAI calculator ( umbc . edu/codon/cai/cais . php ) to generate CAI values . Codon pair bias was analysed using the ANACONDA ( bioinformatics . ua . pt/software/anaconda/ ) software package ( Moura et al . , 2011 ) . The T . brucei proteome analysis set comprised those non-redundant genes encoding proteins that registered >3 unique peptides in the proteome dataset ( Urbaniak et al . , 2012 ) . The T . brucei transcriptome analysis set comprised all non-redundant genes; both sets excluded genes transcribed by RNA polymerase I . The highly expressed T . vivax reference set comprised those genes encoding proteins that registered >20 unique peptides and >15 peptides/kbp in the proteome dataset; data merged from the three life cycle stages ( Jackson et al . , 2015 ) . The T . vivax proteome analysis set comprised those non-redundant genes encoding proteins that registered >3 unique peptides in the proteome dataset and that were also orthologous to genes in the T . brucei set . The T . vivax transcriptome analysis set comprised genes that were orthologous to those in the T . brucei set and that also registered >10 tags . The highly expressed L . mexicana reference set comprised orthologues of the equivalent T . brucei set that also registered >100 FPKM; data merged from the three life cycle stages ( Fiebig et al . , 2015 ) . The L . mexicana transcriptome analysis set comprised genes that were orthologous to those in the T . brucei set and that also registered >5 FPKM . Excel functions were used to derive Pearson correlation coefficients and trend-lines .	Genes are made up of DNA , which contains all the information and instruction needed to build an organism . This information is stored as a genetic code consisting of four bases: adenine ( A ) , cytosine ( C ) , guanine ( G ) and thymine ( T ) . The order of these bases and their different combinations serves as a blueprint for making thousands of different proteins and to assemble living cells . Proteins are vital for almost every process that keeps cells alive . They are made up of chains of small molecules called amino acids , via a process that takes several steps . First , the code from the gene needs to be copied into so-called messenger RNA molecules ( mRNA ) . Cells then decode these mRNAs by reading the bases in groups of three , called codons . Most codons specify an amino acid , though some mark the start and end point of the protein . Cells need different proteins in varying amounts . In fact , a given protein may be present in only a few copies in one cell , while another may be present in millions of copies , and in both cases the amount of protein present often needs to be tightly controlled . One way to control the number of copies of a protein that are made from a mRNA is via the specific choice of codons . Most amino acids are encoded by more than one codon and using a different codon can lead the mRNA to yield more or less protein without actually changing the protein that is made . Researchers are particularly interested in protein production in parasites known as trypanosomes , which cause a range of diseases in humans and other animals . Although most aspects of protein production are similar in humans and in trypanosomes , the differences indicate that codon choice would be particularly important in the parasites . However , until now it was not known how codon choice contributes to the control of the copy number of proteins in these parasites . To test this , Jeacock , Faria and Horn engineered fully synthetic genes for fast or slow protein production and inserted them into the parasite’s genome . The genes generated the proteins at the predicted levels . An analysis of several datasets of protein abundance showed that the number of proteins in related parasites could also be predicted by using only the DNA code of the gene . Jeacock et al . could even predict the number of mRNAs in a parasite cell . A separate study by Nascimento , Kelly et al . also showed that codon identity predicts mRNA levels in trypanosomes . These findings will help researchers to predict mRNA and protein levels in these parasites using simple computer models . A better understanding of gene expression in these parasites will also improve our knowledge of the fundamental aspects of protein production in relation to evolution , biotechnology and disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2018	Codon usage bias controls mRNA and protein abundance in trypanosomatids
Heme metabolism is central to blood-stage infection by the malaria parasite Plasmodium falciparum . Parasites retain a heme biosynthesis pathway but do not require its activity during infection of heme-rich erythrocytes , where they can scavenge host heme to meet metabolic needs . Nevertheless , heme biosynthesis in parasite-infected erythrocytes can be potently stimulated by exogenous 5-aminolevulinic acid ( ALA ) , resulting in accumulation of the phototoxic intermediate protoporphyrin IX ( PPIX ) . Here we use photodynamic imaging , mass spectrometry , parasite gene disruption , and chemical probes to reveal that vestigial host enzymes in the cytoplasm of Plasmodium-infected erythrocytes contribute to ALA-stimulated heme biosynthesis and that ALA uptake depends on parasite-established permeability pathways . We show that PPIX accumulation in infected erythrocytes can be harnessed for antimalarial chemotherapy using luminol-based chemiluminescence and combinatorial stimulation by low-dose artemisinin to photoactivate PPIX to produce cytotoxic reactive oxygen . This photodynamic strategy has the advantage of exploiting host enzymes refractory to resistance-conferring mutations . Malaria is an ancient and deadly scourge of humanity , with hundreds of millions of people infected by Plasmodium malaria parasites and more than 600 , 000 deaths each year ( WHO , 2014 ) . Substantial progress has been made in reducing the global malaria burden , due in part to the success of artemisinin-based combination drug therapies ( ACTs ) . Recent identification of artemisinin-tolerant parasites in southeast Asia , however , has raised concerns that the broad potency of ACTs against all parasite strains may be waning , which could lead to a resurgence in malaria deaths ( Dondorp et al . , 2009; Ariey et al . , 2014 ) . These concerns motivate continued efforts to deepen understanding of basic parasite biology in order to identify new drug targets and facilitate development of novel therapies . Heme is a ubiquitous biological cofactor required by nearly all organisms to carry out diverse redox biochemistry ( Ponka , 1999 ) . Heme metabolism is a dominant feature during Plasmodium infection of erythrocytes , the most heme-rich cell in the human body and the stage of parasite development that causes all clinical symptoms of malaria . Parasites sequester and biomineralize the copious heme released during large-scale hemoglobin digestion in their acidic food vacuole ( van Dooren et al . , 2012; Sigala and Goldberg , 2014 ) ; they also require heme as a metabolic cofactor for cytochrome-mediated electron transfer within mitochondria ( Painter et al . , 2007; van Dooren et al . , 2012; Sigala and Goldberg , 2014 ) . Sequencing of the Plasmodium falciparum genome over a decade ago and subsequent studies have revealed that parasites encode and express all of the conserved enzymes for a complete heme biosynthesis pathway ( Figure 1A ) , but the role and properties of this pathway have been the subject of considerable confusion and uncertainty ( Gardner et al . , 2002; van Dooren et al . , 2012; Sigala and Goldberg , 2014 ) . This pathway was originally proposed to be essential for blood-stage parasite development and thus a potential drug target ( Surolia and Padmanaban , 1992 ) , as host heme was thought to be inaccessible for parasite utilization in mitochondria . Recent studies , however , have clarified that de novo heme synthesis is not required by intraerythrocytic parasites and therefore is unlikely to be a viable target for therapeutic inhibition ( Nagaraj et al . , 2013; Ke et al . , 2014 ) . The parasite-encoded ferrochelatase ( FC ) can be knocked out to ablate heme biosynthesis but parasite growth is unaffected , suggesting that parasites can scavenge host heme to satisfy metabolic requirements during blood-stage infection . 10 . 7554/eLife . 09143 . 003Figure 1 . Exogenous ALA stimulates protoporphyrin IX ( PPIX ) biosynthesis in Plasmodium-infected erythrocytes . ( A ) Schematic depiction of the heme biosynthesis pathway in Plasmodium falciparum parasites . Enzymes abbreviations are in red and pathway substrates and intermediates are in black: ALAS ( aminolevulinic acid synthase ) , ALAD ( aminolevulinic acid dehydratase ) , PBGD ( porphobilinogen deaminase ) , UROS ( uroporphyrinogen synthase ) , UROD ( uroporphyrinogen decarboxylase ) , CPO ( coproporphyrinogen III oxidase ) , PPO ( PPIX oxidase ) , FC ( ferrochelatase ) . For simplicity , all organelles are depicted with single membranes . Succinylacetone ( SA ) inhibits ALAD . ( B ) Bright field and fluorescence microscopy images of untreated and 200 µM aminolevulinic acid ( ALA ) -treated parasites . Fluorescence images were acquired with a Zeiss filter set 43 HE ( excitation 537–562 nm , emission 570–640 nm ) . ( C ) Growth of asynchronous 3D7 parasites in the presence or absence of 200 µM ALA and 50 µM SA , with 2-min exposures to white light on an overhead projector on days 0–2 . Parasitemia ( percentage of total erythrocytes infected with parasites ) as a function of time was fit with an exponential growth equation . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 00310 . 7554/eLife . 09143 . 004Figure 1—figure supplement 1 . Fluorescence excitation and emission spectrum of PPIX in aqueous buffer . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 00410 . 7554/eLife . 09143 . 005Figure 1—figure supplement 2 . Transmission electron microscopy images of untreated and 500 µM ALA-treated P . falciparum-infected erythrocytes after light exposure on an overhead projector light box . The white arrow identifies the digestive vacuole . Scale bar equals 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 00510 . 7554/eLife . 09143 . 006Figure 1—figure supplement 3 . ALA dose dependence of photosensitivity by blood-stage Plasmodium parasites . Asynchronous parasites were cultured in the indicated concentration of ALA and exposed to 2 min light daily . Parasitemia was measured after 48 hr and normalized to a reference culture grown without ALA . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 00610 . 7554/eLife . 09143 . 007Figure 1—figure supplement 4 . Giemsa-stained blood smear of P . falciparum culture after 3 days of treatment with 200 µM ALA and 2-min daily light exposure on an overhead projector light box . Black arrows identify dead parasite remnants . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 007 Here we use chemical and physical probes to decipher the role of upstream enzymes in heme biosynthesis by parasite-infected erythrocytes . Contrary to simple predictions , genetic disruption of the parasite porphobilinogen deaminase ( PBGD ) and coproporphyrinogen III oxidase ( CPO ) had no effect on the ability of Plasmodium-infected erythrocytes to convert exogenous aminolevulinic acid ( ALA ) into heme , as monitored by mass spectrometry and photodynamic imaging of porphyrin intermediates . Biochemical fractionation revealed that vestigial host enzymes remaining in the erythrocyte cytoplasm contribute to ALA-stimulated heme biosynthesis , explaining why disruption of parasite enzymes had no effect on biosynthetic flux . We used small-molecule probes to show that ALA uptake by erythrocytes , which are normally impermeable to ALA , requires the parasite nutrient-acquisition pathways established after invasion . Finally , we show that latent host enzyme activity can be exploited for antimalarial photodynamic therapy ( PDT ) using ( 1 ) ALA to stimulate production of the phototoxic intermediate protoporphyrin IX ( PPIX ) , ( 2 ) luminol-based chemiluminescence to photoactivate PPIX cytotoxicity within infected erythrocytes , and ( 3 ) luminol stimulation by combinatorial low-dose artemisinin . Production of 5-ALA by ALA synthase ( ALAS ) is the regulated step in heme biosynthesis . Exogenous ALA bypasses this step and stimulates biosynthetic flux , leading to accumulation of the final intermediate , PPIX , since conversion of PPIX to heme by FC becomes rate limiting ( Kennedy et al . , 1990 ) . PPIX is fluorescent ( Figure 1—figure supplement 1 ) and its cellular accumulation can be directly visualized by fluorescence microscopy using a phenomenon called ‘photodynamic imaging’ that has been exploited for visualizing cancerous tumor boundaries during surgical resection ( Celli et al . , 2010 ) . Prior work indicated that exogenous ALA stimulates PPIX production in P . falciparum parasites ( Smith and Kain , 2004; Ke et al . , 2014 ) . We therefore posited that ALA treatment could serve as a probe of heme biosynthesis activity in Plasmodium-infected erythrocytes by enabling direct visualization of PPIX production and the cellular consequences of light activation . Untreated parasites imaged on an epifluorescence microscope display only background auto-fluorescence from hemozoin crystals in the parasite digestive vacuole ( Figure 1B ) . Parasites grown in 200 µM ALA , however , display bright red fluorescence distributed throughout the infected erythrocyte , as expected for accumulation of PPIX ( Figure 1B ) . PPIX is also known to generate cytotoxic reactive oxygen species when photo-illuminated , owing to formation of an excited triplet state that can undergo collisional energy transfer with ground state triplet oxygen to produce highly reactive singlet oxygen ( Celli et al . , 2010 ) . The ability to kill ALA-treated cells by light activation of accumulating PPIX has been exploited to selectively target cancerous tumors via a strategy known as 'photodynamic therapy' ( Kennedy et al . , 1990; Celli et al . , 2010 ) . The cytotoxic effects of light-activating PPIX were readily apparent by monitoring the motion of hemozoin crystals that dynamically tumble within the digestive vacuole of individual parasites . Although the origin and physiological significance of this motion remain unknown , hemozoin dynamics serve as an internal biomarker of food vacuole integrity and parasite viability ( Sigala and Goldberg , 2014 ) . Hemozoin motion in untreated parasites was unaffected by exposure to light that overlaps the excitation wavelength of PPIX ( Videos 1 , 2 ) . In contrast , the hemozoin dynamics in ALA-treated parasites were rapidly ablated by light ( Videos 3 , 4 ) . Ultra-structural analysis by electron microscopy revealed a loss of electron density within the digestive vacuole of ALA-treated and illuminated parasites ( Figure 1—figure supplement 2 ) , suggesting disruption of the food vacuole membrane and outward diffusion of the vacuolar protein contents . These changes are consistent with a photodynamic mechanism of PPIX-mediated generation of reactive oxygen species that cause pleiotropic cytotoxic damage , including loss of lipid bilayer integrity . 10 . 7554/eLife . 09143 . 008Video 1 . Hemozoin dynamics in the digestive vacuole of untreated parasites prior to light exposure . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 00810 . 7554/eLife . 09143 . 009Video 2 . Hemozoin dynamics in the digestive vacuole of untreated parasites after light exposure . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 00910 . 7554/eLife . 09143 . 010Video 3 . Hemozoin dynamics in the digestive vacuole of 200 µM ALA-treated parasites prior to light exposure . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 01010 . 7554/eLife . 09143 . 011Video 4 . Hemozoin dynamics in the digestive vacuole of 200 µM ALA-treated parasites after light exposure . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 011 To test the effects of these changes on bulk parasite culture , we monitored the growth of untreated vs ALA-treated parasites over several days with 2-min daily exposures to broad-wavelength white light on an overhead projector light box . Whereas untreated parasite cultures grew normally in the presence of light , the growth of ALA-treated cultures was strongly attenuated by light ( Figure 1C and Figure 1—figure supplement 3 ) , consistent with a prior report ( Smith and Kain , 2004 ) . Microscopic examination by blood smear revealed that ALA-treated cultures predominantly contained karyolytic and pyknotic parasites ( Figure 1—figure supplement 4 ) , indicative of widespread cell death . The photosensitivity of parasite growth in ALA was fully rescued by 50 µM succinylacetone ( SA ) , an ALA dehydratase ( ALAD ) inhibitor shown in previous work to substantially reduce PPIX biosynthesis from ALA in parasite-infected erythrocytes ( Nagaraj et al . , 2013; Ke et al . , 2014 ) . This chemical rescue confirmed that parasite photosensitivity in ALA requires biosynthetic conversion of ALA to PPIX . The constituent enzymes of the parasite's heme biosynthesis pathway are distributed between three sub-cellular compartments: the mitochondrion , the cytoplasm , and the chloroplast-like but non-photosynthetic apicoplast organelle ( Figure 1A ) ( van Dooren et al . , 2012 ) . Prior work indicated that the parasite FC gene could be knocked out , resulting in a complete ablation of de novo heme synthesis but with no effect on parasite growth ( Nagaraj et al . , 2013; Ke et al . , 2014 ) . This observation provided strong evidence that parasites do not require heme biosynthesis for growth in erythrocytes , where they can presumably scavenge abundant host heme to meet metabolic needs . We therefore reasoned that upstream enzymes in the parasite-encoded heme biosynthesis pathway should also be amenable to genetic disruption and that such ablations would block production of PPIX from exogenous ALA ( Figure 1A ) and thus prevent parasite photosensitivity . We successfully disrupted the parasite genes encoding the apicoplast-targed PBGD ( Figure 2—figure supplement 1 ) and the cytosolic CPO ( Figure 2—figure supplement 2 ) using single-crossover homologous recombination to truncate the open reading frame for each gene . Southern blot and polymerase chain reaction ( PCR ) analysis confirmed correct integration and gene disruption in clonal parasite lines ( Figure 2—figure supplements 3 , 4 ) . Contrary to simple predictions , however , these genetic disruptions had no effect on the ability of parasite-infected erythrocytes to incorporate 13C-labelled ALA into heme , PPIX , or coproporphyrinogen III ( CPP ) ( Figure 2A ) , as monitored by a previously developed LC-MS/MS assay ( Ke et al . , 2014 ) , and clonal growth of both parasite lines remained fully photosensitive in ALA ( Figure 2B , C ) . 10 . 7554/eLife . 09143 . 012Figure 2 . Heme biosynthesis in infected erythrocytes persists despite disruption of parasite enzymes or the apicoplast . ( A ) Liquid chromatography-tandem mass spectrometry ( LC-MS/MS ) detection of 13C-labelled heme , PPIX , and coproporphyrinogen III ( CPP ) in parasites grown in 200 µM 5-[13C4]-ALA . Parasites were extracted in dimethyl sulfoxide ( DMSO ) , supplemented with deuteroporphyrin as an internal standard , and analyzed by LC-MS/MS . Integrated analyte peak areas were normalized to PPIX in each sample . RBC: uninfected red blood cells , WT: parental clone 3D7 , IPP/dox: isopentenyl pyrophosphate/doxycycline-treated 3D7 parasites . ( B , C ) Growth of asynchronous ∆PBGD ( B ) and ∆CPO ( C ) 3D7 parasites in the presence or absence of 200 µM ALA , with 2-min light exposures on an overhead projector on days 0–2 . WT growth was fit to an exponential equation . ( D ) Bright field and fluorescence images of live 3D7 parasites expressing ALAD-GFP from a plasmid before or after 2-week treatment with IPP and doxycycline . ( E ) Growth of asynchronous IPP/doxycycline-treated parasites in the presence or absence of 200 µM ALA and 50 µM SA , with 2-min light exposures on an overhead projector on days 0–2 . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 01210 . 7554/eLife . 09143 . 013Figure 2—figure supplement 1 . Immunofluorescence microscopy ( IFM ) images of fixed 3D7 parasites expressing full-length PBGD tagged at its endogenous locus with C-terminal GFP confirm targeting of the native protein to the parasite apicoplast . Parasites were stained with αGFP and αACP ( ACP , apicoplast marker ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 01310 . 7554/eLife . 09143 . 014Figure 2—figure supplement 2 . Fluorescence microscopy images of live 3D7 parasites episomally expressing full-length CPO with a C-terminal GFP tag confirm protein localization to the parasite cytoplasm . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 01410 . 7554/eLife . 09143 . 015Figure 2—figure supplement 3 . Disruption of the P . falciparum PBGD gene ( PF3D7_1209600 ) by single-crossover homologous recombination . ( A ) Schematic depiction of the PBGD gene locus before and after incorporation of the donor plasmid via single-crossover recombination . Red arrows indicate the primers used to selectively polymerase chain reaction ( PCR ) amplify either the intact WT or disrupted ∆PBGD locus . SacI and BglII were used to digest the donor plasmid and genomic DNA ( gDNA ) of WT and ∆PBGD parasites to give the expected fragment sizes indicated in parentheses for hybridization to a PBGD bp 601–960 oligonucleotide probe . ( B ) PCR analysis of gDNA from WT and ∆PBGD clones using the primers indicated in ( A ) to selectively amplify the 1 . 3 kb WT gene or the 960 bp truncated gene . ( C ) Southern blot analysis of gDNA from 3D7 WT and ∆PBGD clonal parasites after digestion with SacI and BglII , hybridization with a PBGD bp 601–960 oligonucleotide probe , and detection using the Amersham AlkPhos Labeling and CDP-Star chemiluminescent reagents . ( D ) Homology model of P . falciparum PBGD ( human PBGD template structure 3EQ1 ) to indicate the portion of the protein retained ( cyan ) or lost ( green ) by single-crossover truncation . The deleted sequence comprises half of the active site binding pocket , identified by the dipyrromethane cofactor shown in red . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 01510 . 7554/eLife . 09143 . 016Figure 2—figure supplement 4 . Disruption of the P . falciparum CPO gene ( PF3D7_1142400 ) by single-crossover homologous recombination . ( A ) Schematic depiction of the CPO gene locus before and after incorporation of the donor plasmid via single-crossover recombination . Red arrows indicate the primers used to selectively PCR-amplify either the intact WT or disrupted ∆CPO locus . XhoI and KpnI were used to digest the donor plasmid and gDNA of WT and ∆CPO parasites to give the expected fragment sizes indicated in parentheses for hybridization to a CPO bp 610–1080 oligonucleotide probe . ( B ) PCR analysis of gDNA from WT and ∆CPO clones using the primers indicated in ( A ) to selectively amplify the 1 . 6 kb WT gene or the 1 . 1 kb truncated gene . ( C ) Southern blot analysis of gDNA from 3D7 WT and ∆CPO clonal parasites after digestion with XhoI and KpnI , hybridization with a CPO bp 610–1080 oligonucleotide probe , and detection using the Amersham AlkPhos Labeling and CDP-Star chemiluminescent reagents . ( D ) Homology model of P . falciparum CPO ( human CPO template structure 2AEX ) to indicate the portion of the protein retained ( blue ) or lost ( green ) by single-crossover truncation . The deleted sequence comprises half of the active site binding pocket , identified by the citrate molecule shown in red . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 01610 . 7554/eLife . 09143 . 017Figure 2—figure supplement 5 . IFM images of fixed 3D7 parasites expressing full-length ALAD tagged at its endogenous locus with C-terminal GFP confirm targeting to the parasite apicoplast . Parasites were stained with αGFP and αACP ( ACP , apicoplast marker ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 01710 . 7554/eLife . 09143 . 018Figure 2—figure supplement 6 . PCR analysis of genomic DNA from untreated WT parasites or parasites cultures ≥7 days in 1 µM doxycycline and 200 µM IPP . ACPL refers to the 385 bp leader sequence ( with introns ) of the nuclear-encoded ACP . Rps8 ( 387 bp ) and ORF91 ( 276 ) are two genes encoded by the Plasmodium apicoplast genome . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 01810 . 7554/eLife . 09143 . 019Figure 2—figure supplement 7 . Western blot analysis of parasites episomally expressing ALAD-GFP and cultured 7 days in IPP and doxycycline . Blots confirm disrupted proteolytic processing of the ALAD leader sequence that results in retarded migration by SDS-PAGE relative to parasites cultured in normal conditions . Extracts from WT and IPP/doxycycline parasites were loaded separately in lanes 2 and 5 , respectively , and were loaded together in lanes 3 and 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 019 To further probe the functional contributions to heme biosynthesis by the parasite apicoplast and its constituent enzymes , we stably disrupted this organelle by treating parasites with doxycycline and isopentenyl pyrophosphate ( IPP ) . Doxycycline inhibits the prokaryotic translation machinery of the apicoplast , which blocks replication of the small apicoplast genome , prevents organelle segregation , and results in unviable apicoplast-deficient parasite progeny ( Dahl et al . , 2006 ) . The lethal effects of doxycycline , however , can be chemically rescued by IPP , which enables parasites to make essential isoprenoids despite apicoplast disruption . This rescue leads to a stable metabolic state in which parasites lack the intact organelle such that nuclear-encoded proteins ordinarily targeted to the apicoplast become stranded in small vesicles ( Yeh and DeRisi , 2011 ) . We confirmed apicoplast loss in doxycycline and IPP-treated parasites by using microscopy to verify disrupted apicoplast targeting of the nuclear-encoded ALAD enzyme ( Figure 2D and Figure 2—figure supplement 5 ) , PCR to confirm loss of the small apicoplast genome ( Figure 2—figure supplement 6 ) , and western blot ( WB ) to visualize disruption of post-translational processing of ALAD ( Figure 2—figure supplement 7 ) . Despite apicoplast disruption , these parasites retained their capacity for heme biosynthesis ( Figure 2A ) and remained fully photosensitive in ALA ( Figure 2E ) . Together with the gene disruption results above , these observations strongly suggested that parasite-infected erythrocytes have a parallel biosynthetic pathway that bypasses functional disruption of the parasite enzymes targeted to the apicoplast and cytoplasm . The heme biosynthesis pathway in human cells is distributed between mitochondria and the cytoplasm , with four cytosolic enzymes that correspond to the four apicoplast-targeted enzymes in Plasmodium parasites ( Ponka , 1997 , 1999; van Dooren et al . , 2012 ) . During human erythropoiesis , precursor reticulocytes carry out prolific heme biosynthesis , but this activity is absent in mature erythrocytes due to loss of mitochondria and their constituent heme biosynthesis enzymes , including ALAS and FC ( Ponka , 1997 ) . Proteomic studies have confirmed that mature erythrocytes retain the cytosolic enzymes ( ALAD , PBGD , uroporphyrinogen synthase [UROS] , and uroporphyrinogen decarboxylase [UROD] ) ( Pasini et al . , 2006; D'alessandro et al . , 2010 ) , but this vestigial pathway is ordinarily quiescent due to the lack of ALA synthesis or uptake in erythrocytes . We hypothesized that exogenous ALA taken up by parasite-infected erythrocytes might stimulate the latent activity of these cytosolic human enzymes , resulting in biosynthetic flux through this truncated host pathway and production of downstream tetrapyrrole intermediates that could be taken up by the parasite via hemoglobin import or other mechanisms and converted into heme within the parasite mitochondrion . The cytosolic human enzymes remaining in the mature erythrocyte would be expected to produce CPP from ALA . Our observation that disruption of the parasite CPO had no effect on conversion of ALA into heme by intraerythrocytic parasites suggests that a soluble fraction of human CPO , which is thought to be predominantly targeted to the mitochondrial intermembrane space ( IMS ) ( Ponka , 1997; van Dooren et al . , 2012 ) , persists in the erythrocyte cytoplasm after maturation of precursor reticulocytes and mitochondrial loss . Indeed , other mitochondrial IMS proteins , such as cytochrome c , are known to partition into the cytoplasm under certain conditions ( Liu et al . , 1996; Soltys and Gupta , 2000 ) . CPO catalyzes production of protoporphyrinogen IX , which in the oxygen-rich environment of erythrocytes can spontaneously oxidize to form PPIX ( Wang et al . , 2008 ) . In support of this model , we noted that red porphyrin fluorescence in ALA-treated infected erythrocytes was not limited to the parasite but was detectable throughout the erythrocyte cytoplasm ( Figure 1B ) , as expected for host enzyme activity and production of PPIX in this compartment . To directly test the model that enzymes remaining in the erythrocyte cytoplasm could catalyze PPIX biosynthesis from ALA , we permeabilized uninfected erythrocytes using the detergent saponin , clarified lysates by centrifugation followed by sterile filtration , and then used LC-MS/MS to monitor heme and porphyrin biosynthesis from 5-[13C4]-ALA added to the filtered lysate supernatant . We detected formation of 13C-labelled PPIX and CPP but not heme , and this biosynthetic activity was fully blocked by SA ( Figure 3A ) . These observations provide direct support for our model that erythrocytes retain a vestigial and partial biosynthesis pathway capable of converting exogenous ALA into PPIX but are unable to convert PPIX to heme due to lack of mitochondria and FC . 10 . 7554/eLife . 09143 . 020Figure 3 . Erythrocytes have latent porphyrin biosynthesis activity that requires exogenous ALA and parasite permeability mechanisms to enable ALA uptake . ( A ) LC-MS/MS detection of 13C-labelled PPIX and CPP in erythrocyte lysate supernatants incubated with 200 µM 5-[13C4]-ALA without or with 50 µM SA . Erythrocytes were lysed in 0 . 04% saponin , centrifuged at 25 , 000×g for 60 min , and 0 . 2 µM syringe filtered prior to ALA addition . ( B ) Bright field and fluorescence ( Zeiss filter set 43 HE ) images of uninfected erythrocytes incubated in 500 µM ALA before or after electroporation . ( C ) Bright-field and fluorescence images of parasites cultured in 500 µM ALA with normal ( +TMP ) or blocked ( −TMP ) establishment of parasite permeability pathways in the erythrocyte membrane . Infected erythrocyte permeability was modulated using a 3D7 parasite line expressing HSP101 tagged at its endogenous locus with a TMP-dependent destabilization domain ( Beck et al . , 2014 ) . TMP was maintained or washed out from synchronous schizont-stage parasites , which were allowed to rupture and invade new erythrocytes . 500 µM ALA was added to both cultures after invasion , and parasites were imaged 8 hr later . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 02010 . 7554/eLife . 09143 . 021Figure 3—figure supplement 1 . Furosemide blocks ALA uptake and PPIX biosynthesis in parasite-infected erythrocytes . Asynchronous wild-type 3D7 parasites were incubated in the absence or presence of 100 µM furosemide for 1 hr to block nutrient acquisition pathways , followed by addition of 500 µM ALA and further incubation for 8 hr . Parasites were then imaged by live microscopy on the bright field or red fluorescence ( Zeiss filter set 43 HE ) channels . In lower panel , parasite nuclei ( blue ) were visualized with Hoechst DNA stain . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 021 In contrast to parasite-infected erythrocytes , uninfected erythrocytes showed no detectable porphyrin fluorescence in the presence of ALA ( Figure 3B ) , consistent with reports that the erythrocyte membrane has a low permeability to amino acids ( Ginsburg et al . , 1985; Kirk et al . , 1994; Desai et al . , 2000 ) . Electroporation of uninfected erythrocytes in the presence of ALA , however , resulted in robust intracellular porphyrin fluorescence ( Figure 3B ) , supporting our model that host enzymes have latent biosynthetic activity that requires a mechanism for ALA uptake across the normally impermeable erythrocyte membrane . Upon invasion , Plasmodium parasites export hundreds of effector proteins into host erythrocytes . These proteins dramatically alter the architecture of the infected erythrocyte and establish new permeability pathways ( NPPs ) that enhance host cell uptake of amino acids and other nutrients from host serum ( Spillman et al . , 2015 ) . We hypothesized that selective uptake of ALA by parasite-infected erythrocytes was mediated by NPP mechanisms . To test this model , we utilized a recently published parasite line in which the export of parasite proteins into host erythrocytes , including establishment of nutrient permeability pathways , can be conditionally regulated with the synthetic small-molecule ligand trimethoprim ( TMP ) ( Beck et al . , 2014 ) . In these parasites , protein export and NPP mechanisms are functional in the presence of TMP but are blocked in its absence . We maintained or washed out TMP from a synchronized culture of late schizonts , allowed parasites to rupture and reinvade new erythrocytes , and then incubated both sets of parasites in ALA for 8 hr . Whereas TMP-treated parasites with normal protein export and permeability displayed robust PPIX fluorescence indistinguishable from that of wild-type parasites , parasites incubated without TMP showed no detectable PPIX fluorescence ( Figure 3C ) . We observed similar results in wild-type parasites using the small-molecule drug furosemide , which blocks parasite-derived NPP mechanisms directly ( Figure 3—figure supplement 1 ) ( Staines et al . , 2004 ) . We conclude that ALA is selectively taken up by infected erythrocytes via parasite-dependent nutrient-acquisition pathways and metabolized to PPIX by vestigial host enzymes within the erythrocyte cytoplasm ( Figure 4 ) . 10 . 7554/eLife . 09143 . 022Figure 4 . Schematic depiction of ALA-uptake and porphyrin biosynthesis pathways in Plasmodium-infected erythrocytes . For simplicity , all membranes are depicted as single . Porphyrins synthesized in the infected erythrocyte cytoplasm from exogenous ALA may be transported across the parasite membrane via unspecified mechanisms or may be taken up via hemoglobin import mechanisms . SA inhibits ALAD and blocks porphyrin synthesis from ALA . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 02210 . 7554/eLife . 09143 . 023Figure 4—figure supplement 1 . Growth of parasites in 1 µM 5-[13C4]-ALA results in detectable biosynthesis of 13C-labeled PPIX . Parasites were grown overnight in 1 µM 5-[13C4]-ALA , permeabilized with saponin , extracted in DMSO , and analyzed by LC-MS/MS for 13C-labeled heme , PPIX , and CPP . Integrated peak areas measured for each analyze were normalized to that measured for the deuteroporphyrin internal standard added to each sample . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 023 Our observation that disruption of the parasite-encoded PBGD and CPO does not affect biosynthetic production of heme and PPIX from ALA suggests that host enzyme activity in the erythrocyte cytoplasm provides the dominant contribution to PPIX biosynthesis in ALA-treated intraerythrocytic parasites . To dissect these parallel pathways and test whether the parasite pathway alone , in the absence of host enzymes , can support heme biosynthesis from ALA , we fractionated parasite-infected erythrocytes using saponin to selectively permeabilize host cell membranes while leaving parasite membranes intact ( Figure 5A ) . Under these conditions , soluble erythrocyte proteins can be washed away to leave the intact live parasite natively embedded within the resulting erythrocyte ghost ( Figure 5B , C ) . Parasites treated in this fashion remain metabolically active for 5–6 hr or longer , retain a membrane potential , accumulate fluorescent dyes such as MitoTracker Red ( Figure 5B , C ) , and carry out DNA synthesis ( Izumo et al . , 1987; Cobbold et al . , 2011 ) . After saponin treatment and washout , we placed parasites back into culture medium containing 13C-ALA and incubated them overnight before extracting them for analysis by LC-MS/MS . We failed to detect biosynthesis of heme , PPIX , or CPP in fractionated asexual and sexual blood-stage parasites ( Figure 5D ) , suggesting that the apicoplast-localized portion of the parasite heme biosynthesis pathway is largely or completely inactive during blood-stage infection . 10 . 7554/eLife . 09143 . 024Figure 5 . Analysis of heme biosynthesis activity in parasite-infected erythrocytes after saponin permeabilization and culture in 13C-labelled ALA . ( A ) Parasite-infected erythrocytes were permeabilized in 0 . 02% saponin , washed to remove the erythrocyte cytoplasm , and placed back into culture medium containing 200 µM 5-[13C4]-ALA for 12 hr prior to DMSO extraction and analysis by LC-MS/MS . Bright field and fluorescence image of live ( B ) asexual trophozoite and ( C ) stage IV sexual gametocyte treated with 0 . 02% saponin and stained with 20 nM MitoTracker Red . ( D ) LC-MS/MS quantification of 13C-labelled heme , PPIX , and CPP in DMSO extracts of intact WT 3D7 asexual parasites , saponin-released asexual parasites , and saponin-released gametocytes cultured overnight in 200 µM 5-[13C4]-ALA . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 024 To test this conclusion within undisrupted parasites , we created a transgenic parasite line expressing an apicoplast-targeted version of the uroporphyrinogen III methyltransferase ( cobA ) protein from Propionibacterium freudenreichii to serve as a biosensor of heme biosynthesis pathway activity within the parasite apicoplast . The cobA protein catalyzes conversion of the heme biosynthesis intermediate uroporphyrinogen III into the red fluorescent products sirohydrochlorin and trimethylpyrrocorphin ( Figure 6A ) and has been shown to function when heterologously expressed in bacteria ( Figure 6B ) , yeast , and mammalian cells ( Sattler et al . , 1995; Wildt and Deuschle , 1999 ) . We targeted a cobA-green fluorescent protein ( GFP ) fusion protein to the apicoplast in ALA-treated parasites ( Figure 6C ) but were unable to detect any red fluorescence in this organelle indicative of cobA-mediated conversion of uroporphyrinogen III to the fluorescent products . This result suggests that parasite enzymes targeted to this organelle are inactive in both asexual and sexual blood stages ( Figure 6D , E ) . We also noted that growth of ∆FC parasites ( Ke et al . , 2014 ) , which would be expected to accumulate PPIX during native pathway activity ( Figure 1A ) , was insensitive to light in the absence of ALA ( Figure 6—figure supplement 1 ) . These observations suggest that heme biosynthesis in blood-stage P . falciparum parasites is only operative when exogenous ALA is present to stimulate PPIX production by remnant host enzymes in the erythrocyte cytoplasm , with subsequent PPIX uptake and conversion to heme by FC in the parasite mitochondrion ( Figure 4 ) . 10 . 7554/eLife . 09143 . 025Figure 6 . Analysis of heme biosynthetic flux within the apicoplast of live parasites using the cobA biosensor . ( A ) Fluorescence excitation ( black ) and emission ( red ) spectra of clarified lysates from Escherichia coli bacteria expressing the cobA gene from Propionibacterium freudenreichii ( shermanii ) , showing the expected peaks for conversion of uroporphyrinogen III to sirohydrochlorin and trimethylpyrrocorphin . ( B ) Fluorescence microscopy images of live bacteria expressing the cobA gene , acquired on the bright field and red ( Zeiss filter set 43 HE ) channels . ( C ) Immunofluorescence ( IFM ) images of fixed 3D7 parasites episomally expressing an ACPleader-cobA-GFP fusion confirm targeting to the parasite apicoplast . Parasites were stained with αGFP and αACP ( acyl carrier protein [ACP] , apicoplast marker ) . The αACP antibody recognizes an epitope that is different from the ACP leader sequence . ( D ) Fluorescence microscopy images of live asexual parasites episomally expressing ACPleader-cobA-GFP without or with 500 µM exogenous ALA . ( E ) Fluorescence microscopy images of live stage III–IV sexual gametocytes episomally expressing ACPleader-cobA-GFP without or with 500 µM exogenous ALA . Fluorescence images in ( D ) and ( E ) were acquired on the GFP ( Zeiss filter set 38 ) and red ( Zeiss filter set 43 HE ) channels . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 02510 . 7554/eLife . 09143 . 026Figure 6—figure supplement 1 . Disruption of the Plasmodium FC gene in ∆FC parasites does not photosensitize parasites . WT or ∆FC D10 parasites were cultured under normal growth conditions ( without exogenous ALA ) with 2-min light exposures on an overhead projector light box on days 0–2 . Culture parasitemia as a function of time was fit to an exponential growth model . Both WT and ∆FC parasites had parasitemia doubling times of 1 . 1 days under these conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 026 Our results and those of prior studies suggested that stimulation of heme biosynthesis might be exploited for antimalarial PDT . Conventional PDT requires an external light source to photoactivate and kill ALA-stimulated cells ( Juzeniene et al . , 2007; Celli et al . , 2010; Wachowska et al . , 2011 ) . While such approaches can successfully target localized shallow-tissue tumors ( Celli et al . , 2010; Wachowska et al . , 2011 ) , they are impractical for treating malaria due to the dispersed nature of blood-stage infection , the sequestration of mature P . falciparum parasites along the vascular endothelium , and the requirement to illuminate every infected erythrocyte . To bypass the need for external light , chemiluminescence has been proposed as an alternative means to photoactivate the cytotoxicity of PPIX in ALA-stimulated cells . Trial studies in cancer cell lines have shown modest success using the small molecule luminol ( Laptev et al . , 2006; Chen et al . , 2012 ) , whose iron-activated chemiluminescence ( Rose and Waite , 2001 ) overlaps the absorbance spectrum of PPIX ( Figure 7—figure supplement 1 ) . Cancer cells tightly sequester iron , which may limit luminol activation in these environments . Plasmodium parasites , however , expose abundant iron during large-scale digestion of erythrocyte hemoglobin and liberation of heme . We therefore hypothesized that intraerythrocytic parasites might show heightened susceptibility to chemiluminescence-based PDT ( CL-PDT ) ( Figure 7A ) . 10 . 7554/eLife . 09143 . 027Figure 7 . Targeting blood-stage Plasmodium parasites by chemiluminescence-based photodynamic therapy ( CL-PDT ) . ( A ) Schematic depiction of a CL-PDT mechanism for targeting blood-stage malaria . ( B ) Effect of 100 µM ALA , 750 µM luminol ( lum ) , 50 µM 4-iodophenol ( ph ) , 50 µM SA and their combination ( all 0 . 25% DMSO ) on the growth of asynchronous 3D7 parasites . ( C ) Effect of 100 µM ALA , 750 µM luminol , 0 . 5 nM dihydroartemisinin ( DHA ) , 50 µM SA and their combination ( all 0 . 25% DMSO ) on the growth of asynchronous 3D7 parasites . Parasite media was changed twice daily , and parasitemia increases were fit to an exponential growth equation . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 02710 . 7554/eLife . 09143 . 028Figure 7—figure supplement 1 . Spectral compatibility of luminol and PPIX . ( A ) Overlay of normalized absorbance spectrum of PPIX ( black ) and chemiluminescence spectrum of luminol ( red ) . ( B ) Chemiluminescence of luminol ( solid red ) is attenuated in the presence of PPIX ( solid black ) , giving rise to a difference spectrum ( dashed black ) that is similar to the absorbance spectrum of PPIX in the spectral region that overlaps luminol chemiluminescence . Solutions contained 25 mM luminol in 100 mM NaOH ( aq ) , 0 . 5% ( wt/vol ) ammonium persulfate , and/or 70 µM PPIX . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 02810 . 7554/eLife . 09143 . 029Figure 7—figure supplement 2 . Effect of combinatorial ALA , luminol , and 4-iodophenol treatment on parasite growth . ( A ) Luminol concentrations as high as 750 µM ( in 0 . 2% DMSO ) have no effect on 3D7 parasite growth over 48 hr . ( B ) Effect of 750 µM luminol , 100 µM ALA , and their combination ( all in 0 . 2% DMSO ) on the growth of asynchronous 3D7 parasites . ( C ) Concentration dependence of growth inhibition of 3D7 parasites by 4-iodophenol ( in 0 . 2% DMSO ) over 48 hr . ( D ) Effect of 100 µM ALA , 50 µM 4-iodophenol , and their combination ( all in 0 . 2% DMSO ) on the growth of asynchronous 3D7 parasites . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 02910 . 7554/eLife . 09143 . 030Figure 7—figure supplement 3 . Giemsa-stained blood smear of 3D7 parasite culture after 3 days of treatment in 100 µM ALA , 750 µM luminol , and 50 µM 4-iodophenol . Arrows point to dead parasite remnants . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 03010 . 7554/eLife . 09143 . 031Figure 7—figure supplement 4 . Efficacy of CL-PDT with drug-resistant parasites . Effect of 100 µM ALA , 750 µM luminol , 50 µM 4-iodophenol , 50 µM SA ( all 0 . 2% DMSO ) and their combination on the growth of asynchronous ( A ) Dd2 parasites or ( B ) Kelch-13 I543T ( MRA-1241 ) parasites . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 03110 . 7554/eLife . 09143 . 032Figure 7—figure supplement 5 . Concentration dependence of growth inhibition of 3D7 parasites by DHA over 48 hr ( in 0 . 2% DMSO ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 03210 . 7554/eLife . 09143 . 033Figure 7—figure supplement 6 . Effect of 500 pM DHA , 100 µM ALA , and their combination ( all in 0 . 2% DMSO ) on the growth of asynchronous 3D7 parasites . DOI: http://dx . doi . org/10 . 7554/eLife . 09143 . 033 To test the efficacy of a CL-PDT strategy for targeting blood-stage P . falciparum , we incubated intraerythrocytic parasites with twice-daily media changes in combinatorial treatments of ALA , luminol , and 4-iodophenol , a small molecule that has been shown to enhance the intensity and duration of luminol chemiluminescence ( Thorpe et al . , 1985 ) . Parasites treated with optimized concentrations of each compound in single or double combination showed little effect on parasite growth ( Figure 7—figure supplement 2 ) . The combination of all three compounds , however , potently inhibited parasite growth ( Figure 7B ) , and microscopic examination by blood smear revealed widespread parasite death ( Figure 7—figure supplement 3 ) . The growth effects of this combination could be fully rescued by SA ( Figure 7B ) , supporting a photodynamic mechanism requiring PPIX biosynthesis from ALA and enhanced chemiluminescence by luminol . The development of parasite resistance to frontline antimalarial drugs continues to hamper malaria treatment and eradication efforts worldwide . To test whether a CL-PDT mechanism remains effective against parasites with diverse resistance to distinct drugs , we turned to studies with Dd2 parasites , which have multidrug resistance to antifolate and quinolone antibiotics ( Sidhu et al . , 2002 ) , and a clinical isolate containing a kelch-13 protein mutation that confers artemisinin tolerance ( Ariey et al . , 2014 ) . In both parasite lines , combination treatment with ALA , luminol , and 4-iodophenol potently ablated parasite growth ( Figure 7—figure supplement 4 ) . Artemisinin and its derivatives are current frontline drugs used for treatment of malaria , usually in combination with a partner drug . Artemisinin , whose endoperoxide moiety is reductively cleaved by intracellular iron to generate reactive oxygen radicals ( Meshnick , 2002 ) , has been shown to potently stimulate luminol chemiluminescence in vitro ( Green et al . , 1995 ) , suggesting the possibility of combining CL-PDT with artemisinin for antimalarial treatment . To test the efficacy of an artemisinin-enhanced CL-PDT strategy , we incubated drug-sensitive 3D7 parasites with twice-daily media changes in ALA , luminol , and sub-therapeutic doses ( 500 pM , 10% of EC50 , Figure 7—figure supplement 5 ) of dihydroartemisinin ( DHA ) . Single and double combinations had little effect ( Figure 7—figure supplement 6 ) , but all three compounds together potently ablated parasite growth , and growth inhibition could be fully rescued with SA ( Figure 7C ) . Malaria parasites , like nearly all other organisms , have a metabolic requirement for heme ( Painter et al . , 2007; van Dooren et al . , 2012; Sigala and Goldberg , 2014 ) . Despite access to abundant host heme during intraerythrocytic infection , parasites retain a complete heme biosynthesis pathway that was regarded for two decades as essential and a potential drug target ( Surolia and Padmanaban , 1992 ) . Our work and recent studies have now successfully disrupted four of the eight pathway enzymes , providing firm evidence that de novo heme synthesis is dispensable during blood-stage infection . These data strongly suggest that parasites have mechanisms to scavenge host heme to meet metabolic needs and clarify that heme biosynthesis is not a viable target for classical drug inhibition ( Nagaraj et al . , 2013; Ke et al . , 2014 ) . Our results further suggest that the parasite pathway is not active during intraerythrocytic infection . This inactivity may reflect the low availability of the succinyl-CoA precursor due to limited tricarboxylic acid cycle flux ( MacRae et al . , 2013 ) and additional regulatory mechanisms that suppress the activity of apicoplast-targeted enzymes during blood-stage parasite development , such as feedback inhibition by host heme or active-site inhibition by endogenous metabolites ( Park et al . , 2015 ) , . The functional quiescence of this specific biosynthetic pathway , despite expression of the constituent enzymes , may reflect a general survival strategy and adaptation by parasites to closely match their metabolic requirements with the nutritional availability of specific host environments , such as those previously described for parasite acquisition of fatty acids ( Yu et al . , 2008 ) . Indeed , prior studies suggest that the parasite heme synthesis pathway is required for development within the mosquito and human liver ( Nagaraj et al . , 2013; Ke et al . , 2014 ) . These stages involve host environments with lower heme availability compared with erythrocytes and in which parasite heme requirements appear to be elevated due to enhanced reliance on mitochondrial electron transport for ATP synthesis ( van Dooren et al . , 2012; Sigala and Goldberg , 2014; Sturm et al . , 2015 ) . A prior study reported that blood-stage P . falciparum parasites adopt distinct physiological states in vivo , including a state with heightened oxidative metabolism and mitochondrial activity that may arise during host starvation ( Daily et al . , 2007 ) . It remains a future challenge to test whether heme biosynthesis by the parasite-encoded pathway can be stimulated by host nutritional status during intraerythrocytic infection . Despite the dispensable nature and apparent inactivity of the parasite-encoded heme biosynthesis pathway , infected erythrocytes retain a paradoxical ability to synthesize heme from exogenous ALA . This biosynthetic activity requires the Plasmodium FC but not the upstream enzymes in the parasite pathway . Indeed , knock-out of the Plasmodium FC prevented conversion of PPIX to heme ( Ke et al . , 2014 ) , but our disruption of the parasite PBGD and CPO genes , as well as the entire apicoplast organelle , had no effect on ALA-stimulated heme synthesis ( Figure 2 ) . Our work resolves this paradox by identifying a latent contribution to heme biosynthesis in parasite-infected erythrocytes from vestigial host enzymes remaining in the erythrocyte cytoplasm . Since erythrocytes lack mitochondria , they are missing the initial and terminal pathway enzymes , are unable to synthesize ALA , and thus retain only a truncated and normally inactive heme synthesis pathway . Analytical studies have reported that human serum contains 0 . 1–0 . 2 µM ALA ( Zhang et al . , 2011 ) , but the low permeability of the erythrocyte membrane to ALA means that extracellular ALA is largely inaccessible to vestigial host enzymes within uninfected erythrocytes . Our study clarifies that NPP mechanisms of parasite-infected red blood cells enable efficient uptake of exogenous ALA . After uptake , ALA can be metabolized by vestigial human enzymes within the oxygen-rich erythrocyte cytoplasm to produce PPIX that can be converted to heme by FC within the parasite mitochondrion ( Figure 4 ) . Thus , the 0 . 2 µM ALA natively present in human serum may be sufficient to stimulate low-level heme biosynthesis within parasites in vivo . Indeed , we detect 13C-labeled PPIX in parasites grown in 1 µM 5-[13C4]-ALA in vitro ( Figure 4—figure supplementary 1 ) , supporting a model that serum ALA stimulates heme biosynthesis during malaria infection in vivo . Such activity , however , is not required to support blood-stage parasite growth . We note that prior work suggested a role for remnant host enzymes in heme biosynthesis by malaria parasites , but this prior proposal differed by positing that human enzymes such as ALAD were somehow imported and active within the parasite cytoplasm ( Padmanaban et al . , 2007 ) . The ability to photosensitize parasites with exogenous ALA and then kill them with light introduces exciting possibilities for developing new photodynamic treatment strategies , exemplifies how a deep understanding of fundamental parasite metabolism can be leveraged for designing novel therapies , and highlights that non-essential pathways can still serve as therapeutic targets . We have shown that luminol-based chemiluminescence , when stimulated by combinatorial delivery of low-dose 4-iodophenol or artemisinin , can circumvent the conventional PDT requirement for external light and potently ablate parasite growth ( Figure 7 ) . We note that ALA , luminol , and DHA have excellent toxicity profiles , favorable pharmacokinetic properties , and have each been used clinically ( Bissonnette et al . , 2001; Dalton et al . , 2002; Gordi and Lepist , 2004; Larkin and Gannicliffe , 2008; Wachowska et al . , 2011 ) . These results suggest the possibility of including ALA and luminol with therapeutic doses of artemisinin ( or its clinical derivatives ) as a novel form of artemisinin combination therapy for treating malaria . Multidrug-resistant parasites remain susceptible to this photodynamic strategy , which relies on host enzyme activity outside the genetic control of parasites and thus is refractory to the development of resistance-conferring mutations . This CL-PDT strategy may also be effective against other intraerythrocytic parasites , such as Babesia . As for any new therapy suggested by in vitro studies , additional in vivo experiments will be required to optimize treatment regimens for our proposed therapy and to confirm efficacy and safety . Finally , we note that this strategy of using artemisinin to stimulate intracellular light emission by luminol for ALA-based PDT may be applicable to treating deep-tissue cancers , for which poor accessibility to external light also limits current PDT approaches in development for cancer therapy ( Celli et al . , 2010 ) . Artemisinin has shown promising anti-cancer properties on its own ( Nakase et al . , 2008 ) , and thus its combination with ALA and luminol may provide potent synergy for cancer chemotherapy . All reagents were of the highest purity commercially available . SA , 5-ALA , 4-iodophenol , TMP , furosemide , saponin , IPP , doxycycline , luminol , and DHA were purchased from Sigma ( St . Louis , MO , USA ) . 5-[13C4]-ALA was purchased from Cambridge Isotope Laboratories , Inc ( Tewksbury , MA , USA ) . Images of live or fixed parasites were acquired on an Axio Imager M1 epifluorescence microscope ( Carl Zeiss Microimaging , Inc . ) equipped with a Hamamatsu ORCA-ER digital CCD camera and running Axiovision 4 . 8 software , as described previously ( Sigala et al . , 2012 ) . Live parasite nuclei were stained with 5 µM Hoechst 3342 added immediately prior to image acquisition . For photodynamic imaging studies , parasites were cultured in 200–500 µM ALA in the absence or presence of 50 µM SA for 6–12 hr prior to visualization . Hemozoin movement in parasite digestive vacuoles was imaged by acquiring 10–20 sequential frames at 1 s intervals on the bright-field channel . Images were cropped and superimposed in Adobe Photoshop , exported as video files with a 0 . 1 s frame delay , and looped for 15 seconds of playback . Immunofluorescence ( IFM ) images were acquired by fixing and staining parasites as previously described ( Tonkin et al . , 2004; Ponpuak et al . , 2007 ) . Electron microscopy images of parasites subjected to light or 500 µM ALA + light were obtained as previously described ( Beck et al . , 2014 ) . Uninfected erythrocytes were washed and resuspended in 1× cytomix containing 500 µM ALA , electroporated in a manner identical to parasite transfections ( see below ) , washed in phosphate-buffered saline ( PBS ) , incubated overnight at 37°C , and imaged as described above for live parasites . Images acquired on different channels for common samples were processed with identical brightness and contrast settings . Parasite growth was monitored by diluting asynchronous parasites to 0 . 5% parasitemia and allowing culture expansion with daily or twice-daily media changes . Parasitemia was measured daily by diluting 10 µl of each resuspended culture in 200 µl acridine orange ( 1 . 5 µg/ml ) and analyzing by flow cytometry , as previously described ( Muralidharan et al . , 2012 ) . To assess the light sensitivity of ALA-treated parasites , asynchronous parasites were cultured in 200 µM ALA in the absence or presence of 50 µM SA and subjected to 2 min daily exposures to broad-wavelength white light on an overhead projector . Daily parasitemia measurements were plotted as a function of time and fit to an exponential growth equation using GraphPad Prism 5 . 0 . For chemiluminescence experiments , asynchronous parasites were diluted to 0 . 5% parasitemia and incubated in ±100 µM ALA and ±50 µM SA for 8 hr . After 8 hr , parasite media was changed to also include 750 µM luminol , 50 µM 4-iodophenol , and 0 . 5 nM DHA in the indicated combinations . Parasite cultures were allowed to expand over 5 days , with twice-daily ( 7 am and 4 pm ) media changes in the indicated combinations . Parasitemia was measured daily on replicate samples , as indicated above . Experiments were performed using 3D7 ( drug sensitive ) parasites , Dd2 ( multidrug resistant ) parasites ( Sidhu et al . , 2002 ) , and a clinical isolate ( Ariey et al . , 2014 ) ( MRA-1241 ) bearing the I543T mutation in the Kelch-13 gene locus responsible for artemisinin tolerance . Exposure of parasite cultures to ambient light was minimized by changing media within darkened TC hoods and covering parasite culture dishes during brief ( 5–10 s ) transits to and from incubators . The effect of 100 µM ALA on parasite growth varied slightly between experiments , possibly due to differences between distinct batches of donated erythrocytes . For IC50 determinations , asynchronous parasites were diluted to 1% parasitemia and incubated with variable drug concentrations for 48 hr without media change . After 48 hr , parasitemia was determined in duplicate samples for each drug concentration , normalized to the parasitemia in the absence of drug , plotted as a function of the log of the drug concentration , and fit to a sigmoidal growth inhibition curve using GraphPad Prism 5 . 0 . Parasite culture and transfection were performed in Roswell Park Memorial Institute medium ( RPMI ) supplemented with Albumax , as previously described ( Sigala et al . , 2012 ) . Cloning was performed using either restriction endonuclease digestion and ligation or the In-Fusion system ( Clontech Mountain View , CA , USA ) . For episomal expression of P . falciparum ALA dehydratase ( PF3D7_1440300 ) and CPO ( PF3D7_1142400 ) fused to a C-terminal GFP tag ( ALAD-GFP and CPO-GFP ) , cDNA inserts encoding the complete ALAD and CPO genes ( exons only ) were reverse transcription ( RT ) -PCR amplified from total parasite RNA using the Superscript III system ( Life Technologies ) and primers CACTATAGAACTCGAGATGTTAAAATCAGATGTA GTGCTTTTATTGTATATAC and CTGCACCTGGCCTAGGTAGAGTTAATTCTATATT AAAATTATTATTTGAATTATCATC ( ALAD ) or ACGATTTTTTCTCGAGATGAAA GATGAGATAGCTCCTAATGAATATTTTAGAAATTTATG and CTGCACCTGGCCT AGGGTAGTCCACCCACTTTTTGGGATAC ( CPO ) . These were digested with XhoI/AvrII and ligated into the XhoI/AvrII sites of a digested pTEOE vector that was identical to a previously described pTyEOE vector ( Beck et al . , 2014 ) except that the pTEOE plasmid contained human dihydrofolate reductase ( DHFR ) in place of yeast dihydroorotate dehydrogenase ( yDHOD ) as the positive selection marker . Plasmid-based expression of the ALAD−GFP fusion was driven by the HSP86 promoter . This plasmid ( 50 µg ) was co-transfected into 3D7 parasites along with plasmid pHTH ( 10 µg ) for transient expression of the piggyback transposase that mediates integration of the pTEOE plasmid into the parasite genome ( Balu et al . , 2005 ) . Parasites were selected with 5 nM WR99210 . For episomal expression of P . freudenreichii ( shermanii ) uroporphyrinogen III methyltransferase ( cobA ) ( Genbank: CBL55989 . 1 ) targeted to the parasite apicoplast , the cobA gene was PCR amplified from plasmid pISA417 ( Sattler et al . , 1995 ) using the primers ACGATTTTTTCTCGAGATGACCACCACACTGTTGCCCGGCACTGTC and CTGCACCT GGCCTAGGGTGGTCGCTGGGCGCGCGATGG , digested with XhoI/AvrII and ligated into the XhoI/AvrII-cut pTEOE vector described above . An insert encoding the P . falciparum acyl carrier protein ( ACP ) leader sequence ( residues 1–60 ) with 5′- and 3′-XhoI sites , previously PCR amplified from parasite cDNA ( Sigala et al . , 2012 ) , was digested with XhoI and ligated into the XhoI-cut cobA/pTEOE vector to generate an in-frame ACPL-cobA-GFP fusion gene . This plasmid was co-transfected with pHTH into 3D7 parasites as described above . For disruption of the P . falciparum genes encoding PBGD ( PF3D7_1209600 ) and CPO ( PF3D7_1142400 ) , primer pairs ( PBGD: CACTATAGAACTCGAGGATCATAATAA TGATACATTATGTACTATTGGGACATCGTCC and CTGCACCTGGCCTAGGAA CTGCTATAATGCCTTGACCTAAGGCAGGATAAATCAGG; CPO: CACTATAG AACTCGAGTTTTTTCAAATATTTATAAAAACAGGAAAAAAGAAGAAAAAATA and CTGCACCTGGCCTAGGATAACATTTACAATCCTTATTATTATTATTATTATTGTTG ATGG ) were used to PCR amplify 360 bp and 471 bp sequences from the middle of the 1 . 3 kb PBGD and 1 . 6 kb CPO genes , respectively . These inserts were cloned by In-Fusion into the XhoI/AvrII sites of the pPM2GT vector ( Klemba et al . , 2004 ) , which encodes a C-terminal GFP tag after the AvrII site and also contains a human DHFR marker for positive selection with 5 nM WR99210 . This vector was further modified to introduce a 2A peptide sequence ( Straimer et al . , 2012 ) followed by the yeast DHOD sequence ( Ganesan et al . , 2011 ) , after the 3′ end of the GFP cassette , to enable positive selection for integration with the parasite DHOD inhibitor DSM1 . The yDHOD marker , however , was not used for selection in this study , and use of the GFP-2A-yDHOD/PM2GT vector for positive selection of plasmid integration into the genome will be described elsewhere . Plasmids ( 50 µg ) were transfected into 3D7 parasites by electroporation . Parasites were subjected to three rounds of positive selection with 5 nM WR99210 . After the first and second selections , cultures were maintained for 3 weeks in the absence of drug pressure prior to the subsequent round of positive selection . After the third round of positive selection , parasites were cloned by limiting dilution . Clonal parasites that had integrated the plasmid at the desired locus to disrupt the target genes by single cross-over homologous recombination were verified by PCR and Southern blot , as previously described ( Klemba et al . , 2004 ) , and retained for further analysis . To introduce a C-terminal GFP tag into the endogenous locus of the P . falciparum PBGD gene ( PF3D7_1209600 ) , primer pairs CACTATAGAACTCGAGGATCATAATAA TGATACATTATGTACTATTGGGACATCGTCC and CTGCACCTGGCCTAGGTTTATTA TTTAAAAGGTGCAATTCAGCCTCCGCTTTTATTTTG were used to PCR amplify the complete PBGD coding sequence . This insert was cloned by In-Fusion into the 2A-yDHOD/PM2GT vector described above and transfected into 3D7 parasites by electroporation as above . Stable integration was achieved after three rounds of positive selection in WR99210 , and clones were isolated by limiting dilution and verified by PCR and Southern blot . ∆FC D10 parasites , described in a prior study ( Ke et al . , 2014 ) , were obtained from Akhil Vaidya ( Drexel University ) and were cultured as described above , including 2-min daily exposures to broad-wavelength white light using an overhead projector light box . For apicoplast disruption experiments , parasites were cultured in 1 µM doxycycline and 200 µM IPP for 7–21 days ( Yeh and DeRisi , 2011 ) . After 7 days , genomic DNA was harvested and analyzed by PCR for the nuclear-encoded ACP gene ( primers ATGAAGATCTTATTACTTTGTATAATTTTTC and TTTTAAAGAGCTAGATGGGTTTTT ATTTTTTATC ) and apicoplast-encoded rps8 ( primers ATGATTATTAAATTTTTAAATAATG and TTACAAAATATAAAATAATAAAATACC ) and ORF91 ( primers ATGACTTT ATATTTAAATAAAAATTT and TTACATATTTTTTTTATTGAAGAACG ) genes to confirm selective loss of the apicoplast genome . 3D7 ∆PM1 ( plasmepsin 1 ) parasites ( Liu et al . , 2005 ) and 3D7 parasites expressing ACPleader-cobA-GFP from a pTEOE plasmid ( described above ) were synchronized using 5% sorbitol and cultured using gentamycin-free media . Gametocytogenesis was stress-induced in mid-trophozoite parasites ( 5–7% parasitemia ) by increasing hematocrit to 4% ( by removing half of the culture medium volume ) for 12 hr . Parasites were maintained in culture for 4–6 days , at which point mostly stage III–IV gametocytes were visible . ACPleader-cobA-GFP gametocytes were incubated overnight in the presence or absence of 500 µM ALA and imaged by live parasite microscopy as described above . For 13C-ALA labeling experiments , gametocytes were passaged over a magnetic column to remove uninfected erythrocytes and concentrate the gametocyte-infected cells , lysed in 0 . 02% saponin ( 0 . 2-µM filtered ) , placed back into culture medium containing 200 µM 5-[13C4]-ALA , and incubated overnight at 37°C . Parasites were then isolated by centrifugation , extracted , and analyzed by tandem mass spectroscopy as described below . Parasites were cultured in 200 µM 5-[13C4]-ALA for 12–24 hr , harvested by centrifugation , lysed in 0 . 05% cold saponin , washed in PBS , and extracted with dimethyl sulfoxide ( DMSO ) . Deuteroporphyrin was added as an internal standard and extracts were analyzed for 13C-labeled heme , PPIX , and coproporphyrin III ( CPPIII ) using a previously published LC-MS/MS assay ( Ke et al . , 2014 ) . Detection of CPPIII serves as a biomarker for detecting coproporphyrinogen III , which is rapidly oxidized to CPPIII upon exposure to air during cell lysis and extraction ( Wang et al . , 2008 ) . To assess whether parasites could synthesize heme from ALA in the absence of host enzymes in the erythrocyte cytoplasm , parasite-infected red blood cells were lysed with 0 . 05% saponin ( 0 . 2 µM filtered ) , spun briefly ( 3 min , 850×g ) to pellet , washed in PBS , and resupended in 12 ml of RPMI/albumax growth media supplemented with 200 µM 5-[13C4]-ALA . Parasites were incubated overnight at 37°C , harvested by centrifugation , and extracted and analyzed by tandem mass spectrometry as described above . To assess the heme biosynthesis capacity of the erythrocyte cytoplasm , 500 µl packed red blood cells ( described previously [Sigala et al . , 2012] ) were lysed in 20 ml of 0 . 04% saponin/PBS ( 0 . 2 µM filtered ) and centrifuged at 25 , 000×g for 60 min to pellet unlysed cells and any organelles . The superficial 15 ml of the lysate supernatant was removed and 0 . 2 µM filtered , supplemented with 200 µM 5-[13C4]-ALA in the absence or presence of 50 µM SA , incubated overnight at 37°C , and analyzed by tandem mass spectrometry as described above . The 3D7 parasite line expressing the endogenous HSP101 from its genomic locus and bearing a C-terminal Escherichia coli DHFR degradation domain ( DDD ) fusion tag was previously published ( Beck et al . , 2014 ) . To test whether ALA uptake by parasite-infected erythrocytes depends on parasite-established nutrient-acquisition pathways in the host cell membrane , we split a synchronous culture of HSP101-DDD into two populations of late schizonts ( purified over a magnetic column ) and washed out TMP from one of the two cultures . Both cultures ( ±TMP ) were permitted to lyse and reinvade fresh erythrocytes overnight . The following morning , parasites were placed in 200 µM ALA as early rings , incubated for 8 hr at 37°C , and imaged by live parasite microscopy as described above . Alternatively , asynchronous wild-type 3D7 parasites were incubated in the absence or presence of 100 µM furosemide for 1 hr to block nutrient-acquisition pathways , followed by the addition of 500 µM ALA and further incubation for 8 hr . Parasites were then imaged by live parasite microscopy as described above . The following antibodies were used for IFM and WB analysis at the indicated dilutions: goat anti-GFP ( Abcam 5450 ) ( IFM 1:500 , WB 1:1000 ) , rabbit anti-ACP ( Waller et al . , 2000; Ponpuak et al . , 2007 ) ( IFM: 1:500 ) , Alexa Fluor 488-conjugated chicken anti-goat ( IFM: 1:500 ) , Alexa Fluor 555-conjugated donkey anti-rabbit ( IFM: 1:500 ) , donkey anti-goat-IRDye 800 ( Licor Biosciences Geneva Lincoln , NE , USA ) ( WB: 1:10 , 000 ) . MitoTracker Red was used at 50 nM final concentration and was added to parasite cultures 30 min prior to harvest and analysis . Fluorescence excitation and emission spectra of pure PPIX in aqueous solution ( excitation 400 nm , emission 620 nm ) and of clarified lysates of E . coli bacteria expressing cobA ( excitation 360 nm , emission 600 nm ) and chemiluminescence spectra of luminol were obtained on a Cary Eclipse Fluorescence Spectrophotometer ( Varian , Inc . ) in either fluorescence or luminescence mode using 10-nm slit widths and a PMT detector voltage setting of 600 . Luminol solutions contained 25 mM luminol in 100 mM NaOH ( aq ) to which 0 . 1% ( wt/vol ) ammonium persulfate ( aq ) was added to stimulate light emission . Time-lapse bright field images of parasites were acquired with 1 s delays , composed into video files using Adobe Photoshop , played back with 0 . 1 s delays between frames , and looped for 15 sec . Bright field images were acquired before and after 4 s image acquisitions with the Zeiss filter sets 38 ( excitation 450–490 nm ) and 43 HE ( excitation 537–562 nm ) . Homology models of PBGD and CPO were generated using the SWISS–MODEL interface on the Expasy website ( http://swissmodel . expasy . org ) . The template structures for modeling were human PBGD ( PBD: 3EQ1 ) and human CPO ( PDB: 2AEX ) .	Malaria is a devastating infectious disease that is caused by single-celled parasites called Plasmodium that can live inside red blood cells . Several important proteins from these parasites require a small molecule called heme in order to work . The parasites have enzymes that make heme via a series of intermediate steps . However , it remains unclear exactly how important this ‘pathway’ of enzymes is for the parasite , and whether this pathway could be targeted by drugs to treat malaria . Now Sigala et al . have used a range of genetic and biochemical approaches to better understand the production of heme molecules in Plasmodium-infected red blood cells . First , several parasite genes that encode the enzymes used to make heme molecules were deleted . Unexpectedly , these gene deletions did not affect the ability of the infected blood cells to make heme . This result suggested that the parasites do not use their own pathway to produce heme while they are growing in the bloodstream . Sigala et al . then showed that human enzymes involved in making heme , most of which are also found within the infected red blood cells , are still active . These human enzymes provide a parallel pathway that can link up with the final parasite enzyme to generate heme . Further experiments revealed that the activity of the human enzymes could be strongly stimulated by providing the pathway with one of the building blocks used to make heme . This stimulation led to the build-up of an intermediate molecule called PPIX . This intermediate molecule can kill cells when it is exposed to light—a property that is called ‘phototoxicity’ . Sigala et al . showed that treating infected red blood cells with a new combination of non-toxic chemicals that emit light can activate PPIX in the bloodstream and can selectively kill the malaria parasites while leaving uninfected cells intact . These findings suggest a new treatment that could be effective against blood-stage malaria . Furthermore , the parasite will be unable to easily mutate to avoid the effects of this treatment because it relies on human proteins that are already made . Future work is now needed to optimize the dosage and the combination of drugs that could provide such a treatment .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "microbiology", "and", "infectious", "disease" ]	2015	Deconvoluting heme biosynthesis to target blood-stage malaria parasites
Transporters from bacteria to humans contain inverted repeat domains thought to arise evolutionarily from the fusion of smaller membrane protein genes . Association between these domains forms the functional unit that enables transporters to adopt distinct conformations necessary for function . The small multidrug resistance ( SMR ) family provides an ideal system to explore the role of mutations in altering conformational preference since transporters from this family consist of antiparallel dimers that resemble the inverted repeats present in larger transporters . Here , we show using NMR spectroscopy how a single conservative mutation introduced into an SMR dimer is sufficient to change the resting conformation and function in bacteria . These results underscore the dynamic energy landscape for transporters and demonstrate how conservative mutations can influence structure and function . Membrane transport proteins catalyze the movement of ions and molecules across the membrane by binding substrates on one side of the bilayer and undergoing conformational changes ( Jardetzky , 1966; Zhang et al . , 2016 ) . Structural and functional experiments have shown support for the alternating access model in which a transporter samples at least two different conformations that expose the substrate to the inside and outside environment of the cell membrane . The presence of inverted structural repeats within a single polypeptide chain is widespread in membrane protein transporters ( Shimizu et al . , 2004; Forrest and biology , 2013 ) and has been proposed as a mechanism to enable alternating access exchange ( Forrest et al . , 2008 ) . Proteins with structural repeats are thought to arise evolutionarily from the fusion of smaller membrane protein genes , such as three or four transmembrane ( TM ) domain transporters that associate into oligomers to achieve the functional state ( Xu et al . , 2014; Bay and Turner , 2009 ) . One class of these proteins is the small multidrug resistance ( SMR ) family ( Schuldiner , 2014; Schuldiner , 2009; Paulsen et al . , 1996 ) found in bacteria and archaea that contain four TM domains and assemble into antiparallel homodimers or heterodimers which resemble the inverted repeat structure in larger transporters . Homodimer transporters such as the multidrug efflux pump EmrE are able to insert into two opposing directions in the membrane ( i . e . dual topology ) whereas heterodimers are comprised of two genes that each insert into the membrane in a single orientation ( i . e . single topology ) . The latter form the paired SMR ( pSMR ) subfamily and are thought to arise from gene duplication and evolution of dual topology SMR genes ( Bay and Turner , 2009; Rapp et al . , 2007; Rapp et al . , 2006; Kolbusz et al . , 2010 ) . The similarity of the quaternary structure displayed by the SMR family with the pseudo two-fold symmetry seen in larger transporters suggests a role for gene duplication and fusion of smaller genes to give rise to transporters commonly found in higher level organisms ( Yan , 2013; Lolkema et al . , 2008 ) . The widespread prevalence of inverted repeats and divergent evolution from homo-oligomeric proteins suggests that asymmetry between functional subunits may have a fitness advantage , such as for evolving substrate specificity and direction of transport . This work aims to establish a structure-function correlation for how single mutations introduced into one subunit of the EmrE dimer – a minimal heterodimer – induce a shift in the conformational equilibrium between inward-open and outward-open states . EmrE is a native E . coli transporter that couples drug efflux of cationic/aromatic compounds with the proton gradient ( Schuldiner , 2009 ) . It has recently been shown that the coupling stoichiometry can proceed in a 1:1 or 2:1 proton:drug ratio , where the protons are bound by the glutamic acid residues at position 14 in the dimer ( Robinson et al . , 2017 ) . Since EmrE consists of an asymmetric and antiparallel homodimer , the relative energies between inward-open and outward-open states are identical for the wild-type protein . NMR spectroscopy has been a sensitive technique to probe this asymmetry by revealing a separate set of signals in the spectrum for monomers A and B within the dimer ( Cho et al . , 2014; Gayen et al . , 2013; Morrison et al . , 2012 ) . Determining the effect of a single mutation in the dimer would provide insight into the energy landscape by mimicking a primitive evolutionary event such as those that may have led to the pSMR subfamily . Previously , we discovered that conservative mutations located within or close to loop 2 of EmrE modulated the overall rate of conformational exchange between inward-open and outward-open states ( Gayen et al . , 2016 ) . Loop 2 is comprised of a short stretch of residues ( approximately Tyr53 , Ile54 , and Pro55 ) and adjoins TM2 and TM3 of EmrE ( Figure 1A ) . Loop 2 of monomer A is located on the open side of the transporter and does not make intermolecular contacts with monomer B in the tetraphenylphosphonium bound form , while loop 2 from monomer B is proximal to loop 3 of monomer A and near the bottom of the substrate cavity formed by TM1-3 ( Figure 1A ) ( Chen et al . , 2007; Vermaas et al . , 2018; Ovchinnikov et al . , 2018 ) . The differential contacts in the structure and distinct chemical shifts for each monomer suggest a minimal heterodimer might influence the conformational equilibrium . Indeed , our preliminary experiments showed that a heterodimer formed between wild-type and a mutant from loop 2 ( I54L ) or a mutant from the N-terminal region of TM3 ( I62L ) could induce a change in the conformational equilibrium between inward-open and outward-open states when bound to the high affinity compound tetraphenylphosphonium ( Gayen et al . , 2016 ) . While these measurements support the presence of differential contacts of loop 2 within monomers A and B , no biological significance was offered for these findings or whether the conformational equilibrium was perturbed for other essential forms of the transporter needed to accomplish transport ( i . e . , proton-bound states , apo states ) . Here , we report a conservative mutation ( L51I ) located in the C-terminal end of TM2 that has the greatest extent of altering the conformational equilibrium when paired with a wild-type monomer . The L51I/wild-type heterodimer was systematically investigated in different states of the transport cycle , including conditions where the essential anionic residue Glu14 was protonated or deprotonated and bound to drugs . These measurements revealed that both protonated and deprotonated drug-free forms of the L51I/wild-type heterodimer displayed a greater change in the conformational equilibrium relative to the drug-bound state . More importantly , growth inhibition experiments and efflux assays established that the conformational bias observed in vitro correlated with the functional output in E . coli . These findings suggest a non-negligible role for conservative mutations in the duplication-divergence evolutionary theory ( Ohno , 1967; Taylor and Raes , 2004 ) thought to govern the creation of larger membrane proteins from smaller genes . Prior to investigating whether mutations in one subunit of an EmrE dimer could induce a change in the conformational equilibrium in vitro , NMR experiments were carried out on homodimer samples of wild-type EmrE and mutants ( L51I , I62L ) . EmrE samples were prepared in magnetically aligned bicelles consisting of a 3 . 5/1 molar ratio of long chain ( O-14:0-PC ) and short chain ( 6:0-PC ) phospholipids at a pH value of 5 . 8 that corresponds to the Glu14 proton-bound form of EmrE . Solid-state NMR experiments were carried out using the PISEMA experiment ( Wu et al . , 1994 ) , since this technique gives the largest frequency separation between peaks corresponding to the two monomers within the asymmetric dimer ( Gayen et al . , 2013 ) . PISEMA spectra of homodimer samples of wild-type , L51I , and I62L selectively labeled with 15N-tyrosine are shown in Figure 1B . Each spectrum displayed 10 peaks , which confirms the presence of twice the number of peaks as tyrosine residues in the primary sequence of EmrE ( Tyr4 , Tyr6 , Tyr40 , Tyr53 , Tyr60 ) , and is consistent with previous observations ( Gayen et al . , 2013; Morrison et al . , 2012; Gayen et al . , 2016; Dutta et al . , 2014b ) . To analyze whether heterodimers might lead to a preferred conformation , we prepared mixtures of wild-type EmrE with the L51I or I62L mutant . In each experiment , only the wild-type or mutant was isotopically enriched with 15N-tyrosine and mixed with its partner protein at natural abundance . Using this approach , only one protein was NMR active , while the other was NMR silent . PISEMA spectra corresponding to wild-type/L51I or wild-type/I62L where wild-type was isotopically enriched showed a primary set of intense peaks that corresponded to monomer B ( Figure 2A ) . These signals were superimposable onto those peaks in the wild-type EmrE PISEMA spectrum ( Figure 2—figure supplement 1A ) . Next , we carried out the reverse experiments where the mutants were isotopically enriched and wild-type was NMR silent . In these spectra , the isotopically enriched mutant ( L51I or I62L ) in the mixed dimer also showed an intense set of peaks ( Figure 2B ) , yet these signals did not overlap with those of isotopically enriched wild-type in the mixed dimer ( Figure 2—figure supplement 1B ) . The peak positions did however match onto the corresponding peaks in the L51I and I62L homodimer PISEMA spectra ( Figure 2—figure supplement 1A ) . These data indicate that wild-type EmrE has a preference for monomer B in the heterodimer while L51I or I62L prefers monomer A for the proton-bound form of EmrE ( Figure 2C ) . It is important to note that the mixture of wild-type EmrE and the L51I or I62L mutant produces a statistical fraction of heterodimers and homodimers in the samples . While our analysis focused on the most intense set of signals , we were able to resolve a weaker set of signals that likely corresponded to homodimers at a lower contour level ( Figure 2—figure supplement 2 ) . Nevertheless , these observations demonstrate that a minimal heterodimer comprised of a single conservative mutation can strongly influence the conformational equilibrium and perturb the energy landscape of EmrE . All experiments in Figure 2 were performed on the proton-bound and drug-free form of EmrE . However , substrate transport requires EmrE to sample additional conformations within its catalytic cycle . To investigate how other states might impact the conformational equilibrium , we turned to solution NMR spectroscopy since it is a more sensitive technique to probe structure and conformational dynamics . To confirm the observations from PISEMA experiments in Figure 2 , we collected solution NMR experiments of mixed dimers under proton-bound sample conditions in 2H-14:0-PC/2H-6:0-PC ( 1/2 mol/mol ) isotropic lipid bicelles . The mixed samples of wild-type EmrE and L51I were prepared in a similar manner as for solid-state NMR experiments with two notable differences: ( 1 ) the NMR active protein was isotopically enriched with 13C at the Cδ position of isoleucine residues and ( 2 ) the ratio of isotopically labeled wild-type EmrE to L51I at natural abundance was 1/3 . The latter was possible due to the increased sensitivity of methyl detection in solution NMR spectroscopy . Figure 3A shows the 1H/13C HMQC spectrum of isotopically enriched wild-type EmrE in a mixed dimer sample with L51I . From this spectrum , we observed an intense set of signals corresponding to monomer B , which confirms the results of our solid-state NMR experiments in aligned lipid bicelles . Thus , under sample conditions used to collect solution and solid-state NMR spectra ( i . e . isotropic and aligned bicelles , respectively ) , the skewed conformational equilibrium is preserved . Next , we sought to investigate whether other states of the transport cycle , such as deprotonation of Glu14 and drug bound forms , may alter the conformational equilibrium . Previously , we reported pH induced chemical shift perturbations within EmrE’s NMR methyl spectrum that were centered at a pH value of 7 . 0 and corresponded to a biologically relevant pKa of Glu14 ( Gayen et al . , 2016 ) . Thus , we acquired a 1H/13C HMQC spectrum with a mixed sample of 13C enriched EmrE and natural abundance L51I at a pH value of 9 . 1 . Similar to the proton-bound dataset , this spectrum showed intense peaks that corresponded to the signals stemming from monomer B ( Figure 3B ) . This is in stark contrast to the spectrum of wild-type EmrE at high pH that showed approximately equal intensities of monomer A and B peaks ( Figure 3B ) . This result supports the conclusion that the skewed conformational equilibrium is preserved under conditions in which Glu14 is deprotonated . Note that we also observed monomer A signals at a lower contour level , which arises due to the statistical nature of mixing as discussed above ( Figure 3—figure supplement 1A ) . Using the relative peak intensities of A and B signals in the spectrum , we estimated the equilibrium population of wild-type EmrE to assume monomer B in the heterodimer at ~96% ( range of 90% to 99% based on standard deviation ) . This value corresponds to a free energy of ~1 . 8 kcal/mol induced by the L51I mutation in the heterodimer . To determine the effect of drug binding on the conformational equilibrium , we acquired NMR spectra for the mixed dimers in the presence of ethidium and tetraphenylphosphonium . These drug substrates are commonly used in the EmrE literature for measuring resistance , transport , and binding ( Yerushalmi et al . , 1995; Curnow et al . , 2004; Robinson et al . , 2018 ) . Addition of ethidium induced a significant amount of spectral broadening and ablation of peak intensities in both the wild-type sample and that of the mixed dimer where wild-type was isotopically enriched ( Figure 3C ) . These data suggest intermediate chemical exchange , which likely stems from motion corresponding to exchange between inward-open and outward-open states ( Cho et al . , 2014 ) . From these data , it is likely that the equilibrium is less skewed in the mixed sample upon ethidium binding . Namely , if the populations were maintained to the same extent as in the mixed dimer sample at pH 9 . 1 ( Figure 3B ) , the effect of intermediate chemical exchange would not induce peak broadening beyond detection . Hence , the effect of ethidium binding suggests that the conformational bias is reduced when the heterodimer is bound to a drug substrate . Furthermore , the line-broadening supports the presence of drug-induced dynamics in the heterodimer , which would enable conformational exchange needed for drug transport . Finally , we investigated the effect of tetraphenylphosphonium binding on the conformational equilibrium . Unlike ethidium , tetraphenylphosphonium binds with greater affinity to EmrE and significantly reduces the rate of conformation exchange ( i . e . slow chemical exchange regime ) ( Cho et al . , 2014; Morrison and Henzler-Wildman , 2014 ) . We carried out the same NMR experiment by adding tetraphenylphosphonium to a mixed dimer sample of 13Cδ-Ile labeled EmrE mixed with natural abundance L51I ( Figure 3D ) . Unlike the effect of ethidium binding , the spectra were nicely resolved in the presence of tetraphenylphosphonium . We observed that monomer B peaks were more intense relative to those of monomer A; however , the signal intensities stemming from monomer A were significantly greater compared to any of the drug-free samples ( see I58A , I62A , and I88A peaks in Figure 3D ) . Specifically , the ratio of monomer B to A peak intensities were reduced upon addition of tetraphenylphosphonium relative to the deprotonated sample . Using the relative peak intensities of A and B , we estimated the equilibrium population of wild-type to assume monomer B in the heterodimer at ~86% ( standard deviation range of 79% to 92% ) , which corresponds to ~1 . 1 kcal/mol induced by the single mutant in the heterodimer . This result is in agreement with the ethidium binding experiment . These data also provide evidence that drug binding would not trap the transporter into a single conformation in the transport cycle and is consistent with previous work proposing the substrate controls the rate of conformational exchange ( Morrison and Henzler-Wildman , 2014 ) . E . coli growth inhibition assays against ethidium bromide were carried out to determine the biological significance of our NMR observations . Initially , we tested whether the mutations L51I and I62L were able to confer resistance when expressed individually . Figure 4A shows a growth inhibition assay using serial 10-fold dilutions on an LB agar plate with wild-type , L51I , I62L , and a control vector . From these data , L51I and I62L displayed a strong phenotype toward ethidium and were indistinguishable relative to wild-type EmrE . Thus , each mutation appears to be fully functional and able to couple to the electrochemical potential to accomplish active drug efflux . To probe the effect of heterodimer induced changes to the conformational equilibrium , we designed transporter complementation experiments by expressing two genes: ( 1 ) gene 1 consisted of wild-type EmrE , L51I , or I62L and ( 2 ) gene 2 consisted of single topology variants of EmrE . The latter utilized similar constructs as previous work showing insertion topology can be controlled by mutating arginine and lysine residues within the primary sequence ( Rapp et al . , 2007 ) ( i . e . the positive inside rule; Heijne , 1986 ) . Using this technology , gene 2 consisted of N- and C-termini facing the cytoplasmic direction ( EmrEin ) or periplasmic direction ( EmrEout ) , respectively . Resistance assays verified literature results ( Rapp et al . , 2007 ) that EmrEin and EmrEout expressed individually were unable to confer a phenotype to ethidium , whereas co-expression displayed a strong phenotype toward this compound ( Figure 4B ) . Next , we carried out resistance assays using mixtures of wild-type , L51I , or I62L with EmrEin or EmrEout . The experiments where wild-type EmrE was co-expressed with EmrEin or EmrEout revealed identical levels of conferred resistance relative to wild-type EmrE alone ( Figure 5A ) . To ensure that wild-type was forming a heterodimer with EmrEin or EmrEout , we designed a control experiment in which an additional mutation ( E14Q ) was engineered into the EmrEin or EmrEout constructs ( i . e . E14Qin or E14Qout ) . E14Q was selected since EmrE requires two glutamic acid residues to confer drug resistance ( Rapp et al . , 2007 ) . Thus , if a dimer of wild-type and E14Qin or E14Qout formed , we would observe a reduced growth phenotype . Indeed , the results of these experiments displayed growth inhibition when wild-type was co-expressed with E14Qin or E14Qout ( Figure 5A ) . Taken together with the robust phenotype of wild-type co-expressed with EmrEin or EmrEout , these results confirmed that the transporter complementation approach leads to heterodimer formation in the cell membrane . We next carried out experiments with L51I and I62L co-expressed with EmrEin or EmrEout . Remarkably , when EmrEin was co-expressed with L51I or I62L , we observed a significant reduction in ethidium resistance on LB agar plates ( Figure 5B ) . To the contrary , EmrEout co-expressed with L51I or I62L retained the same phenotype as the single-site mutant alone . To provide additional support , we carried out resistance assays in liquid media . These experiments were performed by growing E . coli cultures containing plasmids to an optical density at 600 nm of 1 . 0 , then adding ethidium and recording the culture density as a function of time . Similar to the LB agar plate results , we found that the L51I or I62L mutant expressed with EmrEin displayed a faster drop in the optical density , which signified a reduced ability to confer resistance relative to the same mutant expressed with EmrEout ( Figure 5C ) . It is important to note that if heterodimers of L51I or I62L and EmrEin did not form , we would have observed the same resistance phenotype of L51I or I62L alone . However , the reduced phenotype and the control experiments discussed above ( Figure 5A; Figure 5—figure supplement 1 ) indicate a specific association in the transporter complementation experiments that are influenced by the conformational equilibrium within the assembled heterodimer in the cell membrane . To provide direct evidence that the overall rate of ethidium transport was different between the mutant co-expressed with EmrEin and EmrEout , we performed an ethidium efflux assay by measuring the intrinsic fluorescence of ethidium . E . coli were treated with ethidium bromide and the ionophore carbonyl cyanide m-chlorophenylhydrazone ( CCCP ) , which causes the cytoplasmic ethidium concentration and fluorescence to increase . Upon addition of glucose and removal of CCCP , the membrane potential is reestablished , and the fluorescence decreases as ethidium is transported out of the cytoplasm . This assay was performed with L51I co-expressed with EmrEin or EmrEout . Immediately following the addition of glucose , the fluorescence dropped ~3 fold faster for L51I co-expressed with EmrEout than L51I co-expressed with EmrEin ( Figure 5D ) . Furthermore , the fluorescence value at steady state ( 3600 s ) was significantly lower for L51I co-expressed with EmrEout compared to L51I co-expressed with EmrEin ( Figure 5E ) . This indicates a reduced cytoplasmic ethidium concentration for the L51I/EmrEout sample . Taken together , these data are consistent with the resistance assay results and support the conclusion that a change in the conformational equilibrium can influence the overall rate of drug efflux . Note that a previous study reported an ethidium efflux rate difference of 1 . 6-fold between wild-type E . coli and an AcrAB-TolC efflux pump knockout strain that led to a 30-fold difference in the minimum inhibitory concentration ( Paixão et al . , 2009 ) . This shows that somewhat modest transport rate differences can have deleterious effects on bacterial growth , which stems from increased ethidium accumulation that leads to irreversible binding to DNA ( Jernaes and Steen , 1994; Lambert and Le Pecq , 1984 ) . How does the NMR observed conformational change of the heterodimer correlate with the functional results ? The transporter complementation experiments create topologically defined heterodimers due to the single topology of EmrEin and EmrEout and the antiparallel association of the dimer quaternary structure . Using this information and knowledge from NMR experiments that the mutant had a preference for monomer A in the heterodimer , we can make inferences about how a shift in the conformational equilibrium influences function . Specifically , the mutant paired with EmrEout in a heterodimer would favor the inward-open/cytoplasmic-facing conformation ( Figure 5F ) . Likewise , the mutant paired with EmrEin will bias toward the outward-open/periplasmic-facing conformation under the same conditions ( Figure 5F ) . Hence , the resistance assay results suggest that the conformational equilibrium favoring the outward-open/periplasmic state has a deleterious impact on the overall transport cycle . In contrast , a heterodimer with a conformation favoring the inward-open/cytoplasmic facing conformation gives no measurable reduction in phenotype . This finding is in harmony with our prior observation that the role of the pH gradient is to favor the inward-open conformation that is poised for drug binding ( Gayen et al . , 2016 ) . Thus , reducing the effect of the pH gradient by biasing the equilibrium in the opposite direction apparently is sufficient to alter the net transport as observed in growth inhibition experiments and efflux assays . It is important to underscore that the equilibrium constant between inward-open and outward-open states of the heterodimer ultimately stems from differential kinetic rate constants pertaining to the inward-open to outward-open and outward-open to inward-open transitions ( see kinetic model [Robinson et al . , 2017] in Figure 5G ) . Thus , we hypothesize that the primary kinetic steps in the transport cycle that influence the observed phenotype are the inward-open to outward-open and outward-open to inward-open rates for the drug-free heterodimer ( Figure 5G , see double asterisks ) . We make this conclusion based on two observations . First , the populations were more skewed for the drug-free states than those for the drug bound states . Second , NMR experiments with ethidium indicate that the conformational exchange was not halted in the heterodimer . The latter means that the conformational bias does not result in a locked-state that would ultimately lead to loss-of-function ( i . e . no substrate turnover ) . To provide support for this conclusion , we performed a numerical simulation of pH-driven transport into liposomes using the kinetic model and parameters introduced by Robinson et al . ( 2017 ) . The differences were to change the inward-open to outward-open and outward-open to inward-open conformational exchange rates based on the populations observed in our heterodimer experiments and to account for the fact that drug binding affinities to wild-type ( A ) /mutant ( B ) and mutant ( A ) /wild-type ( B ) are not necessarily the same ( see Figure 5—figure supplement 2 ) . Using these simulations , we found that the rate of drug transport was ~5 . 6 fold faster for a heterodimer of mutant and EmrEout versus a heterodimer of mutant and EmrEin ( Figure 5—figure supplement 2 ) . In addition , the former achieved a small but significant increase in the final concentration of drug within the vesicles , which further suggests that differences in kinetic rates in EmrE’s transport cycle give rise to a change in the cytoplasmic drug concentration between heterodimers with opposite insertion topologies in E . coli . These calculations correlate with the enhanced phenotype and increased rate of ethidium efflux and provide evidence that shifting conformational equilibria via specific rate constants in a transport cycle has a direct impact on functional output . Despite the prominence of inverted repeat domains within transporters , there are only a few examples of mutational studies to influence conformational equilibria . These experiments involve the creation of loss-of-function , non-conservative mutations ( e . g . Trp to Gly ) at key structural contacts with the goal of crystallographically trapping particular conformations in the transport cycle ( Smirnova et al . , 2013; Latorraca et al . , 2017 ) . Our observations demonstrate that a minimal heterodimer comprised of a single conservative mutation is sufficient to disrupt the preferred resting conformation and underscore the presence of a dynamic energy landscape where relatively small energy differences exist among conformations in the transport cycle . From an evolutionary perspective , we hypothesize there might be fitness advantages for having separate genes that form a hetero-oligomer or a single polypeptide chain where the structural repeats are contained within one protein . In these cases , a single mutation can be introduced within the functional assembly that is not possible for a homo-oligomeric protein . Functional differences between closely related proteins stemming from one mutation in a hetero-oligomer versus multiple mutations in a homo-oligomer could potentially include influencing the propensity of a transporter to carry out antiport or symport . In fact , in addition to EmrE’s known antiport activity , it is also able to carry out symport under certain conditions ( Robinson et al . , 2017 ) , including import of polyamines upon mutation ( Brill et al . , 2012 ) . Therefore , the observation that conservative mutations can display skewed conformational equilibria and influence function opens the possibility of somewhat modest evolutionary paths for achieving symport or antiport . Protein expression and purification was carried out as previously reported ( Gayen et al . , 2013 ) . EmrE is expressed as a fusion construct with maltose-binding protein in the pMAL vector ( New England Biolabs Inc ) in E . coli BL21 ( DE3 ) cells . For oriented sample NMR experiments , 15N-tyrosine labeled EmrE was expressed with IPTG for 4 hr at 25°C in an amino acid mixture consisting of 19 amino acids at natural abundance ( 300 mg/L ) and 15N-tyrosine ( 120 mg/L ) . Unlabeled EmrE used in the heterodimer experiments was expressed in LB medium at natural abundance . For solution NMR experiments , wild-type and L51I were expressed in a fully perdeuterated background as previously described ( Gayen et al . , 2016 ) . Wild-type protein was isotopically enriched with 13C at the Cδ position of isoleucine methyl groups with the addition of 50 mg/L 2-ketobutyric acid-4-13C , 3 , 3-2H2 sodium salt hydrate 1 hr before induction . The bacterial cells following expression were lysed and the fusion protein was purified with amylose affinity chromatography . The fusion protein was cleaved with tobacco etch virus protease ( TEV ) and EmrE was further purified by size exclusion chromatography using a Superdex 200 10/300 column ( GE Healthcare ) in 0 . 06% w/v n-dodecyl-β-D-maltopyranoside ( DDM ) . For oriented sample solid-state NMR studies , purified EmrE in DDM detergent was reconstituted into 1 , 2-di-O-tetradecyl-sn-glycero-3-phosphocholine/dihexanoyl-sn-glycero-3-phosphocholine ( O-14:0-PC/6:0-PC ) bicelles at a molar ratio of 3 . 5/1 ( i . e . q = 3 . 5 ) . The bicelles were made with a protein concentration ~2 mM with a combined lipid concentration of 25% ( w/v ) in 80 mM HEPES and 20 mM NaCl at pH = 5 . 4 . The heterodimer samples were prepared by mixing the two proteins ( wild-type/L51I or wild-type/I62L ) with a molar ratio 1/1 . 2 , where the protein in excess was at natural abundance ( i . e . NMR silent ) . The two proteins used to prepare the heterodimer sample were incubated together at 37°C in the presence of 50 mM DTT for 1 hr immediately prior to reconstitution into lipids . The reconstitution into the bicelles was carried out in the same manner as those for homodimer samples ( Leninger and Traaseth , 2018 ) . For solution NMR studies , purified EmrE in DDM detergent was reconstituted into dimyristoyl-sn-glycero-3-phosphocholine ( 14:0-PC ) and dihexanoyl-sn-glycero-3-phosphocholine ( 6:0-PC ) bicelles with a molar ratio of 1/2 . The acyl chains of 14:0-PC and 6:0-PC were perdeuterated ( Avanti Polar Lipids ) to reduce the lipid signals in 1H/13C heteronuclear correlation experiments . The heterodimer samples were prepared by mixing isotopically enriched wild-type with natural abundance L51I in a 1/3 molar ratio . The two proteins used to prepare the heterodimer sample were incubated together at 37°C for 1 hr immediately prior to reconstitution into lipids . The heterodimers contained 0 . 533 mM total protein ( 0 . 133 mM wild-type , 0 . 4 mM L51I ) . The long-chain lipid ( 14:0-PC ) to total protein ratio of the solution NMR samples was ~150/1 ( mol/mol ) . Control homodimer samples with isotopically enriched wild-type EmrE only were prepared in a similar fashion . The sample buffer was 150 mM Na2HPO4 and 20 mM NaCl . The experiments with ethidium bromide and tetraphenylphosphonium were carried out at concentrations of 2 . 1 mM and 1 . 05 mM , respectively . In order to assess whether a 1 hr incubation time at 37°C was sufficient to achieve complete mixing of the heterodimer , a control experiment was carried out by comparing NMR spectra after 1 hr and 19 hr of incubation at 37°C . The NMR spectra collected of these experiments showed that shorter and longer incubation times gave statistically the same ratios of monomer A and B peak intensities , which demonstrates that complete mixing was achieved with 1 hr ( Figure 3—figure supplement 1B ) . Note that the 1 hr incubation time in DDM is consistent with the reduced dimer stability of EmrE in DDM detergent micelles compared to that in lipid bicelles or lipid bilayers ( Dutta et al . , 2014a ) . Oriented sample solid-state NMR experiments were acquired using an Agilent DD2 spectrometer at a 1H frequency of 600 MHz equipped with a 1H/15N double resonance probe manufactured by Revolution NMR ( design of Peter Gor’kov; Gor'kov et al . , 2007 ) . Experiments were carried out using magnetically aligned lipid bicelle samples that were flipped with the addition of 3 mM YbCl3 to orient the bicelle normal parallel to the magnetic field . PISEMA ( Wu et al . , 1994 ) spectra were acquired with SPINAL-64 1H decoupling during acquisition and phase modulated Lee-Goldberg ( Vinogradov et al . , 1999 ) ( PMLG ) 1H-1H decoupling in the indirect dimension . The radiofrequency field for SPINAL decoupling was 50 kHz , while the effective field for PMLG was 41 . 7 kHz . Spectra were acquired with ~1500 scans and 14 increments in the indirect dimension with a recycle delay of 3 s . The indirect dimension axis was corrected with the scaling factor of 0 . 82 . The 15N direct dimension was referenced to 15NH4Cl ( solid ) at 41 . 5 ppm . Solution NMR experiments were acquired using a Bruker Avance III spectrometer at a 1H frequency of 600 MHz equipped with a TCI cryogenic probe . 1H/13C HMQC experiments were acquired at 25°C using 1H and 13C spectral widths of 10 , 000 Hz and 4 , 000 Hz , respectively . The total acquisition ( 1H ) and evolution times ( 13C ) corresponding to t2 and t1 were 59 . 9 msec and 18 . 5 msec , respectively . Spectra were acquired with 12 or 24 scans with a recycle delay of 1 s . The equilibrium shown in Equation 1 corresponds to a heterodimer composed of wild-type EmrE and mutant: ( 1 ) WTA⋅mutantB ⇌ mutantA⋅WTB Subscripts 'A' and 'B' correspond to monomers A and B in the asymmetric dimer . A simple calculation of the peak intensity ratio for A and B resonances in the heterodimer samples does not correspond to the equilibrium constant since the peak intensities contain a statistical fraction of homodimers ( fhomo ) and heterodimers ( fhet ) based on the ratio of isotopically labeled and natural abundance proteins present for mixing . Note that fhomo and fhet consider only homodimers or heterodimers that contain the isotopically enriched protein . The observed peak intensities for monomer A ( IA , obs ) and B ( IB , obs ) signals in a heterodimer sample are given by Eqns . 2 and 3: ( 2 ) IA , obs=IA fhomo+fhet pA ( 3 ) IB , obs=IB fhomo+fhet pB Equation 2 divided by Equation 3 gives: ( 4 ) IA , obsIB , obs=IA IB fhomo+fhet pA fhomo+fhet pB IA/IB is the ratio of ‘intrinsic’ intensities of monomer A and B peaks obtained from the homodimer spectrum . This is needed since A and B peaks in the homodimer spectrum are not exactly the same . fhomo and fhet were set to 1/7 and 6/7 , respectively , by assuming statistical mixing of the molar ratio of isotopically labeled wild-type to natural abundance L51I mutant in solution NMR experiments ( i . e . , 1 part labeled to 3 parts unlabeled ) . pA and pB are populations of wild-type in monomer A or B conformations when in a heterodimer with a mutant ( see Equation 1 ) . The addition of these populations is given by Equation 5 and their ratio is the equilibrium constant ( K ) in Equation 6: ( 5 ) pA+pB=1 ( 6 ) K=pBpA Substitution of pB from Equation 5 into Equation 4 gives the following: ( 7 ) IA , obsIB , obs=IA IB fhomo+fhet pA fhomo+fhet ( 1-pA ) pA was solved from Equation 7 for drug-free ( pH = 9 . 1 ) and tetraphenylphosphonium bound forms of wild-type/L51I heterodimers using solution NMR HMQC spectra of wild-type EmrE and isotopically labeled wild-type mixed with natural abundance L51I . The average and error range ( derived from the standard deviation ) of the populations were determined from the following residues that displayed well-resolved signals in each spectrum: Ile11 , Ile54 , Ile58 , Ile62 , Ile68 , Ile88 , and Ile101 . Growth inhibition assays were performed in pET Duet-1 vectors with constructs designed similar to that of Rapp et al . ( 2007 ) . All of the constructs are shown in Figure 4—figure supplement 1 . EmrEin consists of three mutations relative to wild-type EmrE ( R29G , R82G , S107K ) and induces the N- and C-termini to face the cytoplasm . EmrEout consists of three mutations from wild-type ( T28R , L85R , R106A ) and induces the N- and C-termini to face the periplasm . EmrEin and EmrEout were placed in the second cloning site . An additional mutation to EmrEin and EmrEout constructs were made by mutating E14 to Q14 . These constructs are also single topology and are referred to as E14Qin and E14Qout . The L51I and I62L constructs are single site mutants of wild-type EmrE ( UniProt P23895 ) . Luria-Bertani ( LB ) media and Luria agar powder for resistance assay were purchased from Research Products International . Each construct in the pET Duet-1 vector was transformed into BL21 ( DE3 ) and grown at 37°C up to an OD600 of ~1 . 5 . The cultures were diluted to an OD600 of 1 . 0 with fresh LB supplemented with carbenicillin ( 100 μg/mL ) and were then serially diluted by 10-fold to achieve final dilutions of 100 to 106 . All of the dilutions used fresh LB media containing 100 μg/mL carbenicillin . 3 μl of each dilution were pipetted onto plates containing 20 μM IPTG , 100 μg/mL carbenicillin , and 94 μg/ml ethidium bromide . Control experiments were carried out by plating cells onto LB agar plates containing 20 μM IPTG and 100 μg/mL carbenicillin . All resistance assays on solid medium were repeated at least two times . Liquid assays were performed using the same constructs as for the solid media resistance assay . E . coli BL21 ( DE3 ) were transformed with the pET Duet-1 vectors and grown at 37°C until the OD600 was ~1 . 5 . The cultures were then diluted to an OD600 of 1 . 0 with fresh LB supplemented with carbenicillin ( 100 μg/mL ) and ethidium bromide ( 240 μg/mL ) . Cultures were incubated for 11 . 5 hr and the OD600 was measured every hr up to 6 hr and one additional time 11 . 5 hr after the initial exposure to ethidium bromide . A full set of resistance assays in liquid medium were repeated three times . The error bars reflect the standard deviation for replicate trials . Mutant constructs used in resistance assay were transformed into BL21 ( DE3 ) and grown to mid log phase at 37°C ( OD600 ~1 . 0 ) . The cells were spun down and diluted to the OD600 of 0 . 1 in minimal media A ( 40 mM K2HPO4 , 22 mM KH2PO4 , 2 mM sodium citrate , 0 . 8 mM MgSO4 , and 7 . 6 mM ( NH4 ) 2SO4 , pH 7 . 0 ) . The resuspended cells were treated with 80 μM carbonyl cyanide m-chlorophenyl hydrazine ( CCCP ) for 5 min . Ethidium bromide ( 10 μg/mL ) was added to the cells and incubated for 30 min at 37°C while shaking . Cells were spun down for 10 min and the pellet was kept on ice until the florescence experiment . For the efflux assay , cells were resuspended in minimal media A with 10 μg/mL ethidium bromide . To initiate the assay , 0 . 2% ( w/v ) glucose was added . Control experiments were carried out by treating the cells in the same manner except without adding glucose . The fluorescence decays were measured with a Molecular Devices FlexStation 3 instrument using an excitation wavelength of 530 nm and an emission wavelength of 600 nm . The fluorescence was recorded over a time period of 3600 s . Each experiment was acquired in duplicate and repeated at least two times . The error bars reflect the duplicate experiments carried out on the same day .	Cells are bound by a thin membrane layer that protects the cell’s interior from the outside environment . Within this layer are various transporter proteins that control which substances are allowed in and out of the cell . These transporters actively move substances across the membrane by loading cargo on one side of the layer , then changing their structure to release it on the other side . Membrane transporters are typically made up of multiple repeating units . In more complex transporters , the genetic sequence for each of these structural units is fused together into a single gene that codes for the protein . It is thought that the repeated pattern evolved from smaller membrane protein genes that had duplicated and fused together . But , what are the evolutionary advantages of having more complex transporters being produced from a single , fused gene ? To investigate this , Leninger , Sae Her , and Traaseth examined a simple transporter found in Escherichia coli bacteria , called EmrE , which contains two identical protein subunits that associate together to transport toxic molecules across the membrane . Experiments revealed that changing a single amino acid ( the building blocks that make up proteins ) in one of the two subunits to make them minimally different from each other , dramatically modified the transporter’s structure and function . The subtle amino acid change disrupted the balance of inward- and outward-facing proteins . This altered the transporter’s ability to remove toxic chemicals from E . coli and reduced the bacteria’s resistance to drugs . The effects of a minor change to one of the identical halves of the EmrE transporter demonstrates how sensitive membrane transporters are to mutations . Furthermore , this observation could help explain why evolution favored more complex transporters comprised of fused genes in which single amino acid changes can greatly alter how the transporter operates .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics" ]	2019	Inducing conformational preference of the membrane protein transporter EmrE through conservative mutations
Proper organogenesis depends upon defining the precise dimensions of organ progenitor territories . Kidney progenitors originate within the intermediate mesoderm ( IM ) , but the pathways that set the boundaries of the IM are poorly understood . Here , we show that the bHLH transcription factor Hand2 limits the size of the embryonic kidney by restricting IM dimensions . The IM is expanded in zebrafish hand2 mutants and is diminished when hand2 is overexpressed . Within the posterior mesoderm , hand2 is expressed laterally adjacent to the IM . Venous progenitors arise between these two territories , and hand2 promotes venous development while inhibiting IM formation at this interface . Furthermore , hand2 and the co-expressed zinc-finger transcription factor osr1 have functionally antagonistic influences on kidney development . Together , our data suggest that hand2 functions in opposition to osr1 to balance the formation of kidney and vein progenitors by regulating cell fate decisions at the lateral boundary of the IM . Organs arise from precisely defined territories containing progenitor cells with specific developmental potential . Distinct progenitor territories often abut one another , and communication at the interfaces between neighboring territories acts to refine their boundaries ( Dahmann et al . , 2011 ) . This process delineates the final dimensions of each territory and , subsequently , influences the sizes of the derived organs . Boundary refinement is generally thought to be mediated by interplay between opposing inductive and suppressive factors ( Briscoe and Small , 2015 ) . In many cases , however , the identification of and interactions among these factors remain elusive . Kidney progenitor cells originate from the intermediate mesoderm ( IM ) , a pair of narrow bilateral stripes within the posterior mesoderm , flanked by lateral mesoderm that gives rise to vessels and blood and by paraxial mesoderm that gives rise to bone , cartilage , and skeletal muscle . The mechanisms that determine the dimensions of the stripes of IM are not fully understood . Several conserved transcription factors are expressed in the IM and are required for its development , including Lhx1/Lim1 , Pax2 , and Osr1/Odd1 ( Dressler and Douglass , 1992; James et al . , 2006; Krauss et al . , 1991; Toyama and Dawid , 1997; Tsang et al . , 2000; Wang et al . , 2005 ) . Studies in chick have indicated essential roles for the lateral mesoderm , paraxial mesoderm , and surface ectoderm in regulating the expression of these transcription factors ( James and Schultheiss , 2003; Mauch et al . , 2000; Obara-Ishihara et al . , 1999 ) . Furthermore , TGF-beta signaling acts in a dose-dependent manner to pattern the medial-lateral axis of the posterior mesoderm: for example , low levels of BMP signaling promote IM formation while high levels of BMP signaling promote lateral mesoderm formation ( Fleming et al . , 2013; James and Schultheiss , 2005 ) . Beyond these insights , the pathways that set the boundaries of the IM and distinguish this territory from its neighbors are largely unknown . Several lines of evidence have suggested that a carefully regulated refinement process is required to sharpen the boundary between the IM and the lateral mesoderm . In chick , mouse , and Xenopus , Osr1 and Lim1 are expressed in both the lateral mesoderm and the IM before becoming restricted to the IM , implying the existence of a mechanism that acts to exclude IM gene expression from the neighboring lateral territory ( Carroll and Vize , 1999; James et al . , 2006; Mugford et al . , 2008; Tsang et al . , 2000 ) . Additional data have hinted at an antagonistic relationship between the IM and the blood and vessel lineages ( Gering et al . , 2003; Gupta et al . , 2006 ) : for example , overexpression of vascular and hematopoietic transcription factors ( tal1 and lmo2 ) induces ectopic vessel and blood specification while inhibiting IM formation ( Gering et al . , 2003 ) . Along the same lines , zebrafish osr1 morphants exhibit disrupted pronephron formation together with expanded venous structures ( Mudumana et al . , 2008 ) . Despite these indications of interconnections between IM and vessel development , the network of factors that link these processes has not been fully elucidated . Here , we establish previously unappreciated roles for the bHLH transcription factor Hand2 in both IM and vessel formation . Prior studies of Hand2 have focused on its functions in other tissues , including the heart , limb , and branchial arches ( e . g . Charité et al . , 2000; Fernandez-Teran et al . , 2000; Funato et al . , 2009; Miller et al . , 2003; Srivastava et al . , 1997; Yanagisawa et al . , 2003; Yelon et al . , 2000 ) . Although Hand2 is also expressed in the posterior mesoderm ( Angelo et al . , 2000; Fernandez-Teran et al . , 2000; Srivastava et al . , 1997; Thomas et al . , 1998; Yelon et al . , 2000; Yin et al . , 2010 ) , its influence on patterning this tissue has not been extensively explored . Through both loss-of-function and gain-of-function studies , we find that hand2 limits the size of the kidney by repressing IM formation while promoting venous progenitor formation . hand2 is expressed laterally adjacent to the IM , and a set of venous progenitors arise at the interface between the hand2-expressing cells and the IM . Ectopic expression of IM markers within the hand2-expressing territory in hand2 mutants suggests that hand2 establishes the lateral boundary of the IM through direct inhibition of IM fate acquisition . Finally , genetic analysis indicates that hand2 functions in opposition to osr1 to control kidney dimensions . Together , our data demonstrate a novel mechanism for defining territory boundaries within the posterior mesoderm: hand2 represses IM formation to establish its lateral boundary while promoting venous progenitor formation in this region . These important functions of Hand2 help to define the precise dimensions and components of the kidneys and vasculature . Moreover , these findings have implications for understanding the genetic basis of congenital anomalies of the kidney and urinary tract ( CAKUT ) and for developing new approaches in regenerative medicine . Our interest in the role of hand2 during kidney development began with an observation arising from our previously reported microarray analysis of hand2 mutants ( Garavito-Aguilar et al . , 2010 ) . We compared gene expression profiles at 20 hr post fertilization ( hpf ) in wild-type embryos and hans6 mutant embryos , which contain a deletion that removes the entire coding region of hand2 ( Yelon et al . , 2000 ) . Surprisingly , 11 of the 26 transcripts that were increased in hans6 relative to wild-type were expressed in the pronephron , the embryonic kidney . ( For a full list of differentially expressed genes , see Garavito-Aguilar et al . , 2010 . ) We therefore sought to understand the effect of hand2 function on the pronephron . The six genes with the most elevated expression in hans6 mutants ( Garavito-Aguilar et al . , 2010 ) are all expressed in the pronephron tubules . We examined the expression of two of these genes that are expressed throughout the tubules -- atp1a1a . 4 , which encodes a subunit of the Na+/K+ ATPase ( Thisse et al . , 2004 ) , and cadherin17 ( cdh17 ) ( Horsfield et al . , 2002 ) -- and observed an increase in tubule width in hans6 ( Figure 1A–C , E–G ) . A similar increase in width of expression was seen for another gene upregulated in hans6 but only expressed in a single tubule segment ( Wingert et al . , 2007 ) , slc12a3 ( Figure 1I–K ) , which encodes the thiazide-sensitive sodium-chloride cotransporter solute carrier 12a3 . Notably , in contrast to its widened expression , the anterior-posterior extent of slc12a3 expression appeared unaltered . Last , by examining expression of lhx1a and pax2a , which mark glomerular precursors at 24 hpf ( O'Brien et al . , 2011 ) , we found that the populations of glomerular precursors , like the tubules , were expanded in hans6 ( Figure 1M–O and Figure 8B , C ) . Thus , the elevated gene expression detected by microarray analysis in hans6 mutants corresponded with broadened expression of genes throughout the pronephron . 10 . 7554/eLife . 19941 . 003Figure 1 . hand2 inhibits pronephron formation . ( A–P ) Dorsal views , anterior to the left , of pronephron schematics ( A , E , I , M ) , wild-type embryos ( B , F , J , N ) , hans6 mutant embryos ( C , G , K , O ) , and hand2-overexpressing embryos ( D , injected with hand2 mRNA; H , L , P , carrying Tg ( hsp70-hand2-2A-mcherry ) , abbreviated hs:hand2 ) at 24 hpf . In schematics ( A , E , I , M ) , colored regions correspond to area of pronephron gene expression . In situ hybridization demonstrates that atp1a1a . 4 ( A–D ) and cdh17 ( E–H ) are expressed throughout the pronephron tubules , slc12a3 ( I–L ) is expressed in the distal late segments of the pronephron tubules , and lhx1a ( M–P ) is expressed in the glomerular precursors ( arrows , N ) , as well as overlying spinal neurons ( asterisks , N ) . Compared to wild-type ( B , F , J , N ) , gene expression is expanded in hans6 mutants ( C , G , K , O ) and reduced in hand2-overexpressing embryos ( D , H , L , P ) . Of note , injection of a hand2 translation-blocking morpholino caused effects on pronephron formation similar to those seen in hans6 mutants ( data not shown ) . Scale bars represent 100 μm . ( Q , R ) Transverse sections through wild-type ( Q ) and hans6 mutant ( R ) pronephron tubules at 24 hpf . Dashed lines outline the tubule and asterisks indicate individual tubule cells . ( S , T ) Bar graphs indicate the average number of tubule cells per cross-section ( S ) and the average tubule area per cross-section ( T ) in wild-type and hans6 mutant embryos; error bars indicate standard deviation . Asterisks indicate statistically significant differences compared to wild-type ( p<0 . 0001 , Student’s t test; n = 18 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 00310 . 7554/eLife . 19941 . 004Figure 1—source data 1 . Number of tubule cells per cross-section . Number of tubule cells observed in transverse sections from embryos at 24 hpf were quantified at three different anterior-posterior levels along the pronephron , with two sections analyzed per level in each embryo . Three animals for each genotype ( wild-type and hans6 ) were examined . Average number of cells and standard deviation are represented in Figure 1S . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 00410 . 7554/eLife . 19941 . 005Figure 1—source data 2 . Area of pronephron tubule in cross-section . Pronephron tubules in transverse sections from embryos at 24 hpf were measured at three different anterior-posterior levels along the pronephron , with two sections analyzed per level in each embryo . Three animals for each genotype ( wild-type and hans6 ) were examined . Average tubule area and standard deviation are represented in Figure 1T . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 005 To determine if this broadened expression was due to an increase in cell number , we next analyzed the structure of the pronephron in more detail . We found that the number of cells seen in cross-section of the tubule was increased in hans6 ( Figure 1Q–S ) . This increase in cell number was observed at multiple regions along the anterior-posterior axis , suggesting that the entire pronephron is expanded . Concomitant with this increase in the number of cells , the total tubule area in cross-section was increased in hans6 mutants ( Figure 1T ) . Together with the expanded expression of pronephron genes , these findings implicate hand2 in inhibiting the accumulation of pronephron cells . Does hand2 simply prevent excessive expansion of the pronephron or is it potent enough to repress pronephron formation ? To differentiate between these possibilities , we examined the effects of hand2 overexpression on the pronephron . Injection of hand2 mRNA inhibited pronephron formation , as assessed by atp1a1a . 4 expression ( Figure 1D ) . To bypass any potential effects of overexpression on gastrulation , we also utilized our previously characterized Tg ( hsp70:FLAG-hand2-2A-mCherry ) transgenic line to drive hand2 expression at the tailbud stage ( Schindler et al . , 2014 ) . Heat shock-induced overexpression resulted in pronephron defects comparable to those caused by mRNA injection , as assessed by atp1a1a . 4 , cdh17 and slc12a3 expression ( Figure 1H , L and Figure 3A , B ) . Furthermore , like tubule gene expression , lhx1a and pax2a expression in glomerular precursors was dramatically inhibited by hand2 overexpression ( Figure 1P and data not shown ) , again emphasizing the broad effect of hand2 function on pronephron formation . In contrast to these effects of hand2 overexpression on the pronephron , most other regions of expression of these markers , such as expression of atp1a1a . 4 in mucus-secreting cells and expression of lhx1a in spinal neurons , were unaffected ( Figure 1D , P ) . Thus , the effect of hand2 overexpression seems to reflect its particular impact on pronephron development , as opposed to a general influence of hand2 on the expression of each of these genes . Of note , however , we did observe a dramatic reduction in otic vesicle expression of both atp1a1a . 4 and pax2a ( data not shown ) , suggesting a shared susceptibility to hand2 overexpression in the otic vesicles and the pronephron . Taken together , our loss-of-function and gain-of-function studies show that hand2 is necessary to constrain the size of the pronephron , likely through an ability to repress pronephron formation . We next sought to determine the origin of the effect of hand2 on the pronephron , and we hypothesized that the pronephron defects observed in hans6 mutants might reflect a requirement for hand2 to limit IM dimensions . Indeed , using two established IM markers , pax2a and lhx1a , we observed that the width of the IM was expanded in hans6 mutants and hand2 morphants ( Figure 2A , B and Figure 2—figure supplement 1A , B ) , at stages shortly after gastrulation ( i . e . at the 6 , 10 and 12 somite stages ) . In contrast , there was no substantial difference in the length of the IM seen in wild-type and hans6 mutant embryos ( Figure 2A , B ) . Furthermore , compared to wild-type embryos , hans6 mutants typically had ~50% more Pax2a+ IM cells ( Figure 2D , E , G ) . This accumulation of IM cells did not appear to be associated with heightened proliferation within the IM ( Figure 2—figure supplement 2 ) . Thus , our data suggest that hand2 restricts the initial formation of the IM . 10 . 7554/eLife . 19941 . 006Figure 2 . hand2 inhibits IM production . ( A–C ) Dorsal views , anterior to the left , of the posterior mesoderm at the 11 somite stage . In situ hybridization depicts normal expression of pax2a in the IM ( arrows ) of wild-type embryos ( A ) , widened expression in hans6 mutants ( B ) , and lack of expression in hand2-overexpressing embryos ( hs:hand2 ) ( C ) . Expression in the spinal neurons ( asterisk ) is unaffected by altered hand2 function . Scale bar represents 100 μm . ( D–F ) Pax2a immunofluorescence in the posterior mesoderm of wild-type ( D ) , hans6 mutant ( E ) , and hs:hand2 ( F ) embryos at the 12 somite stage . Dorsal views , anterior to the left , are three-dimensional reconstructions of flat-mounted embryos from which the yolk and anterior tissues have been dissected away . Scale bar represents 100 μm . ( D'–F' ) Magnification of 250 µm long regions from ( D–F ) used for quantification of the number of Pax2a+ cells in wild-type ( D' ) , hans6 mutant ( E' ) , and hs:hand2 ( F' ) embryos . White dots indicate Pax2a+ nuclei . Intensity of staining varied from strong ( for example , green arrows ) to weak ( for example , yellow arrows ) . Scale bar represents 25 μm . ( G ) Bar graph indicates the average number of Pax2a+ cells per 100 µm of IM in wild-type , hans6 , and hs:hand2 mutant embryos; error bars indicate standard deviation . Asterisks indicate a statistically significant difference compared to wild-type ( p<0 . 0001 , Student’s t test; n=13 for wild-type , n=10 for hans6 , and n=19 for hs:hand2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 00610 . 7554/eLife . 19941 . 007Figure 2—source data 1 . Pax2a+ cells in wild-type , hans6 , and hs:hand2 intermediate mesoderm . The number of Pax2a+ cells was quantified on the indicated dates of analysis . For wild-type and hans6 embryos , representative 250 μm long regions of IM were analyzed , while for hs:hand2 embryos , 500 μm long regions of IM were analyzed . For hs:hand2 embryos , the IM on both the left and right sides of the embryo were analyzed independently when the dissection and preparation of the sample allowed . All values were normalized to represent the number of cells per 100 μm . Average number of Pax2a+ cells per 100 μm and standard deviation are represented in Figure 2G . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 00710 . 7554/eLife . 19941 . 008Figure 2—figure supplement 1 . hand2 inhibits IM production . ( A–C ) Dorsal views , anterior to the top , of the posterior mesoderm at the 10 somite stage . In situ hybridization depicts normal expression of lhx1a in the IM ( arrows ) of wild-type embryos ( A ) , widened expression in hans6 mutants ( B ) , and reduced expression in hand2-overexpressing embryos ( C ) . Expression in the notochord ( asterisk ) is unaffected by altered hand2 function . Scale bar represents 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 00810 . 7554/eLife . 19941 . 009Figure 2—figure supplement 2 . Comparable proliferation in the IM of wild-type and hans6 mutant embryos . ( A , B ) Immunofluorescence for Pax2a and phospho-Histone H3 ( pH3 ) in wild-type ( A ) and hans6 mutant ( B ) embryos at the 6 somite stage; dorsal views , anterior to the left , are three-dimensional reconstructions of 400 µm long regions used for quantification of the numbers of Pax2a+ ( A' , B' ) and pH3+ ( A'' , B'' ) cells . Arrows indicate Pax2a+ pH3+ cells . Scale bar represents 50 μm . ( C ) Table compares the number of Pax2a+ pH3+ cells per 100 µm of IM , the number of Pax2a+ cells per 100 µm of IM , and the proliferation index ( Pax2a+ pH3+ cells / Pax2a+ cells X 100 ) in wild-type and hans6 mutant embryos . Data shown represent the average and standard deviation for each value . While there was a statistically significant increase in the number of Pax2a+ cells in hans6 mutants compared to wild-type ( p<0 . 0001 ) , no significant differences were observed for the number of Pax2a+ pH3+ cells ( p=0 . 1461 ) or the proliferation index ( p=0 . 0654 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 00910 . 7554/eLife . 19941 . 010Figure 2—figure supplement 2—source data 1 . Pax2a+ pH3+ cells in wild-type and hans6 intermediate mesoderm . The numbers of Pax2a+ cells and Pax2a+ pH3+ cells were quantified on the indicated dates of analysis . 400 μm long regions of IM were analyzed , and values were normalized to represent the number of cells per 100 μm . A proliferation index ( percentage of Pax2a+ pH3+ cells per Pax2a+ cells ) was calculated . Average numbers with standard deviation and proliferation index are represented in Figure 2—figure supplement 2C . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 010 To determine whether hand2 is sufficient to repress IM formation , we examined the effects of hand2 overexpression . Both hand2 mRNA injection and heat shock-induced overexpression of hand2 resulted in loss of IM , as determined by pax2a and lhx1a expression ( Figure 2C and Figure 2—figure supplement 1C ) . More specifically , hand2-overexpressing embryos exhibited phenotypes ranging from a significantly reduced number of Pax2a+ IM cells to a complete absence of Pax2a+ IM ( Figure 2F , G ) . Taken together with our loss-of-function analyses , these results suggest that hand2 limits the dimensions of the IM by repressing its initial specification . To define the time window during which hand2 overexpression can inhibit IM formation , we utilized Tg ( hsp70:FLAG-hand2-2A-mCherry ) embryos to induce hand2 expression at different times after gastrulation ( Figure 3 ) . Unlike overexpression at the tailbud , 2 somite , or 6 somite stages ( Figure 3A–D ) , overexpression at the 10 somite stage failed to inhibit pronephron development ( Figure 3E ) . Furthermore , we found progressively less severe defects as heat shock induction was performed at successively later stages between tailbud and 10 somites . For example , while inhibition at the tailbud or 2 somite stages resulted in the loss of the majority of pronephric atp1a1a . 4 expression ( Figure 3A , B ) , inhibition at the 6 somite stage resulted in only a mild reduction ( Figure 3D ) . Overall , our analysis suggests that there is a hand2-sensitive phase of IM specification prior to the 10 somite stage . 10 . 7554/eLife . 19941 . 011Figure 3 . Pronephron development is susceptible to hand2 overexpression prior to the 10 somite stage . ( A–E ) Dorsal views , anterior to the left , of in situ hybridization for atp1a1a . 4 at 24 hpf depict a range of severity of pronephron defects , ranging from absence of the pronephron ( A ) to unaffected ( E ) . Tg ( hsp70:FLAG-hand2-2A-mCherry ) embryos were subjected to heat shock at the tailbud , 2 somite , 6 somite , or 10 somite stages , and the consequences on pronephron development were scored at 24 hpf . Percentages indicate the distribution of phenotypes produced by each treatment; the number of embryos examined is in the right-hand column . Heat shock at later stages resulted in more mild loss of atp1a1a . 4 expression in the tubule , and heat shock at the 10 somite stage did not disrupt atp1a1a . 4 expression in the tubule . Scale bar represents 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 011 Considering the strong effect of hand2 function on repressing IM formation , we sought to define the location of hand2 expression relative to the IM . Prior studies had demonstrated bilateral hand2 expression in the posterior mesoderm of zebrafish , mouse , chick , and Xenopus embryos ( Angelo et al . , 2000; Fernandez-Teran et al . , 2000; Srivastava et al . , 1997; Thomas et al . , 1998; Yelon et al . , 2000; Yin et al . , 2010 ) , but the precise localization of this expression relative to the IM had not been determined . During the hand2-sensitive phase of IM formation , we found hand2 to be expressed in bilateral regions immediately lateral to the IM ( Figure 4A , B ) . At these stages , hand2 was also expressed lateral to multiple markers of blood and vessel progenitors , including etv2 , tal1 , and gata1 ( Figure 4C and data not shown ) . However , a gap lies between these markers and the hand2-expressing cells , consistent with the IM residing between the lateral hand2-expressing cells and the medial blood and vessel progenitors ( Figure 4C , D ) . 10 . 7554/eLife . 19941 . 012Figure 4 . hand2 expression in the posterior lateral mesoderm . ( A–D , F–I ) Two-color fluorescent in situ hybridization ( A , C , F , H ) and immunofluorescence ( B , D , G , I ) label components of the posterior mesoderm in dorsal views , anterior to the left , of three-dimensional reconstructions , as in Figure 2D–E . In embryos containing transgenes in which GFP expression is driven by the regulatory elements of hand2 ( B , G ) or etv2 ( D , I ) , anti-GFP immunofluorescence was used to enhance visualization . Tg ( etv2:egfp ) expression was also observed in the midline neural tube , as previously reported ( Proulx et al . , 2010 ) . Scale bar represents 100 μm . ( E , J ) Schematics depict posterior mesoderm territories , dorsal views , anterior to the left; hand2-expressing cells , IM , and medial vessel/blood progenitors are shown at 2–10 somites ( E ) and at 11–13 somites ( J ) , together with lateral vessel progenitors . ( A–D ) hand2 is expressed lateral to the IM at the 2–10 somite stages . Embryos shown are at the 10 somite stage; similar expression patterns were seen at earlier stages . ( A , B ) hand2 is expressed lateral to the IM markers lhx1a ( A ) and Pax2a ( B ) . ( C ) hand2 is expressed lateral to tal1 , a marker of blood and vessel progenitors; note the unlabeled gap between expression territories . ( D ) A marker of vessel progenitors , Tg ( etv2:egfp ) , lies medially adjacent to Pax2a . ( F–I ) Vessel progenitors arise at the interface between hand2-expressing cells and the IM at the 11–13 somite stages . ( F , G ) hand2 is expressed lateral to the IM marker pax2a . ( H ) hand2 is expressed lateral to a second territory of tal1 expression; note the presence of lateral tal1-expressing cells ( arrows ) immediately adjacent to hand2 expression . ( I ) The IM lies between two territories of etv2 expression; the more lateral etv2-expressing cells ( arrows ) are lateral to the IM . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 012 We also found that the relationship between hand2 expression , the IM , and blood and vessel progenitors changed after the completion of the hand2-sensitive phase of IM formation . At approximately the 11 somite stage , a second , lateral population of vessel progenitors arises at the interface between the IM and the hand2-expressing cells ( Figure 4F–I ) . These lateral vessel progenitors are likely to be venous progenitors , based on the results of a prior study that suggested that the medial , earlier-forming vessel progenitors contribute to the dorsal aorta and that the lateral , later-forming vessel progenitors contribute to the primary cardinal vein ( Kohli et al . , 2013 ) . This set of results defines the general location of hand2 expression relative to other territories within the posterior mesoderm ( Figure 4E , J ) . However , these observations do not exclude the possibility of transient overlapping expression at the boundaries of each territory ( e . g . overlapping etv2 and pax2a expression ) . Furthermore , we note that gene expression patterns are not necessarily uniform within each territory ( e . g . tal1 and etv2 expression patterns are neither uniform nor equivalent within their territory [Kohli et al . , 2013] ) . Nevertheless , our data suggest that , during the hand2-sensitive phase of IM formation , hand2 may exert its repressive effect on IM formation by constraining the lateral boundary of the IM . Furthermore , our findings suggest the possibility of close interactions between hand2 , the IM , and the lateral population of venous progenitors . The appearance of lateral venous progenitor cells at the interface between the IM and the hand2-expressing cells raised the question of whether hand2 regulates the development of these venous progenitors . To address this possibility , we first assessed the early expression of etv2 , tal1 , and gata1 in the medial population of vessel and blood progenitors , and we observed no differences between wild-type and hans6 mutant embryos at either the 6 or 10 somite stages ( Figure 5—figure supplement 1 and data not shown ) . In contrast , after the 11 somite stage , expression of both etv2 and tal1 in the lateral venous progenitor population was absent in hans6 mutants ( Figure 5A , B , F , G ) . Two-color fluorescent in situ hybridization confirmed the medial location of the remaining territory of etv2 and tal1 expression in hand2 loss-of-function embryos ( Figure 5D , E , I , J ) . Thus , the formation of the lateral venous progenitors , the appearance of which coincides with the end of the hand2-responsive phase of IM formation , requires hand2 . 10 . 7554/eLife . 19941 . 013Figure 5 . hand2 promotes vessel progenitor development . ( A–C , F–H , K–M ) In situ hybridization depicts etv2 ( A–C ) , tal1 ( F–H ) and gata1 ( K–M ) expression in wild-type ( A , F , K ) , hans6 mutant ( B , G , L ) and hand2-overexpressing ( C , H , M ) embryos; dorsal views , anterior to the left , at the 12 somite stage . ( A , F ) etv2 and tal1 are expressed in relatively medial and lateral ( arrows ) territories on each side of the wild-type embryo . In hans6 embryos ( B , G ) , only the medial territory is present . In hand2-overexpressing embryos ( C , H ) , expression of both etv2 and tal1 is increased , but it is not possible to distinguish whether this represents an increase in the medial or the lateral vessel progenitor populations , since no markers exist that distinguish these two groups of progenitors . ( K–M ) gata1 expression is equivalent in wild-type ( K ) and hans6 ( L ) embryos , but it is decreased in hand2-overexpressing embryos ( M ) . Scale bar represents 100 μm . ( D , E , I , J ) Fluorescent in situ hybridization depicts the relationship of etv2 and pax2a expression in wild-type ( D , D' ) and hand2 morphant ( hand2 MO ) embryos ( E , E' ) , and the relationship of tal1 and hand2 expression in wild-type ( I , I' ) and hand2 MO ( J , J' ) embryos; dorsal views , anterior to the left , at the 12 somite stage . Medial and lateral ( arrows ) territories of etv2 expression flank pax2a in wild-type embryos ( D , D' ) . The lateral territory of expression is absent in hand2 morphants ( E , E' ) . The lateral territory ( arrows ) of tal1 expression is located at the medial border of hand2 expression in wild-type embryos ( I , I' ) , but is absent in hand2 morphants ( J , J' ) . Note that we observed variable thickness of tal1 expression in its medial territory of expression . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 01310 . 7554/eLife . 19941 . 014Figure 5—figure supplement 1 . Presence of medial vessel and blood progenitors is unaffected in hans6 mutants . ( A–F ) In situ hybridization depicts expression of etv2 ( A , B ) , tal1 ( C , D ) , and gata1 ( E , F ) in wild-type ( A , C , E ) and hans6 mutant ( B , D , F ) embryos; dorsal views , anterior to the top , of the posterior mesoderm at the 6 somite stage . Expression of all three markers appears unperturbed by loss of hand2 function at this stage . Scale bar represents 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 014 To gain insight into whether hand2 directly promotes the formation of vessel progenitors , we assessed the consequences of hand2 overexpression . Overexpression of hand2 resulted in an expansion of both etv2 and tal1 expression in the posterior mesoderm ( Figure 5C , H ) . In contrast , hand2 overexpression resulted in a reduction of gata1 expression ( Figure 5M ) . Thus , in the posterior mesoderm , hand2 overexpression can inhibit the development of some tissues , such as IM and blood , while promoting vessel progenitor formation . To understand the consequences of the influence of hand2 function on the lateral venous progenitors , we examined the vasculature in hans6 mutants . As noted above , a prior lineage tracing study suggested that the lateral venous progenitors contribute to the posterior cardinal vein ( Kohli et al . , 2013 ) . Furthermore , this study found that inhibition of etv2 function at the time when these lateral cells arise results in loss of expression of mannose receptor c1a ( mrc1a ) , a marker of the cardinal vein ( Kohli et al . , 2013 ) . Similarly , hans6 mutants lack mrc1a and flt4 expression in the posterior cardinal vein , suggesting a defect in venous development ( Figure 6A , B , D , E ) . Using Tg ( flk1:ras-mcherry ) to label the entire vasculature , we identified the presence of the posterior cardinal vein in hans6 mutants ( Figure 6J–L ) , but found that their caudal venous plexus failed to be properly remodeled ( Figure 6M–O ) . In contrast to these changes in the venous system , flt4 and ephrin-b2a ( efnb2a ) expression in the dorsal aorta , as well as mrc1a expression in the posterior blood island , appeared grossly unaffected by hand2 loss-of-function ( Figure 6B , E , H ) . Meanwhile , consistent with the ability of hand2 to promote vessel progenitor formation , hand2 overexpression increased vascular expression of mrc1a , flt4 , and efnb2a ( Figure 6C , F , I ) . 10 . 7554/eLife . 19941 . 015Figure 6 . hand2 promotes proper vein formation . ( A–I ) In situ hybridization depicts expression of mrc1a ( A–C ) , flt4 ( D–F ) and efnb2a ( G–I ) in wild-type ( A , D , G ) , hans6 mutant ( B , E , H ) and hand2-overexpressing ( C , F , I ) embryos; lateral views , anterior to the left , at 24 hpf ( A–C , G–I ) and the 20 somite stage ( D–F ) . ( A–C ) mrc1a expression in the posterior cardinal vein ( blue arrow ) was present in wild-type ( A ) , absent in hans6 mutant ( B ) , and increased in hand2-overexpressing ( C ) embryos . Expression in the posterior blood island ( red arrowhead ) was grossly unaffected . ( D–F ) flt4 expression in the posterior cardinal vein ( blue arrow ) was present in wild-type ( D ) , absent in hans6 mutant ( E ) , and increased in hand2-overexpressing ( F ) embryos . Expression in the dorsal aorta ( red arrow ) was grossly unaffected . ( G–I ) efnb2a expression in the dorsal aorta ( red arrow ) was present in wild-type ( G ) , grossly unaffected in hans6 mutant ( H ) , and slightly increased in hand2-overexpressing ( I ) embryos . ( J–O ) Lateral views of three-dimensional reconstructions of Tg ( flk1:ras-mcherry ) expression in the vasculature of wild-type ( J , M ) and hans6 mutant ( K , N ) embryos at 28 hpf ( J , K ) and 32 hpf ( M , N ) . Pictured region in ( J , K ) is the area of the trunk boxed in the schematic ( L ) ; pictured region in ( M , N ) is the area of the tail boxed in the schematic ( O ) . Both posterior cardinal vein ( blue arrow ) and dorsal aorta ( red arrow ) were present in wild-type ( J ) and hans6 mutant ( K ) embryos . In contrast to wild-type ( M ) , the caudal venous plexus ( arrowhead ) fails to undergo proper remodeling in hans6 mutants ( N ) . Scale bars represent 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 015 Overall , these studies demonstrate a previously unappreciated role for hand2 in regulating trunk vasculature development . More specifically , our data indicate that hand2 is required for the successful execution of certain aspects of venous differentiation , including the expression of characteristic venous markers and the remodeling of the caudal venous plexus . Combined with prior lineage tracing studies and etv2 loss-of-function analysis ( Kohli et al . , 2013 ) , our findings suggest that hand2 implements these functions by promoting the development of the lateral venous progenitors . We cannot , however , rule out the possibility that the roles of hand2 in venous progenitor formation and venous differentiation are independent of one another , such that the differentiation defects represent a later influence of hand2 on vascular development . The lateral localization of hand2 expression and the absence of lateral venous progenitor markers when hand2 function is lost highlighted the possibility that hand2 impacts the IM at its lateral border . We therefore wanted to determine how the expanded IM in hans6 mutants relates to the interface between the IM and the hand2-expressing cells . Do the extra IM cells seem to emerge at this interface ? Do they appear to arise from cells that express hand2 , or , alternatively , at the expense of hand2-expressing cells ? To differentiate among these possibilities , we used two methods to identify the hand2-expressing territory in hand2 loss-of-function embryos . We labeled hand2-expressing cells through in situ hybridization in embryos injected with a hand2 morpholino ( Figure 7A , B ) , and we took advantage of the non-rescuing BAC transgene Tg ( hand2:EGFP ) ( Kikuchi et al . , 2011 ) to interrogate hans6 mutants ( Figure 7C , D ) . Of note , each of these approaches indicated heightened intensity of hand2 expression in hand2 loss-of-function embryos ( Figure 7A , B and Figure 7—figure supplement 1 ) , suggesting the possibility that hand2 acts to inhibit its own expression . Importantly , however , we observed comparable dimensions of the hand2 expression territory in wild-type and morphant embryos ( Figure 7A , B ) . More specifically , when using Tg ( hand2:EGFP ) , we found no significant difference between the numbers of GFP+ cells in the posterior mesoderm of wild-type and hans6 mutant embryos ( Figure 7C–E ) . Thus , it seems that the expansion of the IM in hans6 mutants does not come at the expense of the formation of cells that normally express hand2 . 10 . 7554/eLife . 19941 . 016Figure 7 . Increased presence of Pax2a in hand2-expressing cells of hand2 mutants . ( A–B ) In situ hybridization depicts presence of hand2 expression in hand2 morphants ( B ) ; dorsal views , anterior to the left , at the 10 somite stage . There is no evident loss of hand2-expressing cells in hand2 morphants; moreover , hand2 expression levels appear higher in the context of hand2 loss-of-function . Scale bar represents 100 μm . ( C–D ) Immunofluorescence for Pax2a and GFP in wild-type ( C ) and hans6 mutant ( D ) embryos , both carrying Tg ( hand2:EGFP ) ; dorsal views , anterior to the left , of three-dimensional reconstructions at the 12 somite stage , as in Figure 4G . ( C'–D''' ) Magnification of 250 µm long regions from ( C ) and ( D ) used for quantification of the numbers of GFP+ and Pax2a+ cells in wild-type and hans6 mutant embryos . Yellow dots indicate GFP+ cells , white dots indicate Pax2a+ nuclei , and examples of GFP+ Pax2a+ cells are indicated by white arrows . Intensity of Pax2a+ staining varied from strong ( for example , green arrows ) to weak ( for example , yellow arrows ) . Scale bars represent 100 μm ( C , D ) and 25 μm ( C'–D''' ) . ( E ) Bar graph indicates the average numbers of GFP+ cells and GFP+ Pax2a+ cells per 100 µm of IM in wild-type and hans6 mutant embryos; error bars indicate standard deviation . Asterisk indicates a statistically significant difference compared to wild-type ( p<0 . 0001 , Student’s t test; n = 13 for wild-type and n = 10 for hans6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 01610 . 7554/eLife . 19941 . 017Figure 7—source data 1 . GFP+ and Pax2a+ GFP+ cells in wild-type and hans6 intermediate mesoderm . The numbers of GFP+ and Pax2a+ GFP+ cells were quantified on the indicated dates of analysis . Representative 250 μm long regions of IM were analyzed , and values were normalized to represent the number of cells per 100 μm . Average numbers and standard deviation are represented in Figure 7E . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 01710 . 7554/eLife . 19941 . 018Figure 7—figure supplement 1 . Increased hand2:gfp expression in hans6 . Immunofluorescence for GFP in wild-type ( left ) and hans6 mutant ( right ) embryos , both carrying Tg ( hand2:EGFP ) ; dorsolateral views , anterior to the top , at 12 somite stage . Levels of hand2:gfp expression appear higher in the context of hand2 loss-of-function . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 018 We next assessed whether the increased number of IM cells in hans6 mutants could arise at least in part from aberrant IM marker expression in cells that normally express hand2 . Careful examination of wild-type embryos carrying the Tg ( hand2:EGFP ) transgene revealed that ~6% of the GFP+ cells in their posterior mesoderm also displayed the IM marker Pax2a ( Figure 7C , E ) . In dramatic contrast , ~60% of the GFP+ cells in the hans6 mutant posterior mesoderm were also Pax2a+ ( Figure 7D , E ) . Intriguingly , the GFP+ Pax2a+ cells were typically observed at or near the border where the hand2-expressing cells meet the IM ( Figure 7C , D ) . These data suggest that hand2 limits IM dimensions by constraining the lateral boundary of the IM through repression of IM gene expression within hand2-expressing cells . The expansion of Pax2a expression in the hand2-expressing territory in hans6 mutants suggests that factors that promote IM development are present within hand2-expressing cells . We hypothesized that osr1 could be one of these factors , because of the correspondence between the hand2 and osr1 expression patterns and loss-of-function phenotypes . First , osr1 was previously reported to be expressed in the lateral posterior mesoderm , adjacent to the IM ( Mudumana et al . , 2008 ) . Through direct comparison of hand2 and osr1 expression patterns , we found that the genes are co-expressed in the lateral posterior mesoderm ( Figure 8A ) . Additionally , prior studies demonstrated that loss of osr1 function results in phenotypes opposite to those caused by loss of hand2: in contrast to the expanded pronephron and defective venous vasculature in hans6 mutants , osr1 morphants were found to exhibit inhibited pronephron development together with expansion of venous structures ( Mudumana et al . , 2008; Tena et al . , 2007 ) . 10 . 7554/eLife . 19941 . 019Figure 8 . hand2 and osr1 act in opposing , parallel pathways to regulate pronephron development . ( A–A'' ) Fluorescent in situ hybridization depicts overlap ( A ) of hand2 ( A' ) and osr1 ( A'' ) expression in wild-type embryos; dorsal views , anterior to the left , of three-dimensional reconstructions at the 6 somite stage . ( B–I ) In situ hybridization depicts pax2a ( B–E ) and atp1a1a . 4 ( F–I ) expression at 24 hpf in wild-type embryos ( B , F ) , hans6 mutant embryos ( C , G ) , osr1 morphant ( osr1 MO ) embryos ( D , H ) and hans6 mutant embryos injected with osr1 morpholino ( hans6 + osr1 MO ) ( E , I ) ; dorsal views , anterior to the left . ( B–E ) Compared to wild-type ( B ) , pax2a expression in the glomerular and neck precursors ( arrow ) was expanded in 100% of hans6 mutants ( C , n=11 ) , absent ( 48% ) or reduced ( 48% ) in osr1 morphants ( D , n=25 ) , and relatively normal in hans6 + osr1 MO embryos ( E , n=11 ) . While the extent of marker expression was generally comparable to wild-type in the double loss-of-function embryos , the stereotypic patterning of this population was often somewhat disrupted . Expression in overlying spinal neurons ( asterisk ) was unaffected . ( F–I ) Compared to wild-type ( F ) , atp1a1a . 4 expression in the pronephric tubules was wide in 85% of hans6 mutants ( G , n=13 ) , while many osr1 morphants ( H , n=133 ) had tubules with shortened anterior expression ( 18% ) or tubules with segmental losses ( 35% ) , and 47% of osr1 morphants had a wild-type appearance . 85% of hans6 + osr1 MO embryos ( I , n=46 ) resembled wild-type , whereas 11% had a shortened anterior tubule and 4% had segmental losses in the tubules . Scale bars represent 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 01910 . 7554/eLife . 19941 . 020Figure 8—figure supplement 1 . Expression patterns of osr1 and hand2 appear unaffected by loss of each other's function . ( A–D ) In situ hybridization depicts expression of osr1 ( A , B ) and hand2 ( C , D ) in the posterior mesoderm of wild-type ( A , C ) , hans6 mutant ( B ) , and osr1 MO ( D ) embryos; dorsal views , anterior to the top , of the posterior portion of the embryo at the 4 somite ( A , B ) and 6 somite ( C , D ) stages . ( A , B ) Loss of hand2 function does not appear to affect osr1 expression . ( C , D ) Loss of osr1 function does not appear to affect hand2 expression . Scale bar represents 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19941 . 020 To assess the functional relationship between hand2 and osr1 , we investigated double loss-of-function embryos . Dramatically , removal of hand2 function in the setting of osr1 loss-of-function largely restored normal pronephron development . This interaction was most notable in the glomerular precursor population ( Figure 8B–E ) . While this population was expanded in hand2 mutants , it was absent in embryos injected with an osr1 morpholino ( Figure 8B–D ) . In hand2 mutants injected with the osr1 morpholino , however , the precursor population appeared comparable to wild-type ( Figure 8E ) . Consistent with prior studies ( Mudumana et al . , 2008 ) , osr1 morpholino injection had a less dramatic effect on the pronephron tubule than on the glomerulus ( Figure 8D , H ) . Nevertheless , a comparable interaction between hand2 and osr1 was found in tubule formation ( Figure 8F–I ) . While hans6 mutants had wide tubules and many osr1 morphants had disrupted tubules , hans6 mutantsinjected with osr1 morpholino displayed relatively normal tubules . We wondered whether the observed genetic interaction between hand2 and osr1 might reflect roles for these genes in regulating one another's expression . We found , however , that loss of hand2 function did not seem to affect the intensity or dimensions of osr1 expression ( Figure 8—figure supplement 1A , B ) . Similarly , osr1 loss-of-function did not appear to alter hand2 expression ( Figure 8—figure supplement 1C , D ) . Therefore , hand2 and osr1 appear to affect downstream gene expression independently , without affecting one another’s expression . Furthermore , our data suggest that hand2 and osr1 function in parallel genetic pathways that act in opposition in order to regulate the precise dimensions of the pronephron . More generally , our findings implicate hand2 within a genetic network that regulates progenitor cell fates at the lateral edge of the IM , at the interface with the hand2- and osr1-expressing territory . Together , our data present a new perspective on how the posterior mesoderm is divided into defined territories that create distinct tissues . Importantly , our work provides the first evidence that hand2 restricts the size of the kidney and promotes vein development . Specifically , our data indicate that hand2 limits the specification of pronephric progenitors by restraining the lateral boundary of the IM . Furthermore , at the end of the hand2-sensitive phase of IM formation , hand2 supports the formation of venous progenitors just beyond the lateral edge of the IM , in the same region where it represses IM development . Finally , our data reveal a novel genetic interaction in which the inhibitory function of hand2 is balanced by the inductive role of osr1 during the establishment of IM dimensions . Our findings highlight the importance of regulating the refinement of the lateral boundary of the IM . It is intriguing to consider how hand2 , given its expression in the lateral posterior mesoderm , could influence the restriction of the adjacent IM . Perhaps hand2 acts cell-autonomously to inhibit IM specification; the ectopic appearance of Pax2a+ cells within the hand2-expressing territory in hand2 mutants makes this an attractive model . Alternatively , hand2 could have a non-autonomous influence on IM dimensions , perhaps by regulating the production of a lateralizing signal , the absence of which would result in the lateral expansion of IM gene expression . In this scenario , it is tempting to speculate that the role of hand2 could interface with the BMP signaling pathway , considering its prior implication in posterior mesoderm patterning ( James and Schultheiss , 2005 ) . Wherever its precise location of action may be , this function of hand2 -- preventing lateral hand2+osr1+ cells from acquiring more medial characteristics , such as expression of pax2a or lhx1a --suggests a possible mechanism for distinguishing IM gene expression from that of neighboring lateral territories , a distinct process required during both in vivo ( James et al . , 2006 ) and in vitro ( Takasato et al . , 2014 ) IM development . Additionally , our data imply a close connection between the specification of the IM and the formation of the lateral venous progenitor population . Are the roles of hand2 during IM and venous development coupled to each other , or do these represent two independent functions of hand2 ? We can envision a model in which the lateral venous progenitor cells emerge from a hand2-expressing lineage , and hand2 influences a cell fate decision between IM and venous identities . In particular , we suggest that hand2 normally inhibits IM specification within a portion of the hand2-expressing territory , while also promoting the differentiation of these same hand2-expressing cells into lateral venous progenitors . On the other hand , it remains feasible that hand2 could have completely separable effects on the kidney and vein lineages . Currently , the lineage relationships between the IM , the lateral venous progenitors , and the hand2-expressing lateral mesoderm remain undetermined; future work on the generation of appropriate tissue-specific lineage tracing tools will facilitate progress toward the resolution of these open questions . In future studies , it will also be valuable to elucidate the effector genes that act downstream of Hand2 in the posterior mesoderm . It is plausible that Hand2 could act directly to repress transcription of IM genes , especially considering that ChIP-seq analysis in the mouse limb bud has revealed Hand2-binding peaks associated with Pax2 and Lhx1a ( Osterwalder et al . , 2014 ) . Alternatively , Hand2 could repress IM genes without necessarily engaging Hand2-binding sites in their regulatory regions , since Hand2 can influence transcription through DNA-binding-independent mechanisms ( Funato et al . , 2009; Liu et al . , 2009 ) . Of course , Hand2 could also regulate IM gene expression indirectly . It will be particularly interesting to determine whether the Hand2 targets that influence the IM are also relevant to venous progenitor formation . Along those lines , it is noteworthy that our findings have demonstrated a difference between the functions of hand2 during vessel progenitor development in the anterior and posterior mesoderm: while we have previously shown that hand2 overexpression inhibits the expression of vessel progenitor markers in the anterior mesoderm ( Schindler et al . , 2014 ) , here we observed expanded expression of the same markers in the posterior mesoderm of hand2-overexpressing embryos . These distinct responses to hand2 activity in the anterior and posterior mesoderm may reflect regional differences in in the utilization of different bHLH binding partners or in the implementation of the different molecular functions ascribed to Hand2 ( Firulli et al . , 2005; Funato et al . , 2009; Liu et al . , 2009; McFadden et al . , 2002; Schindler et al . , 2014 ) . Future studies will need to distinguish among these possibilities in order to define how Hand2 functions in different embryonic contexts . By demonstrating that hand2 functions in opposition to osr1 , our data provide the first key step toward incorporating hand2 into the genetic regulatory network that patterns the posterior mesoderm . It is important to note that prior studies have suggested a non-autonomous role for osr1 during IM development ( Mudumana et al . , 2008 ) . The endoderm is expanded in osr1 morphants , and mutations that disrupt endoderm development abrogate the pronephron defects caused by osr1 loss-of-function ( Mudumana et al . , 2008; Terashima et al . , 2014 ) . In contrast , we suspect that hand2 affects IM formation from its position in the lateral mesoderm , rather than from the endoderm . We observe ectopic Pax2a+ cells emerging within hand2-expressing mesoderm in hand2 mutants; additionally , whereas osr1 morphants have an expansion of the endoderm ( Mudumana et al . , 2008; Terashima et al . , 2014 ) , hans6 mutants exhibit normal amounts of endoderm ( Wendl et al . , 2007; Yelon et al . , 2000 ) . Regardless of whether hand2 and osr1 both function within the lateral posterior mesoderm , our data suggest that these two genes control antagonistic and parallel genetic pathways that balance each other's influence in order to regulate the size of the IM . It will be interesting to determine whether the relationship between hand2 and osr1 function is broadly conserved . Loss-of-function and gain-of-function studies in mouse , chick , and Xenopus have demonstrated a role for Osr1 in IM and kidney development through focused analyses of histology and gene expression patterns ( James et al . , 2006; Mugford et al . , 2008; Tena et al . , 2007; Wang et al . , 2005 ) . In contrast , while mice lacking Hand2 have been examined for developmental defects in some posterior structures , such as the limb buds and enteric nervous system ( e . g . Charité et al . , 2000; Lei and Howard , 2011 ) , the influence of Hand2 on IM and kidney development has not yet been carefully interrogated . Moreover , future investigations in mouse may need to evaluate potential redundancy between the functions of Hand2 and Hand1: both of these genes are expressed in the murine lateral posterior mesoderm ( Thomas et al . , 1998 ) , whereas zebrafish have only a single Hand gene . Overall , our new insights into the genetic regulation of IM specification have the potential to impact our understanding of the origins and treatment of kidney disease . Prior studies have suggested a genetic component to the origins of congenital anomalies of the kidney and urinary tract ( CAKUT ) ( Bulum et al . , 2013; Vivante et al . , 2014 ) . In some cases of CAKUT , disease-causing mutations have been identified , many of which disrupt genes that influence early kidney development , including mutations in PAX2 and OSR1 ( Madariaga et al . , 2013; Thomas et al . , 2011; Zhang et al . , 2011 ) . By implicating hand2 in the control of IM specification , we enrich the gene regulatory network that is relevant to the etiology of CAKUT . In this regard , it is interesting to note that there are case reports of chromosomal duplications containing HAND2 in patients with renal hypoplasia ( Otsuka et al . , 2005 ) . Furthermore , the identification of genes that control IM specification can inspire new directions in the design of stem cell technologies and regenerative medicine strategies . Alteration of HAND2 function may allow for enhancement of current protocols for generating kidney progenitor cells in vitro ( Kumar et al . , 2015; Mae et al . , 2013; Taguchi et al . , 2014; Takasato et al . , 2014; Takasato et al . , 2015 ) , potentially facilitating the requisite control that will be essential for future deployment in patient applications . We generated embryos by breeding wild-type zebrafish , zebrafish heterozygous for the hand2 mutant allele hans6 ( Yelon et al . , 2000 ) ( RRID: ZFIN_ZDB-GENO-071003-2 ) , and zebrafish carrying Tg ( hand2:EGFP ) pd24 ( Kikuchi et al . , 2011 ) ( RRID: ZFIN_ZDB-GENO-110128-35 ) , Tg ( hsp70:FLAG-hand2-2A-mCherry ) sd28 ( Schindler et al . , 2014 ) ( RRID: ZFIN_ZDB-GENO-141031-2 ) , Tg ( flk1:ras-mcherry ) s896 ( Chi et al . , 2008 ) ( RRID: ZFIN_ZDB-GENO-081212-2 ) , or Tg ( etv2:egfp ) ci1 ( Proulx et al . , 2010 ) ( RRID: ZFIN_ZDB-GENO-110131-58 ) . For induction of heat shock-regulated expression , embryos were placed at 37°C for 1 hr and then returned to 28°C . Following heat shock , transgenic embryos were identified based on mCherry fluorescence; nontransgenic embryos were analyzed as controls . All heat shocks were performed at tailbud stage unless otherwise indicated . In embryos older than 20 hpf , hans6 mutants were identified based on their cardiac phenotype ( Yelon et al . , 2000 ) . In younger embryos , PCR genotyping of hans6 mutants was conducted as previously described ( Yelon et al . , 2000 ) , with the exception of embryos containing Tg ( hand2:EGFP ) pd24 , in which we used primers that amplify the first exon of hand2 ( 5’– CCTTCGTACAGCCCTGAATAC – 3’ and 5’ – CCTTCGTACAGCCCTGAATAC – 3’ ) . This exon is present in the wild-type allele but is absent in both hans6homozygotes and the TgBAC ( hand2:EGFP ) pd24 transgene . Synthesis and injection of capped hand2 mRNA was performed as described previously ( Schindler et al . , 2014 ) . To knock down hand2 function , we injected 3–6 ng of a previously characterized translation-blocking hand2 morpholino at the one-cell stage ( Reichenbach et al . , 2008 ) . To knock down osr1 function , we used a previously characterized osr1 translation-blocking start site morpholino , osr1 ATG ( Tena et al . , 2007 ) . 16–23 ng of this morpholino was injected with or without 1 . 6 ng of p53 morpholino ( Robu et al . , 2007 ) . Standard and fluorescent whole-mount in situ hybridization were performed as previously described ( Brend and Holley , 2009; Thomas et al . , 2008 ) , using the following probes: atp1a1a . 4 ( ZDB-GENE-001212-4 ) , cdh17 ( ZDB-GENE-030910-3 ) , efnb2a ( ZDB-GENE-990415-67 ) , etv2 ( etsrp; ZDB-GENE-050622-14 ) , flt4 ( ZDB-GENE-980526-326 ) , gata1 ( ZDB-GENE-980536-268 ) , hand2 ( ZDB-GENE-000511-1 ) , lhx1a ( lim1; ZDB-GENE-980526-347 ) , mrc1a ( ZDB-GENE-090915-4 ) , osr1 ( ZDB-GENE-070321-1 ) , pax2a ( ZDB-GENE-990415-8 ) , slc12a3 ( ZDB-GENE-030131-9505 ) , and tal1 ( scl; ZDB-GENE-980526-501 ) . To prepare an osr1 probe , we amplified and subcloned the osr1 cDNA using the primers 5’ – GAGTTTCTACCCCGAGTAACCA – 3’ and 5’ – TTTTCAAAAATAAGTTTAAGGAATCCA – 3’ . Whole-mount immunofluorescence was performed as previously described ( Cooke et al . , 2005 ) , using polyclonal antibodies against Pax2a at 1:100 dilution ( Genetex , Irvine , CA , GTX128127 ) ( RRID: AB_2630322 ) , GFP at 1:1000 dilution ( Life Technologies , Carlsbad , CA , A10262 ) ( RRID: AB_2534023 ) or phospho-Histone H3 ( Ser10 ) at 1:500 dilution ( EMD Millipore Corporation , Temecula , CA , 05-1336-S ) ( RRID: AB_1977261 ) , and the secondary antibodies goat anti-chick Alexa Fluor 488 ( Life Technologies , A11039 ) ( RRID: AB_2534096 ) , goat anti-rabbit Alexa Fluor 488 ( Life Technologies , A11008 ) ( RRID: AB_143165 ) , goat anti-mouse Alexa Fluor 546 ( Life Technologies , A11003 ) ( RRID: AB_2534071 ) , goat anti-rabbit Alexa Fluor 594 ( Life Technologies , A11012 ) ( RRID: AB_10562717 ) , or goat anti-rabbit Alexa Fluor 647 ( Life Technologies , A21245 ) ( RRID: AB_2535813 ) , all at 1:100 dilution . Samples were then placed in SlowFade Gold anti-fade reagent ( Life Technologies ) . To count Pax2a+ or GFP+ cells in Figures 2 and 7 , we flat-mounted and imaged embryos after dissecting away the yolk and the anterior portion of the embryo . In wild-type and hans6 embryos , we examined a representative 250 µm long region in roughly the middle of the IM on one side of the embryo . This strategy allowed us to select contiguous regions that were unaffected by dissection artifacts . In hs:hand2 embryos ( Figure 2 ) , we examined a 500 µm long region in order to minimize any selection bias that could be introduced from the variability associated with the relative dearth of Pax2a+ cells . To count Pax2a+ or pH3+ cells in Figure 2—figure supplement 2 , we chose to examine 400 µm long regions in undissected embryos , due to the low frequency of Pax2a+ pH3+ cells . In all cases , positive cells were determined through examination of both three-dimensional reconstructions and individual optical sections . Cell counts in Figure 2 , Figure 2—figure supplement 2 , and Figure 7 are presented as the average number of positive cells per 100 µm . Statistical analysis of data was performed using Microsoft Excel to conduct unpaired t-tests . Histological analysis was performed on tissues embedded in Spurr Low-Viscosity embedding mixture . Embryos were fixed in 2% paraformaldehyde and 2 . 5% glutaraldehyde in 0 . 1 M sodium phosphate buffer ( pH 7 . 2 ) , post-fixed with 2% osmium tetroxide , dehydrated in an ethanol gradient , infiltrated with a propylene oxide/resin gradient , and then embedded . Samples were oriented as desired and incubated for 24 hr at 60° C for polymerization . 2–4 μm sections were cut using glass blades , collected on standard slides , and stained with 1% toluidine blue . Transverse sections from 24 hpf embryos were examined at three different anterior-posterior levels along the pronephron , with two sections analyzed per level in each embryo . We processed three embryos per genotype . Cell counting and area measurements were performed using Zeiss ZEN software . Data variance homogeneity and normal distribution were confirmed using IBM SPSS Statistics 22 software . Bright-field images were captured with a Zeiss Axiocam on a Zeiss Axiozoom microscope and processed using Zeiss AxioVision . Images of histological sections were captured using a Zeiss AxioImager A2 microscope coupled to a Zeiss Axiocam camera . Confocal images were collected by a Leica SP5 confocal laser-scanning microscope and analyzed using Imaris software ( Bitplane , Switzerland ) . All assessments of phenotypes and expression patterns were replicated in at least two independent experiments with comparable results . Embryos were collected from independent crosses , and experimental processing ( injection , heat shock , and/or staining ) was carried out on independent occasions . Two exceptions to this include data presented in Figure 1Q–T and Figure 8C–E . In each of those cases , multiple embryos were processed , and the n is reported in the associated figure legends .	The human body is made up of many different types of cells , yet they are all descended from one single fertilized egg cell . The process by which cells specialize into different types is complex and has many stages . At each step of the process , the selection of cell types that a cell can eventually become is increasingly restricted . The entire system is controlled by switching different genes on and off in different groups of cells . Balancing the activity of these genes ensures that enough cells of each type are made in order to build a complete and healthy body . Upsetting this balance can result in organs that are too large , too small or even missing altogether . The cells that form the kidneys and bladder originate within a tissue called the intermediate mesoderm . Controlling the size of this tissue is an important part of building working kidneys . Perens et al . studied how genes control the size of the intermediate mesoderm of zebrafish embryos , which is very similar to the intermediate mesoderm of humans . The experiments revealed that a gene called hand2 , which is switched on in cells next to the intermediate mesoderm , restricts the size of this tissue in order to determine the proper size of the kidney . Switching off the hand2 gene resulted in zebrafish with abnormally large kidneys . Loss of hand2 also led to the loss of a different type of cell that forms veins . These findings suggest that cells with an active hand2 gene are unable to become intermediate mesoderm cells and instead go on to become part of the veins . These experiments also demonstrated that a gene called osr1 works in opposition to hand2 to determine the right number of cells that are needed to build the kidneys . Further work will reveal how hand2 prevents cells from joining the intermediate mesoderm and how its role is balanced by the activity of osr1 . Understanding how the kidneys form could eventually help to diagnose or treat several genetic diseases and may make it possible to grow replacement kidneys from unspecialized cells .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2016	Hand2 inhibits kidney specification while promoting vein formation within the posterior mesoderm
In all living organisms , ribosomes translating membrane proteins are targeted to membrane translocons early in translation , by the ubiquitous signal recognition particle ( SRP ) system . In eukaryotes , the SRP Alu domain arrests translation elongation of membrane proteins until targeting is complete . Curiously , however , the Alu domain is lacking in most eubacteria . In this study , by analyzing genome-wide data on translation rates , we identified a potential compensatory mechanism in E . coli that serves to slow down the translation during membrane protein targeting . The underlying mechanism is likely programmed into the coding sequence , where Shine–Dalgarno-like elements trigger elongation pauses at strategic positions during the early stages of translation . We provide experimental evidence that slow translation during targeting and improves membrane protein production fidelity , as it correlates with better folding of overexpressed membrane proteins . Thus , slow elongation is important for membrane protein targeting in E . coli , which utilizes mechanisms different from the eukaryotic one to control the translation speed . The rate of translation elongation varies considerably between different codons of the same gene , with some codons translated slower than the others ( Ingolia et al . , 2009; Li et al . , 2012 ) . Various factors can modulate the elongation rate . Some of them are sequence elements within the translated mRNA or protein that directly interact with the translational machinery ( Tuller et al . , 2010; Li et al . , 2012; Charneski and Hurst , 2013; Pechmann and Frydman , 2013 ) , such as charged amino acids that interact with the ribosome ( Charneski and Hurst , 2013 ) or codons that interact with tRNAs ( Tuller et al . , 2010 ) . Other factors act in trans and require the engagement of additional factors that can in turn affect ribosome speed ( Walter and Blobel , 1981; Shalgi et al . , 2013 ) . Several studies indicate that the elongation rate can modulate protein folding and function ( Kimchi-Sarfaty et al . , 2007; Zhang et al . , 2009; Yona et al . , 2013; Zhou et al . , 2013 ) , but the underlying mechanisms are not yet well understood . One of the best understood speed control mechanisms is found in the pathway that targets membrane proteins to the membrane translocon . Membrane proteins are extremely hydrophobic and aggregation-prone and are therefore inserted into the membrane co-translationally ( White and von Heijne , 2008; Luirink et al . , 2012 ) . To do so , they must be targeted to the translocon early in translation before large polypeptide portions are synthesized ( Lakkaraju et al . , 2008 ) . Targeting of ribosomes translating membrane proteins is mediated by the ubiquitous signal recognition particle ( SRP ) system ( Bibi , 2011; Akopian et al . , 2013 ) . The core machinery is conserved in all organisms and is comprised of the ribonuceoprotein complex SRP and its receptor . The pathway works such that the SRP binds to the first transmembrane segment ( TM ) that emerges from the ribosome and directs it together with the translocon-associated SRP receptor to the translocon . In eukaryotes , the SRP arrests the elongation upon binding to ribosome-nascent chain complexes , in order to prevent the protein synthesis before targeting is complete ( Walter and Blobel , 1981; Mason et al . , 2000 ) . The arrest is mediated through the SRP Alu domain that binds to the elongation factor binding site ( Halic et al . , 2004 ) and is kept until the ribosome arrives at the translocon , causing the SRP dissociation . Puzzlingly , however , the bacterial world displays a dichotomy in the structure of the SRP , wherein Gram-positive bacteria typically possess an Alu domain and Gram-negatives , such as E . coli , are typically devoid of it ( Regalia et al . , 2002 ) . Accordingly , in vitro translation studies suggested that the Escherichia coli SRP cannot arrest translation ( Raine et al . , 2003 ) . Thus , other mechanisms that compensate for the lack of Alu-mediated arrest in Gram-negative bacteria could be required , especially considering their faster translation rates compared to eukaryotes ( Young and Bremer , 1976; Bonven and Gulløv , 1979 ) . One such putative mechanism has recently been described ( Yosef et al . , 2010 ) , which calls for further investigation . Recently , our ability to measure translation rates was revolutionized by the technology of ribosome profiling , enabling the estimation of elongation rates on a genome-wide level and at a single codon resolution ( Ingolia et al . , 2009 ) . Studies conducted so far revealed remarkable heterogeneity in ribosome density between different codons within genes and suggested that translational pausing or slowing are actually quite common events ( Ingolia et al . , 2011; Li et al . , 2012 ) . These experiments also greatly advanced our understanding of how sequence elements modulate elongation rate ( Li et al . , 2012; Charneski and Hurst , 2013 ) . A notable example is found in E . coli , where many if not most of the pause sites are mediated by internal Shine–Dalgarno-like elements , which are capable of transiently binding to the ribosome as it translates the mRNA ( Li et al . , 2012 ) . Despite this progress , it is still not fully understood what could be the function of such commonly observed pauses in cell physiology . To meet this challenge , we explored where Shine–Dalgarno-mediated pause sites are preferentially located across the E . coli transcriptome , by analyzing the existing data from ribosome profiling ( Li et al . , 2012 ) . Our analyses revealed significant enrichment of pause events during the initial stages of membrane proteins translation . This location coincides with the stage in translation in which membrane proteins are thought to be targeted to the membrane translocons . We provide evidence that Shine–Dalgarno-mediated pauses in E . coli possibly compensate for the lack of SRP Alu domain-mediated arrest during membrane protein targeting . Recently , the ribosome occupancy across E . coli mRNAs was profiled in a genome-wide manner , revealing the locations of ribosome slowing as peaks of high ribosomal density . In that study , Li et al . ( 2012 ) found that ribosome pause sites are often explained by a Shine–Dalgarno ( SD ) -like sequence located 8–11 nucleotides upstream of the currently decoded codon . In order to understand the potential function of pauses better , we sought to investigate where such mRNA-programmed pauses are preferentially located . To identify pause sites , we searched for codons exhibiting high local ribosome density compared to other codons in the same coding sequence . Ribosome density profiles for well-expressed genes ( i . e . , with reliable ribosome profiling data , ‘Materials and methods’ ) were calculated by normalizing the ribosome density in codons to the median density of the gene ( Figure 1—figure supplement 1A ) . Unless otherwise noted , all analyses were based on these normalized profiles . Assuming that ribosomes flow along mRNAs is in steady-state , ribosome density in these profiles is inversely proportional to translation speed ( Tuller et al . , 2010 ) . The first and last four codons were excluded since they contain high density related to an initiation ‘ramp’ and translation termination , respectively ( Tuller et al . , 2010; Oh et al . , 2011 ) . We considered ‘pause codons’ as those that constitute the 5% of the most dense codons genome-wide , corresponding to ∼sixfold higher density than the median density of genes in which they reside ( Figure 1—figure supplement 1A ) . To generate a dependable SD-mediated set of pause sites , we filtered out pause codons for which we could not identify an upstream SD-like element ( Figure 1—figure supplement 1B , see ‘Materials and methods’ ) . Codons that passed both filters ( high local density and SD-like element ) constitute ∼2 . 5% of all codons and we refer to them as ‘programmed pauses’ . We first compared the pause occurrence of three protein classes: cytosolic , ( inner ) membrane , and periplasmic proteins . The ribosome density profiles of all coding sequences were aligned at their start codon , and the fraction of programmed pause codons in each position was analyzed . The results show that membrane proteins are enriched with pauses between codons 16 and 60 , after which they level-down to values similar to cytosolic proteins ( Figure 1A ) . Pauses occur in roughly 4% of the proteins at each codon position in membrane proteins within the codon range 16–60 ( Figure 1A ) . Indeed , over the entire range of codons from position 16 to 60 , we find that 69% of membrane proteins have at least one programmed pause position , significantly higher compared to cytosolic ( 50% , p = 1 . 5 × 10−7 ) and periplasmic proteins ( 41% , p = 6 × 10−8 , Z-test for 2 proportions ) ( Figure 1A , inset ) . The occurrence of pauses in this codon range indicates that pauses are specifically enriched in membrane proteins only in the early stages of translation elongation , in a coding sequence region which coincides with the stage of membrane protein targeting . Two distinct peaks of elevated pauses were observed , the first in codons 16–36 and the second in codons 40–60 ( defined here as ‘region I’ and ‘region II’ , respectively , Figure 1A ) . In the following sections , we study each of these peaks separately . 10 . 7554/eLife . 03440 . 003Figure 1 . Frequency of programmed pauses across coding sequences . ( A ) Percentage of the programmed pauses in every codon across the coding sequences in membrane , cytoplasmic , and periplasmic proteins . Regions of elevated pause in membrane proteins are marked by I and II . Pause codon positions refer to codons in the A-site of the ribosome . Inset: percentage of proteins from different classes having at least one pause in the codon range 16–60 . ** indicates p < 10−6 . Error bars in figure and inset indicate s . e . for proportion . ( B ) Histogram of the distribution of the positions of the first codon of TM2 in E . coli membrane proteins that were analyzed in ( A ) . N , number of proteins in which TM2 starts at the indicated position . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 00310 . 7554/eLife . 03440 . 004Figure 1—figure supplement 1 . Detection of mRNA-programmed pauses . ( A and B ) dskA gene as an example . ( A ) Median-normalized ribosome density profile for the E . coli gene dskA based on data from Li et al . ( 2012 ) . Y axis scales to the left and right indicate the median-normalized and un-normalized values , respectively . Blue dashed line indicates median density of the gene . Red dashed line indicates the threshold for pause calling ( density > ∼sixfold over the median ) . Orange indicates codons with programmed pauses ( i . e . , having upstream SD-like elements ) . ( B ) Calculated affinity of sequences in the coding sequences for the ribosome anti-Shine–Dalgarno sequence ( aSD ) . Red dashed line indicates the energetic threshold for SD-like element calling ( ΔG < −3 . 1 ) . ΔG values above 0 are shown in light gray . Asterisk marks SD-like elements that explain the programmed pauses indicated in orange in ( A ) . ( C ) Venn diagrams representing the effect of modifying the energetic threshold for SD calling . A total of 868 , 507 codons from well-expressed genes were analyzed . Blue , fraction of slowly translated codons ( density > 6 . 06 ) ; pink , fraction of codons with at least one upstream sequence element predicted to bind the ribosome aSD with the indicated ΔG ( ‘Materials and methods’ ) . The intersection of blue and pink makes up the ‘programmed pause’ codons in each energetic threshold . Energetic thresholds analyzed represent the bottom 1% , 5% , or 10% genome-wide . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 00410 . 7554/eLife . 03440 . 005Figure 1—figure supplement 2 . Amino acid and codon biases are not mediating pauses in membrane proteins' codons 16–60 . ( A and B ) Frequencies of positively charged residues ( R+K ) that were shown to slow-down translation ( Charneski and Hurst , 2013 ) in regions of programmed pauses . ( A ) Frequencies of R+K along the coding sequences in membrane ( mem ) , cytosolic ( cyt ) , and periplasmic ( per ) proteins . Membrane proteins are depleted of K and R in codons 16–60 ( and also throughout the sequence ) compared to cytosolic proteins . ( B ) Frequency of R+K in regions immediately preceding programmed pauses that occur in codons 16–60 . Programmed pause events that occur in codons 16–60 were analyzed for each protein class separately , and the peptide sequences directly preceding pauses were aligned such that position 0 constitutes the pause codon position . The frequency of K+R in every position is shown as solid line . Dashed lines indicate the average frequencies of K+R in the entire protein sequences . No significant excess of positively charged amino acids preceding pauses is observed . ( C ) Codon rarity along coding sequence position , as indicated by mean relative synonymous codon usage ( RSCU ) ( Sharp et al . , 1986 ) . Most of the bias for rare codons occurs before codon 15 ( dashed line ) , while the pause-enrichment in membrane proteins occurs only later . ( D ) Frequencies of G ( Gly ) residues , whose G-nucleotide-rich codons may mediate pause ( Li et al . , 2012 ) , along the coding sequences . Membrane proteins show only a slight elevation of Gly occurrence , which is not specific to the region of codons 16–60 . ( E ) Analysis of coding reading frames of SD-like sequences preceding pauses indicates that most such motifs do not occur in frames that would encode glycines . Left: the canonical SD sequence ( red ) and definition of frames . Frame 1 was defined as the frame in which the canonical SD encodes Gly–Gly dipeptide . Right: distribution of pause-preceding SD motifs in each frame . The analysis was done separately for membrane proteins in the codon range 16–60 , for the other codons in membrane proteins , or for cytosolic proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 005 An important concern was that the pauses may occur as by-products of amino-acid or codon composition biases near the N-terminus of membrane proteins . In Figure 1—figure supplement 2 , we exclude several such biases for membrane proteins , by showing that positively-charged residues , glycince codons , or rare codons are not significantly enriched in codons 16–60 in membrane proteins . Notably , however , such abundant biases in periplasmic proteins may explain the apparent depletion of pause in their N-terminal codons ( Figure 1A , Figure 1—figure supplement 2 ) , likely due to the presence of N-terminal signal peptides . We first investigated the programmed pause enrichment in region II , around codons 40–60 ( Figure 1A ) ; region I is discussed in a later section . We noticed that region II coincides with the location of the beginning of the second transmembrane segment ( TM2 ) in many membrane proteins ( Figure 1B ) . To test a potential link , we aligned all membrane proteins according to their TM2 start ( Figure 2A ) and calculated the frequency of pauses along the aligned coding sequences at various distances from that point . Indeed , we found a peak of high pause frequency just before the start of TM2 ( Figure 2C ) ( p = 3 × 10−6 , 2-sample t test ) , extending from five codons before TM2 to the second codon of the TM . This observation suggests that membrane protein translation is programmed to pause just before TM2 residues are added to the elongating nascent chain ( Figure 2B ) . Curiously , the peak is much sharper when considering only proteins with intracellular N-termini ( Figure 2D ) ( p = 4 × 10−9 , 2-sample t test ) , in which TM2 occurs in ‘Type-I topology’ ( the N-terminus of TM2 is extracellular ) , suggesting that proteins with extracellular N-termini do not pause before TM2 . To explore the difference between the two groups , we analyzed amino acid biases before TM2 ( Figure 2—figure supplement 1A ) . The main difference was excess of arginines and lysines before the TM in proteins with extracellular N-termini , in accordance with the ‘positive inside rule’ ( von Heijne , 2006 ) . This difference is predicted to increase pauses in this location ( Charneski and Hurst , 2013 ) , unlike the actually observed decreased pauses in this group of proteins . Thus , differences in amino-acid sequences cannot account for the difference in the pausing behavior . We attempted to explore if proteins with extracellular N-termini may instead have pauses at other locations , however , analyses of these proteins were precluded by low statistical power due to small sample size ( only 88 proteins in our data set are predicted to have extracellular N-termini , as opposed to 237 with intracellular N-termini ) . Therefore , with the current limitations , we cannot conclude if these proteins may have pauses elsewhere or if they do not utilize the pause strategy altogether . We therefore excluded these proteins , which constitute a minority of the membrane proteome , from further analysis in sections that deal with pauses in region II . 10 . 7554/eLife . 03440 . 006Figure 2 . Pause before translation of TM2 . ( A ) Membrane proteins were aligned such that the first residue of TM2 is aligned to position 0 ( red line ) . ( B ) Visual scheme of the stage at which pause occurs during translation . Red marks SD-like element in the mRNA . Blue line and cylinders depict nascent polypeptide and possible TM location in the tunnel , respectively . ( C and D ) The frequency of codons having programmed pauses in every position was analyzed , either in all membrane proteins ( C ) or only in proteins having intracellular N-termini ( D ) . Profiles of proteins aligned to TM2 are in blue . Profiles of proteins aligned to all other TMs are in black . ( E ) Average ribosome density profiles of proteins with ( blue ) or without ( black ) identifiable programmed pauses in codons −5 to +1 relative to TM2 start . ( F ) Same as ( E ) but only in mRNA nucleotide positions lacking upstream SD-like elements . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 00610 . 7554/eLife . 03440 . 007Figure 2—figure supplement 1 . Specificity of programmed pause before TM2 . ( A ) Amino-acid frequencies in various positions relative to TM2 start . Black and blue dashed lines depict proteins with intracellular or extracellular N-temini , respectively . ( B ) Programmed pause profiles of coding sequences aligned to different TMs . Membrane proteins with cytosolic N-termini were aligned such that the first residue of TM-X ( X = 1–5 ) is aligned to position 0 . The frequency of codons having programmed pauses in every position was analyzed , similar to Figure 2D . Light blue dashed lines indicate 95% confidence intervals . Red dashed lines depict the average frequency over the range . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 00710 . 7554/eLife . 03440 . 008Figure 2—figure supplement 2 . Comparison between proteins with and without identifiable pauses before TM2 . ( A ) Mean expression levels of genes in both groups . The gene expression was calculated for each gene as the logarithm of the median ribosome density in its coding sequence , reflecting the level of protein translation . Error bars indicate SEM . ( B ) Mean TM1 hydrophobicity in both protein groups . For each protein , TM1 and 5 flanking residues down- and up-stream were taken . Hydrophobicity was calculated by applying a sliding window of 18 residues and calculating the average Kyte–Doolitle score . The maximal score was taken to represent the TM hydrophobicity . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 008 Notably , when considering only the remaining proteins set ( i . e . , with intracellular N-termini ) , the peak of pause before TM2 could not be explained by amino-acid biases at this location since the slightly enriched amino acids there ( Figure 2—figure supplement 1A ) , are not encoded by G-nucleotide-rich codons that resemble the SD motif . Furthermore , the peak of pause in the proteins was specifically related to TM2 , as we could not find similar peaks when aligning the coding sequences according to other TMs ( Figure 2—figure supplement 1B ) . The pauses occur 35 ± 13 codons after the beginning of the first TM . Given that the exit tunnel of the ribosome admits ∼28 amino acids of the extended polypeptide ( Bornemann et al . , 2008 ) , pausing of the ribosome at these positions means that in most proteins the first TM has partially emerged out of the ribosome during pausing . The pause location is reminiscent of the location at which the translation arrest occurs in the eukaryotic case , when the SRP arrests translation after TM1 was synthesized and exposed out of the ribosome . This suggests a common physiological role of slowing elongation during targeting . Interestingly , only ∼25% of membrane proteins ( 59 out of 237 ) have identifiable pauses at codons −5 to +1 relative to TM2 start , and indeed only these proteins display high local ribosome density in this region ( Figure 2E ) . The quantitative difference in density between proteins with and without pause in Figure 2E suggests that the pause duration is approximately 1 . 5 s ( roughly the time it takes to translate 28 codons , see ‘Materials and methods’ ) . To check the dependence of translation rate on SD-like elements , we repeated the density analysis of Figure 2E with a modified data set in which all nucleotide positions with even weak upstream SD-like motif were excluded from the analysis ( ‘Materials and methods’ ) . The results show that upon SD exclusion , the slow-down in this region was nearly abolished ( Figure 2F ) suggesting that the SD motif is necessary to generate the observed pauses . It is important to note that several important pauses probably escaped our detection . First , the pauses may occur at subtly different locations from codon −5 to +1 relative to TM2 and our thresholds for pause consideration might be too stringent . Secondly , the TMs and their boundaries are not accurately predicted by the existing methods ( in about 25% of proteins , we found disagreements of at least 10 residues in the location of TM2 between two databases , Uniprot and ExTopoDB ) . Nevertheless , it seems that SD-programmed pauses before TM2 occur only in a subset of the membrane proteome and are not crucial for the production of all membrane proteins . We thus hypothesized that proteins with or without a pause before TM2 may have different characteristics . Our analyses revealed two interesting differences . First , proteins with pause had higher expression on an average ( p = 0 . 004 , two sample t test ) ( Figure 2—figure supplement 2A ) . This is consistent with a role of the pauses in promoting the production fidelity , since highly expressed proteins and their mRNAs require more tight quality control and are hence better adapted for fidelity ( Drummond and Wilke , 2008; Tuller et al . , 2010; Levy et al . , 2012 ) . Secondly , we find that TM1 in proteins with pause is less hydrophobic compared to proteins without identifiable pauses , suggesting that proteins with pause may have compromised affinity for SRP ( Lee and Bernstein , 2001 ) , ( p = 0 . 006 , two sample t test ) ( Figure 2—figure supplement 2B ) . This is consistent with a recent study showing that slow elongation rates may rescue defects in the SRP-mediated targeting of membrane proteins with sub-optimal targeting signals , likely by extending the time window for productive targeting ( Zhang and Shan , 2012 ) . The existence of pauses before TM2 in only a subset of membrane proteins suggests the existence of alternative solutions in other proteins . We hypothesized that the role of the pause before TM2 is to delay the synthesis of TM2 after TM1 was synthesized , in order to complete membrane-targeting before synthesis of additional hydrophobic protein segments . If this is correct , then a long loop between these TMs whose translation requires more time could be just as beneficial . Indeed , our analysis revealed that first loops ( i . e . , those between TM1 and TM2 ) are enriched for long lengths of ≥60 residues and depleted of short lengths of ≤10 residues ( Figure 3A , Figure 3—figure supplement 1A ) ( p = 6 × 10−10 and p = 3 × 10−3 respectively , hypergeometric test ) . Consecutive loops locations between the other TMs showed no significant enrichment for longer loop ( Figure 3—figure supplement 1B ) . Although there could be alternative explanations for why membrane proteins evolved to have larger first loops , such loops can certainly delay the translation of TM2 after TM1 was synthesized . Indeed , we found that proteins having long first loops ( ≥60 residues , 34 proteins ) almost never had pauses before TM2 compared to proteins with shorter loops ( Figure 3B ) ( p = 0 . 01 , hypergeometric test ) . These findings suggest that the coding sequences of membrane proteins are designed to delay the translation of TM2 after TM1 was synthesized , either by extending the loops connecting these TMs or by pausing . 10 . 7554/eLife . 03440 . 009Figure 3 . Enrichment of long first loops and its effect on pausing . ( A ) Enrichment of various loop lengths in first loops compared to all other loops . Shown are Log2 values of ratios . Inset: occurrence of long loops of length ≥60 in first loops or other loops . Error bars indicate s . e . for proportion . ( B ) Occurrence of pause before TM2 in proteins with long or short first loops . Error bars indicate s . e . for proportion . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 00910 . 7554/eLife . 03440 . 010Figure 3—figure supplement 1 . Analysis of loop lengths . ( A ) Occurrence of short loops of length ≤10 in first loops or other loops . ( B ) Occurrence of loops with length ≥60 in increasing loop locations or all other loops . Similar to the inset in Figure 3A . The sample sizes of the test loops and the p-values ( hypergeometric test ) for enrichment are given above the bars . Error bars indicate s . e . for proportion . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 010 We next investigated the pause in codons 16–36 ( region I , Figure 1A ) . The frequency of programmed pause codons and the mean codon density in this range were computed separately for the membrane and cytosolic proteins . Programmed pauses were enriched in membrane proteins compared to cytosolic proteins ( p = 4 × 10−8 , Z-test for two proportions ) ( Figure 4A ) . Consequently , local ribosome densities in membrane proteins mRNAs were significantly higher than those of cytosolic proteins ( p = 1 × 10−36 , 2-sample t test ) ( Figure 4C ) , suggesting slower translation . A mean difference of 0 . 56 density units per codon ( Figure 4C ) over a range of 21 codons indicates that translation is slowed by ∼0 . 6 s ( the time required to translate 12 codons , see ‘Materials and methods’ ) compared to the average cytosolic protein . 10 . 7554/eLife . 03440 . 011Figure 4 . Pause in codon range 16–36 . ( A and C ) Comparison of the frequency of programmed pauses ( A ) or mean ribosome density ( C ) in codon range 16 to 36 between cytoplasmic and membrane proteins . Error bars indicate s . e . for proportion ( A and B ) or SEM ( C and D ) . ( B and D ) Same comparisons between membrane proteins in which the first residue of TM1 occurs after or before position 80 of the polypeptide . ( E ) Visual scheme of the stage at which pause occurs during translation , similar to Figure 2B . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 011 A tendency to pause between codons 16 and 36 indicates that membrane proteins are programmed to pause before the nascent polypeptide emerges out of the ribosome exit tunnel ( Figure 4E ) . In this state , the SRP cannot directly interact with the nascent peptide . However , recent studies showed that E . coli's SRP can efficiently target such ribosome-nascent-chain complexes to the membrane ( Bornemann et al . , 2008; Holtkamp et al . , 2012 ) and that this is done regardless of the class of translated proteins ( cytosolic/membrane/periplasmic ) . Thus , increased pausing at this codon range may perhaps give membrane proteins an advantage over cytosolic proteins in being targeted even before direct SRP–nascent peptide interaction is possible . To further test this hypothetical link to the SRP pathway , we considered also a later stage of the pathway in which the peptide emerges from the exit tunnel . A previous study showed that once a sufficiently long peptide emerges , targeting by the E . coli SRP becomes substrate-specific; namely , if the exposed peptide contains a hydrophobic transmembrane segment , then SRP will remain bound to the ribosome-nascent-chain complex and direct its targeting . Otherwise , hydrophilic peptides cause SRP-dissociation ( Bornemann et al . , 2008 ) . We therefore reasoned that membrane proteins having long hydrophilic N-terminal tails before TM1 may constitute a unique case and may not benefit from such pauses , since the emerging hydrophilic peptide would cause SRP-dissociation and productive targeting would occur only later , when the first TM emerges . Indeed , unlike other membrane proteins , proteins with at least 80 residues before TM1 ( n = 16 ) did not show an elevated pause frequency at codons 16–36 ( p = 0 . 01 , Z-test for two proportions ) , nor did they show elevated ribosome density ( p = 0 . 013 , one-tailed two-sample t test ) ( Figure 4B , D ) . Thus , although further studies are necessary , our results suggest that pauses that occur when the peptide is still unexposed may aid in the targeting of membrane proteins without long hydrophilic N-terminal tails . Our findings suggest that slow translation might be beneficial during the targeting of E . coli membrane proteins , but what is the function of these pauses ? A role in promoting production fidelity is suggested by the function of translation arrest in eukaryotes . In order to test this hypothesis directly , we chose a set of 100 membrane proteins to overexpress in E coli . Based on the results above , proteins having long polypeptide stretches ( >50 residues ) before the first TM or between TM1 and TM2 were excluded from the expression set . Overexpressed proteins were hybrids with a C-terminal GFP-His8 , which enabled the separate detection of both the unfolded and folded protein levels , as shown for several membrane proteins ( Waldo et al . , 1999; Geertsma et al . , 2008; Linares et al . , 2010; Müller-Lucks et al . , 2012; Schlegel et al . , 2012 ) . The discrimination is possible because unfolding or aggregation of fused proteins hinders GFP-folding and -fluorescence , whereas folded and fluorescent GFP is indicative of the functional and folded fused protein ( Geertsma et al . , 2008; Linares et al . , 2010; Müller-Lucks et al . , 2012; Schlegel et al . , 2012 ) . The resistance of folded GFP to SDS-denaturation allows the resolution of both GFP forms by SDS-PAGE and Western blotting , since folded GFP migrates faster ( Figure 5—figure supplement 1A ) . To confirm the method , we assessed the detergent-solubility of five different overexpressed membrane proteins . The results of fractionation studies showed that only the folded GFP-coupled forms were soluble with mild detergent , while the unfolded forms were insoluble and probably aggregated ( Figure 5—figure supplement 1A ) . Next , we expressed the entire set of proteins and quantified both the folded and unfolded forms ( see an example in Figure 5—figure supplement 1B ) . Eleven proteins were excluded from the original set due to problematic detection and two additional ones were outliers ( Figure 5—source data 1 ) . Our set of 87 tested proteins covers a wide range of translation speeds at regions I and II , that is , at codons 16–36 and before TM2 ( Figure 5A ) . We divided the set to three quantiles according to translation speed: from the 33% of genes having the fastest translation speed ( lowest ribosome density ) to those having the slowest speed ( high density ) in these codon ranges ( Figure 5A , colors ) . Remarkably , genes with slower translation at these codons showed a 95% elevation in expression of the folded protein form , compared with genes having fast translation ( p = 3 × 10−4 , two-sample t test ) ( Figure 5B ) . This difference was not observed for expression of the unfolded/aggregated form of the proteins ( Figure 5C ) , suggesting that translation speed during targeting specifically influences the extent of correct folding rather than affecting the overall expression level . 10 . 7554/eLife . 03440 . 012Figure 5 . Effect of translation rate at codons 16–36 and before TM2 on membrane protein overexpression . ( A ) Histogram of translation speeds . For each gene , the ribosome density at codons 16–36 and before TM2 was averaged . The distribution of values in the different genes is shown ( N , number of genes with indicated density ) . Colors indicate groups corresponding to different translation speeds at these codon ranges and accompany the rest of the figure . ( B and C ) Mean ( ±SEM ) expression values of the folded ( B ) and unfolded ( C ) forms of proteins from the three different groups . ( D ) Effect of the total expression level on the percentage of unfolding in the different protein groups . ( E ) Folded vs total levels of expression in the three protein groups . Each dot represents a different protein . Dashed line depicts a 1:1 ratio wherein all of the expressed protein is detected in its folded form . The experiment was repeated three times . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 01210 . 7554/eLife . 03440 . 013Figure 5—source data 1 . Quantification of protein overexpression . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 01310 . 7554/eLife . 03440 . 014Figure 5—figure supplement 1 . Detection of folded and unfolded membrane protein–GFP-His8 hybrids . SDS-PAGE analysis of various membrane proteins overexpressed in E . coli . In-gel fluorescence detects only folded GFP . Western blotting with detection of all His-tagged proteins reveals both the folded and unfolded forms . The Western and fluorescence measurements were done on the same gels . ( A ) Fractionation and solubility of selected membrane proteins overexpressed in E . coli . Cells expressing the indicated proteins were disrupted by sonication ( total ) and ultracentrifuged to separate the cytosolic fraction ( sup ) from the membranes and aggregates ( pellet ) . The pellet fraction was then solubilized with the mild detergent DDM which only solubilizes non-aggregated membrane proteins . Detergent-solubility was determined by ultracentrifugation ( DDM-sol and insoluble ) . Only the lower , folded band is soluble with DDM in all cases . ( B ) SDS-PAGE analysis of various membrane proteins overexpressed in E . coli . Upper panel: in-gel fluorescence . Middle panels: low and high exposure of Western blotting . Lower panel: Coomassie stain of all cellular proteins , as a control for protein amounts . Two first lanes in all gels were from independent cell cultures expressing EmrD and served as standard . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 01410 . 7554/eLife . 03440 . 015Figure 5—figure supplement 2 . Effect of translation rates at various positions on protein overexpression . ( A ) Membrane proteins were divided to 3 groups , similar to Figure 5A , according to the rate of translation at codon range 16–36 ( upper panel ) or before TM2 ( codons [−5]–[+1] relative to TM2 start ) ( lower panel ) . The folded vs total levels of expression in the three protein groups are plotted , similar to Figure 5E . ( B ) Translation rates in different codon ranges and how they affect expression . Left panels: distributions of average translation rates in different codon ranges , as estimated from mean codon densities in different proteins ( similar to Figure 5A ) . In increasing downstream codon regions , the number of slow proteins decreases . Right panels: folded vs total levels of expression in the three protein groups ( similar to Figure 5E ) . Slowly translated proteins appear to benefit from better folding , regardless of where the pauses occur . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 015 We then turned to investigate how expression level affects folding and its potential interplay with the speed of translation in determining the quality of folding . The percentage of unfolded protein was significantly correlated with the total level of the protein ( folded + unfolded ) ( Spearman's r = 0 . 65 , p = 6 × 10−12 , Figure 5D ) . This indicates that regardless of the identity of the overexpressed protein , the cell's capacity to produce high amounts of folded membrane proteins is limited , and increased expression levels would result in increased aggregation . We propose that this dependency reflects saturation of the membrane protein biogenesis machineries in the cell , which becomes less efficient upon high membrane protein overexpression . This finding is in agreement with previous indications that membrane proteins are often better expressed in functional form under lower induction regimes ( Wagner et al . , 2008 ) . Consistent with this notion , we observe that as total protein levels increase , the levels of folded proteins approach saturation ( Figure 5E ) . However , the saturation curves show remarkable differences between proteins that are translated with different speed regimes in our codon range of interest . Proteins that are translated fast saturated at relatively low amounts of folded protein ( apart from a single protein ) , while slowly translated proteins converged to higher saturation levels , if they saturated at all ( Figure 5E ) . This effect is contributed by translation slowing at both codons 16 to 36 and at codons before TM2 ( Figure 5—figure supplement 2A ) . The trend is also reflected by the difference between the groups in Figure 5D and suggests that slow translation during targeting supports less aggregation of the overexpressed proteins . Notably , slow translation at other more downstream codon locations may also be beneficial for folding ( Figure 5—figure supplement 2B ) . However , slow translation is much more common at codons 16–60 , that is at the codon region that covers regions I and II , than downstream ( Figure 5—figure supplement 2B ) , suggesting that slowing during targeting is of particular importance . Overall , the results suggest that regardless of the protein identity , slow translation during targeting is associated with better folding . Yet , the effect appears to depend on expression level and is possibly required only when membrane protein biogenesis becomes saturated . We propose that the presence of pauses in many proteins may alleviate the burden on cellular quality control under endogenous expression levels . In order to test the potential causative relationship between slow translation and better folding , we set out to modify the rate of translation of genes by modifying their SD-like elements . In our limited set of genes , we found that engineering SD-like motifs in genes that are translated fast would be easier than tampering with existing SDs in slow genes . Three genes that are translated fast in our codon range of interest were selected , and 12 or 14 mutations were designed for each gene in order to maximize the affinity of the mRNA for the ribosomal anti-SD , while maintaining the amino-acid sequence unaltered ( Figure 6—figure supplement 1 ) . The silent mutations enabled robust increases of anti-SD binding affinity in only two genes , ygdD and brnQ , while for the third gene , ybjJ , it only permitted a moderate increase ( Figure 6A ) . In order to test the effect of the mutations on protein folding , we implemented the GFP-folding assay . Indeed , upon overexpressing the wild-type and mutant genes , we observed that the two mutants with robust increase in SD motifs folded better ( Figure 6B , C , YgdD and BrnQ ) , whereas the third gene showed no difference ( Figure 6B , C , YbjJ ) . Thus , SD-like motifs improve membrane protein folding and prevent aggregation , likely by slowing-down translation . 10 . 7554/eLife . 03440 . 016Figure 6 . Effect of silent mutations introducing SD-like motifs on protein folding and aggregation . ( A ) Effect of mutations on affinity for ribosomal anti-Shine–Dalgarno sequence . Lower panels show calculated binding energies to anti-Shine–Dalgarno sequence , similar to Figure 1—figure supplement 1B . wt , wild-type; mut , mutant . Upper panel indicates the nucleotide positions in the coding sequence that code for TM1 and TM2 . Note that all proteins and coding sequences are longer than the range of 70 amino-acids and 210 nucleotides shown . ( B ) Western blot analysis of protein–GFP-His8 hybrids , similar to Figure 5—figure supplement 1 . ( C ) Densitomentry-based quantification of the percentage of the folded state from the total protein . Error bars represent SEM from three independent experiments . * indicates p < 0 . 01 ( one tailed , paired t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 01610 . 7554/eLife . 03440 . 017Figure 6—figure supplement 1 . Silent mutations introducing SD-like motifs . Amino-acid ( AA ) and coding sequences of test genes and mutants . Only the range of amino-acids whose codons were mutated is shown . Amino-acid range is indicated in parenthesis . Mutated nucleotides are indicated in red . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 017 Our findings indicate that , like in eukaryotes , slow translation during targeting promotes the fidelity of membrane protein production and folding , yet the underlying mechanisms are different . It seems that E . coli compensates for the lack of SRP Alu domain-mediated arrest , at least in part , by strategically positioned SD-mediated pauses . To test the notion of compensation more directly , we turned to analyze the data on B . subtilis , an organism that can potentially utilize both mechanisms . SD-like elements were shown to control translation pauses in B . subtilis , like in E . coli ( Li et al . , 2012 ) . Additionally , although SRP-mediated arrest was never demonstrated in B . subtilis , unlike E . coli , this organism may have all components required for SRP-mediated arrest: the Alu domain of SRP RNA is present and this domain binds a protein ( HBsu ) that may function analogously to the eukaryotic SRP14 ( Nakamura et al . , 1999 ) . In order to examine the pausing behavior of B . sublitis , we analyzed ribosome profiling data of this organism ( Li et al . , 2012 ) . We deliberately analyzed pause events , irrespective of the presence of SD-like sequences , since we suspected that in this bacterium pauses might not necessarily be programmed into the mRNA sequence . Thus , all codons having high local densities ( within the top 5% ) were counted as pauses . First , we compared membrane and cytosolic proteins by aligning their coding sequences to their start codons and analyzing the fraction of pause codons in each position . Figure 7A , B show that in both B . subtilis and E . coli , pauses were enriched in membrane proteins compared to cytosolic proteins; however , the precise positioning at which the pauses occurred was different . While in E . coli the enrichment occurred around codons 15–65 , in B . subtilis it occurred only downstream around codons 45–85 , with a peak at codon 64 . Notably , the positioning of transmembrane segments along the sequence was comparable between E . coli and B . subtilis ( Figure 7A , B , orange bars and Figure 7—figure supplement 1 ) . 10 . 7554/eLife . 03440 . 018Figure 7 . Comparison of pausing events in B . subtilis and E . coli . ( A and B ) Percentage of pauses in codons across coding sequences in membrane ( mem ) and cytoplasmic ( cyt ) proteins in E . coli ( A ) and B . subtilis ( B ) . Orange histograms at the bottom indicate the distribution of the position of the first codon of TM2 ( similar to Figure 1B ) . ( C ) B . subtilis membrane proteins were aligned such that the first residues of TM1 or TM2 or TM3 are aligned to the position 0 . The percentage of codons having pauses in every position was analyzed . ( D ) Visual scheme of the stage at which pause occurs during translation in B . subtilis , similar to Figure 2B . ( E ) Effect of position relative to TM1 on ribosome densities in positions lacking upstream SD-like elements . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 01810 . 7554/eLife . 03440 . 019Figure 7—figure supplement 1 . Histograms of TM locations in B . subtilis and E . coli . The first TM residue was taken as the TM position , similar to Figure 1B . N , number of proteins in which the TM starts at the indicated position . DOI: http://dx . doi . org/10 . 7554/eLife . 03440 . 019 The different positioning indicates that unlike E . coli , the pauses in B . subtilis were not localized to the regions at or before TM2 ( Figure 7B ) . We suspected that the Alu domain-mediated arrest is functional in B . subtilis , and therefore the pauses are expected to occur upon direct interaction of SRP with the first TM , when the TM emerges out of the ribosome exit tunnel . To examine this , we aligned the coding sequences of B . subtilis membrane proteins to the positions of the first residues of TM1 and the average pause profile was analyzed . This analysis showed a high elevation of pause events that occurred roughly in 45–65 codons after synthesis of the first residue of TM1 ( Figure 7C , D ) . Notable , albeit lower , pause peaks were also observed when aligning the coding sequences to TM2 and TM3 ( Figure 7C ) . However , considering that the distance between TM1 and TM2 or TM3 is not very variable ( median distances 35 and 72 , respectively ) , these peaks most likely reflect shifted and skewed representations of the peak downstream to TM1 . The position of the pause 45–65 codons downstream to TM1 agrees well with the exposure of the first TM out of the ribosome , where it can directly interact with the SRP ( Houben et al . , 2005 ) . If this pause is indeed mediated by the Alu domain , then it should operate regardless of SD-like elements . To test SD-independent pauses , we excluded all positions in the sequence that have even weak upstream SD-like elements ( ‘Materials and methods’ ) . Using this SD-depleted data set , we then checked how the translation rate ( as judged by the local ribosome density ) is affected by the position relative to TM1 . The results show that ribosome density 45–65 codons downstream to TM1 was significantly elevated compared to other positions ( Figure 7E ) , suggesting that B . subtilis utilizes a pause strategy that is independent of the SD-like sequences , possibly involving the SRP Alu domain . This trend was not observed in E . coli ( Figure 7E ) , suggesting that in the absence of SD-like sequences , the ability of the E . coli SRP to arrest elongation is low , if any . Taken together , our results suggest that B . subtilis employs the eukaryotic-like strategy for slowing the elongation of translated membrane proteins , and consequently does not utilize SD-mediated pauses to this end . Membrane proteins constitute roughly 25% of the proteomes in all living organisms ( Wallin and von Heijne , 1998 ) and their hydrophobic nature necessitates co-translational insertion . The ubiquitous SRP system is responsible for the timely delivery of ribosomes translating membrane proteins to membrane translocons . The conservation of the Alu domain in both eukaryotes and Gram-positive bacteria suggests that slow elongation during targeting is universally important ( Mason et al . , 2000; Lakkaraju et al . , 2008 ) . Therefore , the lack of this domain in Gram-negative bacteria , such as E . coli , was puzzling since it was not clear if and how such organisms can slow the elongation during targeting . In principle , the control of ribosome speed could be encoded in cis in the translated mRNA , or in trans , for example as in the case of the SRP Alu domain . Recent in vivo studies indicated that the E . coli SRP can inhibit membrane protein translation , but the precise underlying mechanism and the potential for influencing elongation rate are not yet clear ( Yosef et al . , 2010 ) . Our analyses of in vivo translation rates in E . coli and B . subtilis support the notion that strong SRP-mediated arrest that is SD-independent requires the Alu domain and is therefore probably missing in Gram-negative bacteria . Remarkably , our results suggest that E . coli depends , at least in part , on a cis-encoded means to attenuate translation during targeting , presumably due to the lack of the Alu domain . Other bacteria probably also need to compensate for the lack of Alu domain-mediated translation arrest , and the utilization of SD-mediated pauses might be a general solution . The mRNA-programmed mechanism is reminiscent of the Alu domain-mediated one . Both mechanisms act to slow elongation in the early stages of translation , in order to facilitate membrane protein-folding and avoid aggregation . However , there are also notable differences . First , the position of the pauses in E . coli is different from the position of Alu domain-mediated arrest , with the former occurring earlier in translation . The reason for this is currently unclear , but it might be related to the potential for earlier targeting of ribosomes in E . coli , even before the nascent peptide emerges from the ribosome exit tunnel ( Bornemann et al . , 2008; Holtkamp et al . , 2012 ) . Secondly , while the prokaryotic pausing relies on information embedded in the mRNA , the eukaryotic counterpart reads the protein . Lastly , there is a principal strategic difference between the two mechanisms . The SRP-mediated arrest is only relieved after proper targeting and SRP-dissociation , thus ensuring the fidelity of the process ( Akopian et al . , 2013 ) . In contrast , at present it is reasonable to guess that the mRNA-programmed mechanism probably pauses for a fixed amount of time , without any feedback about the targeting status . In that respect , the SRP-mediated mechanism appears to be more efficient and elegant . Notably , however , it remains to be studied whether SD-mediated pauses may also be influenced by external factors that may report on the translocon arrival , as shown for translation arrest-peptides ( Ismail et al . , 2012 ) . An attractive possibility is that SD-mediated pauses may somehow cooperate with the bacterial SRP and utilize the feedback provided by SRP-dissociation upon completion of targeting . Such a mechanism may also explain how the E . coli SRP regulates translation ( Yosef et al . , 2010 ) . Slow translation is not essential for membrane–protein targeting . In eukaryotes , the arrest function can be abolished by mutation without severely compromising viability ( Mason et al . , 2000; Lakkaraju et al . , 2008 ) . Similarly , in E . coli we observe a notable variability in the rate of translation between different membrane proteins . Our results indicate that slow translation is important especially at elevated protein amounts , when the cell's membrane protein biogenesis capacity might approach saturation . This observation agrees well with recent results in mammalian cells showing that slow translation helps in overcoming rate limitations that arise due to low concentrations of the SRP receptor . It therefore seems that both eukaryotes and prokaryotes possess limiting amounts of membrane protein biogenesis machineries , which may frequently be saturated , creating rate-limiting bottlenecks . These saturation problems are alleviated by slowing down the translation in strategic locations , which allows more time for targeting to complete before further polypeptide synthesis . In a broader perspective , translation pauses are a general strategy that is employed when a process that requires only a portion of the polypeptide is hindered by additional polypeptide segments . Two such processes are the co-translational targeting and the folding of individual domains in the multidomain proteins ( Zhang et al . , 2009; Pechmann et al . , 2013 ) . The time separation provided by the pause allows the completion of these processes without interruption , thus helping to avoid problems in protein folding and aggregation . Finally , we note that our findings may have implications to membrane protein research in general . Heterologous expression of eukaryotic membrane proteins in E . coli is notoriously difficult , often leading to poor expression of functional proteins , creating a serious bottleneck in the field ( Granseth et al . , 2009 ) . Our study adds to recent studies revealing the existence of sequence preferences within coding sequences of membrane proteins in E . coli ( Prilusky and Bibi , 2009; Nørholm et al . , 2012 , 2013 ) , which may need to be obeyed by heterologous coding sequences . For example , SD-mediated pauses are probably lacking a priori in heterologous coding sequences . It therefore seems that in that respect , Gram-positive bacteria such as B . subtilis might be more compatible with eukaryotic membrane protein heterologous expression ( Kunji et al . , 2005 ) , since they contain a translation arrest logic similar to the eukaryotic one . Genome sequences and associated data were downloaded from the NCBI FTP site ( accessions NC_000913 for E . coli and NC_000964 for B . subtilis ) . Ribosome profiling data were taken from NCBI GEO accession GSE35641 ( samples GSM872393 and GSM872397 ) . Predicted locations of TMs for E . coli were from ExTopoDB ( Tsaousis et al . , 2010 ) . For B . subtilis , TMs were predicted using the TMHMM algorithm ( Krogh et al . , 2001 ) . Only proteins having at least two TMs were considered as membrane proteins . Cytosolic and periplasmic protein localizations were taken from Uniprot . Ribosome density data were obtained from Li et al . ( 2012 ) . Only genes with sufficiently high ribosome density reads were analyzed ( median density of reads across the coding sequence >0 . 2 ) . For each gene , the ribosome density in every nucleotide was divided by the median ribosome density across the coding sequence . Codon-wise density was calculated by averaging the nucleotide density in every three nucleotides of each codon . These densities indicate ribosome occupancy when the respective codons are located in the A-site of the ribosome . The first and last four codons were excluded from further analysis since they contain high densities associated with the slow translation initiation and termination . The threshold for pause codon consideration ( top 5% of codon-wise density ) was calculated by collecting codon density from the entire genome ( for E . coli and B . subtilis , the thresholds were 6 . 06 and 23 . 4-fold higher than the median , respectively ) . Programmed pauses in E . coli were defined as codons having density above these thresholds and also having at least one SD-like element 8–11 nucleotides upstream of at least one of the codon nucleotides . Shine–Dalgarno-like elements were identified based on the calculated binding energy of 10 nucleotide-long sequences to the ribosomal anti-SD , as described ( Li et al . , 2012 ) . All positions having calculated binding energy of ΔG < −3 . 1 kcal/mol , which constitute the lowest 5% ( genome-wide ) , were considered as SD-like elements . We note that these thresholds represent a compromise since even higher ΔG values may affect the ribosome density , and in addition many places with lower binding energies show no indication of pause ( see Figure 1—figure supplement 1C ) . To estimate differences in translation time , we calculated how much extra ribosome density is associated with a particular set of genes compared to a control set of genes over a specific range of codons . For codon range 16–36 , we compared membrane and cytosolic genes over codons 16–36 . A mean difference of 0 . 56 density units per codon ( Figure 4C ) over a range of 21 codons yields a difference of 11 . 8 density units . For codons before TM2 , we computed the difference between proteins having or lacking programmed pauses at codons −5 to +1 relative to TM2 start . Mathematically , the difference is equal to the area trapped between the two curves in Figure 2D and equals 28 density units . Since the density units in the genes' profiles are normalized to the median codon density of the gene , it is assumed that each single density unit corresponds to the median time it takes the ribosome to translate a single codon . To estimate translation time , we used a previously determined value of elongation rate for the E . coli ribosome of 1 codon per 0 . 058 s ( Young and Bremer , 1976 ) . All mRNA positions having calculated the binding energy of ΔG < 2 kcal/mol were considered potential SD-like elements . Accordingly , ribosome densities 8–11 nucleotides downstream of them were excluded from analysis . The median ribosome density of each gene from Li et al . ( 2012 ) was taken as expression level . Compared with mRNA levels , this quantity correlates better with protein expression as it reflects both RNA levels and the level of translation from each transcript ( Ingolia et al . , 2009 ) . Sequences corresponding to the first TM were analyzed . TMs shorter than 24 residues were increased to 24 by symmetrically including flanking residues from both sides . A sliding window average of hydrophobicity was calculated with a window length of 18 residues using the Kyte–Doolittle hydrophobicity scale ( Kyte and Doolittle , 1982 ) . Only the highest hydrophobicity score was kept for each TM . Plasmids encoding membrane proteins–GFP-His8 hybrids were previously reported ( Daley et al . , 2005 ) . Only proteins having cytosolic N- and C-termini ( Daley et al . , 2005 ) were analyzed . The plasmids were transformed to E . coli BL21 ( DE3 ) pLysS . Overnight cultures grown in LB with kanamycin ( 30 μg/ml ) and chloramphenicol ( 34 μg/ml ) were diluted 1:100 into fresh ( same ) medium and allowed to grow for 2 . 75 hr at 37°C , until they reached a typical OD600 of 0 . 4–0 . 6 . Cultures were then induced with 0 . 4 mM IPTG and allowed to grow for additional 2 hr and transferred to ice . Cells were harvested by centrifugation at 4°C , washed once with ice-cold buffer A ( 50 mM Tris–HCl pH 8 , 100 mM NaCl , 1 mM EDTA ) , resuspended in the same buffer at an OD600/ml of 1 . 5–2 , and kept at −20°C . For disruption , cells were thawed , supplemented with 1 mM phenylmethanesulfonyl fluoride ( PMSF ) , and sonicated well until the solution cleared completely . The procedure was carried at 4°C , and the samples were kept from all fractions . The crude sonicated cells ( 0 . 8 ml ) were ultracentrifuged at 200 , 000×g for 30 min . The pellet was thoroughly resuspended in 0 . 72 ml of buffer A supplemented with 1 mM PMSF and then solubilized by the addition of 80 μl 10% β-dodecylmaltoside ( Anatrace ) and homogenized extensively . Samples were incubated for 30 min on ice and then re-centrifuged at 200 , 000×g for 30 min . Portions corresponding to equal amount of the starting material were taken from each fraction , supplemented with 1:4 of 5× SDS sample buffer ( 120 mM Tris–HCl pH 6 . 8 , 50% glycerol , 100 mM DTT , 2% SDS , and 0 . 1% bromophenol blue ) , and loaded to 10% SDS-PAGE for analysis . In-gel fluorescence was recorded using Molecular Dynamics Typhoon 9400 ( GE Amersham ) . The gels were subsequently transferred to nitrocellulose membranes for Western analysis . His-tagged proteins were detected with His probe-HRP ( Pierce ) . Cultures were grown in 96-well plates with each clone in a single well . Each growth experiment included two independently grown cultures harboring the emrD gene as internal standards . Sonicated cells ( 0 . 2 ml ) overexpressing membrane proteins were mixed with 0 . 05 ml 5× SDS sample buffer and loaded to 10% SDS-PAGE gels . Each gel included the two emrD standard clones . The gels' proteins were detected for in-gel fluorescence and Western blotting ( see above ) . Densitometric quantifications were carried using ImageQuant TL software ( GE Healthcare , Tel-Aviv , Israel ) . First , the amount of the folded protein was extracted from fluorescence measurements and normalized to that of the folded EmrD ( the amount of folded EmrD protein was defined as 1 arbitrary unit ) . The fraction of folded and aggregated protein for each clone was calculated based on quantification of Western blots . The amounts of aggregated and total protein were calculated by combining the knowledge on amount of the folded protein ( from fluorescence ) with the fraction of the folded protein ( from Western blotting ) . Quantification values are reported in the Figure 5—source data 1 . Mutagenesis was achieved by PCR with 5′-phosphorylated mutagenic primers ( Sigma-Aldrich ) that amplify the entire plasmid . The PCR products were blunt-end ligated and transformed to E . coli DH5α . Mutant plasmids were sequenced to verify that they contain only the desired mutations . The primers sequences were: ygdD mutagenesis ( GGGCGGTGGAGATGGGGTGGATACAGACGGGGCTCGAATACCAGGCGTTTC and CCATCGTCTTACTCAACACATGCGCCCCAAACGCCCCCAGAGCCACAAAAATGAAG ) ; brnQ mutagenesis ( CCTCCAATGGTGGGGTTGCAGGCGGGGGAGCACGTGTGGACTGCGGCATTCG and GAAAATAATGTTCCCCGCCCCCACAAACAACGCAAATGTCATAAAGCC ) ; ybjJ mutagenesis ( GGGCGACGAGGACGCCGGCGATAAGGGATATTCTCTCTGTCTCG and ACGACGCCATCAACAACCCCGGCAAAAAGAAGAACATAA ) .	Proteins are built as long chain-like molecules . First , a length of DNA is copied to make a messenger RNA ( or mRNA ) molecule , which then binds to a large molecular complex called a ribosome . The ribosome reads and translates the code in the mRNA sequence to build a protein chain , which then folds into a specific three-dimensional shape to allow the protein to perform its function . Many proteins also need to be targeted to the right location within the cell in order to carry out their role . Some proteins have to be inserted into the membranes of cells and these proteins are directed , as they are being built , to the membrane by another molecular complex called the signal recognition particle ( or SRP for short ) . The SRP binds to the new protein as it emerges from the ribosome and helps to direct it to the membrane . To make sure that membrane proteins fold correctly , their translation is paused whilst the protein is being targeted to the membrane . Plants , animals , and other eukaryotes do this via a unique part of the SRP complex that only allows the translation to continue once the translating ribosome has been brought close to the membrane . However , most bacteria lack this part of the SRP complex , and yet they are still able to accurately insert new , correctly folded , proteins into their membranes . This suggests that an alternative mechanism must exist in bacterial cells . Fluman et al . looked at an existing data set that had measured how many ribosomes are found at different points along the length of mRNA molecules at any given time in the bacterium E . coli . If ribosomes are consistently found at specific sites in given mRNA molecules , it suggests that these are the sites where a pause in translation occurs . Specific short mRNA sequences—that bind to a ribosome and hold it in place—are often found in these pause sites . These sequences are similar to another sequence , called the Shine–Dalgarno sequence that is often also found at the very start of an mRNA molecule , where it functions to recruit a ribosome and begin the translation process . Fluman et al . reveal that mRNAs of membrane proteins contained these similar sequences early on in their coding region . Some looked likely to pause the translation before the newly formed protein chain emerged from the ribosome , which could give the ribosome time to be targeted to the membrane . Other Shine–Dalgarno-like sequences were found slightly later on in the mRNA molecules for protein chains that span back-and-forth through the membrane several times . Fluman et al . show that slowing translation in this manner—which is different to that used by eukaryotes—helps to ensure that membrane proteins are folded correctly in E . coli . Although these pauses occur frequently , mainly in the early stages of the translation of membrane proteins , there are many other translation pause sites that are known to exist in other mRNAs . The next challenge is to understand the function of these other pause sites , and how they work together with other mechanisms to regulate translating ribosomes inside cells .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "cell", "biology" ]	2014	mRNA-programmed translation pauses in the targeting of E. coli membrane proteins
An extracellular matrix of Fibronectin adheres the neural tube to the two flanking columns of paraxial mesoderm and is required for normal vertebrate development . Here , we find that the bilaterally symmetric interfaces between the zebrafish neural tube and paraxial mesoderm function as optimally engineered adhesive lap joints with rounded edges , graded Fibronectin ‘adhesive’ and an arced adhesive spew filet . Fibronectin is a ‘smart adhesive’ that remodels to the lateral edges of the neural tube-paraxial mesoderm interfaces where shear stress is highest . Fibronectin remodeling is mechanically responsive to contralateral variation morphogenesis , and Fibronectin-mediated inter-tissue adhesion is required for bilaterally symmetric morphogenesis of the paraxial mesoderm . Strikingly , however , perturbation of the Fibronectin matrix rescues the neural tube convergence defect of cadherin 2 mutants . Therefore , Fibronectin-mediated inter-tissue adhesion dynamically coordinates bilaterally symmetric morphogenesis of the vertebrate trunk but predisposes the neural tube to convergence defects that lead to spina bifida . The vertebrate central nervous system develops from the neural tube which is created via the complex morphogenic processes of convergence and closure of the neural ectoderm during neurulation . Human neural tube defects such as spina bifida arise via a failure of the neural tube to converge and fuse along the dorsal midline . Convergent-extension cell movements along with apical constriction of neural tube cells drive neurulation at the cellular level ( Wallingford et al . , 2013; Copp et al . , 2015 ) . In zebrafish , the neural ectoderm converges along the dorsal midline to form a neural tube with a slit-like lumen , and during posterior trunk development studied here , neural convergence is accompanied by extension both posteriorly and along the dorsal-ventral axis but without an increase in tissue volume ( Papan and Campos-Ortega , 1994; Lowery and Sive , 2004; Harrington et al . , 2010; Steventon et al . , 2016 ) . At the tissue level , the underlying mesoderm is required for neural tube convergence in zebrafish ( Araya et al . , 2014 ) . In Xenopus , the epidermal ectoderm is pulled toward the midline by the neural ectoderm and is necessary for neural convergence ( Davidson and Keller , 1999; Morita et al . , 2012 ) . Conversely , studies of chick neurulation suggest that the epidermis pushes the neural tube closed ( Colas and Schoenwolf , 2001 ) . Locally varied tissue mechanics are exhibited during posterior mouse neural tube closure where a supracellular network of F-actin zippers the neural tube , creating spatially restricted zones of positive and negative strain within the tissue ( Galea et al . , 2017 ) . Overall , these studies highlight the tissue level mechanics and inter-tissue interactions involved in neural tube morphogenesis . Cell-cell adhesion is a crucial biomechanical process that regulates neurulation , since loss of n-cadherin/cadherin 2 leads to neural tube convergence defects in mouse and zebrafish ( Luo et al . , 2001; Lele et al . , 2002; Harrington et al . , 2007 ) . cadherin 2 also regulates the morphogenesis and segmentation of the paraxial mesoderm , which flanks the left and right sides of the neural tube and contributes to the vertebrae of the spinal column ( Radice et al . , 1997; Luo et al . , 2001; Horikawa et al . , 1999; McMillen et al . , 2016; Chal et al . , 2017 ) . The presomitic mesoderm ( PSM ) is the posterior paraxial mesoderm within the extending tailbud . The PSM stiffens as it develops , and this tissue solidification requires cadherin 2 and is important for body elongation ( McMillen and Holley , 2015; Zhou et al . , 2009; Mongera et al . , 2018 ) . cadherin 2 is further required for ordered tailbud cell migration and mutants exhibit shortened body axes ( Lawton et al . , 2013 ) . An extracellular matrix of Fibronectin coats the paraxial mesoderm and mediates inter-tissue adhesion between the paraxial mesoderm and the neural tube and notochord ( Dray et al . , 2013; Araya et al . , 2016 ) , Fibronectin matrix assembly is dependent upon its Integrin receptors , principally Integrin α5β1 and αVβ3 ( Schwarzbauer and DeSimone , 2011 ) . Soluble Fibronectin is readily available to cells in the developing zebrafish trunk , but Fibronectin matrix assembly only occurs when Integrins adopt an active conformation in which they can bind Fibronectin ( Jülich et al . , 2009 ) . Cadherin 2 represses Integrin α5β1 activation by physically associating with and stabilizing a complex of inactive Integrins on adjacent cells ( Jülich et al . , 2015; McMillen et al . , 2016 ) . On the surface of the paraxial mesoderm and along the somite boundaries , there is little Cadherin 2 , and therefore , Integrin α5β1 is activated and a Fibronectin matrix is assembled ( Jülich et al . , 2015; McMillen et al . , 2016 ) . Surprisingly , loss of the Fibronectin receptors leads to a shortened body axis but does not affect cell migration ( Dray et al . , 2013; Yang et al . , 1999 ) . Thus , it is unclear how the Fibronectin matrix functions during body elongation and neural tube morphogenesis . Fibronectin fibrillogenesis is an intrinsically mechanical process involving the unfolding of soluble Fibronectin dimers , non-covalent crosslinking and progressive assembly into larger fibers . Fibronectin fiber assembly is promoted by cell actomyosin contractility , and fiber orientation aligns with cell traction forces in 2D cell culture ( Zhong et al . , 1998; Lemmon et al . , 2009 ) . Similarly , constitutively active myosin regulatory light chain activates Integrin α5 , colocalizes with the activated Integrin , and induces Fibronectin matrix assembly in the zebrafish paraxial mesoderm ( Jülich et al . , 2015; McMillen et al . , 2016 ) . In 3D in vitro micro-tissues , Fibronectin fibers and F-actin co-localize with tissue stress ( Legant et al . , 2009 ) . These results parallel the finding that tissue tension promotes Fibronectin matrix assembly in the Xenopus gastrula ( Dzamba et al . , 2009 ) . Here , we examined the role of Fibronectin-mediated inter-tissue adhesion in neural tube convergence and paraxial mesoderm morphogenesis in zebrafish ( Figure 1A ) . We first performed morphometric analysis of the neural tube and paraxial mesoderm in different genetic backgrounds with reduced cell-cell and/or cell-Fibronectin adhesion . These experiments revealed that inter-tissue adhesion resists neural tube convergence . We developed a simple computational model that predicts several morphological changes due to loss of cell-cell or cell-Fibronectin adhesion . We find that the interfaces between the neural tube and paraxial mesoderm recapitulate features of an ‘adhesive lap joint’ which is commonly used in engineering and is comprised of partially overlapping components bound via an adhesive . Excess adhesive that can ooze from the edges of the overlapping domains is called an ‘adhesive spew’ which can be filleted or sculpted to strengthen the lap joint . Here , the Fibronectin matrix functions as the adhesive in the lap joint formed by the neural tube and left and right paraxial mesoderm . Our computational model , as well as lap joint theory , predicts different mechanical properties for the interface between the paraxial mesoderm and the neural tube compared to the interface between the paraxial mesoderm and epidermis . Morphometric analysis of the Fibronectin matrix and the actomyosin cytoskeleton are consistent with the prediction that there is an increasing medial to lateral gradient of tension along the paraxial mesoderm and neural tube interface . Lastly , we created a photoconvertible Fibronectin transgenic that enables us to paint the extracellular matrix and quantify its remodeling in live embryos . These experiments indicate the Fibronectin matrix continually remodels to the lateral interface of the neural tube and paraxial mesoderm where tension is highest . This remodeling is dependent upon inter-tissue adhesion and convergence of the neural tube , implying that there is shear stress along this tissue interface that drives the remodeling . Moreover , these experiments revealed that Fibronectin matrix remodeling on one side of the body is sensitive to morphogenic variation on the contralateral side of the body due to mechanical coupling via inter-tissue adhesion . Altogether , the data indicate that Fibronectin-mediated inter-tissue adhesion acts as an adhesive lap joint that mechanically coordinates bilaterally symmetric morphogenesis but predisposes the neural tube to convergence defects that lead to spina bifida . We quantified neural tube ( NT ) morphologies in wild type , cadherin 2 mutants ( cdh2-/- ) , maternal zygotic integrin α5 mutants ( MZ itgα5-/- ) and cdh2-/-; MZ itgα5-/- double mutants . We fixed embryos at 12–14 somite-stage and imaged them for Fibronectin and cell nuclei . cdh2 mutants exhibited a wider neural tube than wild-type embryos consistent with its known convergence defect ( Figure 1B–C; Lele et al . , 2002 ) . In contrast , MZ itgα5 mutants exhibit a narrower neural tube ( Figure 1D ) suggesting precocious neural tube convergence after reduction of cell-Fibronectin matrix adhesion . Surprisingly , the convergence defect of cdh2 mutants is rescued in cdh2-/-; MZ itgα5-/- double mutants ( Figure 1E ) . We quantified convergence of the neural tube along the anterior-posterior axis across genotypes . We measured the medial-lateral width of the neural tube ( cartoon in Figure 1F ) on transverse sections every 20 μm along the anterior-posterior axis starting from the last somite boundary until the posterior end of the PSM ( Figure 1—figure supplement 1 ) . In wild-type embryos , the width of the neural tube progressively narrows , comparing posterior to anterior , reflecting convergence over time ( Figure 1F and Video 1 ) . This change in the medial-lateral width of the neural tube is shallower in mutants . In cdh2 mutants , the neural tube is wider than wild-type embryos along the full anterior-posterior length of the PSM ( Figure 1F ) . In MZ itgα5 mutants and cdh2; MZ itgα5 mutants , the neural tube has a narrower medial-lateral width in the posterior PSM compared to wild type ( Figure 1F ) . Similarly , wild-type embryos exhibited a posterior to anterior decrease in the dorsal-ventral length and the cross-sectional area of the neural tube , while the posterior to anterior decreases in these two metrics were shallower in other genotypes ( Figure 1—figure supplement 2A and B ) . These data suggest that inter-tissue adhesion via cell-Fibronectin matrix adhesion constrains neural tube convergence . We examined the effect of loss of inter-tissue adhesion on the PSM and observed an abrogation of bilaterally symmetric morphogenesis . We measured the left-right differences in PSM cross-sectional areas ( Figure 1G ) and found that both MZ itgα5-/- and cdh2-/-; MZ itgα5-/- double mutants exhibited loss of bilateral symmetry . By contrast , cdh2 mutants did not have symmetry defects . Local asymmetries were also observed in MZ itgα5-/- and cdh2-/-; MZ itgα5-/- double mutants in the length of the interface between PSM and neural tube ( PSM\|NT interface ) and the angle formed at the lateral edge of this interface ( Figure 1—figure supplement 2C and D ) . An explanation of these asymmetries is that reduction of cell-matrix interactions creates local tissue detachments ( arrowheads in Figure 1D and E , Figure 1—figure supplement 2E and F ) which in turn leads to disequilibrium in inter-tissue adhesion along the left and right PSM\|NT interfaces . To understand how variable levels of cell-cell adhesion and inter-tissue adhesion underlie the differences in tissue shapes across genotypes , we designed a computational 2D model with variable cell-cell adhesion and inter-tissue adhesion and tested whether the model can predict tissue shapes . We modeled a 2D transverse section of four connected tissues ( NT , left and right PSM and notochord [NC] ) ( see Materials and methods for details ) . In this coarse-grained model , we do not directly consider individual cells , rather the tissues are considered as soft units with an internal pressure ( blue arrows in Figure 2A ) and an elastic surface made of movable points interlinked by springs ( green surface springs in Figure 2A ) . These connected surface springs model tissue surface tension . This tissue surface tension resists the internal pressure and was shown to be proportional to the cell-cell adhesion strength inside a tissue ( David et al . , 2014; Manning et al . , 2010 ) . The tissues are attached to each other by another set of springs ( red springs in Figure 2A ) , which model inter-tissue adhesion via cell-Fibronectin interactions . Overall , there are three parameters in the model: the surface stiffness ( spring constant KS for surface springs ) , the adhesion stiffness ( spring constant Kadh for adhesive springs ) , and the internal pressure ( P ) . The surface stiffness and the adhesion stiffness positively correlate with the levels of Cadherin-dependent cell-cell adhesion and Fibronectin-mediated inter-tissue adhesion , respectively . To determine how the model parameters affect tissue morphology , we analyzed the simulated tissue shapes . The internal tissue pressure should depend on the net fluid content inside a tissue , and there is no obvious reason to vary this quantity across genotypes . Hence , we fixed the internal pressure ( p=5 ) depending on an earlier model of soft grains ( Åström and Karttunen , 2006 ) in such a way that the simulated shapes qualitatively resemble in vivo tissue shapes in 2D transverse sections ( Figure 2B ) . Irrespective of choices of Ks and Kadh , we found that the PSM\|NT and the PSM and epidermis interface ( PSM\|E ) always have very different curvatures ( Figure 2C ) . The PSM\|NT interface is straight and has much higher radius of curvature than the PSM\|E interface which is curved ( Figure 2C ) . This prediction was confirmed experimentally in vivo ( Figure 2C ) . The parameters Ks and Kadh were systematically varied in simulations to assess their influence on the shape of the interface between NT and PSM ( Figure 2—figure supplement 1 ) . We measured two shape metrics: the length of the interface between the PSM and neural tube ( L PSM\|NT , blue boxes in Figure 2B ) and the angle at the lateral edge of this interface ( angle Θ in Figure 2B ) . We found that increasing surface stiffness decreases the interfacial length , while increasing inter-tissue adhesion increases interfacial length ( Figure 2—figure supplement 1A and B ) . Thus , inter-tissue adhesion acts like a ‘zipper’ between two tissues . In fact , low adhesion stiffness decouples the tissues , which is evident from the local detachments seen in our simulations ( Figure 2B top right , arrowhead ) and similar to detachments observed in MZ itgα5 mutants and cdh2; MZ itgα5 mutants ( Figure 1D and E ) . This simple computational model can account for the gradual narrowing of the neural tube along the medial lateral axis as it develops ( Figure 2—figure supplement 1C ) . We can assume that the NT-PSM interface is locally at steady-state along the anterior-posterior axis , and the observed tissue shape change along the anterior-posterior axis may be caused by a progressive increase in Ks along that axis . Indeed , the zebrafish PSM progressively solidifies from posterior to anterior in a cadherin2-dependent manner ( Mongera et al . , 2018 ) . We found that the length of the medial to lateral interface between the PSM and NT ( L PSM\|NT ) steadily decreases with increasing surface stiffness when other parameters are fixed ( Figure 2—figure supplement 1A ) . Given the above correspondence between in vivo and in silico observations , we next assigned reasonable values of surface stiffness ( KS ) and inter-tissue adhesion stiffness ( Kadh ) that reproduce morphometrics of wild-type embryos . We note that in vivo the value of the interfacial length ( L PSM\|NT ) decreases to 0 . 6 of the maximum value at the anterior end of PSM relative to the posterior end ( Figure 2—figure supplement 1C ) . We also found in silico that L PSM\|NT roughly falls to 0 . 6 of the maximum value at KS ≈ 100 for a fixed Kadh ( Figure 2—figure supplement 1A ) . Since we are interested in steady-state shapes near the anterior PSM , we may take this value to represent the wild type . Next , we consider the value of Kadh . In simulations , we varied Kadh for different fixed values of KS , and measured the medial-lateral length of NT relative to the interfacial length between NT and PSM ( Figure 2—figure supplement 1G ) . In vivo , this ratio ( ML length of NT/L PSM\|NT ) is about two on average for the anterior 140 μm of PSM . In silico , we found that irrespective of the KS value , this ratio saturates around a value of 2 . 7 in the range of Kadh = 8 to 12 ( marked in red in Figure 2—figure supplement 1G ) . Based on the above analysis , we then represented the wild-type embryos by a pair of values: ( KS , Kadh ) = ( 100 , 10 ) . After fixing the wild-type parameters , we then chose the parameter values corresponding to cdh2 mutants and MZ itgα5 mutants by matching the in silico and in vivo lengths of PSM\|NT interface in these mutants relative to its wild-type values ( Figure 2E ) . For cdh2 mutants , the mean length of PSM\|NT is around 1 . 5 times higher than wild type . To reproduce the experimentally observed increase of L PSM\|NT relative to wild-type , we assigned reasonable parameter values to cdh2 mutants depending on biological expectations . First , we lowered the surface stiffness relative to wild type , since low cell-cell adhesion is known to reduce tissue surface tension ( David et al . , 2014; Manning et al . , 2010 ) . Second , Cadherin 2 was shown to inhibit Integrin α5 activation and Fibronectin matrix assembly in the PSM ( Jülich et al . , 2015; McMillen et al . , 2016 ) . Therefore , we may assign a higher adhesion stiffness value to cdh2 mutants compared to wild type . After systematic exploration of the parameter space in simulations , we found a pair of parameter values , ( KS , Kadh ) = ( 55 , 12 ) , that reproduce the in vivo increase of L PSM\|NT relative to wild-type ( 1 . 55 times ) . Following the same procedure for MZ itgα5 mutants , the in vivo data indicate that MZ itgα5 mutants exhibit a mean PSM\|NT interface length 0 . 77 times that of the wild-type value . Since cell-matrix interaction is reduced in MZ itgα5 mutants , it is logical to assume a lower inter-tissue adhesion for this genotype relative to wild type , but the surface stiffness was kept same as wild type . We found that a pair of parameter values , ( KS , Kadh ) = ( 100 , 6 . 5 ) , reproduce the experimentally observed decrease of L PSM\|NT relative to wild-type ( 0 . 78 times ) . Using the parameter values corresponding to cdh2 mutants and MZ itgα5 mutants , we can then predict the interfacial length for cdh2; MZ itgα5 mutants simply by combining these parameter values . The double mutants are represented using the value of inter-tissue adhesion in MZ itgα5 mutants , and the same value of surface stiffness as cdh2 mutants ( Figure 2D ) . Hence , cdh2; MZ itgα5 mutants are represented by the values: ( KS , Kadh ) = ( 55 , 6 . 5 ) . These parameter choices for cdh2 mutants and MZ itgα5 mutants accurately predicted the in vivo interfacial length of cdh2; MZ itgα5 mutants , ( Figure 2E ) . Importantly , although we assigned the parameter values depending on the relative interfacial lengths of cdh2 mutants and MZ itgα5 mutants , these parameter choices also reproduced the experimentally observed trends in the variation of interfacial angle Θ across genotypes ( Figure 2F ) . The observation that the PSM\|NT and PSM\|E interfaces have very different curvatures intuitively suggest that they are under different levels of tension . Therefore , we analyzed the tension distributions at tissue surfaces in our simulations ( Figure 3A ) . Interestingly , the model predicted a gradient of tension at the neural tube side of PSM\|NT interface with a higher tension on the lateral side of the interface and a lower tension on the medial side ( see dashed box , Figure 3A ) . In contrast , the medial-lateral tension distribution at the PSM\|E surface is predicted to be homogeneous . To test these predictions in vivo , we analyzed F-actin intensity as an indicator of cortical tension . Consistent with the model prediction , we observed an increasing medial to lateral gradient of F-actin along the PSM\|NT interface ( Figure 3B , F and L ) . A parallel gradient of Myosin-II localization was also observed ( Figure 3—figure supplement 1A–1C ) . Fibronectin matrix assembly is a force-induced process dependent upon cell actomyosin contractility ( Zhong et al . , 1998; Lemmon et al . , 2009; Dzamba et al . , 2009 ) . We therefore analyzed the Fibronectin matrix to determine whether there was a correlation between levels of F-actin and Fibronectin matrix assembly at the tissue interfaces ( Figure 3C and D ) . Since Fibronectin matrix is assembled from soluble and small matrix aggregates into large fibers , we quantified the topology of the Fibronectin matrix by sorting into small and large matrix elements , color coded in cyan and magenta , respectively ( Figure 3H , K , M and O ) . In correlation with the F-actin gradient , we observed a medial to lateral gradient of Fibronectin matrix assembly as large assembled elements were enriched at lateral edge of the interface , while small matrix elements were evenly dispersed . Lastly , we used a monoclonal antibody that recognizes an epitope in human Fibronectin that is exposed when Fibronectin is under tension ( Cao et al . , 2017 ) . In embryos expressing a chimeric zebrafish/human Fibronectin , we observed a medial to lateral gradient of Fibronectin under tension along the PSM\|NT interface ( Figure 3—figure supplement 1D–1F , J and K ) . All together , these data support the model prediction that there is an increasing medial to lateral gradient of tension along the interface between the PSM and neural tube . These medial-lateral gradients of F-actin and Fibronectin matrix were only observed at the PSM\|NT interface . Both F-actin , Fibronectin matrix assembly and Fibronectin under tension were homogenously distributed medial-laterally at the PSM\|E interface ( Figure 3B–O and Figure 3—figure supplement 1G–1I and L–M ) , which is again consistent with the model predictions . Instead , this interface was characterized by a progressive posterior to anterior increase in F-actin density ( Figure 3I and P ) . Similarly , we observed a posterior to anterior increase in the density of large matrix elements and decrease in the density of small matrix elements ( Figure 3K and Q ) . These opposing gradients suggest that the matrix is progressively assembled from posterior to anterior by crosslinking small Fibronectin fibrils to form large fiber networks ( Figure 3J ) . This posterior-to-anterior increase in matrix density and F-actin at the PSM\|E interface also correlates with the progressive epithelialization of PSM surface cells and a posterior to anterior increase of F-actin signal in both internal cells and surface cells of the PSM ( Figure 3—figure supplement 2 ) . Overall , the intriguing observation here is that the homogeneous medial-lateral distribution in F-actin and Fibronectin density observed at the PSM\|E interface is lost at the PSM\|NT interface , which instead displays a medial-lateral gradation both in F-actin and the Fibronectin fibers . Our simplified model suggests that this behavior may result from the particular distribution of mechanical stress at the PSM\|NT interface . This interface between two converging tissues mechanically resembles an ‘adhesive lap joint’ that is commonly used in engineering and is comprised of partially overlapping components bound via an adhesive . The structure formed by the apposition of neural tube and PSM tissues can be idealized as a particular type of lap joint , a single-sided strapped joint ( Ghoddous , 2017 ) , where the neural tube on top of the PSM acts as a single sided strap bridging the two PSMs via a Fibronectin adhesive ( Figure 3R ) . Shear stress at the overlapping interface of a lap joint is highest at the lateral edges of the overlap and lowest in the middle ( Goland and Reissner , 1944 ) . This nonhomogeneous stress distribution can be compensated for by grading the adhesive in the overlap ( Carbas et al . , 2014 ) . In accordance to this engineering insight , we observe a graded Fibronectin ‘adhesive’ at the PSM\|NT interface . An important distinction is that shear stress is generated by an externally applied load in an engineered lap joint , while at the PSM\|NT tissue interface , shear stress is generated via tissue motion relative to one another due to convergence extension . This shear stress at the adhesive interface resists the convergence of neural tube ( green and black arrows in Figure 3R , right ) . In engineering , excess adhesive can be added to the edges of the overlap and sculpted into an arc to strengthen the joint by distributing the forces over a larger area ( Figure 3R; Lang and Mallick , 1998 ) . These arced adhesive ‘spew fillets’ are also observed in our system at the lateral edges of the PSM\|NT interface ( Figure 3S ) . In addition , the geometry of the adherent materials impacts the strength of a lap joint . A rounded end ( rather than an end with sharp corners ) of adherent materials eliminates singularities in the stress field and strengthens the joint by more evenly distributing the forces ( Adams and Harris , 1987 ) . Here , the tissue edges are rounded at PSM\|NT interface . Thus , the PSM\|NT interface behaves like an optimally engineered adhesive lap joint with rounded edges , a graded adhesive and an adhesive spew fillet . In our computational model , we hypothesized a greater inter-tissue adhesion between the neural tube and the PSM in cdh2 mutants and a lower inter-tissue adhesion in itgα5 mutants compared to wild-type embryos . These assumptions were sufficient to accurately predict the relative in vivo values of both the interfacial angle and length of PSM\|NT interface . To further test these model assumptions , we quantified Fibronectin matrix assembly and F-actin at PSM\|NT interface in cdh2 mutants and MZ itgα5 mutants . We observed relatively high average levels of Fibronectin at the interface in the cdh2 mutants and the gradients of both Fibronectin matrix assembly and F-actin were lost ( Figure 4A–C and G–I ) . In contrast , MZ itgα5 mutants have lower levels of Fibronectin , and the medial-lateral gradients of F-actin and Fibronectin matrix assembly were maintained ( Figure 4D–K ) . Lap joints exhibit graded shear stress at the interface of two overlapping materials ( Goland and Reissner , 1944 ) . In our system , the presence of a shear stress at the interface should depend on the relative motion between the attached converging tissues , and this relative motion can be reduced in the absence of neural tube convergence . If the medial-lateral gradation of Fibronectin matrix assembly and F-actin are shear-driven , we can expect a positive correlation between neural tube convergence and presence of these gradients . Our data for cdh2 mutants strengthen this idea as these mutants exhibit a reduction in neural tube convergence as well as loss of the gradients . On the other hand , itgα5 mutants do not have a defect in convergence and they retain the potential to establish this graded assembly , though we observe an overall reduction in Fibronectin matrix production . To directly examine whether Fibronectin matrix is required for inter-tissue adhesion and restriction of neural tube convergence , we generated double mutants for the two zebrafish fibronectin genes , fn1a and fn1b , using CRISPR/Cas9 ( Figure 5—figure supplement 1B ) . Double homozygous fibronectin mutant embryos ( fn1a-/-;fn1b-/- ) completely lacked Fibronectin matrix as detected by immunohistochemistry ( IHC ) ( Figure 5D ) . fn1a-/-;fn1b-/- embryos exhibited a precociously converged neural tube , local tissue detachments , and displayed local disruptions of bilateral symmetry consistent with the MZ itgα5-/- phenotype ( Figure 5A and B , Figure 5—figure supplement 1C–1E ) . We created a photoconvertible Fibronectin transgenic to enable us to ‘paint’ the extracellular matrix and quantify deformation of the ‘painted spots’ in live embryos . The transgenic line expresses Fibronectin 1a tagged with the green-to-red photoconvertible protein mKikumeGR under the control of the heat-shock promoter ( Tg hsp70:fn1a-mKIKGR , Figure 5—figure supplement 1A ) . We tested the functionality of the transgene by creating fn1a-/-;fn1b-/-; hsp70:fn1a-mKIK embryos . fn1a-/-;fn1b-/- embryos are not viable , and we observed somite border defects consistent with published loss of function studies in zebrafish ( Figure 5G; Jülich et al . , 2005; Koshida et al . , 2005 ) . Heat-shocked fn1a-/-;fn1b-/-; hsp70:fn1a-mKIK embryos exhibited normal Fibronectin assembly and somite boundary morphologies , demonstrating the functionality of the transgene ( Figure 5C–H and Tables 1 and 2 ) . Using this transgene , we compared the dynamics of matrix remodeling at different tissue interfaces in wild-type and mutant embryos . We assayed Fibronectin matrix dynamics at tissue interfaces in vivo via confocal timelapse microscopy . When transgenic embryos were heat-shocked at the 2–3 somite stage , we observed a fluorescently labeled Fibronectin matrix at the 12-somite stage ( Figure 6B and C ) . The topology of the Fn1a-mKikGR matrix was consistent with the matrix topology observed at each interface in fixed wild-type samples subjected to Fibronectin IHC . We photoconverted spots of matrix 25–35 μm in diameter either at the PSM\|NT or PSM\|E interface ( Figure 6A–C ) at a distance of 150–200 μm from the last somite , and imaged every 15 min . One hour after the photoconversion , the spots at the PSM\|NT interface showed significant shrinking along the medial-lateral direction , while the spots at the PSM\|E interface did not shrink ( Figure 6B and C and Video 2 ) . We quantified the spatiotemporal dynamics of photoconverted Fibronectin matrix spots after thresholding to account for photobleaching . The spots were converted into binary images and , we used the standard deviation of the white pixel distribution along the medial-lateral and anterior-posterior axes as measures of the width and height , respectively . This metric quantifies the general deformation of the photoconverted spot over time . At the PSM\|NT interface , the medial-lateral width of the photoconverted region decreased by 40% , while the anterior-posterior height was unchanged ( Figure 6D and Figure 6—figure supplement 1A ) . In contrast , the matrix at the PSM\|E interface did not exhibit medial-lateral shrinking ( Figure 6D and Figure 6—figure supplement 1A ) . Further analysis of the directionality of matrix remodeling revealed that there is medial to lateral bias in the displacement fields of the PSM\|NT interface but no anisotropy in the displacement fields of the PSM\|E interface ( Figure 6—figure supplement 1E–1G ) . Lap joint theory suggests that inter-tissue adhesion is required for shear forces at the PSM\|NT interface . We examined the effect of inter-tissue adhesion on shear driven Fibronectin matrix remodeling utilizing the variable inter-tissue detachment phenotype of MZ itgα5 mutants in three groups of photoconversion experiments ( Figure 6E and F ) . In Group 1 embryos , the neural tube is attached on both sides to the PSM at the level of the photoconverted region ( Figure 6G ) . In Group 2 embryos , tissues are detached ipsilateral to the photoconverted side but attached on the contralateral side ( Figure 6H ) . In Group 3 embryos , tissues are attached ipsilateral to the photoconverted side but detached on the contralateral side ( Figure 6I ) . In the absence of tissue detachment , the photoconverted matrix was remodeled in the medial-lateral direction similar to wild type ( Figure 6G and J ) . In contrast , where tissues were ipsilaterally detached the medial-lateral remodeling was lost ( Figure 6H and J ) . This result demonstrates that inter-tissue adhesion is necessary for the medial-lateral matrix remodeling . Strikingly , tissue detachment contralateral to the photoconverted region also affected matrix remodeling ( Figure 6I and J ) . This contralateral phenotype illustrates a long-range effect of mechanical coupling , namely that ECM remodeling on one side of the embryo is responsive to morphogenic variability on the opposite side of the embryo . We next asked whether the medial-lateral remodeling of the Fibronectin matrix at the PSM\|NT interface was dependent on neural tube convergence by performing the photoconversion experiment in cdh2 mutants . These mutants exhibit variable neural tube convergence as measured by the change in medial-lateral width of the neural tube over time . We also observed bilaterally asymmetric convergence of the neural tube in some embryos , which we quantified as the change in medial-lateral width of the left and right halves ( Figure 6M ) . We divided cdh2 mutants into three groups based the variable neural tube convergence phenotype . In Group 1 , the neural tube is symmetric and slowly converges at the level of the photoconverted region ( Figure 6K and N ) . In Group 2 , the neural tube is symmetric at the level of the photoconverted region but convergence is severely retarded ( Figure 6L and N ) . In Group 3 , the neural tube converges asymmetrically at the level of the photoconverted region ( Figure 6M and P ) . When the neural tube converges symmetrically , the matrix exhibits normal medial-lateral remodeling ( Figure 6K and O ) . However , when the neural tube does not converge , there is no medial-lateral matrix remodeling ( Figure 6L and O ) . Thus , anisotropic matrix remodeling at the PSM\|NT interface depends on the neural tube convergence . An interesting phenotype was observed when the neural tube converges asymmetrically . The side that exhibited a strong neural tube convergence defect displayed normal medial-lateral matrix remodeling , while anisotropic matrix remodeling was lost on the contralateral side where the neural tube converged ( Figure 6M and Q ) . This result indicates that neural tube convergence on one side of the body impacts the Fibronectin matrix remodeling on the opposite side of the body . We next performed the photoconversion experiments in cdh2; MZ itgα5 double mutants . Mutant embryos were sorted into two groups based on the degree of neural tube convergence . We observed medial-lateral Fibronectin matrix remodeling when the neural tube converged , whereas medial-lateral matrix remodeling was lost with diminished neural tube convergence ( Figure 6R–U ) . Thus , these experiments confirmed that the medial-lateral remodeling of the matrix positively correlates with the neural tube convergence . Lastly , we did not observe any change in the anterior-posterior length of the photoconverted spots for any mutants , similar to wild type ( Figure 6—figure supplement 1B–1D ) . We revisited these contralateral effects with our in silico model using different parameter values for the left and right sides . Uniform parameters produce symmetric gradients in tension in the neural tube at the left and right PSM\|NT interfaces ( Figure 7A ) . When the adhesion stiffness ( representing inter-tissue adhesion ) is reduced along the left interface relative to the right interface , the left and right PSM become asymmetric in size and the contralateral tension gradient becomes shallower ( Figure 7B ) . This result mimics both the morphological phenotype as well as the contralateral effects on Fibronectin matrix remodeling observed in MZ itgα5-/- embryos . We computationally implemented asymmetric neural tube morphogenesis observed in some cdh2 mutants by reducing the surface stiffness of one half of the neural tube ( Figure 7C ) . Here , we observed a contralateral reduction in the tension gradient which mimics the contralateral effects on Fibronectin matrix remodeling observed in cdh2-/- embryos . Here , we find that inter-tissue adhesion between the neural tube and left and right paraxial mesoderm ensures bilaterally symmetric morphogenesis . However , this inter-tissue adhesion predisposes the neural tube to convergence defects and spina bifida by requiring the neural tube to pull on the adjacent mesoderm ( Figure 7D ) . We find that the mechanics of this inter-tissue adhesion resemble an adhesive lap joint which is well described in engineering . Fibronectin functions as the adhesive in this lap joint . Due the cellular mechanisms that regulate Fibronectin matrix assembly and remodeling , we find that Fibronectin responds to a gradient of stress by remodeling to the region of highest stress . Thus , Fibronectin functions as a ‘smart adhesive’ that continually remodels to where it is most needed . Multiple lines of evidence suggest that the interface between the neural tube and PSM behaves as an optimally engineered adhesive lap joint . Mutants that reduce or eliminate the Fibronectin matrix exhibit a precociously converged neural tube implicating cell-Fibronectin matrix adhesion in constraining neural tube convergence . In vivo testing of predictions of a computational model suggests that the PSM-neural tube interface is under a medial-laterally graded stress . This stress gradient is characteristic of a lap joint ( Goland and Reissner , 1944 ) . The medial to lateral remodeling of Fibronectin produces a graded adhesive which strengthens the lap joint by accounting for the graded stress ( Carbas et al . , 2014 ) . In addition , there is an arc-shaped domain of Fibronectin matrix at the rounded lateral edges of the PSM-neural tube interface . This domain of Fibronectin matrix is reminiscent of an 'adhesive spew fillet' , which is an excess of adhesive present at the edges of a lap joint that fortifies the joint ( Lang and Mallick , 1998 ) . Moreover , the rounded edges of the tissues strengthen the lap joint by more evenly distributing stress ( Adams and Harris , 1987 ) . The extracellular matrix ( ECM ) can act both as a mediator of mechanical forces and a stress sensor . The latter function arises by the activation of biochemical signaling pathways via Integrins in response to changes in ECM fiber morphology ( Vogel and Sheetz , 2009 ) . Moreover , Fibronectin fibrillogenesis is in a positive feedback loop with actomyosin contractility leading to colocalization of Fibronectin matrix , F-actin and traction forces . The combination of Fibronectin’s function as a mediator of mechanical forces , a stress sensor , and positive feedback with actomyosin contractility make Fibronectin a smart adhesive . This smart adhesive activity is revealed in our Fibronectin photoconversion experiments which show that the Fibronectin matrix continually remodels in response to neural tube convergence to create the medial-lateral gradient of matrix . Folate deficiency is the best-known cause of spina bifida in humans , but it is currently unclear how this deficiency leads to a failure of neural tube convergence and closure . Genetic causes of spina bifida include mutations in the planar cell polarity pathway , which directs convergent-extension , and mutations in genes that regulate cytoskeletal remodeling during apical constriction . These two groups of mutations have more direct effects on neural tube morphogenesis , and thus the etiologies of these defects seem clear ( Wallingford et al . , 2013; Copp et al . , 2015 ) . Cadherin-2-mediated cell-cell adhesion is required for neural tube morphogenesis , but here we find that cadherin 2 mutant neural tubes are capable of convergence as long as they do not have the additional work of coupling the left and right paraxial mesoderm . Thus , the neural tube has to be maximally fit in order to converge and fuse , and any environmental or genetic deficiency that weakens the neural tube may lead to spina bifida . Indeed , in humans , neural differentiation appears to be relatively normal during failure of neural tube closure , but exposure to the amniotic fluid , which becomes toxic later in gestation , causes neural degeneration . This is likely a reason why prenatal surgical repair limits the neurological consequences of spina bifida: in the absence of degeneration more normal neural function can be achieved . Thus , folate deficiency may directly affect cell proliferation or epigenetics without dramatically altering neural differentiation ( Copp et al . , 2015 ) . Folate increases cell traction force in neuronal cell culture ( Kim et al . , 2018 ) , thus folate deficiency may cause spina bifida by diminishing the overall mechanical fitness of the neural tube . In principle , the mechanical consequences of folate deficiency on the neural tube could be tested in an animal model . Moreover , mobilization from the flanking mesoderm could rescue spinal cord closure in folate deficient and other environmentally and genetically weakened neural tubes . Based on systematic analysis of cell motion in the tailbud , we previously hypothesized that PSM tissue assembly may involve a fluid to solid transition ( McMillen and Holley , 2015 ) . Recent experiments identifying gradual increases in tissue stiffness and cell packing during PSM maturation support this idea ( Mongera et al . , 2018 ) . Our data here extend this model as we find that there are posterior to anterior increases in ( 1 ) F-actin around the cell cortices of PSM cells , ( 2 ) epithelization of cells on the surface of the PSM and ( 3 ) assembly of large Fibronectin fibers . This posterior to anterior pattern is consistent with the established relationship between cortical tension , Fibronectin matrix assembly and traction forces . Notably , the anterior-posterior patterns of Fibronectin and F-actin are only observed on the lateral surface of the PSM which interfaces the epidermis . Medially , along the neural tube-PSM interface , the lap joint mechanics predominate , and medial to lateral gradients of F-actin and Fibronectin are observed . While our analysis has focused on the development of the posterior trunk , our model likely applies to the more anterior neural tube because it displays even more convergence . However , there is less convergence during the formation of the more posterior spinal cord in the tail . Thus , ECM-mediated inter-tissue adhesion is likely less of a mechanical impediment to the development of the most caudal spinal cord . However , during these later stages of body elongation , inter-tissue adhesion is important for establishing the chevron shape of the myotome which is the major derivative of the zebrafish somite ( Tlili et al . , 2019 ) . In summary , we find that the Fibronectin matrix ensures symmetric morphogenesis of the spinal column . During development , mechanical forces among attached tissues are intrinsically coupled to the ever-changing geometry , and to preserve symmetries , these forces need to be actively coordinated . In the course of vertebrate body elongation , we find that tissue mechanics are tuned to accomplish the competing goals of neural tube convergence and maintenance of bilateral symmetry . The mechanism that dynamically couples these embryonic tissues resembles adhesive lap joints which are utilized in processes ranging from aerospace manufacturing to woodworking . Thus , microscale embryo biomechanics can function very much like macroscale engineered structures . Zebrafish were raised according to standard protocols ( Nüsslein-Volhard and Dahm , 2002 ) and approved by the Yale Institutional Animal Care and Use Committee . The TLAB and TLF wild-type strain were used . The tm101B allele of cdh2 was used ( Lele et al . , 2002 ) . The MZ itgα5-/- mutant line is a maternal zygotic mutant line using the thl30 allele ( Jülich et al . , 2005 ) . The single mutant lines and double mutant lines were crossed to the Tg ( hsp70:fn1a-mKIKGR ) line to the generate the mutants lines in transgenic background used in photoconversion experiments . Embryos were collected from pair-wise natural matings . Sex-specific data were not collected as zebrafish do not have a strictly genetic sex determination mechanism , and sex is determined after the first 36 hr of development studied here ( Wilson et al . , 2014 ) . To generate a double mutant line for fn1a and fn1b genes , we first created single mutant lines for fn1a and fn1b genes separately using CRISPR . Single cell stage embryos were injected with a mix containing: Cas9 mRNA , a single guide RNA ( sgRNA ) specifically designed to target a chosen sequence in the exon 4 of fn1a gene or in the exon 5 of fn1b gene ( see Figure 5—figure supplement 1 for detailed target sequences , and a stop codon cassette oligonucleotide Gagnon et al . , 2014 ) . Sequences for the fn1a-/- line: sgRNA ( 5’ATTTAGGTGACACTATAGGAGGGCACTCCTACAAGATGTTTTAGAGCTAGAAATAGCAAG3’ ) ; stop codon cassette oligonucleotide ( 5’GAGGGAGGGCACTCCTACAAGTCATGGCGTTTAAACCTTAATTAAGCTGTTGTAGGATTGGAGACACATGGCAGA3’ ) . Sequences for the fn1b-/- line: sgRNA ( 5’ATTTAGGTGACACTATAGGACTGCACATGTTTGGGAGGTTTTAGAGCTAGAAATAGCAAG3’ ) ; stop codon cassette oligonucleotide ( 5’CGTGGACTGCACATGTTTGGGTCATGGCGTTTAAACCTTAATTAAGCTGTTGTAGGAGAGGGAAACGGACGCATC3’ ) . General details regarding the choice of target sequence , the design of the stop codon cassette oligonucleotide , the design and synthesis of the sgRNA and conditions of injection can be found in the detailed supplemental protocol described in Gagnon et al . ( 2014 ) . Individuals positive for the insertion of the stop codon cassette were identified by PCR on genomic DNA using fn1a DiagA1/stopA primers or fn1b DiagA1/stopA primers: fn1a Diag A1 ( 5’GACTGTACTTGCATTGGCTCTG3’ ) fn1b DiagA1 ( 5’GAGCGTTGCTATGATGACTCAC3’ ) stop A ( 5’GCTTAATTAAGGTTTAAACGCC3’ ) . Then single mutant lines were then crossed together to obtain the double mutant line . Information concerning the sequence of the transgene is described in Figure 5—figure supplement 1A . The plasmid construct containing the transgene was generated following the previously described method ( Jülich et al . , 2015 ) with the mKikGR plasmid provided by the Atsushi Miyawaki laboratory ( Habuchi et al . , 2008 ) . 75 ng/μl of plasmid were co-injected with 120 ng/μl of Tol2 mRNA at 1 cell stage . For both in vivo tracking of Fibronectin matrix dynamics and rescue experiments , embryos were heat-shocked by incubation 30 min at 38°C . 12–14 somite stage embryos were fixed overnight ( 4°C ) in 4% PFA ( in 1X PBS ) . After dechorionation in 1X PBS , embryos were rinsed 3 × 5 min in PBSDT ( 1% DMSO , 0 . 1% Triton in 1X PBS ) , permeabilized 3 min with Proteinase K ( 5 μg/ml in PBSDT ) , rinsed 2 × 5 min in PBSDT , post-fixed 20 min at RT in 4% PFA and finally rinsed 3 × 5 min in PBSDT . Blocking was done 2–3 hr at RT in blocking solution ( 1% Blocking reagent from Roche in PBSDT ) . Embryos were incubated O/N at 4°C with the primary antibody against Fibronectin ( rabbit polyclonal SIGMA , F3648 ) ( 1/100 dilution in blocking solution ) . Embryos were rinsed 2 × 15 min in blocking solution , 2 × 15 min and 3 × 1 hr in PBSDT . Embryos were incubated O/N at 4°C with both secondary antibody ( Alexa fluor 555 donkey anti-rabbit IgG , Invitrogen A31572 ) and Alexa Fluor 488 Phalloidin ( Life technologies , A12379 ) respectively diluted 1/200 and 1/100 in blocking solution . Embryos were incubated 15 min with DAPI solution ( 100 pg/ml in PBSDT ) , and then rinsed 2 × 15 min and 3 × 1 hr in PBSDT . For morphometric analysis , whole embryos were mounted in glass capillaries ( 1 . 4 mm diameter ) to preserve tissue morphologies ( 1% agarose in 1X PBS ) and imaged with a Lightsheet Z . 1 ( ZEISS , 20x objective ) . For imaging of fibronectin matrix and F-actin at high resolution , only the tails of the embryos were mounted between standard slides and coverslips ( Fisher Finest Premium Microscope Slides ) in 50% glycerol in 1X PBS and imaged with an inverted confocal microscope ( ZEISS LSM880 , 40x oil objective ) . For the photoconversion assay , 12 s stage embryos were mounted in glass bottom dishes in drops of 1% agarose in 1X E2 covered with E2 medium . An inverted confocal microscope ZEISS LSM880 , 20x objective ) was used for both photoconversion and subsequent imaging of the photoconverted region . Regions of interest were photoconverted using a 405 nm laser ( 30–45 cycles , speed 9 average 1 , 10% of maximum laser power ) . We model the 2D cross-section of the posterior tail ( see Figure 2A in main text ) based on an earlier model of soft grains ( Åström and Karttunen , 2006 ) . The tissue surfaces are modeled by closed loops of mass-spring chains . An isolated tissue cross-section is made of 50 mass-points connected by elastic springs with homogeneous stiffness constant KS . This parameter ( KS ) is named 'surface stiffness' . Thus , the tension force along the i-th spring is: T→i=KS ( li−l0 ) h^i , where li and l0 are the instantaneous length and the equilibrium ( unstretched ) length of the i-th spring respectively , and h^i is a unit vector along the i-th spring . The sum of these tension forces along all the surface springs represents the surface tension of the tissue . An internal pressure , P also acts in the bulk of each tissue and it is directed perpendicularly outward at the tissue surface . Therefore , the net force acting on the i-th mass point of an isolated tissue ( F→itissue ) is given by:F→itissue=T→i−T→i+1+P l02 ( n^i+n^i+1 ) Here , n^i denotes an outward directed unit vector perpendicular to the i-th spring . We next modeled the interaction forces between adjacent tissues . Two mass-points belonging to two different tissues are considered to be interacting only if they are within a distance Radh from each other . If this distance between the mass-points is between Rrep and Radh , there exists a spring-like adhesive force that represents 'inter-tissue adhesion' . In addition , to prevent tissues from penetrating into each other , we implemented 'volume exclusion' by assuming spring-like repulsion forces between the mass-points , when the distance between the mass-points is below Rrep . Therefore , the interaction forces ( F→ijint ) between i-th and j-th mass-points belonging to adjacent tissues can be summarized as:F→ijint=Krep ( Rrep−rij ) r^ij , if rij<Rrepor , F→ijint=−Kadh ( rij−Rrep ) r^ij , if Rrep≤rij≤Radhor , F→ijint=0 , if rij>Radh Here , rij is the metric distance between the mass-points ( i . e . rij=\|r→i−r→j\| ) and r^ij is a unit vector pointing from the j-th to the i-th point . Kadh and Krep represent the strength of adhesive and repulsive interactions respectively . In general , Krep is much higher than Kadh , and it is kept constant in all simulations . On the hand , a variation in Kadh represents a variation in the level of inter-tissue adhesion . This parameter , Kadh is referred as 'adhesion stiffness' that represents the strength of cell-ECM interaction . We finally assumed an over-damped dynamics ( neglecting inertia ) of the mass-points to evolve their positions over time as below:c v→i=F→itissue+∑jF→ijint Here , v→i is the velocity of i-th mass-point and c is the viscous coefficient representing a drag force . By evolving the above dynamics in computer , we produced steady final shapes of the tissues at the limit of very long time . We used the 'Explicit Euler' method to simulate the dynamics with a time step set at Δt=0 . 001 . In all simulations , the equilibrium spring lengths ( l0 ) , the repulsion strength ( Krep ) , the viscous coefficient ( c ) , the range of forces between tissues ( Rrep and Radh ) and the internal pressure ( P ) are kept constant . These fixed parameters are: l0=0 . 1 , Krep=30000 , c=10 , Rrep=0 . 1 , Radh=0 . 2 , and P=5 . To theoretically explore the impact of variations of the surface stiffness ( Ks ) and adhesion stiffness ( Kadh ) in Figure 2B , we used Kadh = 4 and K adh= 12 , while keeping the surface stiffness fixed at Ks = 50 . We also used Ks = 35 and Ks = 100 , keeping the adhesion stiffness fixed at Kadh = 8 . We next systematically varied the surface stiffness ( Ks ) and adhesion stiffness ( Kadh ) to mimic the conditions of different genotypes in Figure 2D , E and F . The chosen values for these parameters are: for wild-type Ks = 100 , Kadh = 10; for cdh2-/- mutant Ks = 55 , Kadh = 12; for itgα5-/- mutants Ks = 100 , Kadh = 6 . 5; and for cdh2-/-;itgα5-/- mutants Ks = 55 , Kadh = 6 . 5 ( also see the parameter space plot in Figure 2D ) . In Figures 3A and 7A , we used Ks = 200 and Kadh = 10 to generate the symmetric tension distributions along the PSM\|NT interfaces . By using Imaris , we made transverse sections of 3D reconstructs of the posterior tail in every 20 um starting from the last somite boundary . Each transverse section was exported into image-J and rotated such that the medial-lateral axis of the notochord is exactly parallel to the horizontal axis of the image . We then traced the contours of neural tube and PSM to define regions of interests ( ROIs ) . We applied the ‘bounding rectangle’ tool to these ROIs and the medial-lateral and dorsal-ventral lengths of the tissues are given by the width and heights of the rectangle . The same ROIs were also used to determine the cross-sectional area of the tissues ( Path in image-J: Analyze - > Set measurements - > select ‘area’ and ‘bounding rectangle’ ) . The interfacial length and angle were measured with standard ‘segmented line tool’ and ‘angle tool’ from image J . The radius of curvature was measured using a online ‘Fit circle’ Image J macro adapted from Pratt V . , 1987 ( see the website: http://www . math . uab . edu/~chernov/cl/MATLABcircle . html , http://imagej . 1557 . x6 . nabble . com/Re-bending-and-radius-of-curvature-td3686117 . html ) . We provide the image-J macro at the end of the methods section . To get the cell aspect ratio , we first prepared cross-sections of the PSM prepared by the Imaris software . In these cross-sections , each cell was manually traced and aspect ratio ( major axis/minor axis ) was extracted using the ‘Shape descriptor’ plugin of ImageJ ( path: analyze - > Set measurements - > click ‘shape descriptor’ ) . To quantify the mean F-actin intensity of the PSM surface cells or PSM internal cells , we first traced a region of interest ( ROI ) encompassing either surface or internal cells . Then the mean grey value within the ROI was extracted after setting a threshold . The threshold was set such that we remove the cytoplasmic background without removing the low membrane signals in the posterior of the PSM . We first oriented the images of photoconverted spot in Imaris to get a view perpendicular to the spot in each time frame , and cropped around the spot . These images were then manually thresholded by Image J ( path: Image - > Adjust - > Threshold ) to account for photobleaching and then converted into binary images . From these binarized images , we used the standard deviation of the white pixel distribution along the medial-lateral and anterior-posterior axes as measures of the medial-lateral width and anterior-posterior height of the photoconverted spot respectively . These metrics were measured by a custom Matlab code provided below:for k = 1:8 filename = strcat ( 'SLICE' , num2str ( k ) , ' . tif' ) ; IMECM = imread ( filename ) ; IMECM = im2bw ( IMECM ) ; nwhite = nnz ( IMECM ) ; [rows , cols]=find ( IMECM == 1 ) ; Time ( k ) = ( k-1 ) 15; area ( k ) =nwhite0 . 20 . 2; width ( k ) =std ( cols ) ; % . /mean ( cols ) ; height ( k ) =std ( rows ) ; % . /mean ( rows ) ; %aspect ( k ) =allaspect ( biggest ) ; %NoElement = CC . NumObjects end output=[Time' , width' , height' , width'/width ( 1 ) , height'/height ( 1 ) ]; plot ( Time' , height'/height ( 1 ) , 'o k-' ) ; hold on plot ( Time' , width'/width ( 1 ) , 'o r-' ) xlswrite ( 'RedSpot_WT-4_PSML . xls' , output ) To determine if there is any directional bias in the medial-lateral remodeling of matrix at the PSM\|NT interface , we analyzed the displacement field of the photoconverted signal using a Particle Image Velocimetry ( PIV ) plugin available in Image J ( downloadable versions and tutorials at https://sites . google . com/site/qingzongtseng/piv#tuto ) . For this analysis , images of photoconverted spot were prepared in Imaris to get a view perpendicular to the spot in each time frame . We used the raw grey scale images with full intensity distribution ( without any thresholding ) . In Image-J , we selected the ‘iterative PIV ( advanced ) ’ method , which determines the correlation between two images at a time and requires three user-defined parameters: the ‘interrogation window size’ ( IW ) , the ‘search window size’ ( SW ) , and the ‘vector spacing’ ( VS ) . The set of two images is analyzed by PIV through three passes with a progressive decrease in parameter values so that a fine-tuned vector field of pixel displacement is finally produced . For our analysis , the set of two images corresponds to two consecutive time frames of the spot and we chose the following parameter values: for 1st pass , IW = 60 , SW = 120 , VS = 15; for 2nd pass , IW = 50 , SW = 120 , VS = 12; and for 3rd pass , IW = 40 , SW = 80 , VS = 10 . The displacement was analyzed over the course of 45 min ( using a frame to frame correlation for the first four consecutive time frames ) as most of the remodeling of the photoconverted region happened in the first hour . All the vectors from the Frame 1–2 , Frame 2–3 and Frame 3–4 correlations were plotted together in the same rose plot ( see Figure 6—figure supplement 1E ) . mRNA coding for myl12 . 1-eGFP was synthesized using mMessage mMachine SP6 kit ( Thermofisher , AM1340 ) according to standard manufacturer instructions using a PCS2+ myl12 . 1-eGFP plasmid as a template ( a gift from Carl-Philipp Heisenberg ) . Embryos were injected at the one-cell stage with 100 pg of myl12 . 1-eGFP mRNA ( 25 ng/μl with a 4 nl droplet ) . Live embryos were imaged at the 12–14 somites stage on a ZEISS LSM880 ( 40x water objective , zoom 2 , with airyscan detector set in a fast mode acquisition ) . Each image stack was then automatically sliced transversally every 5 μm using the ‘Reslice tool’ in Fiji . To quantify medio-lateral differences in myl12 . 1-eGFP localization along the NT border on each transverse slice , we used a custom Fiji code ( see code below ) to divide the NT border into 3 regions of 3–5 μm in width and measured the mean intensity value in each region . roiManager ( "Select" , newArray ( 0 , 1 , 6 , 7 ) ) ; roiManager ( "Combine" ) ; getSelectionCoordinates ( xpoints , ypoints ) ; makeSelection ( "polygon" , xpoints , ypoints ) ; run ( "Measure" ) ; roiManager ( "Deselect" ) ; roiManager ( "Select" , newArray ( 1 , 2 , 5 , 6 ) ) ; roiManager ( "Combine" ) ; getSelectionCoordinates ( xpoints , ypoints ) ; makeSelection ( "polygon" , xpoints , ypoints ) ; run ( "Measure" ) ; roiManager ( "Deselect" ) ; roiManager ( "Select" , newArray ( 2 , 3 , 4 , 5 ) ) ; roiManager ( "Combine" ) ; getSelectionCoordinates ( xpoints , ypoints ) ; makeSelection ( "polygon" , xpoints , ypoints ) ; run ( "Measure" ) ; roiManager ( "Deselect" ) ; roiManager ( "Select" , newArray ( 2 , 3 , 4 , 5 ) ) ; roiManager ( "Combine" ) ; getSelectionCoordinates ( xpoints , ypoints ) ; makeSelection ( "polygon" , xpoints , ypoints ) ; run ( "Measure" ) ; roiManager ( "Deselect" ) ; roiManager ( "Save" , "/image path/ROIn . zip" ) ; roiManager ( "Deselect" ) ; roiManager ( "Delete" ) ; selectWindow ( "Results" ) ; String . copyResults ( ) ; Table . deleteRows ( 0 , 2 ) ; To analyze the tension within the Fibronectin matrix , we used the H5 monoclonal antibody which recognizes a tension dependent epitope on the human Fibronectin 10th FN type III repeat ( Cao et al . , 2017 ) . We generated a chimeric FN1a-mKIKGR13 . 2-hsFNIII10-11 protein in which the zebrafish 10th and 11th FN type III repeats have been replaced by the human ortholog’s 10th and 11th FN type III repeats . ( The 10th FN type III repeat contains the PSHRN motif and was formerly identified as the 9th FN type III repeat . The 11th FN type III repeat contains the RGD motif and was formerly identified as the 10th FN type III repeat . ) The PCS2+ FN1a-mKIKGR13 . 2-hsFNIII10-11 construct was generated by first subcloning the FN1a-mKIKGR13 . 2 CDS into the PCS2+ plasmid . The zebrafish FNIII10-11 repeats were replaced with a zebrafish codon optimized CDS of the human ortholog FNIII10-11 repeats using Gibson assembly . To synthetize the mRNA , we followed the protocol of Prince and Jessen ( 2019 ) . Briefly , 30 μg of plasmid was linearized with Not I-HF ( 200 μl total volume reaction with 8 μl of enzyme ) and purified via phenol chloroform extraction , overnight sodium acetate/ethanol precipitation and eluted in 10 μl RNAse free water . mRNA was synthesized using the mMessage Machine SP6 kit ( Thermofisher , AM1340 ) following manufacturer’s instructions for long mRNA synthesis ( 1 μl of GTP was added to the reaction mix ( total volume 21 μl ) with 1 μg of DNA template at high concentration ( 1–3 μl of DNA in the final reaction mix ) . One-cell stage embryos were injected with 600–800 pg of FN1a-mKIKGR13 . 2-hsFNIII10-11 mRNA . Embryos were sorted for bright fluorescence at the 12–14 somite stage , fixed 40 min in 4% PFA ( gentle lateral shaking ) at RT , washed twice for 5 min in PBST ( 0 . 1% Tween in 1X PBS ) , dechorionated , washed twice for 5 min in PBST , treated for 3 min with 5 μg/ml Proteinase K diluted in PBST , washed 3 × 5 min in PBST , post-fixed 20 min in 4% PFA at RT ( gentle lateral shaking ) , washed 2 × 5 min in PBST , blocked for 3 . 5 hr at RT ( gentle lateral shaking ) in blocking solution ( 2% blocking reagent from Roche , 1X maleic acid buffer ( 2X maleic acid buffer: 200 mM maleic acid 300 mM NaCl pH = 7 . 4 ) ) , incubated 20 hr at 4°C with H5-V5-tag antibody ( a gift from Thomas Barker ) diluted at 50 μg/ml in blocking solution , washed 4 × 15 min in 1X maleic acid buffer ( gentle lateral shaking at RT ) , incubated 20H at 4°C with goat anti-V5 antibody ( Abcam , Ab 9137 diluted 1/400 ) in blocking solution , washed 4 × 15 min in 1X maleic acid buffer ( gentle lateral shaking at RT ) , incubated 20H at 4°C with donkey anti-goat Alexa 555 antibodies ( Thermofisher scientific A32816 ) diluted 1/200 in blocking solution , washed 3 × 10 min in 1X maleic acid buffer ( gentle lateral shaking at RT ) , checked for fluorescence , incubated in 25% and 50% glycerol ( 10 min each ) and then mounted in 50% glycerol for confocal imaging ( Zeiss LSM880 , 40x water objective ) . All antibodies incubations were done with 50 μl antibody solution without shaking . To quantify H5 and mKIKGR levels and the H5/mKIKGR ratio , H5 and mKIKGR signals were imaged in conditions minimizing pixel saturation . Tissue interfaces were then segmented using Imaris software and exported for analysis with display adjustments preventing any distortion of the raw pixel values ( gamma 1 , full dynamic range 0–255 ) . From this step onward , further image treatment and quantification were performed using a custom MATLAB code ( see code provided below ) . To summarize , the H5 and mKIK images were individually thresholded ( based on visual assessment ) to define background pixels corresponding to areas between matrix fibrils and to discard any saturated pixels ( value = 255 , which represented than 0 . 8% of pixels per image , on average ) . We then generated the H5/mKIK ratiometric image by dividing , pixel by pixel , the H5 image by the mKIKGR image , and displayed it with a scaled color based on pixel value ( imagesc MATLAB function ) . For the quantification of the medial-lateral distributions of each signal , images were divided in 10 sections of equal size along the medial-lateral axis , and the average signal in each section was quantified and then normalized to the average signal in the most lateral section . For the H5 and mKIKGR levels analysis , the calculation of the average signal in each section included the background pixels , which were attributed a value of zero . Thus , the tissue scale average incorporates both matrix brightness as well as matrix density . By contrast , for the H5/mKIKGR ratio analysis , the calculation of the average signal in each section excluded the background pixels that were attributed a NaN value . Thus , this metric reflects the average signal within the matrix elements themselves and represents the local level of tension within the Fibronectin matrix . MATLAB code for quantification of medial-lateral distribution of H5/mKIK ratio and H5 , mKIK signalsclr = 'rgbk'; lw = 3; fs = 20; th1 = 33; th2 = 30; % th1 and th2 are manually determined to exclude zones in between matrix elements imgs = {'H5 . tif' , 'mKIK . tif' , 'ratio . tif'}; img1 = double ( imread ( imgs{1} ) ) ; img2 = double ( imread ( imgs{2} ) ) ; img1_255 = img1 ( img1 == 255 ) ; img2_255 = img2 ( img2 == 255 ) ; percent255_1 = numel ( img1_255 ) /numel ( img1 ) percent255_2 = numel ( img2_255 ) /numel ( img2 ) img1 ( img1 == 0 ) =NaN; img2 ( img2 == 0 ) =NaN; % 2 previous lines exclude the zones of picture outside of the tissue img1 ( img1 == 255 ) =NaN; img2 ( img2 == 255 ) =NaN; % 2 previous lines exclude the saturated pixels img1 ( img1 <th1 ) =NaN; img2 ( img2 <th2 ) =NaN; % Values in the 2 previous lines are NaN for an averaging calulating the brightness of the matrix % content itself . Change them to 0 for an avering taking into account the % total surface of the tissue imgRatio = ( img1 . /img2 ) ; imgRatio ( imgRatio == Inf ) =NaN; imgRatio ( imgRatio == 0 ) =NaN; fig = figure ( 'Renderer' , 'painters' , 'Position' , [0 0 size ( imgRatio , 2 ) size ( imgRatio , 1 ) ] ) ; imagesc ( imgRatio ) ; caxis ( [0 4] ) ; %change the upper value to tune the colorbar and colormap colorbar; box off; axis off; saveas ( fig , 'ratio . tif' , 'tif' ) ; nbin = 10; img = {img1 , img2 , imgRatio}; fig = figure; figure ( 2 ) ; hold on; for i = 1:size ( imgs , 2 ) [imgmean , imgstd , imgmeannorm , position , scale]=mlGradient ( img{i} , nbin ) ; plot ( position , imgmeannorm , clr ( i ) , 'LineWidth' , lw , 'DisplayName' , imgs{i} ) ; set ( gca , 'xminorgrid' , 'off' , 'yminorgrid' , 'off' , 'fontsize' , fs , 'xminortick' , 'off' , . . . 'yminortick' , 'off' , 'linewidth' , lw ) ; set ( gcf , 'color' , 'w' ) ; legend ( 'Location' , 'bestoutside' , 'FontWeight' , 'Normal' , 'FontSize' , 12 ) ; set ( gca , 'FontSize' , fs , 'linewidth' , lw ) ; box on; ylabel ( 'Average pixel intensity' ) ; xlabel ( 'Medio-lateral position ( relative ) ' ) ; end Function mlGradient function [imgmean , imgstd , imgmeannorm , position , scale]=mlGradient ( img , nbin ) img ( isinf ( img ) ) =NaN; width = size ( img , 2 ) height = size ( img , 1 ) binsize = floor ( width/nbin ) ; imgmean = zeros ( nbin , 1 ) ; imgstd = zeros ( nbin , 1 ) ; for k = 1:nbin if k ~= nbin imgmean ( k ) =nanmean ( img ( : , ( k-1 ) binsize+1:kbinsize ) , 'all' ) ; imgstd ( k ) =std ( double ( img ( : , ( k-1 ) binsize+1:kbinsize ) ) , 0 , 'all' , 'omitnan' ) ; else imgmean ( k ) =nanmean ( ( img ( : , ( k-1 ) binsize+1:width ) ) , 'all' ) ; imgstd ( k ) =std ( double ( img ( : , ( k-1 ) binsize+1:width ) ) , 0 , 'all' , 'omitnan' ) ; end end scale = [0:1/nbin:1]'; position = [0+ ( 1/ ( nbin2 ) ) :1/nbin:1- ( 1/ ( nbin2 ) ) ]'; imgmeannorm = imgmean/imgmean ( nbin ) ; end Each experiment includes a minimum of three to five biological replicates . A biological replicate is equal one embryo whereas technical replicates represent separate experiments ( e . g . staining reactions or timelapses ) . In Figure 6 , the number of biological and technical replicates are equal . In other figures , we provide the number of biological replicates which derive from at least two technical replicates . In our experience , this sample size combined with detailed quantitative analysis of each sample will reveal any significant differences between experiment samples and controls . Samples were excluded from analysis if the signal to noise in the image data were too low to perform a quantitative analysis . To establish if there is any significant difference between samples in a metric , we used a two-sample T-test provided by the Matlab function’ ttest2’ or by standard ‘ttest’ function in Excel . Significance was defined as: p<0 . 05 , p<0 . 005 , and p<0 . 0005 . requires ( "1 . 43k" ) ; if ( nImages == 0 ) exit ( "No image is open . " ) ; showMessageWithCancel ( "Circle Fit Macro . " , "This macro will reset your ROI Manager . \nDo you want to proceed ? " ) ; roiManager ( "reset" ) ; } setTool ( "multipoint" ) ; waitForUser ( "Circle Fit Macro . " , "Multipoint tool selected . \nSelect points in your image , then click OK . " ) ; roiManager ( "Add" ) ; roiManager ( "Select" , 0 ) ; roiManager ( "Rename" , "points" ) ; getSelectionCoordinates ( x , y ) ; n = x . length; if ( n < 3 ) exit ( "At least 3 points are required to fit a circle . " ) ; sumx = 0; sumy = 0; for ( i = 0;i < n;i++ ) { sumx = sumx + x[i]; sumy = sumy + y[i]; } meanx = sumx/n; meany = sumy/n; X = newArray ( n ) ; Y = newArray ( n ) ; Mxx = 0; Myy = 0; Mxy = 0; Mxz = 0; Myz = 0; Mzz = 0; for ( i = 0;i < n;i++ ) { X[i]=x[i] - meanx; Y[i]=y[i] - meany; Zi = X[i]X[i] + Y[i]Y[i]; Mxy = Mxy + X[i]Y[i]; Mxx = Mxx + X[i]X[i]; Myy = Myy + Y[i]Y[i]; Mxz = Mxz + X[i]Zi; Myz = Myz + Y[i]Zi; Mzz = Mzz + ZiZi; } Mxx = Mxx/n; Myy = Myy/n; Mxy = Mxy/n; Mxz = Mxz/n; Myz = Myz/n; Mzz = Mzz/n; Mz = Mxx + Myy; Cov_xy = MxxMyy - MxyMxy; Mxz2 = MxzMxz; Myz2 = MyzMyz; A2 = 4Cov_xy - 3MzMz - Mzz; A1 = MzzMz + 4Cov_xyMz - Mxz2 - Myz2 - MzMzMz; A0 = Mxz2Myy + Myz2Mxx - MzzCov_xy - 2MxzMyzMxy + MzMzCov_xy; A22 = A2 + A2; epsilon = 1e-12; ynew = 1e+20; IterMax = 20; xnew = 0; //Newton's method starting at x = 0for ( iter = 1;i <= IterMax;i++ ) { yold = ynew; ynew = A0 + xnew ( A1 + xnew* ( A2 + 4 . xnewxnew ) ) ; if ( abs ( ynew ) >abs ( yold ) ) { print ( ‘Newton-Pratt goes wrong direction: \|ynew\| > \|yold\|" ) ; xnew = 0; i = IterMax+1; } else { Dy = A1 + xnew* ( A22 + 16xnewxnew ) ; xold = xnew; xnew = xold ynew/Dy; if ( abs ( ( xnew-xold ) /xnew ) <epsilon ) i = IterMax+1; else { if ( iter >= IterMax ) { print ( ‘Newton-Pratt will not converge’ ) ; xnew = 0; } if ( xnew <0 ) { //print ( ‘Newton-Pratt negative root: x = "+xnew ) ; xnew = 0; } } } } DET = xnewxnew - xnewMz + Cov_xy; CenterX = ( Mxz* ( Myy-xnew ) -MyzMxy ) / ( 2DET ) ; CenterY = ( Myz* ( Mxx-xnew ) -MxzMxy ) / ( 2DET ) ; radius = sqrt ( CenterXCenterX + CenterYCenterY + Mz + 2xnew ) ; if ( isNaN ( radius ) ) exit ( "Points selected are collinear . " ) ; CenterX = CenterX + meanx; CenterY = CenterY + meany; print ( "Radius = "+d2s ( radius , 2 ) ) ; print ( "Center coordinates: ( "+d2s ( CenterX , 2 ) +" , "+d2s ( CenterY , 2 ) +" ) " ) ; print ( " ( all units in pixel ) " ) ; makeOval ( CenterX-radius , CenterY-radius , 2radius , 2*radius ) ; roiManager ( "Add" ) ; roiManager ( "Select" , 1 ) ; roiManager ( "Rename" , "circle fit" ) ; setTool ( "point" ) ; makePoint ( CenterX , CenterY ) ; roiManager ( "Add" ) ; roiManager ( "Select" , 2 ) ; roiManager ( "Rename" , "center" ) ; roiManager ( "Show All without labels" ) ;	In embryos , the spinal cord starts out as a flat sheet of cells that curls up to form a closed cylinder called the neural tube . The folding tube is attached to the surrounding tissues through an extracellular matrix of proteins and sugars . Overlapping strands of a protein from the extracellular matrix called Fibronectin connect the neural tube to adjacent tissues , like a kind of biological glue . However , it remained unclear what effect this attachment had on the embryonic development of the spinal cord . Connecting two overlapping objects with glue to form what is known as an ‘adhesive lap joint’ is common in fields such as woodworking and aeronautical engineering . The glue in these joints comes under shearing stress whenever the two objects it connects try to pull apart . But , thanks to work in engineering , it is possible to predict how different joints will perform under tension . Now , Guillon et al . have deployed these engineering principles to shed light on neural tube development . Using zebrafish embryos and computational models , Guillon et al . investigated what happens when the strength of the adhesive lap joints in the developing spine changes . This revealed that Fibronectin works like a smart adhesive: rather than staying in one place like a conventional glue , it moves around . As the neural tube closes , cells remodel the Fibronectin , concentrating it on the areas under the highest stress . This seemed to both help and hinder neural tube development . On the one hand , by anchoring the tube equally to the left and right sides of the embryo , the Fibronectin glue helped the spine to develop symmetrically . On the other hand , the strength of the adhesive lap joints made it harder for the neural tube to curl up and close . If the neural tube fails to close properly , it can lead to birth defects like spina bifida . One of the best-known causes of these birth defects in humans is a lack of a vitamin known as folic acid . Cell culture experiments suggest that this might have something to do with the mechanics of the cells during development . It may be that faulty neural tubes could close more easily if they were able to unglue themselves from the surrounding tissues . Further use of engineering principles could shed more light on this idea in the future .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cell", "biology" ]	2020	Fibronectin is a smart adhesive that both influences and responds to the mechanics of early spinal column development
The leishmaniases are vector-borne diseases that have a broad global distribution throughout much of the Americas , Africa , and Asia . Despite representing a significant public health burden , our understanding of the global distribution of the leishmaniases remains vague , reliant upon expert opinion and limited to poor spatial resolution . A global assessment of the consensus of evidence for leishmaniasis was performed at a sub-national level by aggregating information from a variety of sources . A database of records of cutaneous and visceral leishmaniasis occurrence was compiled from published literature , online reports , strain archives , and GenBank accessions . These , with a suite of biologically relevant environmental covariates , were used in a boosted regression tree modelling framework to generate global environmental risk maps for the leishmaniases . These high-resolution evidence-based maps can help direct future surveillance activities , identify areas to target for disease control and inform future burden estimation efforts . The leishmaniases are a group of protozoan diseases transmitted to humans and other mammals by phlebotomine sandflies ( Murray et al . , 2005; WHO , 2010 ) . Considered as one of the neglected tropical diseases ( NTD ) ( WHO , 2009 ) , the leishmaniases can be caused by around 20 Leishmania species and include a complex life cycle involving multiple arthropod vectors and mammalian reservoir species ( Ashford , 1996; Ready , 2013 ) . Sandflies belonging to either Phlebotomus spp . ( Old World ) or Lutzomyia spp . ( New World ) are the primary vectors; domestic dogs , rodents , sloths , and opossums are amongst a long list of mammals that are either incriminated or suspected reservoir hosts . Non-vector transmission ( e . g . , by accidental laboratory infection , blood transfusion , or organ transplantation ) is possible , but rare ( Cardo , 2006 ) . Transmission of the leishmaniases can be either anthroponotic or zoonotic . The leishmaniases rank as the leading NTD in terms of mortality and morbidity with an estimated 50 , 000 deaths in 2010 ( Lozano et al . , 2012 ) and 3 . 3 million disability adjusted life years ( Murray et al . , 2012 ) . Symptoms of Leishmania infection can take many different and diverse forms ( Banuls et al . , 2011 ) , the two main outcomes being cutaneous leishmaniasis ( CL ) and visceral leishmaniasis ( VL ) . Cutaneous leishmaniasis typically presents as cutaneous nodules or lesions at the site of the sandfly bite ( localised cutaneous leishmaniasis ) . In some cases , parasites disseminate through the skin and present as multiple non-ulcerative nodules ( diffuse cutaneous leishmaniasis , DCL ) or propagate through the lymphatic system resulting in nasobronchial and buccal mucosal tissue destruction ( mucosal leishmaniasis , ML ) ( Reithinger et al . , 2007; Dedet and Pratlong , 2009 ) . Localised CL may resolve spontaneously and usually responds well to treatment; management of DCL and ML cases is more difficult and cases may take considerably longer to resolve , if at all . Visceral leishmaniasis generally affects the spleen , liver , or other lymphoid tissues , and , if left untreated , is fatal; a fraction of successfully treated VL cases may result in maculopapular or nodular rashes ( post-kala-azar dermal leishmaniasis ) ( Murray et al . , 2005; Dedet and Pratlong , 2009 ) . While the Leishmania species determines which of the main two forms of the leishmaniases will result from infection , establishment , progression , and severity of infection as well as treatment regimen and outcome is dependent on a range of other factors , including parasite strain , characteristics of sandfly saliva , parasite infection with Leishmania RNA virus , host genetics , and immunosuppression , particularly due to HIV co-infection ( Reithinger et al . , 2007; Ives et al . , 2011; Novais et al . , 2013 ) . Species distribution models provide a robust means of mapping these diseases at a global level . These models define a set of conditions , from a selection of environmental covariates , which best categorise known occurrences . Through this categorisation , areas of unknown pathogen presence can be identified and thus a global evaluation of environmental suitability for presence can be made . A variety of factors can influence the distribution of an organism , including an array of environmental and other abiotic characteristics as well as biotic factors ( Peterson , 2008 ) . Whilst many areas may be environmentally suitable for a given species , other factors may prevent the species from being present in all of these locations . This distinction is often referred to as the difference between the fundamental and the realised niche of the species , the former describing a potential distribution based upon specific features of the environment whilst the latter indicates the distribution we observe . Such a framework can be applied just as successfully in the context of pathogens and their vectors as with macroorganisms ( Peterson et al . , 2011 ) and has already been applied to the mapping of malaria vectors ( Sinka et al . , 2010 , 2010 , 2011 ) and dengue ( Bhatt et al . , 2013 ) . The relationships between the leishmaniases and environmental and socioeconomic factors known to influence their distribution at a global scale has not previously been considered in a comprehensive and quantitative manner ( Hay et al . , 2013 ) . This study uses these modelling techniques in order to define the first evidence-based global environmental risk maps of the leishmaniases . For each province or state across the globe ( classed as Admin 1 by the Food and Agriculture Organization's Global Administrative Unit Layers ( FAO , 2008 ) , totalling some 3450 ) evidence was collected regarding CL and VL presence or absence . An assessment of the consensus of this evidence ranging from comprehensive agreement on disease presence ( +100% ) to consensus of disease absence ( −100% ) was made . Figures 1A–4A present these evidence consensus maps , with full reasoning for each administrative unit's score outlined in the associated data set ( Dryad data set doi: 10 . 5061/dryad . 05f5h ) . For Brazil , it was possible to perform this analysis at the district level ( classed as Admin 2 ) totalling some 5510 units . In total , 950 Admin 1 units from 84 countries reported a consensus on CL presence greater than indeterminate ( a score of 0 ) , with 310 Admin 1 units from 42 countries reporting a complete consensus on the presence of CL . In Brazil , 2469 Admin 2 regions recorded CL cases over the period of investigation . Consensus on the presence of VL ( score greater than 0 ) was reported in 793 Admin 1 units from 77 countries , with 88 Admin 1 units from 32 countries reporting complete consensus on VL . In Brazil , 1320 Admin 2 units recorded VL cases . 10 . 7554/eLife . 02851 . 003Figure 1 . Reported and predicted distribution of cutaneous leishmaniasis in the New World . ( A ) Evidence consensus for presence of the disease ranging from green ( complete consensus on the absence: −100% ) to purple ( complete consensus on the presence of disease: +100% ) . The blue spots indicate occurrence points or centroids of occurrences within small polygons . ( B ) Predicted risk of cutaneous leishmaniasis from green ( low probability of presence ) to purple ( high probability of presence ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 00310 . 7554/eLife . 02851 . 004Figure 1—figure supplement 1 . Uncertainty associated with predictions in Figure 1B . Uncertainty was calculated as the range of the 95% confidence interval in predicted probability of occurrence for each pixel . Regions of highest uncertainty are in dark brown , with blue representing low uncertainty . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 00410 . 7554/eLife . 02851 . 005Figure 2 . Reported and predicted distribution of visceral leishmaniasis in the New World . ( A ) Evidence consensus for presence of the disease ranging from green ( complete consensus on the absence: −100% ) to purple ( complete consensus on the presence of disease: +100% ) . The blue spots indicate occurrence points or centroids of occurrences within small polygons . ( B ) Predicted risk of visceral leishmaniasis from green ( low probability of presence ) to purple ( high probability of presence ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 00510 . 7554/eLife . 02851 . 006Figure 2—figure supplement 1 . Uncertainty associated with predictions in Figure 2B . Uncertainty was calculated as the range of the 95% confidence interval in predicted probability of occurrence for each pixel . Regions of highest uncertainty are in dark brown , with blue representing low uncertainty . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 00610 . 7554/eLife . 02851 . 007Figure 3 . Reported and predicted distribution of cutaneous leishmaniasis in the Old World . ( A ) Evidence consensus for presence of the disease ranging from green ( complete consensus on the absence: −100% ) to purple ( complete consensus on the presence of disease: +100% ) . The blue spots indicate occurrence points or centroids of occurrences within small polygons . ( B ) Predicted risk of cutaneous leishmaniasis from green ( low probability of presence ) to purple ( high probability of presence ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 00710 . 7554/eLife . 02851 . 008Figure 3—figure supplement 1 . Uncertainty associated with predictions in Figure 3B . Uncertainty was calculated as the range of the 95% confidence interval in predicted probability of occurrence for each pixel . Regions of highest uncertainty are in dark brown , with blue representing low uncertainty . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 00810 . 7554/eLife . 02851 . 009Figure 3—figure supplement 2 . Reported and predicted distribution of cutaneous leishmaniasis in northeast Africa . ( A ) Evidence consensus for presence of the disease ranging from green ( complete consensus on the absence: −100% ) to purple ( complete consensus on the presence of disease: +100% ) . The blue spots indicate occurrence points or centroids of occurrences within small polygons . ( B ) Predicted risk of cutaneous leishmaniasis from green ( low probability of presence ) to purple ( high probability of presence ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 00910 . 7554/eLife . 02851 . 010Figure 3—figure supplement 3 . Reported and predicted distribution of cutaneous leishmaniasis across the Near East , including Syria , Iran and Afghanistan . ( A ) Evidence consensus for presence of the disease ranging from green ( complete consensus on the absence: −100% ) to purple ( complete consensus on the presence of disease: +100% ) . The blue spots indicate occurrence points or centroids of occurrences within small polygons . ( B ) Predicted risk of cutaneous leishmaniasis from green ( low probability of presence ) to purple ( high probability of presence ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 01010 . 7554/eLife . 02851 . 011Figure 4 . Reported and predicted distribution of visceral leishmaniasis in the Old World . ( A ) Evidence consensus for presence of the disease ranging from green ( complete consensus on the absence: −100% ) to purple ( complete consensus on the presence of disease: +100% ) . The blue spots indicate occurrence points or centroids of occurrences within small polygons . ( B ) Predicted risk of visceral leishmaniasis from green ( low probability of presence ) to purple ( high probability of presence ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 01110 . 7554/eLife . 02851 . 012Figure 4—figure supplement 1 . Uncertainty associated with predictions in Figure 4B . Uncertainty was calculated as the range of the 95% confidence interval in predicted probability of occurrence for each pixel . Regions of highest uncertainty are in dark brown , with blue representing low uncertainty . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 01210 . 7554/eLife . 02851 . 013Figure 4—figure supplement 2 . Reported and predicted distribution of visceral leishmaniasis in northeast Africa . ( A ) Evidence consensus for presence of the disease ranging from green ( complete consensus on the absence: −100% ) to purple ( complete consensus on the presence of disease: +100% ) . The blue spots indicate occurrence points or centroids of occurrences within small polygons . ( B ) Predicted risk of visceral leishmaniasis from green ( low probability of presence ) to purple ( high probability of presence ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 01310 . 7554/eLife . 02851 . 014Figure 4—figure supplement 3 . Reported and predicted distribution of visceral leishmaniasis in the Indian subcontinent . ( A ) Evidence consensus for presence of the disease ranging from green ( complete consensus on the absence: −100% ) to purple ( complete consensus on the presence of disease: +100% ) . The blue spots indicate occurrence points or centroids of occurrences within small polygons . ( B ) Predicted risk of visceral leishmaniasis from green ( low probability of presence ) to purple ( high probability of presence ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 01410 . 7554/eLife . 02851 . 015Figure 4—figure supplement 4 . Population at risk estimates for leishmaniasis . Four scatterplots showing the relationship between non-zero estimated mean annual incidence ( Alvar et al . , 2012 ) and estimated population at risk derived from the cartographic approach for ( A ) New World cutaneous leishmaniasis , ( B ) New World visceral leishmaniasis , ( C ) Old World cutaneous leishmaniasis , and ( D ) Old World visceral leishmaniasis . For each country the bars represent the annual incidence estimate range . DOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 015 Of the 10 countries ( Afghanistan , Colombia , Brazil , Algeria , Peru , Costa Rica , Iran , Syria , Ethiopia , and Sudan ) that contribute 75% of the global estimated CL incidence ( Alvar et al . , 2012 ) , only Algeria did not have regions of complete evidence consensus on presence due to incomplete and non-contemporary case data . Similarly , of the six countries ( Brazil , Ethiopia , Sudan , South Sudan , India , and Bangladesh ) that report 90% of all VL cases ( Alvar et al . , 2012 ) , all six had regions of complete consensus on VL . Figures 1A–4A also show the spatial distribution of occurrence data , defined as one or more reports of leishmaniasis in a given calendar year , collated from a variety of sources . Overall , there is a relatively broad geographic spread and good correspondence with the evidence consensus maps for each disease . Tunisia , Morocco and Brazil report the highest number of unique CL occurrences in any given year , whilst India reported the largest proportion of the VL occurrence data . Table 1 reports the sources and types of data within the occurrence database . Whilst the majority of occurrence records contain accurate point data ( 62% ) , the remainder were recorded at a provincial or district level . Occurrence records for the two diseases were relatively similar in number with a total of 6426 records for CL and 6137 for VL . 10 . 7554/eLife . 02851 . 016Table 1 . Origin and spatial resolution of leishmaniasis occurrence dataDOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 016Origin and resolution of occurrence dataPoint dataProvince level dataDistrict level dataTotalCutaneous leishmaniasis Literature368087912205779 CNR-L5314731609 HealthMap31––31 GenBank6–17 Total424892612526426Visceral leishmaniasis Literature3050150010685618 CNR-L4292429482 HealthMap321–33 GenBank3–14 Total3514152510986137Each cell gives the number of occurrence records added to the data set by considering each additional datasource after removing duplicate records . Occurrence records are separated by spatial resolution—whether they are recorded as points ( typically representing settlements ) or as province level ( admin 1 ) or district level ( admin 2 ) data . Figures 1B–4B show the global predicted environmental risk maps for CL and VL . Table 2 identifies the top five predictor variables in each of the four modelled regions ( since CL and VL were modelled separately in the Old World and New World ) as measured by average contribution to the boosted regression trees ( BRT ) submodels . Peri-urban and urban land cover is an important predictor of the distribution of CL in the Old World and of VL globally . Abiotic factors such as land surface temperature ( LST ) were better predictors of CL than of VL . In total , LST variables ( annual minimum , maximum and mean ) explain 21 . 99% of CL distribution in the Old World and 43 . 65% of CL distribution in the New World ( with maximum LST having the highest relative contribution ) . Abiotic factors combined ( including LST , normalised difference vegetation index ( NDVI ) and precipitation ) accounted for 29 . 02% and 48 . 55% of VL distribution in the Old World and New World , respectively . Validation statistics for all models were high with a mean area under the receiver operator curve ( AUC ) above 0 . 97 and mean correlations above 0 . 85 for all models . 10 . 7554/eLife . 02851 . 017Table 2 . Mean relative contribution of predictor variables to the ensemble BRT models of CL and VL in both the Old and New WorldDOI: http://dx . doi . org/10 . 7554/eLife . 02851 . 017Top predictors of CLRelative contributionTop predictors of VLRelative contributionOld world Peri-urban extents47 . 34Peri-urban extents51 . 50 Minimum LST18 . 36Urban extents17 . 38 Urban extents9 . 01Maximum NDVI7 . 87 G-Econ7 . 33Minimum LST5 . 87 Minimum Precipitation4 . 95Maximum Precipitation4 . 00New World Maximum LST36 . 91Peri-urban extents25 . 90 Peri-urban extents18 . 61Urban extents21 . 24 Maximum precipitation12 . 06Mean LST9 . 18 Minimum precipitation6 . 21Mean NDVI7 . 83 Minimum LST4 . 39Maximum LST6 . 40LST = Land Surface Temperature , G-Econ = Geographically based Economic data , NDVI = Normalised Difference Vegetation Index . In the New World , CL is predicted to occur primarily within the Amazon basin and other areas of rainforest . By contrast , VL is predicted to occur mainly along the coastline of Brazil , with sporadic foci across the rest of Southern and Central America . Outside of their main foci , both diseases are strongly associated with urban and peri-urban areas , resulting in a focal distribution throughout much of the New World . In the Old World , both CL and VL are predicted to be present from the Mediterranean Basin across the Near East to Northwest India , with a few foci in Central China as well as in a thin band of predicted risk across West Africa and in the Horn of Africa . The predicted distribution of VL also extends into Northeast India and China with a large predicted focus in the northwest . The populations living in areas predicted to be subject to environmental risk of CL and VL are estimated to be 1 . 71 billion and 1 . 69 billion , respectively , approximately a quarter of the world's population . Figure 4—figure supplement 4 compares these national estimates to the annual case incidence data from all countries for which at least one case per annum was estimated by Alvar et al . ( 2012 ) . There is a strong positive association between the two measures of disease occurrence . We provide estimates of the populations at risk in 90 countries for which no human cases of CL or VL were regularly reported ( Alvar et al . , 2012 ) . A full table of this information is presented in the associated Dryad data set ( doi: 10 . 5061/dryad . 05f5h ) . For many of these countries , Alvar et al . ( 2012 ) reported a handful of sporadic cases over the years indicating very rare occurrence of infection , whilst the remainder were countries with inconclusive evidence of disease presence or absence . It is important to note that the relationship between environmental risk and true incidence of disease remains to be elucidated; however the association between populations living in areas of environmental risk and national level estimates of incidence suggests that the modelled occurrence–incidence relationship approach used by Bhatt et al . ( 2013 ) for dengue could be applied if the necessary longitudinal cohort study data were available . These maps represent evidence-based estimates of the current global distribution of the leishmaniases incorporating a comprehensive occurrence database and a rigorous statistical modelling framework with associated uncertainty statistics . We estimate that 1 . 71 billion and 1 . 69 billion individuals live in areas that are suitable for CL and VL transmission , respectively . These figures highlight the need for much greater awareness of this disease at a global scale . These maps provide an important baseline assessment and a strong foundation on which to base future burden estimates , target regions for control efforts and inform public health decisions . The methodology used for generating the definitive extents for the leishmaniases was adapted from work on dengue ( Brady et al . , 2012 ) . Four primary evidence categories were used to determine a consensus on the presence or absence of the leishmaniases: ( i ) health reporting organisations; ( ii ) peer-reviewed evidence of local autochthonous transmission; ( iii ) case data; ( iv ) supplementary information . Cutaneous and visceral leishmaniasis were the two symptomatologies investigated: other forms of the disease were subset within these two – whilst VL contained cases of post-kala-azar dermal leishmaniasis , CL included diffuse , disseminated , and mucosal forms of the disease . Although limited amounts of data were available for some of these forms , their epidemiology is similar , and consequently this categorisation was seen as appropriate . Information was collected at provincial level ( termed Admin 1 units by the Food and Agriculture Organization's ( FAO ) Global Administrative Unit Layers ( GAUL ) coding ( FAO , 2008 ) ) to better capture the focal nature of these diseases . Two separate searches using PubMed and Web of Knowledge were undertaken using the search parameter “leish* , ” and including articles up to December 2012 , and their respective abstracts , were filtered for relevance . From these searches , 4845 articles were collated , with data recorded at the resolution of either a point or Admin 1 or 2 polygon . These were then geo-positioned using Google Maps ( https://maps . google . co . uk/ ) . Each entry was evaluated to ensure that non-autochthonous cases and duplicate entries were eliminated . Each occurrence was assigned a start and end date based upon the content of the paper , used to define the time period over which occurrences were reported . In addition to this resource , reports were taken from the HealthMap database ( http://healthmap . org/en/ ) . HealthMap is an online based infectious disease surveillance system that compiles data from informal data sources ranging from online news articles to ProMED reports ( Freifeld et al . , 2008 ) . It parses information from these sources searching for relevant keywords , and then , using crowdsourcing and automated processes , geopositions those relating to the disease of interest . As of December 2012 , a total of 690 leishmaniasis relevant articles were archived . Searches were also performed on GenBank accessions , searching for archived genetic information from Leishmania spp . known to infect humans ( WHO , 2010 ) . If the host was identified as human , geographic indicators were assigned either as point , Admin 1 or Admin 2 , based upon the information in the location tag . Tags at the national level were filtered out of the data set . In total , 563 accessions were associated with sub-national location details and added to the database . Finally , data were provided from the curated strain archives of the Centre National de Référence des Leishmanioses ( CNR-L ) in Montpellier , France . In total , information about 3465 strains isolated from humans was provided , collected from between 1954 and 2013 . All data were geopositioned as precisely as possible , which resulted in both point-level data ( referring to cities , towns or villages ) as well as polygon-level data ( provinces or districts ) with area no greater than one square decimal degree . All data that had been manually geopositioned were checked to ensure coordinates were plausible and then occurrences were standardised annually to remove intra-annual duplicates , so that each individual record used in our model represented an occurrence of leishmaniasis infections in a given 5 km × 5 km location or administrative unit for one given year . As a result , the occurrence data were independent of burden; a location with 200 cases in one year has equal weighting in the model as a location with just one reported case , since it was only the presence of the disease being modelled . Leishmania spp . are known to have anthroponotic , zoonotic , or sylvatic transmission cycles in nature ( WHO , 2010; Ready , 2013 ) which is apparent in the focal nature of the disease; however , there are some key features of the environment that are important in determining the distribution of disease across the globe . Numerous models have been constructed for local transmission scenarios implicating various environmental features from temperature and precipitation to socioeconomic factors relating to standards of living in villages in endemic foci . For the modelling process , a suite of global gridded environmental , biologically plausible , correlates was generated . The boosted regression trees ( BRT ) methodology employed for mapping the leishmaniases is a variant of the model used in a previous analysis of dengue ( Bhatt et al . , 2013 ) . Boosted regression tree modelling combines both regression trees , which build a set of decision rules on the predictor variables by portioning the data into successively smaller groups with binary splits ( De'ath , 2007; Elith et al . , 2008 ) , and boosting , which selects the tree that minimises the loss of function , to best capture the variables that define the distribution of the input data . The core BRT setup followed standard protocol already defined elsewhere ( Elith et al . , 2008; Bhatt et al . , 2013 ) . Population living in areas of risk was estimated by using a threshold probability to reclassify the probabilistic risk maps into a binary risk map , then extracting the total human population in the ‘at risk’ areas using a gridded data set of human population density from 2010 ( Balk et al . , 2006; CIESIN/IFPRI/WB/CIAT , 2007 ) . The threshold value was set such that 95% of the point occurrence records fell within the at risk area . 5% of occurrence points were allowed to fall outside the predicted risk area to account for errors which could have arisen either from errors in the occurrence data set or from inaccuracies in the predicted risk maps . For external validation , this population at risk information was compared to national reported annual cases ( Alvar et al . , 2012 ) to produce Figure 4—figure supplement 4 . In these figures , the points represent the mean value of the estimated annual incidence reported taking into account the authors estimates of underreporting rates ( Alvar et al . , 2012 ) . The upper and lower limits to these estimates are reflected by the bars around each point . Note that these figures use a log-scale on each axis and that only countries with non-zero estimates by Alvar et al . ( 2012 ) are included . The threshold probabilities of occurrence used to define ‘at risk’ were as follows: NW CL—0 . 22 , OW CL—0 . 19 , NW VL—0 . 42 , OW VL—0 . 19 .	Each year 1–2 million people are diagnosed with a tropical disease called leishmaniasis , which is caused by single-celled parasites . People are infected when they are bitten by sandflies carrying the parasite , and often develop skin lesions around the bite site . Though mild cases may recover on their own or with treatment , sometimes the parasites multiply and spread elsewhere causing further skin lesions and facial disfigurement . Furthermore , the parasites can also infect internal organs such as the spleen and the liver , which without treatment can be fatal . The parasites that cause leishmaniasis are found in 88 countries around the world , mainly in South and Central America , Africa , Asia , and southern Europe . However , over 90% of potentially fatal infections occur in just six countries: Brazil , Ethiopia , Sudan , South Sudan , India , and Bangladesh . Although a few studies have looked at the distribution of leishmaniasis within different countries , we still do not have a complete picture of the distribution of the disease on a global scale . To address this , Pigott et al . set out to create detailed maps of the distribution of leishmaniasis and the factors that promote its spread . Similar techniques had been previously used to map dengue fever , another tropical disease . Computer modelling was used to generate the maps based on data about the environment at the locations of known cases of leishmaniasis . This information was then used to infer the likelihood of leishmaniasis being present at other locations across the globe . Based on their maps , Pigott et al . estimate that about 1 . 7 billion people , or one quarter of the world's population , live in areas where they are at potential risk of leishmaniasis . People living in built-up areas outside of cities are at the greatest risk , likely because some sandfly species prefer to live near dwellings , but other social and economic factors also contribute to the risk of catching this disease . The factors driving the transmission of leishmaniasis differed in the Old World ( Europe , Africa and Asia ) and the New World ( the Americas ) : built-up areas were more likely to be at risk in the Old World , while temperature and rainfall were bigger factors affecting risk in the New World . It is hoped that the maps created by Pigott et al . will help inform future estimates of the burden of leishmaniasis and target surveillance and disease control efforts more effectively to combat this tropical disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "epidemiology", "and", "global", "health", "microbiology", "and", "infectious", "disease" ]	2014	Global distribution maps of the leishmaniases
A central question is how specificity in cellular responses to the eukaryotic second messenger Ca2+ is achieved . Plant guard cells , that form stomatal pores for gas exchange , provide a powerful system for in depth investigation of Ca2+-signaling specificity in plants . In intact guard cells , abscisic acid ( ABA ) enhances ( primes ) the Ca2+-sensitivity of downstream signaling events that result in activation of S-type anion channels during stomatal closure , providing a specificity mechanism in Ca2+-signaling . However , the underlying genetic and biochemical mechanisms remain unknown . Here we show impairment of ABA signal transduction in stomata of calcium-dependent protein kinase quadruple mutant plants . Interestingly , protein phosphatase 2Cs prevent non-specific Ca2+-signaling . Moreover , we demonstrate an unexpected interdependence of the Ca2+-dependent and Ca2+-independent ABA-signaling branches and the in planta requirement of simultaneous phosphorylation at two key phosphorylation sites in SLAC1 . We identify novel mechanisms ensuring specificity and robustness within stomatal Ca2+-signaling on a cellular , genetic , and biochemical level . Cytosolic calcium ( [Ca2+]cyt ) functions as key cellular second messenger in a plethora of crucial processes in plants and other eukaryotes ( Hetherington and Woodward , 2003; Clapham , 2007; McAinsh and Pittman , 2009; Berridge , 2012; Charpentier and Oldroyd , 2013; Webb , 2013 ) . Elucidation of the mechanisms mediating specificity in Ca2+ signaling is fundamental to understanding signal transduction ( Berridge et al . , 2003; Hetherington and Woodward , 2003; Clapham , 2007; Webb , 2013 ) . In a few cases , the biochemical and cellular mechanisms mediating Ca2+ signaling specificity have been revealed ( e . g . De Koninck and Schulman , 1998; Dolmetsch et al . , 1998; Oancea and Meyer , 1998; Dolmetsch et al . , 2001; Bradshaw et al . , 2003; Rellos et al . , 2010; Chao et al . , 2011 ) . More than one ( non-exclusive ) mechanism is likely to contribute to specificity in Ca2+ signal transduction ( Berridge et al . , 2003; Dodd et al . , 2010 ) . However , characterization of the combined cellular , biochemical , and genetic mechanisms underlying Ca2+ specificity in a single cell type has not been achieved to our knowledge . The genome of the plant Arabidopsis thaliana encodes over 200 EF-hand Ca2+-binding proteins ( Day et al . , 2002 ) , with many of these genes co-expressed in the same cell types ( Harmon et al . , 2000; McCormack et al . , 2005; Schmid et al . , 2005; Winter et al . , 2007 ) , illustrating the need for Ca2+ specificity signaling mechanisms in plants . Two guard cells form a stomatal pore representing the gateway for CO2 influx , which is inevitably accompanied by plant water loss . The aperture of stomatal pores is consequently tightly regulated by the guard cells . Intracellular Ca2+ represents a key second messenger in stomatal closing ( McAinsh et al . , 1990; MacRobbie , 2000; Hetherington , 2001; Hetherington and Woodward , 2003; Hubbard et al . , 2012 ) , but intracellular Ca2+ also functions in stomatal opening ( Irving et al . , 1992; Shimazaki et al . , 1992; Curvetto et al . , 1994; Shimazaki et al . , 1997; Cousson and Vavasseur , 1998; Young et al . , 2006 ) , raising the question how cytosolic free Ca2+ concentration ( [Ca2+]cyt ) elevations trigger a specific cellular response . The underlying mechanisms mediating specificity in guard cell Ca2+ signaling are not well understood . The development of genetic , electrophysiological , and cell signaling tools for the dissection of Ca2+ signaling within this model cell type renders guard cells a powerful system for the investigation of specificity mechanisms within Ca2+ signal transduction . Recent studies including analyses in intact Arabidopsis ( Young et al . , 2006 ) and Vicia faba ( Chen et al . , 2010 ) guard cells , have shown that stomatal closing stimuli including abscisic acid ( ABA ) and CO2 enhance the [Ca2+]cyt sensitivity of downstream signaling mechanisms , switching them from an inactivated state to an enhanced Ca2+-responsive ‘primed’ state , thus tightly controlling specificity in Ca2+ responsiveness ( Young et al . , 2006; Munemasa et al . , 2007; Siegel et al . , 2009; Chen et al . , 2010; Xue et al . , 2011 ) . A rise of [Ca2+]cyt from resting to elevated levels alone does not trigger the full ion channel regulation and stomatal response ( Young et al . , 2006; Munemasa et al . , 2007; Siegel et al . , 2009; Chen et al . , 2010; Xue et al . , 2011 ) . Similarly , a recent study of pathogen-associated molecular pattern ( PAMP ) signaling suggests that prior PAMP signaling enhances the sensitivity to intracellular Ca2+ during signal transduction ( Kadota et al . , 2014 ) , indicating that this principle for Ca2+ specificity priming may be more widely used in plants . The biological closing stimulus has to be present for the guard cell to react to physiological Ca2+ elevation . However , the biochemical and genetic mechanisms mediating Ca2+ sensitivity priming remain unknown . SLAC1 represents the major anion channel mediating S-type anion currents in guard cells ( Negi et al . , 2008; Vahisalu et al . , 2008 ) and Ca2+ activation of S-type anion currents is an early and crucial step in stomatal closure ( Schroeder and Hagiwara , 1989; McAinsh et al . , 1990; Siegel et al . , 2009; Chen et al . , 2010 ) . Ca2+-independent SnRK2 protein kinases ( Li et al . , 2000; Mustilli et al . , 2002; Yoshida et al . , 2002 ) , most importantly OST1 , have been shown to activate SLAC1 in Xenopus leavis oocytes ( Geiger et al . , 2009; Lee et al . , 2009; Brandt et al . , 2012 ) . The full length Ca2+-dependent protein kinases 6 , 21 , and 23 ( CPK6 , CPK21 , and CPK23 ) also activate SLAC1 in oocytes ( Geiger et al . , 2010; Brandt et al . , 2012 ) . Presently , the Ca2+-dependent and Ca2+–independent branches are considered to function independently ( e . g . Li et al . , 2006; Kim and et al . , 2010; Roelfsema et al . , 2012 ) . The activation of SLAC1 by OST1 or CPK6 is inhibited by the clade A protein phosphatase 2Cs ( PP2Cs ) ABI1 , ABI2 , or PP2CA in oocytes ( Geiger et al . , 2009; Lee et al . , 2009; Brandt et al . , 2012 ) . The cytosolic ABA-receptors pyrabactin resistance ( PYR ) /PYR-like ( PYL ) /regulatory component of ABA receptor ( RCAR ) ( Ma et al . , 2009; Park and et al . , 2009 ) have been shown to inhibit PP2C activity in the presence of ABA ( Ma et al . , 2009; Park and et al . , 2009; Santiago et al . , 2009; Nishimura et al . , 2010; Szostkiewicz et al . , 2010 ) . Reconstitution of ABA activation of SLAC1 in Xenopus oocytes has been shown by co-expression of the ABA-receptor PYR1 together with SLAC1 , PP2Cs , and either Ca2+-independent OST1 or Ca2+-dependent CPK6 protein kinases ( Brandt et al . , 2012 ) . However , whether the Ca2+-dependent and–independent branches in ABA signal transduction are functionally linked and depend on one-another in planta remains to be investigated using higher order genetic mutants . Here we present biochemical , genetic and cellular signaling findings that describe mechanisms underlying specificity and robustness in Ca2+ signaling within a single cell type and demonstrate an unexpected strong dependence of the Ca2+-dependent signal transduction branch on the Ca2+-independent pathway in guard cells . Moreover our results suggest that in contrast to OST1 ( Umezawa et al . , 2009; Vlad et al . , 2009 ) , calcium-dependent protein kinases ( CPKs ) are not directly deactivated by PP2Cs , but these PP2Cs rapidly deactivate both of the Ca2+-dependent and Ca2+–independent branches by directly dephosphorylating the protein kinase target SLAC1 . Previous studies have shown that A . thaliana single or double mutants in CPKs cause partial ABA-insensitivities in guard cell signaling ( Mori et al . , 2006; Zhu et al . , 2007; Hubbard et al . , 2012 ) . We addressed the question whether higher order CPK gene disruption mutant plants display more strongly impaired ABA responses . CPK6 and CPK23 were shown to activate SLAC1 in Xenopus oocytes and disruption of the corresponding genes in plants leads to a partial reduction of S-type anion current activation in guard cells ( Mori et al . , 2006; Geiger et al . , 2010; Brandt et al . , 2012 ) . The closest homolog to CPK6 , CPK5 , is associated with reactive oxygen species signaling ( Boudsocq et al . , 2010; Dubiella et al . , 2013 ) . CPK5 also activates SLAC1 in oocytes ( Figure 1—figure supplement 1A , B ) . Whole-cell patch-clamp analysis showed that mutation of CPK5 alone does not substantially disrupt ABA-activation of S-type anion channels ( Figure 1—figure supplement 1C , D ) , consistent with findings of over-lapping gene functions in this response ( Mori et al . , 2006; Hubbard et al . , 2012 ) . CPK11 is highly expressed in guard cells and involved in ABA responses ( Zhu et al . , 2007; Geiger et al . , 2009 ) . We isolated cpk5/6/11/23 quadruple T-DNA insertion mutant plants and investigated ABA-induced S-type anion channel current regulation . Either ABA treatment ( Siegel et al . , 2009 ) or by-passing ABA signaling by exposure of guard cells to a high external Ca2+ shock ( Allen et al . , 2002 ) renders wildtype ( Col0 ) guard cells sensitive to physiological [Ca2+]cyt increases . Notably , even when previously exposed to ABA or a high external Ca2+ shock , 2 μM [Ca2+]cyt did not result in S-type anion current activation in cpk5/6/11/23 quadruple mutant guard cells in contrast to WT plants ( Figure 1A–D ) . These results show an important role of these calcium sensing protein kinases in ABA-dependent S-type anion channel activation in guard cells . We further investigated ABA-induced stomatal movement responses . Application of 5 μM ABA to WT leaves significantly decreased stomatal apertures compared to mock-treated control stomatal apertures ( Figure 1E; p < 0 . 05 ) . In the cpk5/6/11/23 mutant , however , 5 μM ABA-induced stomatal closing was not significant ( Figure 1E; p = 0 . 51 ) . When the ABA concentration was increased to 10 μM , ABA-induced stomatal closure was weakened in cpk5/6/11/23 mutant leaves ( Figure 1F; p = 0 . 07; 0 min ABA-exposed cpk5/6/11/23 mutant leaves compared to 60 min ABA-exposed cpk5/6/11/23 mutant leaves ) . The partial ABA response at the higher ABA concentration may be linked to parallel activation of R-type anion channels ( see ‘Discussion’ ) . 10 . 7554/eLife . 03599 . 003Figure 1 . Calcium-dependent protein kinase ( CPK ) quadruple loss of function mutants show abscisic acid ( ABA ) and Ca2+ insensitive S-type anion current activation and are impaired in stomatal closing . ( A–D ) Intracellular Ca2+-activation of S-type anion channels enabled by pre-exposure to ABA ( A and C ) or high external Ca2+ pre-shock ( Allen et al . , 2002 ) ( B and D ) is strongly impaired in cpk5/6/11/23 guard cells at 2 μM [Ca2+]cyt . ( E and F ) 5 μM ABA-application to intact leaves shows impaired ABA-induced stomatal closing in cpk5/6/11/23 mutant plants ( E; p = 0 . 51 Mock-treated cpk quadruple mutant vs ABA-treated cpk quadruple mutant stomata; unpaired t-test; n = 6 experiments and >51 total stomata per group ) . Application of 10 μM ABA results in a partially reduced average stomatal response ( F , p = 0 . 07; 0 min ABA-exposed cpk5/6/11/23 mutant leaves compared to 60 min ABA-exposed cpk5/6/11/23 mutant leaves; Student's t-test; n = 3 experiments and >59 total stomata per group ) . Representative whole cell currents ( A and B ) , average steady-state current–voltage relationships ±SEM ( C and D ) , average guard cell apertures ±SEM ( E and F ) are shown . Measurements shown in Figure 1C and Figure 1—figure supplement 1D were acquired under the same experimental condition . Therefore , WT Control and WT + ABA control data are the same in both figures . Several error bars are not visible , as these were smaller than the illustrated symbols . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 00310 . 7554/eLife . 03599 . 004Figure 1—figure supplement 1 . CPK5 activates SLAC1 in Xenopus oocytes and ABA-activation of S-type anion currents in cpk5 single mutant is not impaired . ( A and B ) Whole cell currents were measured in Xenopus oocytes expressing SLAC1 together with CPK5 and , as a control , CPK6 . Large Cl− currents show that CPK5 is capable of activating SLAC1 . ( C and D ) ABA activates S-type anion currents in cpk5 mutant guard cells similar to ABA-activation of S-type anion currents in WT guard cells . Representative current traces ( A and C ) , average steady-state current–voltage relationships ( ±SEM ) , and numbers of individual cells are shown ( B and D ) . Note that experiments shown in Figure 1C and Figure 1—figure supplement 1D were acquired under the same experimental conditions . Therefore WT Control and the WT + ABA control data are the same in both figure panels . Several error bars are not visible , as these were smaller than the illustrated symbols . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 004 Members of the clade A of the PP2C class play important roles as negative regulators of ABA signaling ( Cutler et al . , 2010 ) and were shown to inhibit CPK-activation of SLAC1 in oocytes ( Geiger et al . , 2010; Brandt et al . , 2012 ) . To determine whether these PP2Cs function in the ABA-triggered enhancement of the [Ca2+]cyt-sensitivity in guard cells , we performed whole-cell patch-clamp analysis using a plant line carrying T-DNA insertion mutations in the key ABA signaling PP2Cs ABI1 , ABI2 , HAB1 , and PP2CA ( abi1-2/abi2-2/hab1-1/pp2ca-1 ) . Surprisingly , in abi1-2/abi2-2/hab1-1/pp2ca-1 quadruple mutant guard cells , strong Ca2+-activated S-type anion currents were observed even without pre-exposure to ABA ( Figure 2A–D ) . At low 0 . 1 μM [Ca2+]cyt S-type anion channels did not show significant activation in the pp2c quadruple mutant compared to WT ( Figure 2—figure supplement 1A , B; p = 0 . 294 at −145 mV ) . These findings provide genetic evidence for first genes that are essential for the ABA-triggered Ca2+ sensitivity priming in guard cells and show that these PP2Cs provide a mechanism ensuring specificity in Ca2+ signal transduction . 10 . 7554/eLife . 03599 . 005Figure 2 . In protein phosphatase 2C ( PP2C ) quadruple mutant plants , Ca2+ activation of S-type anion currents is constitutively primed . ( A and C ) 2 μM [Ca2+]cyt activates S-type anion currents in WT if the guard cells were pre-exposed to ABA . ( B and D ) In PP2C quadruple mutant guard cells ABA pre-exposure is not required for 2 μM [Ca2+]cyt-activation of S-type anion currents . Average steady-state current–voltage relationships ±SEM , guard cell numbers ( C and D ) , and representative whole cell currents ( A and B ) are presented . Several error bars are not visible , as these were smaller than the illustrated symbols . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 00510 . 7554/eLife . 03599 . 006Figure 2—figure supplement 1 . Analysis of ABA activation of S-type anion currents in PP2C quadruple mutant guard cells at low [Ca2+]cyt . ( A and B ) ABA application in WT and abi1-2/abi2-2/hab1-1/pp2ca-1 guard cells with [Ca2+]cyt buffered to a resting level of 0 . 1 μM does not result in large S-type anion current activation . Typical current traces ( A ) , average steady-state currents in response to applied voltages ( ±SEM ) , and numbers of individual measured guard cells are shown ( B ) . Several error bars are not visible , as these were smaller than the illustrated symbols . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 006 Based on the above results we sought to determine the biochemical mechanisms by which PP2Cs down-regulate Ca2+ sensitivity in the absence of ABA . The main SLAC1-activating protein kinase in the Ca2+-independent branch , OST1 ( Mustilli et al . , 2002; Yoshida et al . , 2002 ) , is directly inactivated by PP2Cs through de-phosphorylation of the activation loop ( Umezawa et al . , 2009; Vlad et al . , 2009 ) . We tested whether CPKs might be down-regulated by PP2Cs in a similar manner and whether pp2c quadruple mutant plants may also exhibit a constitutive OST1 activity . Our first approach to test whether CPK activity is regulated by ABA through PP2Cs was an in-gel protein kinase assay using protein extracts of Arabidopsis seedlings , which is routinely used to test OST1 activation by ABA ( Mustilli et al . , 2002 ) and also CPK activation by flg22 ( Boudsocq et al . , 2010 ) . Guard cell [Ca2+]cyt ranges from resting levels of ≈0 . 15 μM to stimulus induced elevated levels of above 1 μM ( McAinsh et al . , 1990 ) . Similar to studies reporting the ABA-activation of SnRK2 . 2 , SnRK2 . 3 , and SnRK2 . 6/OST1 ( Mustilli et al . , 2002; Yoshida et al . , 2002; Fujii et al . , 2007 ) , we compared the phosphorylation pattern of a reaction carried out at 0 . 15 μM free Ca2+ with the phosphorylation pattern at 3 μM free Ca2+ ( Figure 3A , B; for intermediate free Ca2+ concentration of 0 . 4 μM Ca2+ see Figure 3—figure supplement 2 ) . Incubating the gels in a reaction buffer with 3 μM free Ca2+ led to strong Ca2+-activated phosphorylation signals compared to resting Ca2+ at 0 . 15 μM ( Figure 3A , B ) . To determine whether these Ca2+-activated signals are CPK-derived we included two distinct quadruple mutants , cpk5/6/11/23 and cpk1/2/5/6 , in the in-gel kinase assays . Several Ca2+-activated bands disappeared or became notably weaker when extracts were tested from cpk5/6/11/23 and cpk1/2/5/6 ( Boudsocq et al . , 2010 ) plants ( Figure 3B and for improved visibility Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 03599 . 007Figure 3 . CPK activity is not changed by ABA or hyper-activated in pp2c quadruple mutants at defined Ca2+ concentrations . ( A and B ) In-gel kinase assays with Histone-III as substrate for whole plant protein extracts show ( B ) 3 μM Ca2+-activated trans-phosphorylation kinase activities independent of application of 50 μM ABA ( lanes 9 and 10 ) . In contrast , ABA activation of OST1 is clearly visible ( lanes 1–2 and 9–10 at ∼41 kDa; lower ‘OST1’ inset shows the same signal optimized autoradiography at the ∼41 kDa region and the corresponding gel regions are indicated by blue lines; see ‘Materials and methods’ ) . Disruption of four PP2C genes ( ABI1 , ABI2 , HAB1 , and PP2CA ) does not result in constitutive Ca2+-activated and OST1 kinase activities ( lanes 3–4 and 11–12 ) . In-gel kinase activities of two independent CPK quadruple mutant lines indicate that the Ca2+-activated kinase signals are CPK-derived ( compare lanes 9–10 with 13–16 in B and see Figure 3—figure supplement 1 ) ; predicted MWs for CPK1 , CPK2 , CPK5 , CPK6 , CPK11 , and CPK23 are 68 . 3 kDa , 72 . 3 kDa , 62 . 1 kDa , 61 . 1 kDa , 55 . 9 kDa , 58 . 7 kDa , respectively . ( C and D ) In-gel protein kinase assays with recombinant proteins show that incubation of the protein kinases with the PP2Cs ABI1 and PP2CA does ( C ) not change CPK6 activity while ( D ) OST1 activity is strongly down-regulated by PP2Cs . Each experiment has been repeated at least three times with similar results . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 00710 . 7554/eLife . 03599 . 008Figure 3—figure supplement 1 . Close up view of Ca2+-activated kinase activities . The region of predicted molecular weights of CPKs ( CPK1 , CPK2 , CPK5 , CPK6 , CPK11 , and CPK23 are 68 . 3 kDa , 72 . 3 kDa , 62 . 1 kDa , 61 . 1 kDa , 55 . 9 kDa , 58 . 7 kDa , respectively ) of the same autoradiograph which is shown in Figure 3B is magnified to increase the visibility of the individual bands . The apparent loss of the prominent bands with high molecular weight in cpk1/2/5/6 plants could correspond to CPK1 and CPK2 protein isoforms which possess the largest predicted molecular weights of all CPKs . Note that bands that run at a higher molecular weight than the predicted mass could be due to post-translational modifications . The faint band indicated with red asterisks might correspond to the closely related CPK5 and CPK6 as it is not clearly resolved in both mutants and also runs at the expected molecular weight . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 00810 . 7554/eLife . 03599 . 009Figure 3—figure supplement 2 . Protein kinase activities are not altered by ABA-application at 150 nM and 400 nM free Ca2+ . ( A and B ) Whole plant protein extracts were analyzed in in-gel protein kinase assays with the free Ca2+ concentration buffered to either 150 nM or 400 nM . No differences in the band pattern could be found in response to ABA at these buffered free Ca2+ concentrations ( A and B ) . The presence of 400 nM free Ca2+ did not enhance ABA activation of OST1 ( lower inset ‘OST1’ in A and B; lower ‘OST1’ inset shows the same signal optimized autoradiography at the ∼41 kDa region and the corresponding gel regions are indicated by blue lines; see ‘Materials and methods’ ) in WT , abi1-2/abi2-2/hab1-1/pp2ca-1 or cpk5/6/11/23 plants . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 00910 . 7554/eLife . 03599 . 010Figure 3—figure supplement 3 . Signals in in-gel kinase assays are largely derived from kinase trans-phosphorylation activities . ( A and B ) In-gel kinase assays with recombinant ( A ) CPK6 and ( B ) OST1 protein kinases in gels with ( left lanes in A and B ) and without ( right lanes in A and B ) the kinase substrate Histone-III were carried out . The signal in the gels without the substrate corresponds to auto-phosphorylation of the respective protein kinase alone and the band intensity of the gel in which Histone-III was immobilized results from auto- and trans-phosphorylation activities . Very low band intensities in the gels without Histone-III for CPK6 and OST1 ( right lanes in A and B ) show that signals in in-gel kinase assays using Histone-III as a substrate were mainly due to kinase trans-phosphorylation activities under the imposed conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 01010 . 7554/eLife . 03599 . 011Figure 3—figure supplement 4 . CPK6 is de-phosphorylated by the PP2Cs ABI1 , ABI2 , and PP2CA . In in vitro protein kinase assays , recombinant CPK6 was incubated in the presence of 5 μM free Ca2+ which results in auto-phosphorylation signals ( lanes 1 and 5 ) . After the initial auto-phosphorylation period the kinase inhibitor staurosporine ( Stau . ) and the PP2Cs ABI1 , ABI2 , and PP2CA were added to the reactions . For the samples displayed in lanes 1–4 , the Ca2+-chelator EGTA , which buffers free Ca2+ concentrations to <10 nM , was added together with staurosporine and the indicated PP2Cs . Subsequent addition of the PP2Cs ABI1 , ABI2 , and PP2CA resulted in decreased auto-phosphorylation signals showing that these PP2Cs de-phosphorylate CPK6 ( lanes 2–4 and 6–8 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 01110 . 7554/eLife . 03599 . 012Figure 3—figure supplement 5 . CPK6 kinase activity is not inhibited in the presence of ABI1 or PP2CA . In vitro protein kinase assays measuring the kinase activity via ATP consumption show that staurosporine ( Stau . ) but not ABI1 or PP2CA inhibited CPK6 kinase activity . The increased ATP-consumption signal in the presence of ABI1 and PP2CA can be explained by higher ATP consumption triggered by kinase auto-phosphorylation of residues removed by the PP2C protein phosphatases . Data shown represent the mean of three experiments ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 012 Exposure of Arabidopsis seedlings to ABA led to OST1 protein kinase activation , confirming functional ABA responses ( Figure 3A , B , lanes 1–2 and 9–10; ‘OST1’ inset ) . However , CPK-derived band intensities did not change in the presence of ABA , indicating that CPK activities may not be directly ABA-regulated , in contrast to OST1 ( Figure 3B ) . These findings were also obtained at an intermediate free Ca2+ concentration of 0 . 4 μM ( Figure 3—figure supplement 2A , B ) . Moreover , in-gel CPK protein kinase activities were not altered with or without ABA in seedling extracts of abi1-2/abi2-2/hab1-1/pp2ca-1 quadruple mutant plants ( Figure 3A , B , lanes 3–4 and 11–12; Figure 3—figure supplement 1A , B ) . Interestingly , the pp2c quadruple mutants did not enable constitutive OST1 activation in vivo , differing from ( Fujii et al . , 2009 ) , but consistent with ( Vlad et al . , 2009 ) ( Figure 3A , B , lanes 3–4 and 11–12 and Figure 3—figure supplement 2A , B; see ‘OST1’ inset ) . Furthermore , OST1-derived band intensities were not changed in the cpk5/6/11/23 and cpk1/2/5/6 mutant plants showing that these cpk quadruple mutants retain ABA-activation of OST1 ( Figure 3A , B , lanes 5–8 and 13–16; see ‘OST1’ inset ) . Initially , we tested whether the signals found in in-gel protein kinase assays are derived from kinase auto-phosphorylation or due to trans-phosphorylation activities of the protein kinases . To distinguish between auto- and trans-phosphorylation activities of recombinant CPK6 and OST1 we compared in-gel band intensities of gels with or without the substrate Histone-III ( Figure 3—figure supplement 3A , B ) . The strong reduction of band intensities for recombinant CPK6 and OST1 when no Histone-III is present ( Figure 3—figure supplement 3A , B ) indicates that the signals observed in in-gel protein kinase assays are largely derived from CPK6 and OST1 kinase trans-phosphorylation activities of Histone-III consistent with previous reports for CPKs involved in pathogen signaling ( Boudsocq et al . , 2010 ) . To determine whether PP2Cs can directly down-regulate CPKs we next investigated whether the SLAC1-activating CPK6 ( Mori et al . , 2006; Brandt et al . , 2012 ) , is negatively regulated by the PP2Cs ABI1 and PP2CA . In-gel protein kinase assays using recombinant proteins were pursued in which kinases and phosphatases are separated by size prior to substrate phosphorylation . CPK6 , and as positive control OST1 , were pre-incubated either alone or with ABI1 or PP2CA with and without ATP before being subjected to in-gel protein kinase assays . Pre-incubation with either ABI1 or PP2CA did not inhibit CPK6 trans-phosphorylation activity ( Figure 3C , lanes 2–3 and 5–6 ) . In contrast , control OST1-derived substrate phosphorylation band intensities strongly decreased when ABI1 or PP2CA proteins were present during the pre-incubation period ( Figure 3D , lanes 2–3 and 5–6 ) . These results indicate that OST1 , but not CPK6 activity , is directly down-regulated by ABI1 and PP2CA . CPKs have been previously reported to interact with ABI1 ( Geiger et al . , 2010 ) . An electro-mobility shift can be observed for OST1 as well as for CPK6 ( Figure 3C , D ) . These shifts could be due to dephosphorylation of CPK6 ( Figure 3—figure supplement 4 ) and OST1 ( Umezawa et al . , 2009; Vlad et al . , 2009 ) by PP2Cs . However , dephosphorylation by PP2Cs did not inhibit CPK6 activity ( Figure 3C ) . An additional independent biochemical assay measuring ATP consumption also did not show down-regulation of CPK6 activity in the presence of ABI1 and PP2CA ( Figure 3—figure supplement 5 ) , further underlining no direct down-regulation of CPK6 activity by these three PP2Cs , in contrast to OST1 controls . Our results suggest that PP2Cs neither down-regulate CPK6 activity directly in vitro ( Figure 3C , D and Figure 3—figure supplement 5 ) nor that CPK activities are strongly ABA-regulated independent of [Ca2+] changes in native plant protein extracts ( Figure 3A , B ) . We next investigated the kinetics and specificity of PP2C down-regulation of SLAC1 activation by CPKs through dephosphorylation of the SLAC1 channel , a mechanism reported for CPK-dependent transcription factor regulation ( Lynch et al . , 2012 ) and consistent with previous findings ( Brandt et al . , 2012 ) . First , we determined whether SLAC1 interacts with the PP2C ABI1 in planta using bimolecular fluorescence complementation ( BiFC ) . We observed clear BiFC signals for full length SLAC1 co-expressed with CPK6 and ABI1 ( Figure 4A , B ) while signal intensities of SLAC1 co-expressed with a control protein phosphatase 2A catalytic subunit 5 ( PP2AC5 ) were very low ( Figure 4B ) . Protein–protein interaction of SLAC1 with PP2CA in BiFC experiments was reported earlier ( Lee et al . , 2009 ) . As shown in Figure 4C , D , the ABI1-mediated dephosphorylation of the N-terminus of SLAC1 ( SLAC1-NT ) previously phosphorylated by CPK6 ( Brandt et al . , 2012 ) occurs very rapidly . Already 1 min after the addition of ABI1 a strong decrease of the phosphorylation signal was observed ( Figure 4D , lane 4 ) . This de-phosphorylation was also found when the PP2C phosphatase PP2CA was added instead of ABI1 ( Figure 4C , E , lane 4 ) . To test whether this is a general phenomenon , we phosphorylated the SLAC1-NT with the SLAC1-activating and -phosphorylating kinases CPK21 , CPK23 , and OST1 ( Geiger et al . , 2009 , 2010; Lee et al . , 2009 ) and analyzed whether ABI1 and PP2CA are able to remove phospho-groups added by these kinases ( Figure 4F–H and Figure 4—figure supplement 1 ) . After inhibiting the kinase with staurosporine , band intensities decreased only after addition of the PP2C protein phosphatases for all combinations , showing that this rapid SLAC1 de-phosphorylation is mediated by PP2Cs ( Figure 4F–H and Figure 4—figure supplement 1 , lanes 5–6 ) . 10 . 7554/eLife . 03599 . 013Figure 4 . PP2Cs interact with and directly and rapidly dephosphorylate the N-terminus of SLAC1 ( SLAC1-NT ) when previously phosphorylated by several SLAC1-activating CPK and OST1 protein kinases . ( A ) Bimolecular fluorescence complementation ( BiFC ) experiments in Nicotiana benthamiana leaves show YFP-derived fluorescence signals of YC-SLAC1 co-expressed with CPK6-YN and YN-ABI1 . ( B ) Quantification of BiFC-mediated YFP-fluorescence shows that SLAC1 interacts with CPK6 and ABI1 but not with the control catalytic protein phosphatase 2A subunit C5 ( PP2AC5 ) . YFP signals of positive control YN-PP2AC5 with protein phosphatase 2A regulatory subunit A3 fused to YC ( YC-PP2AA3 ) confirm expression of PP2AC5 . Data shown in ( B ) represent the average fluorescence intensity of randomly picked leaf areas ( n = 40; ±SEM ) and these data are also included in Figure 6—figure supplement 5 . ( C–E ) CPK6-phosphorylated SLAC1-NT is rapidly de-phosphorylated by ABI1 and PP2CA . SLAC1-NT phosphorylation by CPK6 ( D and E , lane 1 ) is strongly inhibited if the PP2C protein phosphatase was added before starting the reaction ( D and E , lane 2 ) , but remains stable after addition of elution buffer ( Elu . ) and kinase inhibitor staurosporine ( Stau . ) with subsequent 10 min incubation ( D and E , lane 3 ) . If ( D ) ABI1 or ( E ) PP2CA together with staurosporine are added after the initial 10 min CPK6 mediated phosphorylation period , the SLAC1-NT phosphorylation signal rapidly decreases within 1 min ( D and E , lanes 4–7 ) . Staurosporine pre-exposure control inhibits SLAC1-NT phosphorylation by CPK6 ( D and E , lane 8 ) . ( F–H ) PP2Cs de-phosphorylate the SLAC1-NT which was phosphorylated by major SLAC1-activating kinases CPK23 and OST1 . The SLAC1-NT is phosphorylated by CPK23 ( G , lane 1 ) and OST1 ( H , lane 1 ) which is inhibited when the PP2Cs ABI1 and PP2CA are added before starting the reactions ( G and H , lanes 2–3 ) . When adding staurosporine and elution buffer after the initial phosphorylation period and incubating for 10 min the signal does not change ( G and H , lane 4 ) . Addition of ABI1 or PP2CA after supplementing the reaction with staurosporine leads to rapid ( 10 min ) dephosphorylation of the SLAC1-NT previously phosphorylated by the OST1 and CPK23 protein kinases ( G and H , lanes 5–6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 01310 . 7554/eLife . 03599 . 014Figure 4—figure supplement 1 . When previously phosphorylated by CPK21 , the SLAC1-NT is de-phosphorylated by the PP2Cs ABI1 and PP2CA . Recombinant SLAC1-NT phosphorylation by CPK21 ( lane 1 ) is inhibited if the protein phosphatases ABI1 and PP2CA are added before starting the reaction ( lanes 2–3 ) . The phosphorylated SLAC1-NT derived signal is rapidly and strongly decreased if the PP2Cs ABI1 and PP2CA ( lanes 4–7 ) are added after the addition of staurosporine . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 014 The Ca2+-independent and Ca2+-dependent branches of ABA signal transduction are presently considered to be independent ( e . g . , Li et al . , 2006; Kim and et al . , 2010; Roelfsema et al . , 2012 ) , but this model has not been genetically investigated in Arabidopsis . In the cpk5/6/11/23 quadruple mutant , ABA-activation of S-type anion currents and stomatal closure were impaired ( Figure 1A–E ) , providing evidence for a possible interdependence of these signaling branches . The ost1 single gene disruption mutant in the Col ecotype shows intermediate S-type anion current activation by ABA ( Geiger et al . , 2009 ) . Three Ca2+-independent SnRK kinases , SnRK2 . 2 , SnRK2 . 3 , and OST1 can activate SLAC1 in oocytes ( Geiger et al . , 2009 ) and redundantly function in controlling leaf water loss ( Fujii and Zhu , 2009 ) . Interestingly , snrk2 . 2/snrk2 . 3/ost1 triple mutants were strongly impaired in ABA activation and notably also external Ca2+ shock-induced activation of S-type anion channels at 2 μM [Ca2+]cyt ( Figure 5A–D ) . Imposing repetitive cytosolic Ca2+ transients by alternating guard cell incubation buffers induces a fast Ca2+-reactive stomatal closure response ( Allen et al . , 2001 ) . We further analyzed imposed Ca2+ oscillation-induced stomatal closure in snrk2 . 2/snrk2 . 3/ost1 triple mutants . Ca2+ reactive stomatal closure of the snrk triple mutant was impaired compared to wildtype plants ( Figure 5E , p < 0 . 02 for wildtype vs snrk2 . 2/snrk2 . 3/ost1 at 120 min ) . These data show that disruption of Ca2+-independent signaling in snrk2 triple mutants also impairs Ca2+-dependent stomatal responses . Thus these findings investigating S-type anion channel regulation and stomatal movements both provide genetic evidence for an unexpected interdependence of the Ca2+-dependent and -independent branches of the guard cell signaling network . 10 . 7554/eLife . 03599 . 015Figure 5 . Both , ABA- and high external Ca2+-activation of S-type anion currents at elevated [Ca2+]cyt and imposed Ca2+-oscillation-triggered stomatal closure are impaired in snrk2 . 2/2 . 3/ost1 triple mutant guard cells while the ABA-activation of ICa currents is intact . ( A–D ) Whole-cell patch-clamp experiments reveal that [Ca2+]cyt-activation of S-type anion currents is disrupted in snrk2 . 2/2 . 3/ost1 triple mutant guard cells even if pre-incubated with high external Ca2+ shock ( A and B ) or ABA ( C and D ) . Note that pre-incubation with high external Ca2+ shock by passes early ABA signaling ( Allen et al . , 1999a; Allen et al . , 2002 ) . Typical current responses ( A and C ) , average steady-state current–voltage relationships ±SEM , and the number of measured cells are presented ( B and D ) . In ( B ) data for snrk2 . 2/2 . 3/ost1 triple mutants with and without ABA overlap with WT controls . ( E ) Imposed Ca2+ oscillation-induced stomatal closure is impaired in Ca2+-independent protein kinase snrk2 . 2/2 . 3/ost1 triple mutant leaves , providing further evidence for an interdependence of these responses . Four 5-min extracellular Ca2+-pulses were applied in 10-min intervals from time = 0 to 35 min . Average individually tracked stomatal apertures were normalized to the stomatal apertures at time zero . The averages of the normalized apertures ±SEM and the number of independent genotype-blind experiments ( n = 4 ) are shown ( >40 total stomata per group ) . Average stomatal apertures at time zero were 4 . 61 ± 0 . 44 μm in WT ( n = 4 ) and 5 . 51 ± 0 . 87 μm in the snrk2 . 2/2 . 3/ost1 triple mutant ( n = 4 ) . ( F ) Patch clamp experiments reveal that ABA activation of ICa currents is not impaired in snrk2 . 2/2 . 3/ost1 triple mutant guard cells . Average steady-state current–voltage relationships ±SEM , and the number of measured cells are presented in ( F ) . Representative whole cell current traces for ( F ) are presented in Figure 5—figure supplement 1 . Several error bars are not visible , as these were smaller than the illustrated symbols . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 01510 . 7554/eLife . 03599 . 016Figure 5—figure supplement 1 . snrk2 . 2/2 . 3/ost1 triple mutant guard cells show intact ABA activation of Ca2+-permeable ICa currents . ( A and B ) Whole-cell patch-clamp experiments of WT ( A ) and snrk2 . 2/2 . 3/ost1 triple mutant ( B ) guard cells in the absence of ABA ( top traces ) . ABA activation of Ca2+-permeable ICa currents was similar in wildtype guard cells ( A , lower panel ) and snrk2 . 2/2 . 3/ost1 triple mutant guard cells ( B , lower panel ) . Typical current responses of WT ( A ) and the snrk2 . 2/2 . 3/ost1 triple mutant ( B ) are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 016 The Ca2+-independent OST1 protein kinase affects Ca2+ signaling in Landsberg erecta guard cells via regulation of plasma membrane-localized Ca2+-permeable channels ( ICa ) ( Acharya et al . , 2013 ) . To test whether the functional linkage of the Ca2+-dependent and Ca2+-independent branch is due to the regulation of the ICa channels by SnRK2 protein kinases in the Columbia ecotype , we performed patch clamp analyses measuring plasma membrane ICa channel currents in snrk2 . 2/snrk2 . 3/ost1 triple mutant guard cells . However , ABA activation of ICa channels remained intact in snrk2 . 2/snrk2 . 3/ost1 triple mutant guard cells ( Figure 5F and Figure 5—figure supplement 1 ) . In positive control experiments , ABA receptor pyr1/pyl1/2/4 quadruple mutant guard cells showed clear impairment of ABA activation of ICa channels ( Data not shown , n = 5; control vs ABA , p = 0 . 96; Student's t-test ) , consistent with previous findings ( Wang et al . , 2013 ) . In addition to possible direct cross-regulation of CPKs and SnRK2s , another non-mutually exclusive potential mechanism for the requirement of both SnRK and CPK kinases for ABA activation of S-type anion channels could be that SLAC1 serves as coincidence detector through differential phosphorylation by protein kinases of the Ca2+-dependent and -independent branches . The amino acid residue serine 120 of SLAC1 has been shown to be required for OST1 , but not for CPK23 activation of SLAC1 in Xenopus oocytes ( Geiger et al . , 2009 , 2010 ) . A different site , serine 59 , has been shown to be required for SLAC1 activation by CPK6 ( Brandt et al . , 2012 ) . Thus we investigated whether several CPKs can activate the SLAC1 S120A mutant in oocytes and whether the SLAC1 S59A mutant is activated by OST1 and other CPKs in oocytes . CPK5 , CPK6 , and CPK23 activation of SLAC1 S120A was similar to WT SLAC1 activation ( Figure 6A–C and Figure 6—figure supplement 1A , B ) . In contrast , SLAC1 S59A activation by these CPKs was strongly impaired ( Figure 6A–C and Figure 6—figure supplement 1A , B ) . Interestingly however , OST1 was able to activate SLAC1 S59A ( Figure 6D–F ) , which was confirmed in multiple independent experimental sets under the imposed conditions . These results suggest that S59 is required for strong activation by protein kinases of the Ca2+-dependent CPK branch , while S120 represents a crucial amino acid for strong activation by the Ca2+-independent branch of the ABA signaling core . To avoid spurious phosphorylation by high protein kinase concentrations in oocytes , effects of co-expression of CPK6 and OST1 at low levels that do not fully activate SLAC1 were investigated . These experiments show a clear enhanced SLAC1 activation in oocytes when both kinases are co-expressed ( Figure 6—figure supplement 2A–D ) . This enhancement of SLAC1 activation by OST1 became less clear when an inactive OST1 protein kinase ( OST1 D140A ) was analyzed ( Figure 6—figure supplement 2E ) . 10 . 7554/eLife . 03599 . 017Figure 6 . Ca2+-dependent protein kinase and OST1 protein kinase activation of SLAC1 in oocytes requires serine 59 or serine 120 , respectively while in planta ABA-dependent S-type anion current activation and stomatal closing are only impaired in SLAC1 S59A/S120A double amino acid mutants . ( A–C ) SLAC1 activation by CPK6 in Xenopus oocytes was abolished when serine 59 is mutated to alanine ( S59A ) ( A and C ) ( Brandt et al . , 2012 ) but was comparable to wild type SLAC1 activation for the SLAC1 S120A mutated version ( B and C ) . ( D–F ) OST1 activation of SLAC1 was abolished in the SLAC1 S120A mutant ( E and F ) ( Geiger et al . , 2009 ) , while OST1 robustly activated SLAC1 S59A ( D and F ) . ( G ) In whole-cell patch-clamp experiments , slac1-1 guard cells show impaired ABA-activation of S-type anion currents . Expression of SLAC1 WT , S59A , and S120A in slac1-1 plants restores ABA activation of S-type anion currents in guard cells , but expression of SLAC1 S59A/S120A does not . ( H ) The ABA-insensitive phenotype of slac1-1 stomata was recovered by expression of SLAC1 WT , S59A , and S120A , but not by expression of S59A/S120A . Note that SLAC1 WT , S59A , S120A , and S59A/S120A are expressed as C-terminal mVenus fusion proteins under native SLAC1 promoter ( see ‘Materials and methods’ ) . Representative current traces are depicted in ( A , B , D and E ) and average current voltage relationships are shown ( C and F; ±SEM ) . Average steady-state current responses ±SEM at −145 mV are plotted in ( G ) and average stomatal apertures ±SEM in ( H ) . * indicates p < 0 . 05; unpaired Student's t-test . Exact p-values and number of individual experiments for ( G and H ) can be found in Figure 6—figure supplement 4 . Note that WT ( Col0 ) and slac1-1 control measurements shown in ( G and H ) are the same control data as those shown in Figure 6—figure supplement 3A , B as all lines were investigated under the same conditions . Several error bars are not visible , as these were smaller than the illustrated symbols . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 01710 . 7554/eLife . 03599 . 018Figure 6—source data 1 . Statistical data and number of repeats ( n ) for the ( Table 1 ) patch clamp measurements shown in Figure 6G and Figure 6—figure supplement 3A and ( Table 2 ) for measurements of stomatal apertures presented in Figure 6H and Figure 6—figure supplement 3B ( n = 3 experiments and >45 total stomata per group ) . The Student's t-test was used to calculate all p-values . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 01810 . 7554/eLife . 03599 . 019Figure 6—figure supplement 1 . SLAC1 serine 59 but not serine 120 is required for CPK5 or CPK23 activation in Xenopus oocytes . ( A and B ) SLAC1 activation by CPK5 ( A ) and CPK23 ( B ) is comparable to WT when serine 120 is substituted by alanine ( S120A ) while the CPK5 and CPK23 activation of SLAC1 S59A was strongly impaired . Average steady-state current–voltage relationships ( ±SEM ) , and numbers of individual measured cells are depicted ( A and B ) . Several error bars are not visible , as these were smaller than the illustrated symbols . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 01910 . 7554/eLife . 03599 . 020Figure 6—figure supplement 2 . SLAC1 exhibits enhanced activity by co-expression of CPK6 and OST1 in Xenopus oocytes . ( A–D ) If SLAC1 ( 5 ng cRNA ) is expressed alone or with non-BIFC OST1 ( 7 . 5 ng ) , no anion currents can be detected ( C and D ) . If CPK6 ( 0 . 5 ng ) is co-expressed , SLAC1-mediated currents can be seen ( A , C , and D ) which are enhanced when OST1 ( 7 . 5 ng ) is added ( B–D ) . Due to overlapping data of ‘SLAC1’ and ‘SLAC1 + OST1’ alternating data points are shown in ( C ) . ( E ) This enhancement is almost completely impaired when the kinase inactive mutant OST1 D140A is co-injected with SLAC1 and CPK6 . In ( A and B ) typical current responses are shown while in ( C and E ) average current–voltage relationships ±SEM and the number of the measured cells are presented . ( D ) Shows average currents at −140 mV ±SEM ( *** indicates p = 0 . 005 ) . Several error bars are not visible , as these were smaller than the illustrated symbols . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 02010 . 7554/eLife . 03599 . 021Figure 6—figure supplement 3 . ABA-induced S-type anion currents and stomatal closure responses are impaired when both SLAC1 S59 and S120 are substituted with alanine in independent double amino acid mutant line . ( A ) In whole-cell patch-clamp experiments , slac1-1 guard cells show impaired ABA-activation of S-type anion currents . Expression of SLAC1 WT , S59A , and S120A in slac1-1 plants restores ABA activation of S-type anion currents , but expression of SLAC1 S59A/S120A does not . ( B ) ABA-insensitive stomatal closing phenotype of slac1-1 was recovered by expression of SLAC1 WT , S59A , and S120A , but not by expression of S59A/S120A . Note that SLAC1 WT , S59A , S120A , and S59A/S120A are expressed as C-terminal mVenus fusion proteins under the native 1 . 63 kbp of the SLAC1 5′ UTR promoter region ( see ‘Materials and methods’ ) . The results shown here were recorded from independent Arabidopsis slac1-1 transformation lines that differ from the transformation lines shown in Figure 6G , H . Note that Col0 and slac1-1 measurements are the same control data as those shown in Figure 6G , H as all lines were investigated under the same conditions . Average steady-state current responses ±SEM at −145 mV are plotted in ( A ) . In ( B ) average stomatal apertures ±SEM . * indicates p < 0 . 05; t-test . Exact p-values and number of individual experiments for ( A and B ) can be found in Figure 6—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 02110 . 7554/eLife . 03599 . 022Figure 6—figure supplement 4 . Analysis of expression and subcellular localization of SLAC1-WT , SLAC1S59A , S120A , and S59A/S120A in slac1-1 complementation lines . Confocal laser microscopy of guard cells of slac1-1 mutant lines expressing SLAC1-WT-mVenus , SLAC1S59A-mVenus , SLAC1-S120A-mVenus or SLAC1-S59A/S120A-mVenus shows membrane-localized expression of all SLAC1 versions . In addition intracellular YFP fluorescence was also observed in some guard cells , which may be linked to trafficking of SLAC1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 02210 . 7554/eLife . 03599 . 023Figure 6—figure supplement 5 . BiFC fluorescence intensities are altered for CPK6 and ABI1 co-expression with SLAC1-WT , SLAC1S59A , S120A , and S59A/S120A . Quantitative BiFC experiments showed that CPK6-YN + YC-SLAC1-WT derived fluorescence signals are significantly decreased when CPK6-YN was expressed with YC-SLAC1-S59A , YC-SLAC1-S120A , or YC-SLAC1-S59A/S120A ( p < 0 . 0001; unpaired t-test ) . YC-SLAC1-S59A expression with YN-ABI1 did not result in a significant decrease of the average fluorescence intensity in comparison with YC-SLAC1-WT and YN-ABI1 co-expression ( p = 0 . 2647; unpaired t-test ) . Signal intensities significantly decreased when YN-ABI1 was co-expressed with YC-SLAC1-S120A or YC-SLAC1-S59A/S120A vs co-expression with YC-SLAC1-WT ( p < 0 . 005; unpaired t-test ) . The data for CPK6-YN + YC-SLAC1-WT , YN-ABI1 + YC-SLAC1-WT , YN-PP2AC5 + YC-SLAC1-WT and YN-PP2AC5 + YC-PP2AA3 are also shown in Figure 4B . Average fluorescence intensities were determined of leaf areas which were chosen randomly in white light ( ±SEM; n = 40 images per condition ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03599 . 023 To more directly investigate S-type anion channel regulation in planta , we established slac1-1 plant lines which express SLAC1 WT , S59A , S120A , and S59A/S120A fused to mVenus under the native SLAC1 promoter and carried out patch clamp analyses . Expression of wildtype SLAC1-mVenus in slac1-1 guard cells resulted in recovery of S-type anion channels ( Figure 6G and Figure 6—figure supplement 3A ) . Unexpectedly , expression of the single site SLAC1 mutants , SLAC1 S59A or SLAC1 S120A in slac1-1 guard cells restored ABA regulation of S-type anion currents ( Figure 6G and Figure 6—figure supplement 3A ) . However , expression of the double phosphorylation site SLAC1 mutant , SLAC1 S59A/S120A did not restore ABA activation of S-type anion channels ( Figure 6G and Figure 6—figure supplement 3A ) . Furthermore , ABA-induced stomatal closing responses in these complementation lines confirmed the need to mutate both the S59 and S120 sites to alanine to significantly impair ABA-induced stomatal closing in planta ( Figure 6H and Figure 6—figure supplement 3B ) . The above described patch clamp and stomatal movement experiments were conducted with two independent complementation lines ( Figure 6G , H and Figure 6—figure supplement 3A , B ) . To ensure that the impaired ABA-activation of S-type anion currents and stomatal closure in the SLAC1 S59A/S120A mutant was not due to non-expressed protein we investigated the mVenus-derived fluorescence in all complementation lines . All SLAC1 complementation lines expressed SLAC1-mVenus driven by the native SLAC1 promoter to a similar degree ( Figure 6—figure supplement 4 ) . We examined putative roles of the two phosphorylation sites in SLAC1 for interaction of SLAC1 with CPK6 and ABI1 by BiFC analysis . Reconstituted YFP fluorescence intensity of CPK6-YN co-expressed with YC-SLAC1-S59A , YC-SLAC1-S120A , or YC-SLAC1-S59A/S120A was significantly lower than that of CPK6-YN co-expressed with YC-SLAC1-WT ( Figure 6—figure supplement 5 ) . YN-ABI1 co-expression with the YC-SLAC1-S59A mutant did not significantly change the YFP fluorescence intensity while YN-ABI1 co-expression with YC-SLAC1-S120A or YC-SLAC1-S59A/S120A resulted in lower YFP fluorescence intensity when compared to YN-ABI1 co-expression with YC-SLAC1-WT ( p < 0 . 005; unpaired t-test; Figure 6—figure supplement 5 ) . These results point to the need for future research to determine whether these phosphorylation sites in SLAC1 might contribute to promotion of CPK6 kinase and ABI1 phosphatase interaction strength with the SLAC1 channel ( Figure 6—figure supplement 5 ) . In summary , the present study reveals a first genetic mechanism that mediates Ca2+ sensitivity priming . Ca2+ sensitivity is demonstrated here to be constitutively primed in pp2c quadruple mutant guard cells , showing that PP2Cs ensure Ca2+ signaling specificity . Interestingly , PP2Cs do not directly down-regulate CPK activity , in contrast to direct PP2C down-regulation of the SnRK2 protein kinases . Rather PP2Cs very rapidly down-regulate signaling targets downstream of CPKs , which could enable the same CPKs to function in more than one pathway . We have further identified a cpk quadruple mutant here that for the first time strongly abrogates ABA activation of S-type anion channels . This abrogation occurs despite an intact Ca2+-independent SnRK2 signaling branch . Furthermore , disruption of the Ca2+-independent signaling branch in snrk2 protein kinase triple mutant plants abrogates Ca2+ signaling . Thus , unexpectedly genetic analyses reveal a dependence of the Ca2+-sensitive ABA signaling branch on the Ca2+-insensitive branch in planta . The control of ABA-triggered stomatal closure by parallel interdependent Ca2+-dependent and–independent mechanisms could contribute to the robustness of the stomatal ABA signaling network ( Hetherington , 2001 ) . Unexpectedly , in planta studies show that the S59 and S120 phosphorylation sites in SLAC1 are together required for intact ABA-induced stomatal closing in vivo . The Ca2+ sensitivity priming mechanism described here could represent a more general principle present in plants contributing to Ca2+ specificity within cellular signal transduction pathways , while also maintaining the availability of Ca2+ sensors for distinct Ca2+-dependent signaling outputs . All A . thaliana plants used in this study are in the Col0 ecotype . cpk5/6/11/23 quadruple T-DNA insertion mutant plants were established by crossing cpk5/6/11 ( sail_657C06/salk_025460/salk_054495 ) kindly provided by Dr Jen Sheen ( Harvard Medical School ) ( Boudsocq et al . , 2010 ) with cpk23-1 ( salk_007958 ) obtained from ABRC ( Ma and Wu , 2007; Geiger et al . , 2010 ) . Dr Ping He ( Texas A&M University ) shared cpk1/2/5/6 ( salk_096452/salk_059237/sail_657C06/salk_025460 ) mutant seeds ( Gao et al . , 2013 ) . The PP2C quadruple knock-out plants ( abi1-2/abi2-2/hab1-1/pp2ca-1; salk_072009/salk_015166/salk_002104/salk_028132 ) and snrk2 . 2/2 . 3/ost1 ( GABI-Kat_807G04/salk_107315/salk_008068 ) were kindly provided by Dr Pedro L Rodriguez ( University of Valencia ) ( Antoni et al . , 2013 ) . A second independent snrk2 . 2/2 . 3/ost1 ( GABI-Kat_807G04/salk_107315/salk_008068 ) line was established by crossing snrk2 . 2/2 . 3 supplied by Dr Jian-Kang Zhu ( Shanghai Center for Plant Stress Biology ) with ost1-3 . To establish SLAC1 complementation lines a 4 . 4 kb fragment including 1 . 63 kb of the 5′-UTR , the genomic SLAC1 gene region and 0 . 9 kb of the 3′-UTR ( Negi et al . , 2008 ) was amplified using the PfuX7 polymerase ( Norholm , 2010 ) . The fragment was cloned into a modified pGreenII ( Hellens et al . , 2000 ) vector lacking a promoter and being compatible with USER-cloning . Employing USER cloning ( Nour-Eldin et al . , 2006; Bitinaite et al . , 2007; Geu-Flores et al . , 2007 ) the point mutations were introduced and SLAC1 was fused with mVenus ( C-terminally ) ( Nagai et al . , 2004 ) . These pGreenII constructs were transformed into Agrobacterium tumefaciens GV3101 ( pMP90 ) RG ( Koncz and Schell , 1986 ) . slac1-1 mutant plants were then transformed by the floral dipping method ( Clough and Bent , 1998 ) and propagated until the T-DNA insertion was confirmed to be homozygous . Arabidopsis plants were grown on soil in the growth chamber at 21°C under a 16-hr-light/8-hr-dark photoperiod with a photon flux density of 80 μmol/ ( m2 × s ) . The plants were watered from bottom trays with deionized water once or twice per week and sprayed with deionized water every day . The growth chamber humidity was 50–70% . Arabidopsis guard cell protoplasts were isolated enzymatically as previously described ( Pei et al . , 1997 ) . One or two rosette leaves of 4- to 5-week-old plants were blended in a blender with deionized water at room temperature ( RT ) for approximately 30 s . For isolation of guard cell protoplasts from snrk2 . 2/snrk2 . 3/ost1 triple mutants , four or five rosette leaves were used . Epidermal tissues were collected using a 100-μm nylon mesh and rinsed well with deionized water . The epidermal tissues were then incubated in 10 ml of enzyme solution containing 1% ( wt/vol ) Cellulase R-10 ( Yakult , Japan ) , 0 . 5% ( wt/vol ) Macerozyme R-10 ( Yakult , Japan ) , 0 . 1 mM KCl , 0 . 1 mM CaCl2 , 500 mM D-mannitol , 0 . 5% ( wt/vol ) BSA , 0 . 1% ( wt/vol ) kanamycin sulfate , and 10 mM ascorbic acid for 16 hr at 25°C on a circular shaker at 40 rpm . Guard cell protoplasts were then collected by filtering through a 20-μm nylon mesh . Subsequently , the protoplasts were washed twice with washing solution containing 0 . 1 mM KCl , 0 . 1 mM CaCl2 , and 500 mM D-sorbitol ( pH 5 . 6 with KOH ) by centrifugation for 10 min at 200×g . The guard cell protoplast suspension was kept on ice before use . To investigate ABA activation of S-type anion channels , the guard cell protoplast suspension was pre-incubated with 10 μM ( Figure 1A , C , Figure 1—figure supplement 1C , D , Figure 6G , and Figure 6—figure supplement 3A ) or 50 μM ( Figure 2A–D and Figure 2—figure supplement 1A , B as well as Figure 5C , D ) ± ABA ( Sigma , St . Louis , MO ) for 30 min . S-type anion channel currents in guard cell protoplasts were recorded by the whole-cell patch-clamp technique as previously described ( Pei et al . , 1997; Vahisalu et al . , 2008; Siegel et al . , 2009 ) . The pipette solution contained 150 mM CsCl , 2 mM MgCl2 , 5 mM Mg-ATP , 6 . 7 mM EGTA , and 10 mM Hepes-Tris ( pH 7 . 1 ) . To obtain a free [Ca2+]cyt of 2 μM and 110 nM , 5 . 86 mM and 1 . 79 mM of CaCl2 were added to the pipette solution , respectively . Osmolality of the pipette solution was adjusted to 500 mmol/l using D-sorbitol . The bath solution contained 30 mM CsCl , 2 mM MgCl2 , 1 mM CaCl2 , and 10 mM MES-Tris ( pH 5 . 6 ) . Osmolality of the bath solution was adjusted to 485 mmol/l using D-sorbitol . To investigate external Ca2+ activation of S-type anion channels , guard cell protoplasts were pre-incubated with the bath solution containing 40 mM CaCl2 , instead of 1 mM CaCl2 for 30 min . Whole-cell currents were recorded 3–5 min after achieving the whole-cell configuration . The seal resistance was no less than 10 GΩ . The voltage was decreased from +35 mV to −145 mV with 30 mV decrements and the holding potential was +30 mV . To investigate ABA activation of Ca2+-permeable ICa channels , the pipette solution contained 10 mM BaCl2 , 4 mM EGTA , and 10 mM HEPES-Tris ( pH 7 . 1 ) . 5 mM NADPH was freshly added to the pipette solution before experiments . The bath solution contained 100 mM BaCl2 , and 10 mM MES-Tris ( pH 5 . 6 ) . 0 . 1 mM DTT was freshly added to the bath solution before experiments . Osmolarity was adjusted to 500 mmol/l for the pipette solution and 485 mmol/l for the bath solution with D-sorbitol . A ramp voltage protocol from +20 to −180 mV ( holding potential , 0 mV; ramp speed , 200 mV/s ) was used for ICa recordings ( Pei et al . , 2000 ) . The seal resistance was no less than 10 GΩ . Data were filtered at 3 kHz . Initial control whole-cell currents were recorded 10 times with a 1 min interval between each recording 1–3 min after achieving whole-cell configurations . The average current obtained from the 10 current traces per cell at 0 , −30 , −60 , −90 , −120 , −150 , and −180 mV was determined for IV curves . After control current recordings , ABA was added to the bath solution by perfusion , and guard cell protoplasts were incubated with ABA in the bath solution for 3 min . Then , ABA-activated ICa currents were recorded 10 times for another 10 min and the average current obtained from the 10 traces was determined for IV curves . 2-week-old plate-grown plants were transferred to soil and grown in >70% relative humidity under 16 hr light/8 hr dark . Rosette leaves from 4- to 5-week-old plants were detached and incubated in stomatal opening buffer ( 5 mM KCl , 50 μM CaCl2 , 10 mM MES and pH 5 . 6 with Tris base ) for 2 . 5 hr , in 150–180 μmol/ ( m2 × s ) light . Next , leaves were treated with either 5 μM ABA or 0 . 05% ethanol for an additional 1 hr incubation . After the incubation period , leaves were blended and fragments were collected with a 100 μm nylon mesh ( Figure 1E , Figure 6H and Figure 6—figure supplement 3B ) except for Figure 1F . In Figure 1F , epidermal peels were prepared using a perforated-tape epidermal detachment method ( Ibata et al . , 2013 ) . Images of stomata from the abaxial side of the leaves were collected by microscopy . Stomatal aperture analyses were conducted as single-blind experiments in which the experimenter did not know the plant genotypes during measurements ( Figure 1E , F ) or as double-blind experiments in which the experimenter did not know both the ABA concentration and the plant genotypes ( Figure 6H and Figure 6—figure supplement 3B ) . Stomatal aperture analyses for imposed Ca2+ pulses were performed as previously described ( Allen et al . , 2001; Mori et al . , 2006; Siegel et al . , 2009 ) . Stomatal apertures of individually mapped stomata were measured at the indicated time points after the start of imposed Ca2+ pulses . The lower epidermis of rosette leaves from 4- to 5-week-old plants was attached onto a coverslip using medical adhesive ( Hollister ) . Then mesophyll layers of the leaf were carefully removed using a razor blade until only the epidermal layer remained . The lower epidermis was incubated in depolarizing buffer ( 50 mM KCl and 10 mM MES-Tris [pH 5 . 6] ) for 3 hr under white light ( 150–180 μmol/ ( m2 × s ) ) to open stomata . Depolarizing buffer was changed to hyperpolarizing buffer ( 1 mM KCl , 1 mM CaCl2 , and 10 mM MES-Tris at pH 5 . 6 ) . Four 5-min extracellular Ca2+ pulses were applied in 5-min intervals in the first 35 min . Stomatal aperture analyses were conducted as blind experiments in which the experimenter did not know the plant genotypes during measurements ( Figure 5E ) . Over-expression and purification of recombinant proteins were performed as described in Brandt et al . ( 2012 ) with minor adjustments: For the isolation of the PP2C proteins ABI1 , ABI2 , and PP2C additionally 5 mM MgCl2 and 5% Glycerol were added to the buffer in which the bacterial pellet were re-suspended ( buffer W in IBA manual ) . Also , all proteins except SLAC1-NT were eluted in elution buffer supplemented with 20% Glycerol instead of 10% and stored at −80°C instead of −20°C . To assess protein concentrations , several volumes of the eluates were loaded on a gel together with several defined bovine serum albumin ( BSA ) protein amounts . After separating the proteins by SDS-PAGE ( Laemmli , 1970 ) , the proteins were stained with coomassie brilliant blue R-250 , dried between two sheets of cellophane , and then scanned . BSA and recombinant protein band intensities were measured using Fiji ( Schindelin et al . , 2012 ) . After subtracting the background signal , BSA band signal intensities were used to plot a standard curve . Concentrations of isolated recombinant proteins were then calculated based on the equation resulting from the linear regression of the BSA standard curve . Seeds were sterilized by incubation in sterilization medium ( 70% ethanol and 0 . 04% ( wt/vol ) SDS ) for 15 min followed by three washes in 100% ethanol . After drying , the seeds for all genotypes were plated on one plate with ½ Murashige and Skoog Basal Medium ( MS; Sigma–Aldrich , St . Louis , MO ) and 0 . 8% phyto-agar . The plate was then stored at 4°C for >3 days and subsequently transferred to a growth cabinet ( 16/8 light/dark and 22°C ) . After a growth phase of 10–14 days >10 seedlings per genotype were floated on liquid ½ MS and equilibrated for 60–90 min in the growth cabinet . Either ±ABA ( Sigma ) to a final concentration of 50 μM ( indicated by + in the figure ) or the same volume of solvent control ( ethanol; indicated by—in the figure ) was added to the floating seedlings . After 30 min the seedlings were removed from the ½ MS and flash frozen in liquid nitrogen . Plant tissue was disrupted by shaking the frozen seedlings together with steel balls in a shaker ( Retsch ) for three times 30 s at 30 Hz in pre-cooled mountings . Subsequently , extraction buffer: 100 mM HEPES-NaOH pH 7 . 5 , 5 mM EDTA , 5 mM EGTA , 0 . 5% ( vol/vol ) Triton X-100 , 150 mM NaCl , 0 . 5 mM DTT , 10 mM NaF , 0 . 5% ( vol/vol ) protease inhibitor ( Sigma–Aldrich ) , 0 . 5% ( vol/vol ) phosphatase inhibitor 2 ( Sigma–Aldrich ) , 0 . 5% ( vol/vol ) phosphatase inhibitor 3 ( Sigma–Aldrich ) , 5 mM Na3VO4 , and 5 mM β-Glycerophosphate disodium salt hydrate was added . The samples were then treated in a sonication water bath ( Fisher Scientific ) with ice added to the water for 30 s . Cell debris was removed via centrifugation at 20 , 000×g and 4°C for 40 min . Protein concentrations of the supernatants were measured using the BCA Protein Assay Kit ( Pierce ) . 20 μg of total protein for each genotype and treatment were subjected to SDS-PAGE ( Laemmli , 1970 ) under denaturing conditions ( see in-gel kinase assay ) . The reaction buffer consisted of 100 mM HEPES-NaOH pH 7 . 5 , 10 mM MgCl2 , 2 mM DTT , 1 mM EGTA , and CaCl2 was added to get a final concentration of 2 . 5 μM free Ca2+ for all assays except the assay depicted in Figure 3—figure supplement 4 for which free Ca2+ was adjusted to 5 μM ( calculated with http://www . stanford . edu/∼cpatton/webmaxc/webmaxcE . htm ) . Note that the pH of the reaction buffer dropped to pH 7 . 3 after adding all components and free Ca2+ calculations were performed accordingly . The flow charts in the respective figures indicate the components which were added subsequently in sequence ( from top to bottom ) and the respective incubation times . For the reactions shown in Figure 3—figure supplement 4 0 . 5 μg of CPK6 and 1 μg of the PP2Cs ABI1 , ABI2 , and PP2CA were used . The addition of EGTA for reactions shown in Figure 3—figure supplement 4 lanes 2–4 resulted in a free Ca2+ concentration <10 nM ( calculated with http://www . stanford . edu/∼cpatton/webmaxc/webmaxcE . htm ) . For the experiments shown in Figure 4D–H and Figure 4—figure supplement 1 , SLAC1-NT ( 1 . 5 μg ) was mixed together with 200 nM of the protein kinases CPK6 , CPK23 , OST1 , and CPK21 in reaction buffer . Staurosporine was added to a final concentration of 100 μM and the final concentration of the PP2Cs ABI1 and PP2CA was 600 nM . To start all in vitro kinase reactions , 5 μCi of [γ-32P]-ATP ( Perkin–Elmer ) was added and the reactions were incubated at RT for 10 min . The final volumes were 20 μl and the reactions were stopped by the addition of 4 μl of 6× loading dye with subsequent incubation at 95°C for 5 min . The proteins were then separated by SDS polyacrylamide gel electrophoresis ( SDS-PAGE , Laemmli , 1970 ) in 4–20% acryl amide gradient gels ( Biorad ) . After , the proteins were stained with coomassie brilliant blue R-250 ( Sigma ) . To visualize incorporated 32P-derived radioactive signals , gels were exposed to a storage phosphor screen ( Molecular Dynamics; Figure 4D–H and Figure 4—figure supplement 1 ) or HyBlot CL autoradiography films ( Denville Scientific; Figure 3—figure supplement 4 ) . The phosphor storage screen was read out using a Typhoon scanner ( Amersham Bioscience ) . To compare CPK6 activities by measuring ATP consumption ( Figure 3—figure supplement 5 ) with or without the PP2Cs ABI1 and PP2CA , 0 . 5 μM of the protein kinase was incubated at RT for 7 . 5 min either alone or with 1 μM of PP2C protein in the above mentioned reaction buffer supplemented with 10 μM ATP and ∼150 μM Histone III-S ( Sigma ) . The reactions were stopped by the addition of staurosporine . Residual ATP levels were quantified using the Kinase-Glo kit ( Promega ) according to the manufacturer's instructions resulting in luminescence signals measured in a plate reader ( Berthold Mithras LB 940 ) ( Latz et al . , 2012 ) . ATP consumption was calculated by first assessing the maximum range ( ∆Rmax ) of luminescence by subtracting the signal intensity of the background ( no ATP added; RB ) from the maximum signal ( no kinase added; Rmax ) . To calculate the ATP consumption , signal intensities derived from the residual ATP in the reactions ( Rx ) were subtracted from the maximum signal ( Rmax ) and then related to the maximum range ( ∆Rmax ) and plotted in per cent ( [ ( Rmax − Rx ) /∆Rmax] × 100 ) . For in-gel protein kinase assays using recombinant proteins shown in Figure 3C , D , 500 ng of OST1 and CPK6 kinase and the PP2Cs ABI1 and PP2CA in a 1:3 molar ration were mixed in reaction buffer with 2 . 5 μM free Ca2+ ( for buffer composition see in vitro kinase assay section ) . The reactions labelled with ‘ ( ATP ) ’ were additionally supplemented with 100 μM ATP . All samples were incubated at RT for 20 min and stopped by adding SDS loading dye and heating at 95°C for 5 min . For the assay depicted in Figure 3—figure supplement 3 , 1250 ng of OST1 and 500 ng of CPK6 were used . These samples as well as the samples described in the ‘whole plant protein extraction’ section were subjected to SDS-PAGE ( Laemmli , 1970 ) . The 10% SDS acryl amide resolving gels were supplemented with 0 . 25–0 . 5 mg/ml Histone III-S ( Sigma–Aldrich ) ( Figure 3 , Figure 3—figure supplement 2 and Figure 3—figure supplement 3 ) or without Histone III-S ( Figure 3—figure supplement 3 ) . After electrophoresis , the gel was washed three times with washing buffer ( 25 mM Tris–HCl pH 8 . 0 , 0 . 5 mM DTT , 0 . 1 mM Na3VO4 , 5 mM NaF , 0 . 5 mg/ml BSA , and 0 . 1% ( vol/vol ) Triton X-100 ) for 30 min each at RT , followed by two washes with renaturation buffer ( 25 mM Tris–HCl pH 8 . 0 , 1 mM DTT , 0 . 1 mM Na3VO4 , and 5 mM NaF ) for 30 min each at RT and one wash at 4°C overnight . Then , the gels were equilibrated with reaction buffer ( see in vitro kinase assay ) for 30–45 min at RT and incubated in 20 ml of reaction buffer supplemented with 50 μCi [γ-32P]-ATP ( Perkin–Elmer ) . The reaction times were: Figure 3A , B , Figure 3—figure supplement 2 , and Figure 3—figure supplement 3: 90 min; Figure 3C: 60 min; Figure 3D: 120 min . To stop the reactions and to remove background signals the gels were subsequently extensively washed with a solution containing 5% ( vol/vol ) trichloroacetic acid and 1% ( vol/vol ) phosphoric acid for at least six times for 15 min each . The gels were then stained with coomassie brilliant blue R-250 , dried on Whatman 3MM paper , and exposed to a storage phosphor screen ( Molecular Dynamics ) . The storage phosphor screen was scanned with a Typhoon reader ( Amersham Bioscience ) . For the in-gel kinase assays shown in Figure 3A , B and Figure 3—figure supplement 2 all steps except the equilibration in reaction buffer and the reactions were carried out together and exactly the same way which allows the autoradiographs to be compared . The image files given by the Typhoon reader software are automatically adjusted to best display the bands with the highest intensity . Ca2+-activated kinase signals are stronger than OST1-derived bands which renders OST1 bands hardly visible by the Typhoon reader software . To better visualize OST1 activity , the signal intensity of the ∼41 kDa regions ( blue box ) in Figure 3A , B and Figure 3—figure supplement 1 were adjusted as described in the following: The output files ( . gel ) of the Typhoon scanner software was opened using Fiji ( Schindelin et al . , 2012 ) and in order to enhance the visibility of OST1-derived bands the maximum signal was adjusted for the entire image including controls and the two gels which are compared in accordance with publication policies ( http://jcb . rupress . org/content/166/1/11 . full ) . Subsequently , the regions around 41 kDa were saved as . jpg file which was used for the preparation of the figures . The parallel adjustment of the whole image showing both gels which are depicted in either Figure 3A , B or Figure 3—figure supplement 1 allows the comparisons of band intensities within each figure . Additionally , in Figure 3—figure supplement 1 several lanes of the same gel have been cut out indicated by the black line as explained in http://jcb . rupress . org/content/166/1/11 . full . Quantitative BiFC experiments were carried out as described in ( Waadt et al . , 2008 ) with changes explained in the following: BIFC vectors were altered to be USER cloning ( Nour-Eldin et al . , 2006 ) compatible ( indicated by the ‘u’ addition to the vector name ) as described in Nour-Eldin et al . ( 2010 ) . SLAC1-WT , SLAC1 S59A , SLAC1 S120A , SLAC1 S59A/S120A , CPK6 , and ABI1 coding sequences were amplified using the PfuX7 polymerase ( Norholm , 2010 ) . SLAC1-WT , SLAC1 S59A , SLAC1 S120A , and SLAC1 S59A/S120A , were USER-cloned into pSPYCE ( MR ) u while CPK6 and ABI1 were inserted into pSPYNE173u , and pSPYNE ( R ) 173u , respectively . Coding sequences of PP2AA3 ( AT1G13320 ) and PP2AC5 ( AT1G69960 ) were PCR-amplified using Phusion DNA polymerase ( Invitrogen ) and inserted SpeI/XmaI into a modified pUC19 vector including the pUBQ10 promoter ( pUC-pUBQ10; [Waadt et al . , 2014] ) . PP2AA3 was subcloned into the BiFC vectors pSPYCE ( MR ) and PP2AC5 into pSPYNE ( R ) 173 . Spinning disc confocal microscopy was performed using the following setup: Nikon Eclipse TE2000-U microscope with Nikon Plan 20×/0 . 40 ∞/0 . 17 WD; 1 . 3 and Plan Apo 60×/1 . 20 WI ∞/0 . 15–0 . 18 WD; 0 . 22 objectives . Attached were a CL-2000 diode pumped crystal laser ( LaserPhysics Inc . ) , and a LS 300 Kr/Ar laser ( Dynamic Laser ) , a Photometrics CascadeII 512 camera , a QLC-100 spinning disc ( VisiTech international ) , and a MFC2000 z-motor ( Applied Scientific Instruments ) . The software used to acquire the images was Metamorph ( version 7 . 7 . 7 . 0; Molecular Devices ) . Figure 4A images depict maximum projections of z-stacks . Two electrode voltage clamp measurements in X . leavis oocytes were carried out as described previously ( Brandt et al . , 2012 ) with adjustments listed in the following . The recording solution contained 10 mM MES/Tris ( pH 5 . 6 ) , 1 mM CaCl2 , 1 mM MgCl2 , 2 mM KCl , 24 mM NaCl , and 70 mM Na-gluconate . Osmolality was adjusted to 220 mM using D-sorbitol . Oocytes were held at a holding potential of 0 mV , and subjected to voltage pulse from +40 mV to −120 mV , −140 mV , or −160 mV in −20 mV decrements . The amounts of injected cRNA were 10 ng except for Figure 6—figure supplement 2A , E . For the experiments shown in Figure 6—figure supplement 2A , E the amounts of injected cRNA were 5 ng of SLAC1 , 0 . 5 ng of CPK6 , and 7 . 5 ng of OST1 .	Plant leaves have tiny openings or pores called stomata , which allow carbon dioxide , water vapor and other gases to diffuse in and out of the plant . Two cells called guard cells surround each stoma and control the opening and closing of the pore . If a plant is losing excessive amounts of water , for example during a drought , the plant produces a hormone called abscisic acid that promotes the closure of its stomata . When abscisic acid is present , the guard cells are sensitive to changes in their internal concentration of calcium ions so that calcium ions can activate a protein called SLAC1 . This leads to responses in the guard cells that close the stoma . The calcium ions activate SLAC1 by stimulating enzymes called calcium-dependent protein kinases ( CPKs ) . However , abscisic acid can also trigger other enzymes that can activate SLAC1 independently of the calcium ions . Calcium ions are also reported to be involved in the opening of stomata , when abscisic acid is not present . Therefore , it is not clear how abscisic acid works to specifically ‘prime’ guard cells to close the stomata in response to increases in calcium ions during drought . Brandt , Munemasa et al . studied stomata in a plant called Arabidopsis thaliana . The experiments show that , in the presence of abscisic acid , mutant plants that lack four different CPK enzymes are impaired in the activation of SLAC1 and the closing of stomata in response to increases in calcium ions . Further experiments found that other enzymes called the PP2Cs—which are switched off by abscisic acid—are responsible for regulating the Ca2+ sensitivity of guard cells . Switching off PP2Cs enables closing of the stomata in response to calcium ions . It has been suggested previously that the CPKs and the calcium-independent enzymes are involved in two separate pathways that promote the closure of stomata . However , Brandt , Munemasa et al . found that the calcium-independent enzymes are required for calcium ions to activate SLAC1 in guard cells , revealing that these two pathways are linked . Brandt , Munemasa et al . 's findings reveal how abscisic acid is able to specifically prime guard cells to close stomata in response to calcium ions . The next challenge is to understand how the CPKs and calcium-independent enzymes work together during the closure of stomata .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology" ]	2015	Calcium specificity signaling mechanisms in abscisic acid signal transduction in Arabidopsis guard cells
The coordinated action of the auxin-sensitive Aux/IAA transcriptional repressors and ARF transcription factors produces complex gene-regulatory networks in plants . Despite their importance , our knowledge of these two protein families is largely based on analysis of stabilized forms of the Aux/IAAs , and studies of a subgroup of ARFs that function as transcriptional activators . To understand how auxin regulates gene expression we generated a Physcomitrella patens line that completely lacks Aux/IAAs . Loss of the repressors causes massive changes in transcription with misregulation of over a third of the annotated genes . Further , we find that the aux/iaa mutant is blind to auxin indicating that auxin regulation of transcription occurs exclusively through Aux/IAA function . We used the aux/iaa mutant as a simplified platform for studies of ARF function and demonstrate that repressing ARFs regulate auxin-induced genes and fine-tune their expression . Further the repressing ARFs coordinate gene induction jointly with activating ARFs and the Aux/IAAs . The plant hormone auxin plays a central role in plant growth and development . Depending on the context , the cellular response to auxin can be very different , including changes in cell division , cell expansion , and differentiation . The hormone acts by regulating the transcription of auxin responsive genes . Strikingly , auxin-regulated gene sets can vary significantly between different cell types consistent with cell specific cellular responses ( Bargmann et al . , 2013 ) . In the absence of auxin , transcription of auxin-regulated genes is repressed by members of a family of repressors called Auxin/INDOLE-3-ACETIC-ACID ( Aux/IAA ) proteins . The Aux/IAAs are recruited to promoters through an interaction with AUXIN RESPONSE FACTOR ( ARF ) transcription factors . Repression is relieved when auxin binds to a co-receptor complex consisting of a TRANSPORT INHIBITOR RESISTANT 1/AUXIN F-BOX ( TIR1/AFB ) F-box protein and an Aux/IAA protein . The Aux/IAA protein is degraded , permitting ARF-dependent transcription to occur [ ( Calderon Villalobos et al . , 2012; Dharmasiri et al . , 2005a; 2005b; Kepinski and Leyser , 2005; Tan et al . , 2007 ) reviewed in ( Salehin et al . , 2015; Wang and Estelle , 2014 ) ] . The elucidation of the auxin co-receptor mechanism provided the molecular link between auxin perception at the cellular level and subsequent changes in gene expression . However , the mechanisms by which the interactions between TIR1/AFBs , ARFs and Aux/IAAs results in the induction of specific gene sets , as well as the establishment of gene regulatory networks , are not known . The Aux/IAAs contain , in most cases , three domains: an N terminal repression domain required for the recruitment of the co-repressor TOPLESS ( TPL ) ( Szemenyei et al . , 2008 ) domain II , required for interaction with the TIR1/AFB co-receptors; and a C terminal region ( designated domains III and IV ) that forms a Phox and Bem1 ( PB1 ) domain ( Guilfoyle and Hagen , 2012 ) . ARF proteins are generally comprised of an N terminal B3-type DNA binding domain , a variable middle region ( MR ) , and a C terminal PB1 domain . The ARFs have been characterized as activating or repressing based on their behavior in transient protoplast assays ( Guilfoyle and Hagen , 2007; Tiwari et al . , 2003; Ulmasov et al . , 1999 ) . It was recently shown that Arabidopsis ARF5 , an activating ARF , interacts with SWI/SNF chromatin remodeling ATPases through its MR , and that this interaction mediates changes in chromatin required for gene activation ( Wu et al . , 2015 ) . On the other hand the repressing activity of the putative repressing ARFs has not been demonstrated in plants . Similar , the mechanism of repression is unknown . ARF-Aux/IAA interactions occur through their PB1 domains . In addition , the PB1 domain permits homo- and heterodimerization both within and between the Aux/IAA and ARF families . Recent structural studies revealed that ARFs can form higher order ARF and Aux/IAA protein complexes through both their DNA binding domain [ARF-ARF dimerization ( Boer et al . , 2014 ) ] and the acidic and basic faces of their PB1 domains [ARF-ARF and ARF-Aux/IAA multimerization ( Dinesh et al . , 2015; Han et al . , 2014; Korasick et al . , 2014; Nanao et al . , 2014 ) reviewed in ( Korasick et al . , 2015; Wright and Nemhauser , 2015 ) ] . This combinatorial diversity may result in complex regulation of gene expression . Further , tightly regulated negative feedback loops in which the Aux/IAA genes are regulated by the ARFs , present an additional layer of complexity . The body plan of the early-diverged moss Physcomitrella patens ( P . patens ) is relatively simple . The vegetative gametophyte is composed of only a few cell types and developmental stages . Auxin was shown to have a central role in the vegetative growth of P . patens ( Prigge and Bezanilla , 2010 ) . In the filamentous protonemal stage , auxin promotes the differentiation of chloroplast-rich filaments called chloronemata into elongated filaments with fewer chloroplasts called caulonemata ( Ashton et al . , 1979 ) . The development of leafy gametophores is also affected by auxin including stem elongation ( Eklund et al . , 2010; Fujita et al . , 2008 ) , elongation of the gametophore leaves ( Bennett et al . , 2014; Decker et al . , 2006 ) , formation of rhizoids from gametophore epidermal cells ( Ashton et al . , 1979 ) , and gametophore branching ( Coudert et al . , 2015 ) . We previously demonstrated that the molecular mechanism of auxin signaling is conserved between P . patens and Arabidopsis ( Lavy et al . , 2012; Prigge et al . , 2010 ) . This finding , together with its morphological simplicity establishes P . patens as a powerful model for studies of auxin signaling . Here we utilized a P . patens aux/iaa null mutant to determine the effects of the complete loss of Aux/IAA-based transcriptional repression . Our analysis reveals that the Aux/IAAs are essential for auxin regulation of transcription indicating that at least in moss , auxin does not affect transcription independently of the Aux/IAAs . Complete loss of Aux/IAA repression dramatically alters the transcriptome , including expression of a large number of genes that are not affected by auxin treatment . In addition , our studies revealed new features of repressing ARF function . We demonstrate that auxin induction of gene expression is controlled by complex interactions between the Aux/IAAs and both activating and repressing ARFs . Flowering plants possess large families of Aux/IAA genes ( 29 in Arabidopsis ) . Genetic studies of these genes have relied almost entirely on gain-of-function mutations in the degron motif ( Prigge et al . , 2010; Mockaitis and Estelle , 2008 ) . These dominant mutations prevent interaction between the Aux/IAA and the TIR1/AFB co-receptors resulting in stabilization of the affected Aux/IAA and reduced sensitivity to auxin ( Figure 1G ) . Very few loss-of-function mutants have been described in flowering plants , presumably because of gene redundancy and a plant completely lacking the entire Aux/IAA family is not available . In a recent study , the single Aux/IAA gene in the early diverging Liverwort Marchantia polymorpha ( M . polymorpha ) was knocked down using an artificial microRNA ( Flores-Sandoval et al . , 2015 ) . The resulting lines were auxin hypersensitive . However , because these plants retained some auxin responsiveness , there are unlikely to be nulls . The genome of P . patens encodes only three Aux/IAA genes: IAA1A , IAA1B , and IAA2 . To study the role of these genes in auxin response and plant development , and to simplify the study of auxin response mechanisms , we generated a mutant lacking all three genes . To generate the triple mutant , we replaced the IAA1B coding region with a β-glucuronidase ( GUS ) marker gene , followed by deletion of the IAA1A and IAA2 coding regions ( PIAA1B:GUS aux/iaaΔ ) ( Figure 1A and Figure 1—figure supplement 1 ) . Because expression of the IAA1B gene is induced by auxin ( Lavy et al . , 2012; Prigge et al . , 2010 ) , we could use the PIAA1B:GUS gene as an auxin reporter . Following auxin treatment , very low levels of GUS staining was detected in the PIAA1B:GUS line , whereas GUS staining was high in the PIAA1B:GUS aux/iaaΔ line both in the absence and presence of applied auxin , indicating that this line displays a constitutive auxin response ( Figure 1B ) . 10 . 7554/eLife . 13325 . 003Figure 1 . The aux/iaaΔ mutant displays a severe phenotype . ( A ) Scheme representing the auxin-signaling pathway in WT plants ( left panel ) and in the aux/iaaΔ mutant ( right panel ) . ( B ) GUS expression in PIAA1B:GUS and aux/iaaΔ plants carrying the PIAA1B:GUS reporter . Arrows denote GUS expression . ( C ) WT and the aux/iaaΔ mutant grown for one month on BCD medium , stimulating gametophore development without auxin or with different concentrations of 1-naphthalene-acetic acid ( NAA ) . ( D ) Microscopic enlargement of WT gametophores: left panel- leafy gametophores grown without exogenous auxin . Right panel-ectopic rhizoids emerging from a gametophore grown on 12 . 5 μM NAA . ( E–H ) Plants grown for one month on BCDAT medium to promote filamentous growth . ( E , F ) WT plant grown without auxin or with 12 . 5 μM NAA , respectively . ( G ) iaa2-P328S degron-motif mutant . ( H ) The aux/iaaΔ mutant has a variable phenotype . ( I ) WT and aux/iaaΔ mutant grown for one month on BCDAT or BCDAT supplemented with 10 μM L-Kyn . Scale bars: 1 mm ( B ) , 0 . 5 cm ( C , H , I ) , the scale bar in H also corresponds to E–G , 0 . 5 mm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 00310 . 7554/eLife . 13325 . 004Figure 1—figure supplement 1 . The aux/iaaΔ is a null mutant that displays constitutive auxin response . ( A ) PCR validating the absence of the genomic sequence of IAA1A , IAA1B , and IAA2 after CRE recombination . Primer pairs used: PML 749 , 751 ( IAA1A ) , PML 750 , 751 ( IAA1B ) , PML753 , 754 ( IAA2 ) . ( B ) qPCR validating that IAA1A , IAA1B , and IAA2 are not expressed . The expression of the housekeeping gene EF1α is normal in the mutant and is used as a control . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 00410 . 7554/eLife . 13325 . 005Figure 1—figure supplement 2 . Early stages of filamentous growth of WT and aux/iaaΔ mutant . ( A ) Six WT and aux/iaaΔ mutant plants were transferred to BCD medium four days after protoplast recovery and measured for 5 days . Error bars represent s . e . m . ( B ) Protonemal filaments of WT and aux/iaaΔ plants grown on BCD for two weeks following protoplast recovery without auxin or with 0 . 5 μm NAA . ( C ) Confocal images visualizing chloroplasts in WT and aux/iaaΔ plants seven days after protoplast recovery . Chloroplasts are visualized in white . Scale bars: 0 . 25 mm ( B ) , 200 μm ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 005 The resulting aux/iaa triple knockout mutant ( aux/iaaΔ ) displayed a phenotype that was similar to that of WT plants grown on high levels of auxin for one month , but more extreme ( Figure 1C ) . When grown on minimal medium ( BCD ) , WT plants grew leafy gametophores ( Figure 1D-left panel ) , while in the presence of auxin , these plants produced gametophores consisting of shoot structures with ectopic brown-pigmented rhizoids and without leaves ( Figure 1D-right panel ) . The aux/iaaΔ mutant produced gametophores with ectopic rhizoids and was completely insensitive to auxin ( Figure 1C ) . To characterize plants at the filamentous protonemal stage , we compared WT , a highly auxin-resistant IAA2 degron mutant ( iaa2-P328S ) , and the aux/iaaΔ mutant . The plants were grown for one month on medium supplemented with ammonium tartrate to allow for slower protonemata differentiation ( BCDAT medium ) ( Figure 1E–H ) . Under these conditions , WT plants responded to exogenous auxin by developing long caulonemal filaments ( Figure 1E , F ) . Whereas the differentiation of the iaa2-P328S auxin-resistant mutant arrested at the primary chloronemal stage ( Figure 1G ) , aux/iaaΔ plants contained disorganized brown-pigmented filaments ( Figure 1H ) . Interestingly , the phenotype of aux/iaaΔ plants was highly variable . Each plant developed heterogeneous filaments with varying levels of chloroplasts and brown pigment ( Figure 1H , left and right panel illustrates the range of phenotypes ) . Since the aux/iaaΔ mutant displayed a phenotype similar to that of auxin treated plants , we also tested its response to reduced endogenous auxin levels using the auxin biosynthesis inhibitor L-Kynurenine ( L-Kyn ) ( He et al . , 2011 ) . Differentiation of chloronema to caulonemal filaments and gametophore development in WT plants were slower in the presence of L-Kyn , whereas the aux/iaaΔ mutant was insensitive to the inhibitor ( Figure 1I ) . Next , we characterized the aux/iaaΔ mutant at earlier stages of growth following either protoplast recovery or tissue homogenization . Following protoplast recovery , WT plants had a higher growth rate compared to the mutant ( Figure 1—figure supplement 2A ) . After seven days both chloronemal and caulonemal cells were observed in WT plants , whereas clearly differentiated cell types , either chloronemata or caulonemata , were not observed in the aux/iaaΔ mutant ( Figure 1—figure supplement 2B , C ) . A comparison between WT chloronemal cells and the aux/iaaΔ mutant cells , seven days following tissue homogenization , revealed that the mutant cells are significantly wider ( Figure 2A , B ) suggesting that polarized cell growth is impaired . The chlorophyll concentration at that stage was lower in the mutant compared to WT ( Figure 2C ) . Despite these significant morphological and physiological differences , the mutant filamentous growth appeared robust ( Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 13325 . 006Figure 2 . Loss of the Aux/IAA genes results in dramatic changes in gene expression at the filamentous developmental stage . ( A ) Confocal images of WT and aux/iaaΔ mutant protonemata seven days after tissue homogenization ( Left panels-chloroplasts are visualized in green . Right panels-cell structures are visualized with DIC ) . Scale bar: 200 μm . ( B ) Average length and width ( μm ) of WT chloronemal cells and aux/iaaΔ mutant cells proximal to branch points . Error bars represent s . e . m . *p<0 . 001 ( t-test ) , n=30 . ( C ) Total chlorophyll concentration in WT and aux/iaaΔ protonemata seven days after tissue homogenization . Error bars represent s . e . m . p<0 . 05 ( t-test ) , n=3 . ( B ) Venn diagram showing the overlap between the four data sets of differentially expressed genes ( padj <0 . 01 , fold change ≥1 . 5 ) . ( D ) Hierarchical clustering of genes displaying differential expression between auxin treated and untreated WT plant samples with fold change ≥2 compared to their expression levels in the aux/iaaΔ mutant . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 00610 . 7554/eLife . 13325 . 007Figure 2—figure supplement 1 . Protonemal tissue grown under the growth conditions used for the RNAseq experiment . WT and aux/iaaΔ grown on BCDAT plates with cellophane overlays for seven days after tissue homogenization . Scale bar: 0 . 5 cm . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 00710 . 7554/eLife . 13325 . 008Figure 2—figure supplement 2 . qPCR showing the expression levels of auxin responsive genes in WT plant and aux/iaaΔ mutant in mock- or 10 μM IAA-treatment for one hour . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 00810 . 7554/eLife . 13325 . 009Figure 2—figure supplement 3 . Graphical representation of enriched cellular components associated with auxin-responsive genes . Colors of cells represent significance levels . The significance values are designated within each cell . Green arrows represent negative regulation . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 009 The availability of the aux/iaaΔ mutant provided a unique opportunity to fully separate auxin perception from auxin response . Taking advantage of this opportunity , we profiled auxin-responsive transcription in WT and aux/iaaΔ protonemata by RNA sequencing ( RNAseq ) and compared the transcriptomes . Our mutant analysis revealed that the differences between WT and the aux/iaaΔ mutant are more significant in later developmental stages . To minimize these differences , we performed RNAseq on protonemata grown on BCDAT for seven days after tissue homogenization ( Figure 2A ) We first analyzed the WT auxin-responsive gene set following five hours of IAA treatment and identified 723 upregulated and 762 repressed genes [1 . 5 fold or more; p value adjusted ( padj ) <0 . 01] ( Supplementary file 1A–C ) . Among the upregulated genes were several well-characterized early auxin response genes including the Aux/IAA genes as well as several genes homologous to SAURs and LBD/ASLs ( Paponov et al . , 2008 ) . The P . patens genome encodes only two GH3 homologs with a conserved function in IAA-conjugate synthesis ( Ludwig-Muller et al . , 2009 ) . Neither of these genes displayed differential regulation in the auxin-responsive gene set and one of them was downregulated in the aux/iaaΔ mutant . A few ARF genes in Arabidopsis have been shown to respond to auxin ( Lau et al . , 2011; Paponov et al . , 2008 ) . Five out of sixteen P . patens ARFs ( including both activating and repressing ) were upregulated by auxin in the WT protonemata . Nine additional ARFs were upregulated in the aux/iaaΔ mutant compared to WT , indicating the presence of an ARF-dependent feedback loop that may have evolved in the ancestral land plant lineage . The larger number of auxin regulated ARFs in our datasets may reflect the fact that by comparing the aux/iaaΔ mutant to WT , we uncover all Aux/IAA-repressed genes , whereas the Arabidopsis transcriptome data describes the effect of exogenous auxin on a complex tissue and under specific conditions . Strikingly , our results show that the aux/iaaΔ mutant is completely insensitive to auxin , with no changes in gene expression upon auxin treatment under our experimental conditions and statistical threshold ( Figure 2D ) . A less stringent statistical threshold of padj <0 . 05 revealed only a few genes displaying a low fold change further validated this finding ( Supplementary file 1D ) . These results imply that any factors that regulate auxin-dependent changes in gene expression must act through the Aux/IAAs , emphasizing the central role of these proteins in auxin signaling . The aux/iaaΔ transcriptomic analysis revealed that a third of all annotated genes were differentially regulated in the mutant compared to WT ( with 3752 and 4031 up- and downregulated genes , respectively ) , demonstrating the broad role of the TIR1-Aux/IAA pathway in land plant growth and development . Strikingly , approximately 80% of these genes were not differentially regulated by auxin in WT plants whereas most of the regulated genes in WT were presented in the aux/iaaΔ mutant dataset ( Figure 2D ) . Thus , our analysis allowed us to identify genes that were not revealed by auxin treatment . These may include genes that are regulated in a specific developmental , temporal , or environmental context as well as genes that are indirectly affected by the loss of the Aux/IAAs . Additionally , some genes may be revealed only when the Aux/IAAs are completely absent . For example , it is possible that very low levels of Aux/IAAs are sufficient to repress expression of some genes while for others the Aux/IAAs may be protected from auxin dependent-degradation . Finally , some genes may be represented by a small number of transcripts and/or exhibit a low level of differential expression , making them difficult to identify with standard procedures . As illustrated by a hierarchical clustering of differentially expressed genes , the absence of the Aux/IAAs also had a strong effect on transcript levels ( Figure 2E ) . Numerous up- and downregulated genes identified in WT plants displayed higher or lower expression levels in the aux/iaaΔ mutant respectively , indicating that many genes can be expressed over a broad range . For example , while the highest fold-change of differentially expressed genes following auxin treatment was fifteen , some of the same genes were differentially expressed in the aux/iaaΔ mutant with a fold-change of up to 430 compared to WT . These results indicate that significant levels of the Aux/IAA are present even after auxin treatment , further illustrating the robust nature of the auxin system . Several up- and downregulated genes that were previously described ( Lavy et al . , 2012 ) , and genes displaying differential expression in this analysis , were selected as representative auxin-responsive gene markers , and used to validate the transcriptomic analysis ( Figure 2—figure supplement 2 ) . The moss auxin-responsive transcriptome may facilitate discovery of ancestral auxin gene targets . We analyzed Gene Ontology ( GO ) terms associated with auxin responsive genes and found specific enriched categories for up- and downregulated genes . Analysis of upregulated genes revealed overrepresentation of genes involved in regulation of transcription and biosynthesis , whereas downregulated genes were highly enriched for processes occurring in the chloroplast and associated with photosynthesis , including light responses and carbon fixation ( Supplementary file 2A , B and Figure 2—figure supplement 3 ) . The set of downregulated genes differentially expressed in the mutant but not in WT presented a higher level of enrichment of these processes . Thus , our GO analysis suggests that in P . patens photosynthetic tissues , auxin signaling mediates separation of auxin-induced responses from energy production via induction of specific gene sets and the concurrent repression of others . The constitutive auxin response displayed by the aux/iaaΔ mutant and the broad effect of loss of Aux/IAA function on gene regulation highlight the central role of ARF transcription factors on gene expression . Based on sequence analyses and transient activity assays ( Tiwari et al . , 2003; Ulmasov et al . , 1999 ) , the ARF proteins were classified as either activators or repressors of transcription . While it is clear that the activating ARFs do activate transcription in plants and their mode of action is beginning to emerge ( Wu et al . , 2015 ) , this is not the case for the repressing ARFs . It has not been clearly demonstrated that the repressing ARFs act as transcriptional repressors in plants and a conceptual understanding of their role is lacking . For example , it is not known if repressing ARFs require the Aux/IAAs for their function or if the repressing and activating ARFs regulate the same genes . In addition it is not known if the repressing ARFs have a specific role in targeting downregulated genes . The aux/iaaΔ mutant allows us to study the function of the repressing ARF in a relatively simple context lacking Aux/IAA repression and the negative feedback loops associated with their activity . Phylogenetic analyses reported that the P . patens ARF protein family consists of three clades ( Plavskin and Timmermans , 2012 ) shared between all land plants and a fourth , non-seed-plant-specific clade characterized by the lack of the DNA binding domain [ ( Paponov et al . , 2009 ) , Figure 3—figure supplement 1] . One of the clades , comprising four proteins , groups with Arabidopsis repressing ARFs , whereas another clade , comprising seven ARF proteins , groups with Arabidopsis activating ARFs . The M . polymorpha ARF proteins were recently shown to affect transcription in a transient expression assay in accordance with their phylogenetic classification ( Kato et al . , 2015 ) strongly suggesting that ARF function is conserved in land plants . Furthermore , in P . patens , higher levels of putative repressing ARFs resulted in decreased auxin response ( Plavskin et al . , 2016 ) . We selected ARFb4 as a representative repressing ARF , and overexpressed an ARFb4 c-Myc fusion in the aux/iaaΔ mutant background ( ARFb4OE_aux/iaaΔ ) . We found that overexpression of ARFb4 suppressed the constitutive auxin phenotype of the aux/iaaΔ mutant including the formation of green chloronemal-like filaments ( Figure 3A ) . In fact , in a high expressing transgenic line ( Figure 3A #1 and Figure 3—figure supplement 2C ) , the phenotype was quite similar to that of dominant auxin-resistant aux/iaa mutants [Figure 1G and ( Prigge et al . , 2010 ) ] . This result demonstrates that repressing ARFs do function as repressors in plants and that this activity does not require the Aux/IAAs . It is worth noting that ARFb4 , as well ARFb2 and ARFb3 do not contain any of the known motifs that are thought to recruit TPL and therefore may not affect chromatin remodeling directly . To define the mechanism underlying the suppression of the aux/iaaΔ phenotype we assessed the expression of the PIAA1B:GUS marker and other auxin-responsive genes . The GUS marker ( Figure 3B ) , as well as the additional reporter genes ( Figure 3C ) had dramatically lower expression levels compared to the aux/iaaΔ mutant indicating that activating ARFs and the repressing ARFs can affect , either directly or indirectly the same target genes . Further , the auxin response genes did not respond to auxin in the ARFb4OE_aux/iaaΔ line confirming that overexpression of the repressing ARF converted the constitutive auxin response line to an auxin resistant line . Overexpression of Arabidopsis ARF1 resulted in similar effects , indicating that the role of repressing ARFs is conserved between P . patens and Arabidopsis ( Figure 3C and Figure 3—figure supplement 2 ) . 10 . 7554/eLife . 13325 . 010Figure 3 . Repressing ARFs target auxin-induced genes . ( A ) WT , aux/iaaΔ and two aux/iaaΔ lines overexpressing ARFb4 ( ARFb4OE_aux/iaaΔ ) grown for one month on BCDAT . ( B ) GUS expression in aux/iaaΔ and two ARFb4OE_aux/iaaΔ lines carrying the PIAA1B:GUS reporter . Arrows denote GUS expression . ( C ) qPCR showing the expression levels of auxin responsive genes in WT , ARFb4OE_aux/iaaΔ ( line #1 ) , AtARF1_aux/iaaΔ , aux/iaaΔ , and iaa2-P328S in presence of 10 μM IAA-treated for five hours or mock . Error bars represent s . e . m . a/b/c=P<0 . 05 ( t-test ) , n=3 . a=t-test comparing the lines to WT , b= t-test comparing the lines to aux/iaaΔ . c=t-test comparing the lines to iaa2-P328S . ( D ) WT and ARFb4OE_aux/iaaΔ #1 grown on BCDAT without auxin or with 12 . 5 μM NAA for one month . Scale bars: 0 . 5 cm ( A ) , 0 . 5 mm ( B ) , 0 . 5 cm ( D ) DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 01010 . 7554/eLife . 13325 . 011Figure 3—figure supplement 1 . Phylogeny of Land Plant ARFs . The relationships between land plant ARFs were inferred using an alignment of amino-acid sequences and the MrBayes 3 . 2 . 2 program . The text colors indicate the plant whose genome encodes the protein: black , Arabidopsis thaliana; purple , Selaginella moellendorffii; and green , Physcomitrella patens . Numbers inside the nodes indicate posterior probabilities . Distinct clades of ARF genes are highlighted by background shading . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 01110 . 7554/eLife . 13325 . 012Figure 3—figure supplement 2 . The Arabidopsis repressing ARF1 has similar effects on plant growth as ARFb4 . This figure is an extension of Figure 3A , B and includes two independent AtARF1 overexpression lines in aux/iaaΔ mutant background . ( A ) WT , aux/iaaΔ and two ARFb4 overexpression in aux/iaaΔ ( as shown in Figure 3A ) and two Arabidopsis ARF1 overexpression in aux/iaaΔ grown for one month on BCDAT . ( B ) GUS expression of IAA1B transcriptional fusion line , aux/iaaΔ mutant , and two ARFb4 and AtARF1 overexpression lines in aux/iaaΔ mutant background . ( C ) Immunoblot of total protein extract from plants overexpressing PpARFb4 and AtARF1 , detected with a c-Myc antibody . The line numbers #1 and #2 refer to plants in panel A and B . Scale bars: 0 . 5 cm ( A ) , 0 . 5 mm ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 012 Although ARFb4 overexpression resulted in a clear suppression of the aux/iaaΔ constitutive response , the phenotype was not as dramatic as the aux/iaa degron mutants . Expression of the auxin reporter genes was lower in the iaa2-P328S mutant compared to the ARFb4OE_aux/iaaΔ ( Figure 3C #5 ) . This result suggests that the repression conferred by the Aux/IAAs is stronger than ARF-based repression . Our results clearly indicate that auxin cannot stimulate transcriptional response in the ARFb4OE_aux/iaaΔ following several hours of treatment . However , it is possible that auxin can promote protonemal development through other mechanisms or by longer treatment . To explore this possibility we grew the ARFb4OE_aux/iaaΔ #1 line on NAA . Filament differentiation was not observed after one month of auxin treatment ( Figure 3D ) indicating that the Aux/IAA function is required for developmental transition in protonema . We next examined how repressing and activating ARFs interact to regulate the same gene targets . Several attempts to overexpress the activating ARFs P . patens ARFa8 and Arabidopsis ARF7 in the aux/iaaΔ mutant were unsuccessful . We only recovered one ARFa8 transgenic line with weak expression of ARFa8 suggesting that further gene activation , beyond that observed in the aux/iaaΔ mutant , is lethal . As an alternative approach we selected one ARFb4OE_aux/iaaΔ line ( Figure 3A #1 ) and introduced an inducible ARFa8-glucocorticoid receptor ( GR ) fusion gene into this line ( ARFa8-GR_ARFb4OE_aux/iaaΔ ) . Two resulting lines were selected for further analysis ( ARFa8-GR_ARFb4OE_aux/iaaΔ #1 & #2 ) ( Figure 4A–C ) . Following dexamethasone ( DEX ) treatment these lines developed gametophores with ectopic rhizoids indicative of auxin hypersensitivity ( Figure 4A , B ) . Expression of the auxin reporter genes supported the phenotypic analysis since expression levels of the upregulated genes in the DEX treated plants were higher compared to both the ARFb4OE_aux/iaaΔ background and untreated ARFa8-GR_ARFb4OE_aux/iaaΔ transgenic lines ( Figure 4D ) . Similarly , the level of a representative downregulated gene was lower ( Figure 4D-CBS ) . This demonstrates that ARFb4 and ARFa8 , as representatives of repressing and activating factors respectively , display opposing activity on the same gene targets . 10 . 7554/eLife . 13325 . 013Figure 4 . Repressing and activating ARFs display opposite effects on the same target genes . ( A ) ARFb4OE_aux/iaaΔ ( line #1 in Figure 3A ) , and two ARFa8-GR ARFb4OE_aux/iaaΔ lines grown on BCDAT or BCDAT supplemented with 10 μM DEX for a month . Scale bar: 0 . 5 mm . ( B ) Enlargement of gametophores of line #1 . Scale bar: 0 . 5 mm . ( C ) Immunoblot of total protein extracts from plants shown in E , detected with GR and c-Myc antibodies . ( D ) qPCR showing the expression levels of auxin responsive genes in the transgenic lines shown in A , ARFb4OE_aux/iaaΔ ( ARFb4 ) and ARFa8-GR ARFb4OE_aux/iaaΔ ( ARFb4 , ARFa8 ) mock- or 10 μM DEX-treated for overnight . Error bars represent s . e . m . p<0 . 05 ( t-test comparing DEX-induced- to uninduced gene expression levels in ARFb4 , ARFa8 lines ) , n=3 . ( E ) Electrophoretic mobility shift assay with GST-ARFb4 DBD and GST-ARFa8 DBD ( ARFb4 , ARFa8 ) with biotin-labeled DNA probes from DR5 , IAA1A , IAA1B , and ARFb4 promoters . + and – signs denote the presence or the absence of ARF proteins , and unlabeled specific or mutant competitor DNA sequences . Black line and numbers represent the probe locations respectively to the gene transcription start sites . Red triangles: canonical TGTCTC AuxREs . Black triangles: core TGTC AuxREs . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 013 Activating and repressing ARFs may target the same genes by competing for the same promoter elements , by binding to different regions within the same promoters , or through an indirect mechanism . To distinguish between these possibilities we performed an electrophoretic mobility shift assay with DNA sequences from the DR5 reporter and representative auxin responsive gene promoters , IAA1A , IAA1B , and ARFb4 . The results indicated that ARFb4 and ARFa8 DNA binding domains can bind to the same DNA sequences ( Figure 4E ) . Our results demonstrate that both the repressing ARFs and Aux/IAAs are capable of gene repression , albeit to different extents . To learn more about these two types of repression , we determined the effects of loss of repressing ARFs either in the presence or absence of the Aux/IAAs . The coding regions of ARFb2 and ARFb4 were deleted in WT as well as in the iaa1b iaa2Δ double and aux/iaaΔ triple mutants ( Figure 5—figure supplement 1 ) . The loss of ARFb2 and ARFb4 in the aux/iaaΔ line ( arfb2 arfb4Δ aux/iaaΔ ) did not result in any morphological changes or a clear trend with respect to changes in gene expression ( Figure 5—figure supplement 2 ) . Conversely and surprisingly , the arfb2 arfb4 knockout in both the WT and the iaa1b iaa2Δ double mutant ( arfb2 arfb4Δ and arfb2 arfb4Δ iaa1b iaa2Δ , respectively ) resulted in lower expression levels of some of the upregulated auxin responsive genes , and higher levels of downregulated genes compared to their backgrounds ( Figure 5A ) . Although the expression of these genes in both untreated and auxin-treated arfb2 arfb4Δ plants was reduced , the fold change was either similar or even higher compared to the WT and iaa1b iaa2Δ backgrounds ( Figure 5A , RSL4: 7–3 fold higher and Dox: 2–1 . 4 , respectively ) . Phenotypic analysis of the arfb2 arfb4Δ revealed developmental defects including shorter filaments and fewer leafy gametophores consistent with reduced auxin response ( Figure 5—figure supplement 3 ) . These findings further support our hypothesis that repressing ARFs can affect the same genes as activating ARFs . However , they revealed an unexpected trend in which the loss of repressing ARFs affects transcription in the same direction as their overexpression . To further confirm these contradictory results we also used transient RNAi ( Bezanilla et al . , 2005 ) targeting the four repressing ARFs ( arfb1-4 ) . The RNAi construct was expressed in a transgenic plant expressing the synthetic auxin-responsive reporter DR5:DsRED . In agreement with the analysis of the arfb2 arfb4Δ stable lines the resulting transformants displayed a lower expression level of the DR5:DsRED reporter compared to the control plants when treated with auxin , indicative of reduced auxin sensitivity of the arfb1-4 RNAi knockdown lines ( Figure 5B ) . 10 . 7554/eLife . 13325 . 014Figure 5 . Both repressing and activating ARFs affect auxin response in the same direction in the presence of the Aux/IAAs . ( A ) qPCR showing the expression levels of auxin responsive genes in WT , arfb2 arfb4Δ ( arfb2 , 4 ) , iaa1b iaa2Δ ( iaa1b , 2 ) , and arfb2 arfb4 iaa1b iaa2Δ ( iaa1b , 2 arfb2 , 4 ) , mock- or 10 μM IAA-treated for five hours . Error bars represent s . e . m . p<0 . 05 ( t-test comparing the arfb2 arfb4Δ , and arfb2 arfb4 iaa1b iaa2Δ lines to either WT or iaa1b iaa2Δ , respectively ) , n=3 . ( B ) Three independent arfb1-4 RNAi and control lines mock- or 10 μM IAA-treated . Each line is presented by two images ( DsRED fluorescence ( red ) , chlorophyll auto-florescence ( blue ) . Each image is from projection stacks of multiple confocal sections . Scale bar: 100 μm . ( C ) WT and an ARFa8 overexpression line ( ARFa8OE ) grown for one month on BCDAT without auxin or with 12 . 5 μM NAA . Scale bar: 0 . 5 cm . ( D ) qPCR showing the expression levels of auxin responsive genes in WT and ARFa8OE mock- or 10 μM IAA-treated for five hours . Error bars represent s . e . m . *p<0 . 05 ( t-test comparing the ARFa8OE line to WT ) , n=3 . ( E ) Model for auxin regulation of transcription . The expression level of an auxin responsive gene is determined by the interplay between the Aux/IAAs , the repressing ARFs , and the activating ARFs . At low auxin levels the Aux/IAAs provide stringent repression . At high auxin levels , the Aux/IAAs are degraded resulting in increased transcription through the action of the activating ARFs . The repressing ARFs act to buffer gene expression by attenuating the activity of the activating ARFs . The interplay between the three protein groups results in a wide range of gene expression levels that contribute to a dynamic and context specific auxin response . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 01410 . 7554/eLife . 13325 . 015Figure 5—figure supplement 1 . PCR detecting the insertion position of different arfb2 arfb4Δ mutants . ( A ) Validating the absence of the genomic sequence of ARFb2 and ARFb4PCR . Upper panel: validating the transgene integration in arfb2 arfb4Δ mutant . Primer pairs used to confirm the presence of either up or downstream transgene-endogenous sequence junctions , respectively: ARFb2 ( PML703 , 722; PML267 , 704 ) , ARFb4 ( PML195 , 675; PML197 , 676 ) . Lower panel: validating the transgene integration in arfb2 arfb4Δ iaa1b iaa2Δ and arfb2 arfb4Δ aux/iaaΔ mutants after CRE/lox recombination . Primer pairs used to confirm the absence of the endogenous sequences: ARFb2 ( PML703 , 704 ) , ARFb4 ( PML675 , 676 ) . WT sequences were not amplified using these endogenous primers since the resulting fragments are too long and cannot be amplified under the PCR conditions used . ( B ) Validating the absence of the coding region of ARFb2 and ARFb4 in cDNA samples by RT-PCR . Primer pairs used: ARFb2 ( PML551 , 612 ) ARFb4 ( PML512 , 513 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 01510 . 7554/eLife . 13325 . 016Figure 5—figure supplement 2 . Arfb2 and arfb4 Knockout in aux/iaaΔ mutant background ( arfb2 arfb4Δ aux/iaaΔ ) does not result in phenotypes comparable to reduced auxin sensitivity . ( A ) qPCR showing the expression levels of auxin responsive genes in WT , aux/iaaΔ , and arfb2 arfb4Δ aux/iaaΔ . ( B ) arfb2 arfb4Δ aux/iaaΔ and aux/iaaΔ mutants grown for one month on BCD . Scale bar: 0 . 5 cm . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 01610 . 7554/eLife . 13325 . 017Figure 5—figure supplement 3 . Phenotypic analysis of arfb2 arfb4Δ in WT and iaa1b iaa2Δ mutant background . WT , arfb2 arfb4Δ , iaa1b iaa2Δ , and arfb2 arfb4Δ iaa1b iaa2Δ lines grown for one month on BCDAT medium without auxin or with different NAA concentrations . Scale bar: 0 . 5 cm . DOI: http://dx . doi . org/10 . 7554/eLife . 13325 . 017 These contradictory results may be explained by the indirect effect of negative feedback loops , in which the extensive production of Aux/IAA proteins could result in an overall decrease in auxin response . However , this possibility does not appear to explain the phenotype of arfb2 arfb4Δ lines , as the transcript levels of IAA1A ( Figure 5A-IAA1A ) are low . Alternatively , it is possible that both activating and repressing ARFs as well as the Aux/IAAs jointly coordinate gene induction . This hypothesis is consistent with our analysis of the arfb2 arfb4Δ lines . In the presence of the Aux/IAAs , loss of ARFB2 and ARFB4 results in reduced auxin response . However , this effect is suppressed by the deletion of the Aux/IAA genes in the arfb2 arfb4Δ aux/iaaΔ line ( Figure 5—figure supplement 2 compared to Figure 5A ) . Since the activating and repressing ARFs can target the same genes , loss of the repressing ARFs may result in increased levels of activating ARFs on the auxin-responsive promoters . This may have the unexpected effect of recruiting more Aux/IAA repressors to these promoters . To test this idea we overexpressed the activating ARFa8 , in WT plants ( ARFa8OE ) . The resulting lines displayed a phenotype characteristic of auxin-resistant plants and exhibited reduced expression of auxin responsive genes ( Figure 5C , D ) , which supports our hypothesis . Collectively , our results support a model in which repressing ARFs can affect the occupancy of activating ARFs and Aux/IAAs on auxin responsive promoters . These interactions may provide buffering capacity that allows fine-scale regulation of auxin responsive genes ( Figure 5E ) . In this study we exploited the relative simplicity of P . patens to address the complexity of auxin-regulated transcription . The Aux/IAA proteins have a central role in auxin signaling , serving as both auxin co-receptors and transcriptional repressors . However , genetic analysis of these genes has been limited in flowering plants because of genetic redundancy . By deleting all three Aux/IAAs in moss we have determined the effects of the complete absence of Aux/IAA function on plant growth and auxin response . Remarkably , we found that the Aux/IAAs have a broad impact on the expression levels of a great number of genes , the majority of which do not respond to auxin treatment of moss protonemata . Thus our analysis demonstrates that existing transcriptomic studies conducted in flowering plants may significantly underestimate the effects of auxin on the genome and illustrates the central role of the TIR-Aux/IAA pathway on plant growth and development . While our work highlights the remarkably broad impact of the Aux/IAA proteins on gene expression , it also demonstrates an absolute requirement for the Aux/IAAs in auxin-responsive transcription . Some auxin-mediated signaling pathways have been proposed to affect transcription independently of the TIR1/AFB-Aux/IAA signaling cascade . These include a negative role of Mitogen-Activating Protein Kinase ( MAPK ) activity on auxin-regulated genes ( Lee et al . , 2009 ) , and the effect of the F-box protein SKP2A on the stability of cell division transcription factors ( Jurado et al . , 2010 ) . Our findings indicate that at least in moss , any possible effect of these pathways on auxin-regulated transcription requires Aux/IAAs’ function . In P . patens , auxin-dependent modulation of gene expression triggers developmental transitions and differentiation . Our mutant analyses reveal that the transition from chloronema to caulonema is regulated by a dynamic balance of quantitative auxin responses . When auxin response is low , as in the auxin-resistant mutants , protonema growth is not accompanied by developmental transition to the caulonemal stage . In contrast , a constitutive auxin response , as exhibited by the aux/iaaΔ mutant , results in rapid and abnormal maturation . It has been proposed that repressing and activating ARFs might compete for binding to the same promoters ( Vernoux et al . , 2011 ) . Structural studies of Arabidopsis ARF1 and ARF5 revealed that both classes of ARF can potentially bind to similar cis elements and form high order oligomers of ARFs and Aux/IAAs ( Boer et al . , 2014; Korasick et al . , 2014 ) . Our work provides experimental evidence that auxin responsive genes are targeted by both repressing and activating ARFs , which in turn coordinate their levels of expression . The structural studies as well as our results can support two models of ARF activity . Repressing and activating ARFs can either compete on the same binding sites or cooperatively induce transcription by forming heterodimers . Despite its classification as an activating ARF , the Arabidopsis ARF5 was shown to repress some of its identified gene targets ( Zhang et al . , 2014; Zhao et al . , 2010 ) . Given that ARF transcription factors may display opposing roles in different contexts , ARF activity can result in a very complex gene expression pattern . These mechanisms , by which different ARFs coordinate transcriptional responses , can explain the broad effect of auxin on gene expression and the complexity of the resulting transcriptional networks as each gene can display a wide range of expression levels . Recent studies in liverwort support this hypothesis . The M . polymorpha encodes only one Aux/IAA and one member of each ancient ARF clade yet these transcription factors are sufficient to pattern a complete body plan ( Flores-Sandoval et al . , 2015; Kato et al . , 2015 ) . Unlike the activating ARFs , our understanding of the repressing ARFs is limited . For some repressing ARFs , the presence of an EAR domain suggests that repression could involve recruitment of the TPL co-repressor . However , many repressing ARFs , including PpARFb2 , PpARFb4 , and AtARF1 used in this work do not have this motif . We show that repression provided by these repressing ARFs is weaker compared to the repressing effect of the Aux/IAAs . Based on this observation , we suggest that repressing ARFs may fine-tune gene expression . The Aux/IAAs repress gene expression through interactions with co-repressors such as TPL and the subsequent recruitment of chromatin remodeling factors ( Kagale and Rozwadowski , 2011 ) . This may provide stable and long term repression when auxin levels are low . In contrast , when auxin levels are high and/or when auxin response needs to be dynamic , the repressing ARFs may provide a less stable repression that fine-tunes auxin response in the absence of the Aux/IAAs . The fact that the repressing ARF genes are induced by auxin and therefore constitute a negative feedback module is consistent with this idea . In flowering plants , auxin integrates light signals and growth responses ( Halliday et al . , 2009 ) . While we have some knowledge of how light affects auxin synthesis and distribution , the effects on auxin transcriptional responses are less known . In moss , the interplay between light , auxin , and growth has been implicated in the transition from chlronema to caulonema . In addition to auxin , light intensity and the availability of nutrients affect the transition from one cell type to the other . High light and glucose triggers caulonema formation while low light inhibits caulonemal growth and stimulates chloronemal branching ( Thelander et al . , 2005 ) . These observations suggest that favorable conditions promote the differentiation of elongated caulonemal cells , whereas low energy conditions promote the formation of chloroplast-rich chloronemal cells thus increasing photosynthetic capacity and energy production . We found that the auxin downregulated gene set is dominated by genes involved in photoperception and carbon fixation . This finding establishes the molecular basis for auxin signaling in response to light stimuli and suggests that auxin may function as a molecular switch between energy production and growth . Our model cannot explain a direct effect of the ARF transcription factors on downregulated genes . Only a few repressed genes have been identified after a short auxin treatment ( Chapman et al . , 2012; Paponov et al . , 2008 ) . Based on the GO analysis presented here it seems likely that transcription factors that are directly induced by auxin repress many processes downstream of auxin , particularly those associated with photosynthesis . These auxin-induced regulators and their downstream targets are yet to be defined . WT ‘Gransden-2004’ and mutant P . patens strains were grown at 25°C under continuous light at an intensity of 40–70 µmol/m2/s on BCD or BCDAT ( BCD supplemented with 5 mM Ammonium Tartrate ) media . Growth and differentiation of protonema is slower on BCDAT compared to BCD . For each experiment , a medium was selected to allow for analysis of either protonema development or filamentous and gametophyte differentiation , respectively . PCR-amplified DNA fragments were cloned into pENTR-D/TOPO ( Life Technologies ) , and were subsequently cloned either as digested fragments or by LR-Clonase-mediated recombination ( Life Technologies ) into the final vectors as detailed in the following procedures ( Primer pairs used for amplification are listed in Supplementary file 3 . backbone vectors are listed in Supplementary file 4 ) . Gene knockout constructs: Approximately one kilobase of genomic sequence upstream ( 5’ region ) and downstream ( 3’ region ) to the genes coding region were amplified and cloned into either pBNRF , pBNRF-GUS , or pBHRF2 as specified bellow: GenesPrimersCloning procedureBackboneIAA1A5’ regionPML61 , 62BamHI ligationpBHRF23’ regionPML59 , 60SpeI ligationIAA1B5’ regionPML52 , 53BamHI ligationpBNRF-GUS3’ regionPML59 , 60SpeI ligationIAA25’ regionPML63 , 64BamHI ligationpBNRF3’ regionPML65 , 66SpeI ligationARFb45’ regionPML599 , 600BamHI ligationpBNRF-GUS/ pBNRF3’ regionPML601 , 602SpeI ligationARFb25’ regionPML677 , 678BamHI ligationpBHRF23’ regionPML679 , 680SalI ligation Gene replacement construct: To create a mutant degron motif of IAA2 ( iaa2-P328S ) the PpIAA2 genomic region was amplified from the ppiaa2-183 mutant ( Prigge et al . , 2010 ) using the PpIAA2 genomic region-F’ and R’ primer pair . The fragment was subcloned as an AvrII and Bam HI fragment . Overexpression and inducible expression constructs: Constructs for protein-expression were generated by LR-Clonase-mediated recombination between pENTR-D/TOPO plasmids and destination vectors . To create PpARFb4 , PpARFa8 , and AtARF1 c-myc fusions , the genes’ coding regions were amplified using the primer pairs PML455 & 457 , PML461 & 463 , and PML776 & 777 respectively and recombined into pMP1377 . To create ARFa8 and glucocorticoid receptor fusion ( ARFa8-GR ) , the rat glucocorticoid receptor DNA fragment was inserted into the AscI site of ARFa8- pENTR-D/TOPO plasmid to create an ARFa8-GR fusion . The resulting recombinant fusion was recombined into pTHUBiGate . To create a construct carrying the auxin responsive marker ( DR5:DsRED2 ) eight repeats containing the DR5rev element ( ggGAGACAttt ) were inserted ahead of a minimal CaMV 35S promoter ( -50 to +26 ) . This promoter fragment was fused to the DsRED2 coding region by PCR then inserted together as a BamHI fragment into pMP1432 ( replacing the Hsp:Gateway cassette ) . Protoplast isolation and PEG-mediated transformation of P . patens was performed as described in ( Nishiyama et al . , 2000 ) . Three to five days after regeneration , transformants were selected on BCDAT medium containing 20 mg/l of either G418 or hygromycin , or 150 mg/l Gentamycin . Plants transformed with iaa2-P328S fragment were identified based on their phenotype on media containing 20 μM NAA . Transgenic knockout lines were screened by PCR for the presence of both left and right transgene-endogenous sequence junctions to verify the insertion of the transgene to the targeted locus and for the absence of the corresponding coding region . cDNA was detected by RT-PCR to confirm the absence of a transcribed product . Genomic DNA extracted from over- and inducible expression transgenic lines was detected by PCR to confirm the presence of the transgene , and recombinant protein expression was detected by immunoblot as described in Supplemental Experimental Procedures . iaa2-P328S lines were genotyped by derived Cleaved Amplified Polymorphic sequences ( dCAPS ) ( Table S7 ) . All knockout mutants expressed either nptII or Hygromycin selectable marker genes . To create higher order mutants , the selectable marker cassettes were first excised using the site-specific recombination Cre/lox system ( Schaefer and Zryd , 2001 ) to allow recycling of the selectable marker genes . The resulting lines are listed in the following table . DescriptionName in the textSourcePIAA1B:GUS/iaa1bΔPIAA1B:GUSThis workPIAA1B:GUS/iaa1bΔ iaa2Δiaa1b iaa2ΔThis workPIAA1B:GUS/iaa1bΔ iaa2Δ iaa1aΔaux/iaaΔThis workPARFb4:GUS/arfb4ΔThis workPARFb4:GUS/arfb2Δ arfb4ΔArfb2 arfb4ΔThis workarfb2ΔThis workPIAA1B:GUS/iaa1bΔ iaa2Δ arfb2ΔThis workPIAA1B:GUS/iaa1bΔ iaa2Δ arfb2Δ arfb4Δarfb2 arfb4Δ iaa1b iaa2This workPIAA1B:GUS/iaa1bΔ iaa2Δ iaa1aΔ arfb2Δ arfb4Δarf2b arf4b aux/iaaΔThis workPIAA2:iaa2-P328S G4 GH3:GUSiaa2-P328SThis workPUBi:PpARFa8-c-mycARFa8OEThis workPIAA1B:GUS/iaa1bΔ iaa2Δ iaa1aΔ PUBi:PpARFb4-c-mycARFb4OE_aux/iaaΔThis workPIAA1B:GUS/iaa1bΔ iaa2Δ iaa1aΔ PUBi:AtARF1-c-mycAtARF1OE_aux/iaaΔThis workPIAA1B:GUS/iaa1bΔ iaa2Δ iaa1aΔ PUBi:PpARFb4-c-myc PUBi:PpARFa8-GRARFa8-GR_ARFb4OE_aux/iaaΔThis workNLS4 ( Bezanilla et al . , 2005 ) DR5:dsRED NLS4DR5:dsREDThis workG4 GH3:GUS ( Bierfreund et al . , 2003 ) Protonemal tissue comprised mainly of chloronemata from WT and aux/iaa knockout mutant plants was homogenized in a Waring blender and plated in triplicate on BCDAT plates with cellophane overlays for seven days . The cellophane overlays covered with protonemal tissue were transferred into liquid BCD medium containing either 10 μM IAA or the equivalent amount of ethanol solvent and incubated at 25°C under continuous light for 5 hr . Total RNA was isolated using the RNeasy plant mini kit ( Qiagen ) and treated with DNA-freeTM DNAse removal kit ( Life technologies ) . The three biological replicates were sequenced on Illumina HiSeq2500 by New York Genome Center , resulting in 30 million 50 bp paired-end reads per sample . Total reads were mapped to the P . patens genome ( version 1 . 2 . 1 ) using RNAseq aligner STAR ( Dobin et al . , 2013 ) . Genes were quantified with featureCounts ( Liao et al . , 2014 ) , using v6 annotation the GTF file corresponding to the annotation v6 ( obtained from http://phytozome . jgi . doe . gov/ ) . Differential Expression was determined using the DESeq2 ( Love et al . , 2014 ) . The sequences reads have been deposited to the sequence Read Archive ( SRA ) database ( BioProject accession PRJNA317343 ) . GO enrichment analysis was carried out using AgriGO ( Du et al . , 2010 ) with GO annotation obtained from https://www . cosmoss . org . Hierarchical clustering was performed with DNASTAR . RNA interference was carried out as described in Bezanilla et al . ( 2005 ) . Overlapping PCR fragments from PpARFb4 ( PML506 , 507 ) and PpARFb3 ( PML508 , 509 ) were fused by PCR . Overlapping PCR fragments from PpARFb1 ( PML510 , 511 ) and PpARFb2 ( PML512 , 513 ) were fused by PCR . The resulting two fragments were fused together by PCR and cloned into pENTR-D/TOPO . The hairpin expression plasmids were generated by LR-Clonase-mediated recombination between the resulting pENTR plasmid and pUGGi . A pUGi vector lacking the gateway cassettes was used as control . The hairpin expression plasmids were transformed into DR5:DsRED strain which expresses a nuclear-localized GFP-GUS fusion protein . Following protoplast regeneration , the transformants , carried on cellophane overlays , were transferred to BCD media with Hygromycin and grown for eight days and then transferred into liquid BCD medium containing either 10 μM IAA or ethanol and incubated for 32 hr . Transformants lacking GFP fluorescence were selected and photographed . Fluorescence signals were detected for GFP ( excitation 488 nm , emission 493–535 nm ) , DsRED ( 561 nm excitation , 566–623 nm emission ) , and chlorophyll auto-fluorescence ( excitation 633 nm , emission 703–735 nm ) , using laser scanning confocal microscope ( Zeiss MLS 710 ) . Protonemal tissue was grown in triplicate or quadruplicates on BCDAT plates with cellophane overlays for seven days . For IAA or DEX treatment , plant tissue was transferred into liquid BCD medium containing either 10 μM IAA or 10 μM DEX or the equivalent amount of ethanol . Following incubation , the tissue was collected and total RNA was isolated using RNeasy plant mini kit ( Qiagen ) . 500 μg RNA was reverse transcribed using the Superscript III First Strand cDNA Synthesis System ( Life Technologies ) . 20 μl RT reaction was diluted with water to a final volume of 200 μL . PCR samples contained 4 μl diluted cDNA were detected using the CFX ConnectTM Real-Time PCR Detection System ( Bio-Rad ) . The following primer pairs were used to amplify the target genes: ARFa8 ( PML399 , 400 ) , ARFa6 ( PML409 , 419 ) , RSL4; Pp1s164_82V6 . 1 ( PML614 , 615 ) , GCN5; Pp1s20_204V6 . 1 ( PML618 , 619 ) , SAUR; Pp1s4_222V6 . 1 ( PML624 , 625 ) , AUX1; Pp1s56_28V6 . 1 ( PML626 , 627 ) , CBS domain; Pp1s11_325V6 . 1 ( PML810 , 811 ) , Dox; Pp1s15_259V6 . 1 ( PML812 , 813 ) , Epimerase; Pp1s189_34V6 . 3 ( PML814 , 815 ) , bHLH TF; Pp1s231_17V6 . 1 ( PML818 , 819 ) , Ubiquitin Ligase; Pp1s37_28V6 ( PML822 , 823 ) , IAA1A ( IAA1A-F’ , IAA1A-R’ ) , IAA1B ( IAA1B-F’ , IAA1B-R’ ) IAA2 ( IAA2-F’ , IAA2-R’ ) , EF1α; Pp1s84_186V6 . 1 ( EF1α-F’ , EF1α-R’ ) . The sequences are listed in Supplementary file 3 . Normalized expression ( ΔΔC ( t ) method ) was calculated using the Bio-Rad CFX manager software using PpEF1α as a reference gene and plotted as relative values ± SEM . Each analysis included three biological and four technical replicates . qPT-PCR analysis comparing gene expression levels between arfb2 arfb4Δ aux/iaaΔ and aux/iaaΔ lines included four biological replicates . cDNA fragments encoding ARFb4 and ARFa8 DNA binding domains were amplified using primer pairs PML 455 & 851 and PML 461 & 852 respectively , and cloned into pDEST15 ( Invitrogen ) . Recombinant GST fusions were expressed in Escherichia coli strain BL21-AI ( Invitrogen ) and purified by GST-agarose affinity . The eluted proteins were dialyzed in 100 mM Tris pH7 . 5 , 100mM KCl , 5 mM MgCl2 . Electrophoresis mobility shift assay was carried out using LightShift Chemiluminescent EMSA Kit ( Pierce , 20148 ) . DNA oligonucleotides were biotinylated using Biotin 3' End DNA Labeling Kit ( Thermo , 89818 ) , annealed and used as DNA probes . For each binding reaction 150 fmol of DNA probe was incubated with x protein at room temperature for 20 min in a final volume of 20 μl containing binding buffer ( 20 mM Tris pH 7 . 5 , 100 mM KCl , 2 mM DTT , 5% glycerol , 0 . 1 NP40 , 10 MgCl2 ) , 0 . 5 μg Poly ( dI•dC ) , and in the presence or the absence of excess molar ratio of specific or mutated unlabeled DNA competitor . The resulting protein-DNA complexes were electrophoresed on 5% native polyacrylamide gels , and then transferred to a Hybond N+ nylon membrane ( GE Healthcare ) . Biotin labeled DNA detection was carried out according to the LightShift Chemiluminescent EMSA Kit manufacturer’s instructions . Gene promoter regionSpecific DNA probeMutated oligonucleotideDR5PML 839 , 840PML 841 , 842ARFb4PML 1108 , 1109PML 1110 , 1111IAA1APML 1082 , 1083PML 1084 , 1085IAA1BPML 1068 , 1069PML 1094 , 1095 Following protoplast recovery , cellophane overlays were transferred to BCD plates . Same plants were imaged every day for five days using a Nikon SMZ1500 dissecting scope for five days and their length was measured using ImageJ . Following tissue homogenization , plants were grown for seven days on BCDAT plates with cellophane overlays and imaged by confocal microscope ( Zeiss MLS 710 ) . Cells proximal to filament branches were selected . The cell length and width of thirty cells from different confocal photos were measured using ImageJ . Plant tissue grown for seven days on BCDAT plates with cellophane overlays was extracted in 80% Acetone for 24 hr . Light absorbance of chlorophyll a and b were measured at wavelengths 663 and 646 nm using spectrophotometer and total chlorophyll concentration was calculated . Tissue was stained in GUS staining solution ( 50 mM NaH2PO4 ( pH7 . 0 ) , 0 . 5 mM X-Gluc , 0 . 5 mM K3Fe ( CN ) 6 , 0 . 5 mM K4Fe ( CN ) 6 , 0 . 05% Triton X-100 ) at 37°C for 5 hr following the auxin treatment of PIAA1B:GUS background lines . Plants were cleared in 70% ( v/v ) ethanol and imaged using a Nikon SMZ1500 dissecting scope . Proteins from protonemal tissue were extracted in 65 mM Tris pH6 . 8 , 2% SDS , and 10% glycerol . The resulting protein extract was centrifuged for 10 min at 10 , 000 g and the supernatant was collected . Proteins were resolved by SDS-PAGE and transferred onto nitrocellulose membranes . Membranes were stained with Ponceau-S to standardize the input and an HRP-conjugated monoclonal anti-c-Myc 9E10 ( Roche ) or anti-GR P-20 ( Santa Cruz ) antibodies were used for protein detection . Proteins were visualized using ECL Plus Western Blot Detection System ( GE healthcare ) . Full-length ARF protein sequences were extracted from the Physcomitrella patens , Selaginella moellendorffii , and Arabidopsis thaliana protein databases downloaded from the Joint Genome Institute plant genome website ( http://www . phytozome . org , accessed 14 Jan , 2014 ) . The sequences were aligned using T-Coffee ( Notredame et al . , 2000 ) , and poorly aligned regions were removed from the alignment . The tree was inferred using MrBayes [v3 . 2 . 2 x64; ( Ronquist et al . , 2012 ) ] with the following parameters: aamodelpr=mixed , nst = 6 , rates = invgamma , nruns = 2 , nchains = 4 , and ngen = 2000000 . The consensus tree was visualized and exported using FigTree ( http://tree . bio . ed . ac . uk/software/figtree/ ) . Gene identifiers are as follows: AtARF1 , At1g59750; AtARF2 , At5g62000; AtARF3_ETT , At2g33860; AtARF4 , At5g60450; AtARF5_MP , At1g19850; AtARF6 , At1g30330; AtARF7_NPH4 , At5g20730; AtARF8 , At5g37020; AtARF9 , At4g23980; AtARF10 , At2g28350; AtARF11 , At2g46530; AtARF12 , At1g34310; AtARF13 , At1g34170; AtARF14 , At1g35540; AtARF15 , At1g35520; AtARF16 , At4g30080; AtARF17 , At1g77850; AtARF18 , At3g61830; AtARF19 , At1g19220; AtARF20 , At1g35240; AtARF21 , At1g34410; AtARF22 , At1g34390; AtARF23 , At1g43950; PpARFa1 , Pp3c1_14480V3 . 1/Pp1s86_1V5_1; PpARFa2 , Pp3c1_14440V3 . 1/—; PpARFa3 , Pp3c2_25890V3 . 1/Pp1s119_25V5_1; PpARFa4 , Pp3c13_4720V3 . 1/Pp1s133_57V5_1; PpARFa5 , Pp3c26_11550V3 . 1/Pp1s6_230V5_3; PpARFa6 , Pp3c17_19900V3/Pp1s65_225V5 ( edited ) ; PpARFa7 , Pp3c14_16990V3 . 15/Pp1s48_142V5_1; PpARFa8 , Pp3c1_40270V3/Pp1s163_120V5_1 ( edited ) ; PpARFb1 , Pp3c27_60V3 . 1/Pp1s280_7V5_1; PpARFb2 , Pp3c16_6100V3 . 1/Pp1s341_4V5_1; PpARFb3 , Pp3c5_9420V3 . 1/Pp1s64_136V5_1; PpARFb4 , Pp3c6_21370V3 . 1/Pp1s14_378V5_1; PpARFc1A , Pp3c4_12970V3 . 1/Pp1s339_45V6_1; PpARFc1B , Pp3c4_13010V3 . 1/—; PpARFc2 , Pp3c6_26890V3 . 1/Pp1s279_8V5_1; PpARFd1 , Pp3c9_21330V3 . 1/Pp1s316_22V6_1; PpARFd2 , Pp3c15_9710V3 . 1/Pp1s250_62V6_1; SmARFa1 , Selmo117217 ( edited ) ; SmARFa2 , Selmo424114 ( edited ) ; SmARFa3 , Selmo181406 ( edited ) ; SmARFb1 , Selmo437944 ( edited ) ; SmARFb2 , Selmo81992 ( edited ) ; SmARFc1 , Selmo61688 ( edited ) ; SmARFc2 , Selmo51695 ( edited ) ; SmARFd1 , Selmo1d28604 ( edited ) ; and SmARFd2 , Selmo1d115320 ( edited ) .	Auxin is a plant hormone that regulates many aspects of growth and development . It does so by promoting the degradation of proteins called the Aux/IAAs . These proteins normally act to keep genes switched off by interacting with another family of proteins called ARFs that can bind directly to the genes . Some ARFs activate genes , while others repress gene activity . In most plants , large families of genes encode Aux/IAA and ARF proteins and individual members often have overlapping roles , which makes it more difficult to study how they work . To avoid this problem , Lavy et al . chose to study how these proteins work in a moss called Physcomitrella patens , which only has three genes that encode Aux/IAA proteins . The experiments show that the loss of all three Aux/IAA proteins results in the plants becoming completely insensitive to auxin . Furthermore , over a third of known moss genes had altered activity in the mutant plants compared to normal moss . These findings suggest that the Aux/IAA proteins have a bigger role in regulating the activities of genes than previously thought . Further experiments show that repressive ARFs act by directly competing with the activating ARFs for binding to auxin-regulated genes . However , repression of gene activity by the ARFs is weaker than the effect of the Aux/IAAs . Unexpectedly , the loss of repressive ARFs causes moss plants to become less sensitive to auxin , which Lavy et al . suggest is due to the recruitment of additional Aux/IAA proteins to the genes . Thus these findings demonstrate that control of gene activity by auxin involves the coordinated action of both types of ARFs and the Aux/IAAs . Future challenges are to find out which genes directly bind ARFs and the Aux/IAAs , and to use computational approaches to create models of how auxin regulates gene activity in the moss .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "plant", "biology" ]	2016	Constitutive auxin response in Physcomitrella reveals complex interactions between Aux/IAA and ARF proteins
The primary cilium has an important role in signaling; defects in structure are associated with a variety of human diseases . Much of the most basic biology of this organelle is poorly understood , even basic mechanisms , such as control of growth and resorption . We show that the activity of the anaphase-promoting complex ( APC ) , an E3 that regulates the onset of anaphase , destabilizes axonemal microtubules in the primary cilium . Furthermore , the metaphase APC co-activator , Cdc20 , is specifically recruited to the basal body of primary cilia . Inhibition of APC-Cdc20 activity increases the ciliary length , while overexpression of Cdc20 suppresses cilium formation . APC-Cdc20 activity is required for the timely resorption of the cilium after serum stimulation . In addition , APC regulates the stability of axonemal microtubules through targeting Nek1 , the ciliary kinase , for proteolysis . These data demonstrate a novel function of APC beyond cell cycle control and implicate critical role of ubiquitin-mediated proteolysis in ciliary disassembly . The primary cilium is a non-motile microtubule-based organelle , organized by the mother centriole ( basal body ) . Though known for over a century , it received little interest until it was appreciated that it is highly enriched in receptors for various signaling pathways , including PDGF , calcium , receptor tyrosine kinases , Wnt , and Hh ( Satir et al . , 2010 ) . The assembly of primary cilium is coupled to cell cycle exit and entry into quiescence . The primary cilium functions as a tumor suppressor organelle that regulates cell proliferation and differentiation ( Pan et al . , 2012 ) . Dysfunctions of the primary cilium cause a set of human diseases , classified as ciliopathies , such as polycystic kidney diseases ( Hildebrandt et al . , 2011 ) . Cancers are also associated with the loss of cilia , due to misregulated cell proliferation through ciliary signaling ( Plotnikova et al . , 2008 ) . Ciliopathies are highly associated with defective ciliogenesis and abnormal ciliary length . The assembly/disassembly of cilia and maintenance of ciliary length are tightly regulated through control of intraflagellar transport ( IFT ) and axoneme modification ( Avasthi and Marshall , 2011 ) . Ciliary length is regulated by the balance of the bidirectional transport of microtubule subunits and associated protein complexes through IFT ( Pedersen and Rosenbaum , 2008 ) . Importantly , the extension/resorption of primary cilia also involves changes of the stability of axonemal microtubules , which are controlled by post-translational modifications . For example , Aurora kinases stimulate disassembly of cilia and flagella by phosphorylation and activation of tubulin deacetylase ( Pan et al . , 2004; Pugacheva et al . , 2007 ) . Another key post-translational modification during ciliogenesis is ubiquitylation . In Chlamydomonas flagella , there is an increase in the levels of protein ubiquitylation during flagellar resorption ( Huang et al . , 2009 ) . However , the precise mechanism for how ubiquitylation might control the process of ciliogenesis is not known . It is curious that many mitotic regulators have been found to be involved in ciliogenesis , such as Aurora A kinase ( Pugacheva et al . , 2007 ) and NIMA-related kinases ( Moniz et al . , 2011 ) . Recently , it was shown that anaphase-promoting complex ( APC ) , the key ubiquitin E3 ligase that governs cell cycle progression is localized to the basal body of the motile cilia in multiciliated Xenopus epidermal cells , where it has some role in regulating ciliary polarity ( Ganner et al . , 2009 ) . In mitosis , APC sequentially recruits co-activators ( Cdc20 followed by Cdh1 ) and cooperates with two specific E2s ( UBCH10 and Ube2S ) to control proteolysis of key regulators of the metaphase to anaphase transition ( Fujita et al . , 2009; Fujita et al . , 2009 ) . APC also has a less well understood role in regulating other cellular processes , such as neurogenesis , metabolism , and myogenesis ( Kim et al . , 2009; Eguren et al . , 2011 ) . In mitosis , the activity of APC-Cdc20 is tightly controlled at multiple levels . Regulation of phosphorylation of APC subunits and Cdc20 is essential for the timing of APC activation during mitotic progression ( Kramer et al . , 2000 ) ; APC-Cdc20 is also modulated by its inhibitor proteins ( Luo et al . , 2000; Madgwick et al . , 2006 ) . In addition , protein levels of APC cognate E2s , UBCH10 and Ube2S are crucial for the activation of APC-Cdc20 , and play a role in the override of mitotic arrest ( Reddy et al . , 2007; Garnett et al . , 2009 ) . Although the function and regulation of APC-Cdc20 during mitosis and meiosis is thoroughly studied ( Kramer et al . , 2000; Madgwick et al . , 2006 ) , the non-mitotic roles of APC-Cdc20 remain largely unknown . It was recently found that APC-Cdc20 is localized to the centrosome of postmitotic neurons and plays an important role in dendrite morphogenesis ( Kim et al . , 2009 ) . It would therefore be of interest to explore the potential activity of Cdc20 in quiescent and differentiated cells . All of this circumstantial evidence suggests that there may be a value in investigating whether APC has a broad regulatory role in the primary cilium . We report here that APC-Cdc20 is central to the regulation of the primary cilium . The APC is localized to the basal body of primary cilia of human epithelial cells , where it negatively controls the length and stability of the axonemal microtubules . During exit from the quiescent ( G0 ) state APC-Cdc20 is activated and drives ciliary resorption . Given the well-established role for APC in dividing cells , we are interested in whether APC is still active and what role it might have in quiescent ciliated cells . Human hTERT-RPE1 cell line is an established model to study the assembly/disassembly of primary cilia ( Pugacheva et al . , 2007 ) . Greater than 80% of RPE1 cells are ciliated after serum-starvation for 48 hr , and disassembly of primary cilia occurs 1–2 hr after re-adding serum . We found that APC is still active in cytoplasmic extracts of quiescent ciliated RPE1 , using the canonical APC substrate Securin ( Figure 1—figure supplement 1 ) . Immunostaining showed that APC subunit 2 ( APC2 ) is localized to the basal body of the ciliated cells ( Figure 1A and Figure 1—figure supplement 2A ) , consistent with the previous report that APC is localized to the basal body of motile cilia in the Xenopus epidermis ( Ganner et al . , 2009 ) . Co-localization of APC2 with dynactin subunit P150 ( Guo et al . , 2006 ) further confirmed its localization to the mother centriole ( Figure 1—figure supplement 2B and 2C ) . Furthermore , we found that the APC co-activator Cdc20 is also localized to basal body of the primary cilium ( Figure 1B ) , whereas Cdh1 did not show such localization ( Figure1—figure supplement 3 ) . Interestingly , Cdc20 was not observed at the centrosome of cycling interphase non-ciliated cells ( Figure 1B and Figure1—figure supplement 2D ) , indicating that Cdc20 is specifically recruited to the basal body during ciliogenesis , rather than generally binding to the interphase centrosome . Western blotting showed that Cdc20 protein levels are dramatically reduced after serum starvation for 24 hr , due to the exit of cells from proliferation ( Figure 1—figure supplement 2E ) . Importantly , Cdc20 protein levels were significantly elevated at 48 hr after serum starvation , when >80% of the cells are ciliated , consistent with the appearance of this protein at the basal body . Thus , our data suggest that APC-Cdc20 may have a special role at the basal body after initiation of ciliogenesis . 10 . 7554/eLife . 03083 . 003Figure 1 . APC-Cdc20 is localized to the basal body of the primary cilium . ( A ) Subconfluent hTERT-RPE1 cells were serum starved for 48 hr , fixed , and stained for APC subunit 2 ( APC2 , red ) , acetylated tubulin ( green ) and DNA ( blue ) . ( B ) Proliferating interphase RPE1 cells , serum starved cells , and quiescent cells after serum stimulation were fixed and stained for Cdc20 ( red ) , acetylated tubulin ( green ) , and DNA ( blue ) . Boxes in main images indicate structures shown at higher magnification to right . The scale bars in this figure represent 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 00310 . 7554/eLife . 03083 . 004Figure 1—figure supplement 1 . Degradation of securin in RPE1 cell extracts . Cytosolic cell extracts were made from unsynchronized proliferating , serum-starved RPE1 cells respectively . Degradation of securin in these cell extracts at indicated time points was analyzed by western blot . TAME was added to test the dependence of the degradation on APC activity . Cell extract immunodepleted of Cdc20 was used for degradation assay as comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 00410 . 7554/eLife . 03083 . 005Figure 1—figure supplement 2 . Localization of APC-Cdc20 to the basal body of primary cilia . ( A ) Non-ciliated RPE1 cells ( left ) and ciliated cells ( right ) after serum starvation were stained for APC2 ( red ) , acetylated tubulin ( green ) , and DNA ( blue ) . ( B ) Subconfluent hTERT-RPE1 cells were serum starved for 48 hr , fixed , and stained for APC subunit2 ( APC2 , red ) and dynactin subunit P150 ( green ) , and DNA ( blue ) . ( C ) Starved RPE1 cells were fixed and stained for centrin 1 ( red ) , dynactin subunit P150 ( green ) , and DNA ( blue ) . ( D ) Proliferating interphase cells and serum-starved cells were stained for Cdc20 ( red ) and γ-tubulin ( green ) . The scale bars in this figure represent 10 μm . ( E ) RPE1 cells were collected at indicated time points during starvation and post serum stimulation and subjected to western blot for Cdc20 and GAPDH respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 00510 . 7554/eLife . 03083 . 006Figure 1—figure supplement 3 . Cdh1 is not localized to the basal body of primary cilia . Starved RPE1 cells were fixed and stained for acetylated tubulin ( green ) , DNA ( blue ) , and Cdh1 ( red ) . For Immunostaining of endogenous Cdh1 , three different antibodies against Cdh1 were used as depicted . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 006 We then tested whether Cdc20 is required for some aspect of ciliogenesis . Reduction of APC subunit 2 ( APC2 ) by siRNA significantly increased the length of primary cilia marked by acetylated tubulin staining ( Figure 2A–C ) . Consistently , the depletion of Cdc20 by siRNA had no effect on the percentage of ciliated cells , but there was a significant increase in the ciliary length ( Figure 2D–F ) . Negative Ki-67 staining demonstrated that these ciliated cells after the knockdown of either APC2 or Cdc20 are still in quiescent stage ( Figure 2—figure supplement 1A ) . In contrast , knockdown of Cdh1 by siRNA gave no obvious ciliary phenotype ( Figure 2—figure supplement 2A ) . This result is contradictory to a previous study proposing that APC-Cdh1 promote ciliogenesis ( Miyamoto et al . , 2011 ) . Reciprocally , we asked what the effect would be for overexpression of Cdc20 . When we transiently transfected cells with plasmid encoding Cdc20-GFP , we found it localized to the centrosome of quiescent cells , and completely suppressed the formation of primary cilia ( Figure 2G , H ) . This phenotype was confirmed by Immunostaining of Arl13b , another specific marker for ciliary axoneme ( Figure 2—figure supplement 2B ) . These results suggested that Cdc20 negatively controls the length of the primary cilium . 10 . 7554/eLife . 03083 . 007Figure 2 . APC-Cdc20 negatively regulates the length of primary cilia . ( A and D ) RPE1 cells were treated with control siRNA or siRNA targeting APC2 or Cdc20 respectively , serum starved for 48 hr , and stained for acetylated tubulin ( green ) and DNA ( blue ) . The scale bar represents 10 μm . ( B and E ) Western blot showed knockdown efficiency of siRNA treatment against Cdc20 and APc2 . ( C and F ) Ciliary length was measured from ciliated cells based on acetylated-tubulin staining ( n > 30 ) . Data are means ±S . D . p <0 . 005 . ( G ) RPE1 cells were transiently transfected with vectors expressing GFP ( green ) or Cdc20-GFP ( green ) respectively , serum starved for 48 hr , and were stained for acetylated tubulin ( red ) and DNA ( blue ) . The scale bar represents 5 μm . ( H ) Percentage of ciliated cells was obtained from five independent experiments . An average of 100 cells was counted in each of the experiments . Data are means ±S . D . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 00710 . 7554/eLife . 03083 . 008Figure 2—figure supplement 1 . The cells with longer cilia after inhibition of APC-Cdc20 are still in the quiescent stage . ( A ) Proliferating PRE1 cells , starved cells transfected with siRNA against APC2 , Cdc20 , or Ube2S were fixed and stained for ace-tubulin ( red ) , DNA ( blue ) , and Ki-67 ( green , the proliferation marker ) . ( B ) RPE1 cells were serum-starved for 60 hr and further starved for 3 hr in the presence of DMSO or 20 μM proTAME before staining for acetylated-tubulin ( red ) , Ki-67 ( green ) , and DNA ( blue ) . The scale bars represent 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 00810 . 7554/eLife . 03083 . 009Figure 2—figure supplement 2 . The effects of deregulation of APC co-activators on ciliogenesis . ( A ) RPE1 cells were treated with control siRNA or siRNA-targeting Cdh1 respectively , serum starved for 48 hr . Samples were collected for western blot for Cdh1 and GAPDH . Ciliary length was measured from ciliated cells based on acetylated-tubulin staining ( n > 30 ) . Data are means ±S . D . ( B ) RPE1 cells were transiently transfected with control vector or vector expressing Cdc20-GFP ( green ) , serum starved for 48 hr , and stained for Arl13b ( red ) and DNA ( blue ) . The scale bar represents 5 μm . Percentage of ciliated cells was calculated in control and Cdc20-GFP transfected cells . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 009 To further validate the effect of Cdc20 , we used a small molecule inhibitor of APC , proTAME . Previous studies showed that cell-permeable proTAME efficiently inhibits the APC activity through disruption of the association of APC with its co-activators ( Zeng et al . , 2010; Zeng and King , 2012 ) . The advantage of small molecule inhibition over RNAi is that proTAME exerts its inhibitory effect immediately after drug treatment , which helps differentiate the ciliary phenotype from the possible secondary effect caused by impaired cell cycle control . The length of preformed cilia was increased from 4µm to 7µm after treatment of fully ciliated quiescent cells with proTAME for 3 hr ( Figure 3 ) , consistent with the Cdc20 knockdown or APC2 knockdown phenotype . Negative Ki-67 staining shows that the ciliated cells are still in quiescent stage after short-term treatment with this drug ( Figure 2—figure supplement 1B ) . This result supports the view that the activity of APC is required for maintaining the proper length of primary cilia . 10 . 7554/eLife . 03083 . 010Figure 3 . Treatment of ciliated cells with proTAME increased the ciliary length . RPE1 cells were serum-starved for 60 hr and further starved for 3 hr in the presence of DMSO or 20 μM proTAME before staining for acetylated-tubulin ( green ) and DNA ( blue ) . Ciliary length was measured from ciliated cells based on acetylated-tubulin staining ( n > 30 ) . Data are means ±S . D . p <0 . 001 . The scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 010 These observations suggested that activation of APC-Cdc20 might be responsible for the destabilization and shortening of ciliary axoneme when quiescent cells re-enter the cell cycle . To test this hypothesis , we treated serum-starved cells with proTAME when adding serum; serum causes ciliary resorption and subsequent re-entry into the cell cycle . Primary cilia on DMSO-treated control cells significantly shortened and resorbed 4 hr after serum addition . ProTAME treatment delayed the shortening and resorption of cilia ( Figure 4A , B ) , demonstrating that APC activity is required for efficient ciliary resorption before the exit of cells from quiescent phase . Suppression of either Cdc20 or APC2 by siRNA also significantly delayed ciliary shortening ( Figure 4—figure supplement 1 ) . Calcium influx is known to be necessary and sufficient for shortening and resorption of the primary cilium ( Besschetnova et al . , 2010; Plotnikova et al . , 2012 ) , and we found that proTAME treatment also prevented ciliary shortening induced by Calcium ionophore treatment ( Figure 4—figure supplement 2A and 2B ) . Based on these results , APC-Cdc20 at the basal body has two related functions: it maintains cilium length of quiescent cells; it is required for shortening and resorption of the cilium . 10 . 7554/eLife . 03083 . 011Figure 4 . Inhibition of APC activity prevented proper ciliary resorption post serum stimulation . ( A ) Quiescent ciliated RPE1 cells were treated with serum containing either DMSO or 20 μM proTAME for 4 hr , fixed and stained for acetylated tubulin ( green ) and DNA ( blue ) . Scale bar: 10 μm . ( B ) Quantitation of ciliary length and ciliation of DMSO or proTAME treated cells after serum stimulation . Ciliary length was measured from ciliated cells ( n > 30 ) ( p <0 . 001 ) ; the percentage of ciliated cells was obtained from five independent experiments . Data are means ±S . D . ( C ) Cells were treated with control siRNA or Ube2S targeting siRNA and were serum starved for 2 days in the presence of siRNA . Cilia were quantitated after release of cells from starvation for 4 hr . Ciliary length was measured from ciliated cells ( n > 30 ) ( p <0 . 001 ) ; the percentage of ciliated cells was obtained from five independent experiments . Data are means ±S . D . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 01110 . 7554/eLife . 03083 . 012Figure 4—figure supplement 1 . Suppression of endogenous Cdc20 or APC2 by siRNA delayed cilium resorption . Cells were transfected with control siRNA or Cdc20/APC2 targeting siRNA and serum starved for 2 days in the presence of siRNA . 4 hr after serum stimulation , these cells were fixed and stained for acetylated tubulin ( red ) , γ-tubulin ( green ) , and DNA ( blue ) . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 01210 . 7554/eLife . 03083 . 013Figure 4—figure supplement 2 . Inhibition of APC activity prevented cilia resorption caused by Calcium influx . ( A and B ) Starved RPE1 cells were treated with DMSO , 1 μg/ml calcium ionophore ( ion ) , or 1 μg/ml calcium ionophore +20 μM proTAME without serum for 3 hr . Cilia were visualized by acetylated tubulin staining ( green ) and DAPI staining of DNA ( blue ) . Scale bar: 10 μm . Ciliary length was measured from ciliated cells ( n > 30 ) . ( C ) RPE1 cells were transfected with control siRNA or Ube2S targeting siRNA , serum starved for 2 days , treated with 1 μg/ml calcium ionophore for 3 hr . Ciliary length was measured from ciliated cells ( n > 30 ) . Data are means ±S . D . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 01310 . 7554/eLife . 03083 . 014Figure 4—figure supplement 3 . RNAi suppression of endogenous Ube2S increased the ciliary length . ( A ) siRNA-transfected cells were subjected to western blot for Ube2S . ( B ) siRNA-transfected starved cells were stained for acetylated tubulin ( green ) and DNA ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 014 Regulation of APC in mitosis employs a special E2 enzyme , Ube2S , which elongates ubiquitin chains on a substrate through specific linkages at the K-11 position of ubiquitin ( Williamson et al . , 2009; Wu et al . , 2010 ) . Ube2S keeps APC partially active during mitotic arrest and is required to fully activate APC-Cdc20 ( Garnett et al . , 2009 ) . We were interested in whether Ube2S is also involved in regulation of APC-Cdc20 during ciliogenesis . We found that depletion of Ube2S increased the cilium length in quiescent cells ( Figure 4—figure supplement 3 ) and delayed ciliary shortening and resorption upon either serum stimulation or calcium influx ( Figure 4C; Figure 4—figure supplement 2C ) . These data imply that the cell utilizes the general processes for regulation of APC-Cdc20 activity for both mitosis and ciliogenesis . In mitosis APC ubiquitylates specific regulatory molecules like cyclin B and securin that control the processes of the metaphase–anaphase transition . Similarly , there should be substrates of APC that might regulate the stability of the ciliary axoneme . Aurora-A , a known APC substrate during late mitosis , plays an important role in regulating disassembly of primary cilium ( Pugacheva et al . , 2007 ) . We showed that protein levels of Aurora-A do not decrease upon serum stimulation when APC-Cdc20 is activated ( Figure 5—figure supplement 1A ) , consistent with the previous report ( Pugacheva et al . , 2007 ) . This may be explained by previous studies showing that during ciliogenesis , Aurora-A is autophosphorylated at S51 ( Plotnikova et al . , 2012 ) , which protects it from APC-mediated degradation ( Littlepage and Ruderman , 2002; Littlepage et al . , 2002 ) . Accordingly , the knockdown of Cdc20 in quiescent cells did not affect Aurora-A protein levels ( Figure 5—figure supplement 1B ) . It was indicated by recent studies that APC may regulate ciliogenesis through degradation of dishevelled protein ( Ganner et al . , 2009; Miyamoto et al . , 2011 ) . However , both specific ubiquitylation assay and in vitro degradation assay showed that dishevelled is not an authentic substrate of APC ( Figure 5—figure supplement 1C and 1D ) , contradictory to their argument . Nek1 , the founding member of the NIMA-related kinase family was a good candidate . Genetic mutation in Nek1 was highly associated with short-rib polydactyly syndrome Type Majewski , a new ciliopathy ( Thiel et al . , 2011 ) , and Nek1 is localized to the basal body region ( Shalom et al . , 2008 ) . Both human and mouse fibroblast cells with Nek1 mutations have cilia with abnormal branched structures ( Shalom et al . , 2008; Thiel et al . , 2011 ) . We found that the depletion of endogenous Nek1 by siRNA in serum-starved RPE1 cells generated cilia with branched structures , revealed by acetylated-tubulin staining ( Figure 5A–C and Figure 5—figure supplement 2A ) , indicating that Nek1 controls the stability and integrity of axonemal microtubules . Nek1 knockdown in proliferating RPE1 cells did not affect the structure of either cytoplasmic-acetylated tubulin in microtubules or centrosomal gamma-tubulin ( Figure 5—figure supplement 2C ) , suggesting a specific and confined role of Nek1 in ciliogenesis . After serum addition , the branched structure can still be efficiently resorbed ( Figure 5—figure supplement 2B ) , consistent with the dynamics of ciliary disassembly , indicating that the branched structure is ciliary axoneme . We found that Nek1 levels decreased after serum stimulation ( Figure 5D ) , suggesting that the degradation of Nek1 could be essential for triggering the destabilization of the primary cilium . Finally , Nek1 also has a canonical destruction box ( RXXLXXXN ) , the APC recognition motif , right after its coiled-coil domain . We showed by degradation assays in vitro that Nek1 is degraded in HeLa cell extracts , and that this degradation can be suppressed by Emi1 , a specific APC inhibitor protein ( Figure 5E ) . Nek1 is efficiently ubiquitylated by APC in in vitro ubiquitylation assay involving purified APC ( Figure 5F ) . Furthermore , overexpression of Cdc20 leads to reduction of Nek1 levels in quiescent RPE1 cells ( Figure 5G ) . These results validate Nek1 as an APC substrate . Moreover , simultaneous depletion of Cdc20 and Nek1 partially rescued the branched cilia phenotype caused by Nek1 knockdown ( Figure 5C ) , suggesting that APC-Cdc20 regulates the stability of ciliary axoneme at least in part through antagonizing Nek1 activity . 10 . 7554/eLife . 03083 . 015Figure 5 . Nek1 is the substrate of APC-Cdc20 for the regulation of the structure of primary cilia . ( A ) RPE1 cells were treated with control siRNA or Nek1 targeting siRNA , serum starved for 48 hr , and stained for acetylated tubulin ( red ) , γ-tubulin ( green ) and DNA ( blue ) . Scale bar represents 5 μm . ( B ) siRNA-treated cells were subjected to Western blot for endogenous Nek1 and GAPDH . ( C ) Cells were transfected with control siRNA , Nek1 siRNA , or Nek1 siRNA together with Cdc20 siRNA respectively , serum starved for 2 days , and stained for acetylated tubulin . The percentage of cells with normal cilia or branched abnormal cilia was calculated from three independent experiments . Data are means ±S . D . ( D ) Western blot showed the decrease of Nek1 protein levels after serum stimulation . ( E ) Degradation of 35S-labelled Nek1 was examined in cell extracts prepared from HeLa S3 cells . Emi1 was added to test if the degradation was dependent on APC activity . The data were analyzed by autoradiography . ( F ) HA-tagged Nek1 was expressed and labeled with 35S in reticulocyte lysate . 35S-labelled Nek1 was purified through HA tag and subjected to in vitro ubiquitylation catalyzed by purified APC . ( G ) RPE1 cells were transfected with plasmid expressing HA-Nek1 and/or Myc-Cdc20 , starved for 2 days and subjected to western blot for Myc tag , HA tag , and GAPDH . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 01510 . 7554/eLife . 03083 . 016Figure 5—figure supplement 1 . Test of ciliary proteins as APC substrates . ( A ) RPE1 cells were starved for 48 hr , samples were collected at indicated time point after serum addition , and were subjected to western blot for Aurora-A and GAPDH . ( B ) RPE1 cells were transfected with control siRNA or siRNA against Cdc20 , samples were collected after serum starved for 48 hr or 1 hr post serum addition and subjected for western blot for Aurora-A , Cdc20 , and GAPDH . ( C ) 35S-labelled dishevelled protein ( DVL ) and securin were incubated in the HeLa S3 cell extracts . Samples were collected at indicated times and analyzed by autoradiography . Emi1 was added in the reaction to test whether the degradation was APC dependent . ( D ) 35S-labelled dishevelled and securin were subjected to in vitro ubiquitylation assay catalyzed by purified APC , recombinant Cdh1 and UBCH10 . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 01610 . 7554/eLife . 03083 . 017Figure 5—figure supplement 2 . Depletion of Nek1 caused specific microtubule defects in primary cilia . ( A ) RPE1 cells were transfected with control siRNA and Nek1 targeting siRNA respectively , serum starved for 2 days , and stained for acetylated tubulin ( red ) , Arl13b ( green ) , and DNA ( blue ) . ( B ) RPE1 cells transfected with Nek1 siRNA were starved for 2 days , fixed at 0 and 3 hr post serum addition , and stained for DNA ( blue ) and acetylated tubulin ( red ) . ( C ) Proliferating RPE1 cells were transfected with siRNA in the presence of serum and stained for acetylated tubulin ( red ) , γ-tubulin ( green ) , and DNA ( blue ) . Scale bars represent 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03083 . 017 The assembly , maintenance , and disassembly of the primary cilium are regulated by the control of intraflagellar transport and protein posttranslational modifications ( Besschetnova et al . , 2010 ) . Impairment of ciliary length control and ciliary assembly can lead to inappropriate cell proliferation and defects in signal transduction , causing ciliopathies and perhaps tumorigenesis . Though more and more is understood about which proteins control ciliary function , much of it through human genetics , much less is known about regulatory features that control the timing of ciliogenesis , and with respect to our work here , ciliary resorption . Surprisingly , some of the machinery for regulation is very similar to that used in regulation of the mitotic cycle , which depends significantly on posttranslational modifications . In particular the mitotic process in cells is controlled from the top by the activation of a kinase , Cdk1-cyclin B and an E3 component of the ubiquitin pathway , APC-Cdc20 and APC-Cdh1 . We have found that the same Cdc20 is recruited to the basal body of the primary cilium , where in complex with APC it negatively regulates ciliary length , by destabilizing axonemal microtubules . Following serum stimulation of quiescent cells , APC-Cdc20 is fully activated leading to ciliary shortening and resorption . A previous study in this area claimed that APC-Cdh1 promoted the assembly of primary cilia; a conclusion completely contrary to our results . Although they found , as we did ( Figure 2D–F ) , that Cdc20 knockdown promotes ciliogenesis , they explained it by assuming that Cdc20 is an inhibitor of APC-Cdh1 , and that Cdc20 needs to be degraded for the activation of APC-Cdh1 to promote the assembly of cilia ( Miyamoto et al . , 2011 ) . They also proposed that APC-Cdh1 , but not Cdc20 , specifically degrades dishevelled during ciliogenesis . There are several problems with their data and conclusions . We found that Cdc20 is recruited to the basal body of fully formed primary cilia , and there is no evidence for the presence of Cdh1 at the basal body ( Figure 1B and Figure 1—figure supplement 3 ) . Knockdown of Cdh1 does not inhibit the assembly of cilia , contradictory to their model ( Figure 2—figure supplement 2A ) . In addition , depletion of APC subunit ( APC2 ) or treatment of ciliated cells with proTAME , which inhibits APC through promoting dissociation of both Cdc20 and Cdh1 , increases ( and not decreases ) ciliary length ( Figure 2A–C and Figure 3 ) , consistent with the Cdc20 knockdown phenotype . There is no precedent for Cdc20 functioning as an inhibitor of Cdh1 in mitosis . Instead , APC-Cdh1 ubiquitylates Cdc20 , leading to its rapid disappearance during anaphase ( Pfleger and Kirschner , 2000 ) . Finally , we showed that dishevelled is not an authentic substrate of APC-Cdh1 using both specific ubiquitylation assays and degradation assays in vitro ( Figure 5—figure supplement 1C and 1D ) . Thus , our data argue against their model and instead point to the role of APC-Cdc20 in negatively regulating ciliogenesis . Furthermore , our data show that APC-Cdc20 counteracts the Nek1 activity to regulate the stability of the ciliary axoneme during maintenance and resorption of cilia . Depletion of Nek1 resulted in branched axonemal microtubules . An explanation for this phenotype is that the function of Nek1 is to stabilize axonemal microtubules . Thus , depletion of Nek1 leads to the destabilization and disassembly of axonemal subunits . It would appear that the disassembled axonemal subunits cannot be transported out of cilia efficiently , and instead they are randomly reassembled in the ciliary lumen , resulting in branched and frayed structures . Depletion of Nek1 also results in defects in ciliary signaling , revealed by the absence of Arl13b staining ( Figure 5—figure supplement 2A ) . This indicates that proper regulation of APC-Cdc20 activity is also essential for normal cellular signaling through primary cilia . We note that APC may control the degradation of a series of ciliary proteins besides Nek1 , as this E3 ligase does during mitosis . It will be interesting to try to identify other known and unknown APC substrates in the primary cilium and determine how they are regulated during the process of elongation and resorption . The regulation of APC activity during mitosis and meiosis is controlled by the activity of kinases , such as Cdk1 ( Kramer at al . , 2000; Madgwick et al . , 2006 ) . Our preliminary data showed that suppression of Cdk activity increased the ciliary length ( unpublished data ) , consistent with the role of Cdk1 in stimulating APC activity as in mitosis/meiosis . It is also possible that some protein inhibitor of APC is involved in ciliogenesis , regulating the suppression and activation of APC activity during ciliary maintenance and disassembly respectively . It is worthwhile to further dissect the detailed mechanism underlying this process in the future study . The regulation of the primary cilium has assumed a great importance in overall cellular regulation and disease . The assembly/disassembly of primary cilium is under strict homeostatic control , and it is responsive to cell cycle regulation . The regulatory apparatus bears great similarities to the posttranslational control in the mitotic cycle . Despite the known presence of many of the same components that were thought to be quintessential components of the mitotic engine , it was not clear that their function bears strong similarities to their use in the cell cycle . Central to mitosis is this is the anaphase-promoting complex and its metaphase/anaphase regulator Cdc20 . It acts as a master regulator of downstream kinases and phosphatases but it also targets terminal protein substrates . The primary cilium is thought to be a concentrated signaling system whose length may therefore act as a rheostat controlling a complement of signaling activities in cells . It therefore becomes attractive to think of ways of perturbing ciliary length . As more and more evidence accumulates documenting the connection between defects in ciliogenesis and human diseases , there could be a growing interest need for new therapies that modulate ciliary length and ciliary disassembly . An understanding of the central role of the APC in control of ciliary length could aid in the identification of new drug targets that may have broad applicability . hTERT-RPE1 cells were grown in Dulbecco's Modified Eagle Medium ( DMEM ) with 10% fetal bovine serum ( FBS ) . For analysis of ciliary assembly , cells were plated at subconfluence on glass coverslips and starved for 48 hr in DMEM without serum to induce cilia formation . To induce ciliary resorption , starved ciliated cells at subconfluence were treated with DMEM medium with 10% serum for 1–4 hr . Plasmids were transiently transfected into cells with FuGENE 6 Transfection Reagent ( Promega , Madison , WI ) . For siRNA treatment , cells were transfected with indicated siRNA by Lipofectamine RNAiMAX Transfection Reagent ( Life Technologies , Beverly , MA ) . Plasmids from the laboratory stock are as follows: human securin , human HA-Nek1 , GFP , human Cdc20-GFP , human cMyc-Cdc20 , and human DVL in pCS2 vector . SMARTpool ON-TARGETplus siRNA for Nek1 and Cdc20 were obtained from Thermo scientific , Waltham , MA . Silencer Select siRNA for Ube2S were obtained from Life Technologies . Calcium ionophore was from Sigma , St . Louis , MO . proTAME was from Boston Biochem , Cambridge , MA . Recombinant Emi1 and APC components are purified as previously described ( Wang and Kirschner , 2013 ) . Securin and HA-Nek1 were expressed and radiolabelled with 35S in reticulocyte lysate ( Promega ) . HA-Nek1 was further purified through HA beads ( Sigma ) for ubiquitylation assay . Cell extracts were prepared from HeLa S3 cells and RPE1 cells as indicated . In vitro protein degradation and in vitro ubiquitylation assays were performed as described previously ( Wang and Kirschner , 2013 ) . hTERT-RPE1 cells were lysed in SDS sample buffer , and proteins were separated by SDS-PAGE on 4–12% gel . Data were analyzed with the Odyssey Infra-red imaging system . Primary antibodies for western blot were as follows: anti-GAPDH ( 1:20 , 000; 0411; sc47724; Santa Cruz Biotech , Santa Cruz , CA ) , anti-Nek1 ( 1:100; PAB3282; Abnova , Taiwan ) , anti-c-Myc ( 1:500; 9E10; sc40; Santa Cruz Biotech ) , anti-HA ( 1:500; Y11; sc805; Santa Cruz Biotech ) , anti-Ube2S ( 1:1000; 2257; Strategic Diagnostics , Newark , DE ) , anti-actin ( 1:2000; A2066; Sigma ) , anti-Aurora A ( 1:1000; 3092; Cell Signaling , Beverly , MA ) , anti-Cdc20 ( 1:1000; AR12; K0140-3; MBL International , Woburn , MA ) . Infrared secondary antibodies ( 1:20 , 000; Li-COR Biosciences , Lincoln , NE ) were used for Odyssey imaging . Cells on coverslips were fixed with 3 . 5% paraformaldehyde ( 10 min ) at 37°C and then methanol ( 4 min ) at 4°C , permeabilized with 0 . 2% Triton-X100 in PBS , blocked in 2 . 5% BSA in PBS , and incubated with antibodies . Alternatively , to optimize signals at centrosomes or basal bodies , cells were fixed in methanol at −20°C for 10 min and subjected to blocking and antibody incubation . Primary antibodies included anti-APC2 ( 1:50; home-made ) , anti-APC7 ( 1:50; 21 , 418; Santa Cruz Biotech ) , anti-Cdc20 ( 1:100; AR12; K0140-3; MBL International ) , anti-γ-tubulin ( 1:5000; T5192; Sigma ) , anti-acetylated tubulin ( K40 ) ( 1:2000; 6-11B-1; T7451; Sigma ) , anti-acetylated tubulin ( K40 ) ( 1:2000; D20G3; 5335; Cell Signaling ) , anti-p150Glued ( 1:100; 1/p150Glued; 610473; BD Biosciences , San Jose , CA ) , anti-Arl13b ( 1:500; 17 , 711-1-AP; Proteintech , Chicago , IL ) , anti-ki-67 ( 1:500; ab15580; Abcam , Cambridge , MA ) , anti-centrin 1 ( 1:200; ab11257; Abcam ) , anti-Cdh1 ( 1:100; home-made ) , anti-Cdh1 ( 1:200; Ab-2; Calbiochem ) , anti-Cdh1 ( 1:200; 34-2000; Life Technologies ) . Secondary antibodies labeled with Alexa 488 and Alexa 568 were from Life Technologies . DNA was stained with DAPI . Images were acquired with Nikon 80i Upright Microscope , and processed with MetaMorph image acquisition software and Adobe Photoshop SC3 . In order to measure the length of cilia , during coverslip mounting , the coverslips were pressed hard against microscope slides to flatten most of the cilia . Only those cilia which were flattened into one plane were selected for image acquirement and length measurement .	The majority of cells in the human body have small hair-like structures that project from the cell surface . These structures , known as primary cilia , are involved in sensing light and touch , and they are also required for an organism to develop normally . Defects in cilia result in a wide range of human diseases that are collectively known as ciliopathies . These include polycystic kidney disease and Bardet–Biedl syndrome . Ciliary disorders can also affect almost every organ in the body leading to blindness , obesity , diabetes , and cancer . Cilia are dynamic structures that are dis-assembled when cells start to divide and are then re-assembled when cells are quiescent . The anaphase promoting complex ( APC ) has a critical role during cell division and targets key proteins that need to be degraded at specific times during this process . APC is localized in the basal body , which is found at the bottom of cilia , and it works together with a number of proteins which assist its function . Wang et al . now report that a complex formed by APC and its co-activator protein Cdc20 has two functions at the basal body: it is needed to maintain the optimal length of the cilia in quiescent cells and to shorten the cilia when cells exit from quiescent stage . Wang et al . also investigated the role of Nek1 , an enzyme that is localised in the basal body . It was found that reducing the level of Nek1 in quiescent cells resulted in the formation of defective cilia , suggesting that this enzyme controls the stability and integrity of cilia . Moreover , when cells undergo division , the APC-Cdc20 complex targets the Nek1 enzyme , causing it to be degraded and allowing the cilia to be disassembled . A detailed understanding of how cells maintain the length of cilia could lead to the development of new approaches for the treatment of human ciliopathies .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2014	The master cell cycle regulator APC-Cdc20 regulates ciliary length and disassembly of the primary cilium
Human tactile afferents provide essential feedback for grasp stability during dexterous object manipulation . Interacting forces between an object and the fingers induce slip events that are thought to provide information about grasp stability . To gain insight into this phenomenon , we made a transparent surface slip against a fixed fingerpad while monitoring skin deformation at the contact . Using microneurography , we simultaneously recorded the activity of single tactile afferents innervating the fingertips . This unique combination allowed us to describe how afferents respond to slip events and to relate their responses to surface deformations taking place inside their receptive fields . We found that all afferents were sensitive to slip events , but fast-adapting type I ( FA-I ) afferents in particular faithfully encoded compressive strain rates resulting from those slips . Given the high density of FA-I afferents in fingerpads , they are well suited to detect incipient slips and to provide essential information for the control of grip force during manipulation . The most fundamental requirement for dexterous manipulation is the ability to handle objects without slippage and dropping of the object . To ensure that an object is held safely in the hand , one must apply a sufficient amount of force to the object’s surface to counteract the forces tending to make it slip , for example , the object’s weight and inertia ( Johansson and Flanagan , 2009 ) . Exerting an excessive grip is inefficient and can result in crushing the object . Inversely , a minimal amount of force is required to avoid slip and is dictated by friction: a stronger grip is required for slippery surfaces and a looser grip is sufficient for sticky surfaces . A good strategy is then to adjust the grip to friction with an amount of force slightly above the minimum . Previous work has suggested that humans can quickly and accurately adapt to changes in friction ( Westling and Johansson , 1984; Cadoret and Smith , 1996 ) . Importantly , tactile feedback is necessary for grip adjustments to take place as disruption of this feedback abolishes the normal , fine-tuned grasp control and results in frequent object dropping despite excessive compensatory grip forces ( Westling and Johansson , 1987; Augurelle et al . , 2003; Witney et al . , 2004 ) . However , how information about friction is encoded by tactile afferents is largely unknown . Friction information might be partly available at the initial contact . Anecdotal evidence suggests that when contacting surfaces of different frictions the strength of the initial burst of activity of the afferents varies as the surface is changed ( Johansson and Westling , 1987 ) . In these experiments , however , different frictional conditions were associated with different surface textures and thus do not necessarily imply that the afferent responses specifically represented friction between the fingerpads and contact surface . Friction might not be encoded per se , but instead , tactile afferents might respond to short and localized slip events that imply impending slip . There is indeed evidence that small , short-lasting slips occur during manipulation ( Johansson and Westling , 1984 ) . Those slips trigger strong afferent responses and elicit grip force adjustments ( Johansson and Westling , 1987 ) . In addition to the context of object manipulation , it has also been suggested that tactile slip detection plays an important role in a range of tactile tasks involving movements of surfaces relative to the skin ( Gueorguiev et al . , 2016; Schwarz , 2016 ) . Tactile slip is not instantaneous but develops progressively ( Levesque and Hayward , 2003; Tada et al . , 2006; André et al . , 2011; Terekhov and Hayward , 2011; Delhaye et al . , 2014; Barrea et al . , 2018 ) . As the surface-tangential force , that is , traction , of the fingerpad increases , ‘local’ slips begin at the periphery of the contact and progress toward the center until the last central ‘stuck’ point finally slips , that is , the instant of a ‘full slip’ or a global slip . We refer to the period between the beginning of the tangential loading and the instant of a full slip as the partial slip phase . Such a phenomenon gives rise to substantial local strain patterns in the slipping regions , near the boundary between the stuck and the slipping points ( Delhaye et al . , 2016 ) . We hypothesized that information about these local deformations is carried by tactile afferents that inform the central nervous system about the contact state . Single-unit recordings of primary tactile afferents , both in humans and monkeys , have shown that type I afferents respond strongly to local skin deformation ( Johansson et al . , 1982; Sripati et al . , 2006; Saal et al . , 2017 ) . Those responses contain information about local geometric features such as edge orientation ( Pruszynski and Johansson , 2014; Suresh et al . , 2016; Delhaye et al . , 2019 ) . The most common stimuli used to evoke deformation of the skin are indentations and scanning with embossed geometric patterns or textures . Applying such stimuli makes it possible to relate the strength or the timing of the response to the topography or the statistics of the stimulus itself , but does not provide a mechanistic understanding of the nature of the response with respect to the local skin deformation at the mechanoreceptors themselves . Moreover , it is mostly unknown how tactile afferents respond to surface strains , that is , strains acting tangentially to the surface ( as opposed to features indented perpendicularly to the skin surface ) . Afferent recordings in the hairy skin of the human hand have shown that afferents of all types strongly respond to local skin stretch and that the fast-adapting type I ( FA-I ) afferents also strongly respond when the stretch is released ( Edin , 1992; Edin , 2004 ) . Slowly-adapting type II ( SA-II ) afferents are also known to be sensitive to skin stretch , but relating their response to the exact local stretch pattern is complex given the large size of their receptive fields . How glabrous skin afferents , that is , those engaged in the contact with objects during manipulation , respond to local skin strain has , to our knowledge , never been studied . This is mainly due to the difficulties of applying well-controlled mechanical stimuli and measuring the strain at the same time . To address this , we took advantage of a recently developed imaging system that can measure fingertip skin strain through a transparent material during tangential loading of the fingertip until slip occurs ( Delhaye et al . , 2014; Delhaye et al . , 2016 ) . While recording local strains with this system , we simultaneously recorded the activity of human tactile afferents innervating the fingertip ( Figure 1A ) . This way we were able to relate local strains to responses of afferents with receptive fields inside the fingerpad contact area . We found that all tactile afferents in the fingertip responded to slip events , but that FA-I afferents in particular faithfully signaled local skin compressions related to the progression of slips . We suggest that FA-I afferents are primarily responsible for detecting changes in surface strains and that their discharges are a primary source of information for the central nervous system to , for instance , quickly adjust fingertip forces to different levels of friction . We used microneurography ( Vallbo and Hagbarth , 1968 ) to record the activity of single units whose receptive fields were located at the fingertip ( Figure 1A ) . We focused on FA-I and SA-I afferents , which respond to local deformation events and have small , well-defined receptive fields ( Johansson and Vallbo , 1983 ) . We also recorded from a few type II afferents ( FA-II and SA-II ) . Sufficient recordings for data analysis ( three out of five repetitions of each plate move direction ) were obtained from 22 afferents ( 13 FA-I , 6 SA-I , 2 SA-II , and 1 FA-II ) . The locations of the receptive fields of all afferents are depicted in Figure 2—figure supplement 1 . As expected , the afferents responded vigorously to contact ( Figure 2 , ‘contact’ ) , but also responded in a variety of ways to the tangential loading ( Figure 2 , ‘forward’ and ‘backward’ ) . First , we looked at the overall discharge pattern of the afferents . FA-I units showed a phasic response , with a burst of activity at the instant of contact , and another one during the tangential loading ( e . g . , Figure 2A ) . However , a majority of FA-I afferents responded mostly only during the partial slip phase , showing no or almost no response during the start ( ‘onset’ phase ) and the end ( ‘plateau’ phase ) of the tangential movement ( for U115-04 in Figure 2A , there was one spike at the onset phase during one repeat ) . SA-I units instead presented a rather tonic response beginning at the initial contact that changed during the tangential loading phase by increasing or decreasing their firing rates ( e . g . , Figure 2C ) . The discharge patterns of all recorded afferents for all directions and frictions are reported in Figure 2—figure supplement 1 . Video 1 and Video 2 show image recordings of the fingerpad during one trial , together with spike sound associated with the afferent responses , for the two example units shown in Figure 2 . Note that the tangential movement of the plate led to slight fluctuations in the normal force that could not be suppressed by the force controller ( see Materials and methods ) . Those fluctuations did not evoke strong afferent responses . Indeed , the discharge rates were neither correlated to the normal force nor to its derivative ( Figure 2—figure supplement 2 ) . In fact , we considered a causal relationship untenable observing in Figure 2A that the afferent discharge seemed to follow the normal force fluctuation by ~100 ms in the distal direction but preceded it by ~200 ms in the proximal direction . First , we describe how the tactile afferent responded with respect to ‘global’ stimulus parameters such as the movement phase or direction . Afferents were more active during the partial slip phase than during the two other tangential loading phases ( i . e . , onset and plateau ) , suggesting that this period is key to the afferent responses ( Figure 3A ) . Indeed , for all four afferent types , the fraction of trials during which the afferent responded with at least one spike was higher for the partial slip phase than for the two others ( one-way ANOVA with repeated measures , F ( 2 , 24 ) = 47 . 05 , p<0 . 001 for FA-I units; F ( 2 , 10 ) = 13 . 97 , p=0 . 001 for SA-I units; all p<0 . 05 for paired comparisons; for FA-II and SA-II afferents , we did not have enough data for such analysis ) . FA-I units were silent during the plateau , except when a stick-and-slip phenomenon was observed , in which case the firing phase was locked to the stick-and-slip events ( e . g . , see U109-04 during plateau phase in Figure 2—figure supplement 1 ) . For each afferent , we defined its preferred global direction ( i . e . , ulnar , distal , radial , or proximal ) , namely , the plate movement direction for which the highest mean firing rate was observed during partial slip . Figure 3B shows the mean firing rate in the preferred global direction ( labeled ‘North’ ) and in the remaining other directions with respect to the preferred one . We found that different movement directions elicited different responses , confirming previous observations ( Birznieks et al . , 2001 ) . Some FA-I units tended to have a preferred-opposite pattern ( 5 out of 13 ) , that is , the preferred ( ‘N’ for North in Figure 3B ) and opposite directions ( ‘S’ for South in 3B ) tended to elicit a stronger response than the two other directions . This was not observed for the other afferent types , which tended to discharge less in the direction opposite to the preferred direction . The distribution of the preferred global direction covered the four directions tested; however , we observed overall a slight preference for the proximal-distal axis ( Figure 3C ) . Firing rates during partial slips were consistent across forward and backward ( Figure 3D; correlation r = 0 . 97 , n = 176 , p<0 . 001 ) , suggesting similar directional preference when the skin was already under tension ( backward movement ) . While the firing rates were strongly correlated across friction levels ( Figure 3E; correlation r = 0 . 95 , n = 176 , p<0 . 001 ) , the firing rates across the whole population tended to be slightly lower for low friction ( paired t-test , t ( 171 ) = 7 . 57 , p<0 . 001 ) and the ratio of the mean firing rate during low and high friction condition was 0 . 82 ± 0 . 55 ( mean ± std ) . Next , we sought to relate the afferent discharge to the strain pattern , that is , the local events taking place inside the afferent receptive field . We observed that the discharge of FA-I units was strongly coupled to the compressive strain changes taking place in their receptive fields during the partial slip phase ( Figure 4 ) . This was particularly clear for U104-02 , whose receptive field ( depicted with a gray circle in the fingerpad heatmaps of Figure 4 ) was located on the proximal side of the contact area . Indeed , during a proximal movement ( Figure 4A , left , and B , right ) , the plate movement elicited a compressive wave of strain changes ( in red ) along the y-axis ( proximal-distal axis ) moving from the proximal side of the contact area toward the center and crossing the afferent’s receptive field . This crossing elicited a strong discharge burst . Moreover , the low friction surface exhibited full slip at a lower level of tangential force , and therefore also earlier than the high friction surface , with respect to the movement onset . As a consequence , the compressive strain wave moved across the afferent receptive field earlier in this case and , strikingly , the discharge burst of the afferent also took place earlier , coinciding with the strain changes . This is even easier to observe for the backward movement ( Figure 4B , right ) . In this case , due to the previous loading , the movement of the compressive wave came even earlier when the low friction condition was used and the discharge burst of the afferent coincided . Finally , we observed that the response evoked by a stretch wave , generated by a distal movement ( Figure 4A , right , and B , left ) , was much weaker than its compressive counterpart . Still , the timing of the response was perfectly synchronized with the occurrence of the stretch in the receptive field . Note that a short burst was elicited at the onset of the movement in the distal direction in the backward case ( Figure 4A , right ) . Such transient burst cannot be explained by our strain measurements and occurred in a small fraction of the trials and only in a few afferents ( Figure 3A ) . Also note that part of the unit receptive field lost contact during the partial slip phase in the high friction case . To test the hypothesis that the responses of the tactile afferents are caused by specific ‘local’ strain patterns taking place inside the contact area , we took two different approaches . In the first model-free approach , we looked at the strain pattern observed at the time of each spike across all stimulus directions and frictions and computed a ‘spike-triggered average’ ( STA , see Materials and methods ) for all recorded FA-I and SA-I units . If the afferents with a receptive field in the contact area responded to local strains , we expected to observe a clear strain pattern associated with these units’ discharges . In contrast , for afferents with a receptive field outside the contact area , we expected no clear strain pattern at all . First , we used the strain rate norm ( \|\|e\|\| ) as a variable to estimate the STA . We found that , indeed , the average strains causing spikes in all recorded FA-I afferents had a clear , more or less annular ( ring-like ) pattern ( Figure 5A ) . Such an annular pattern is expected from the stimulus , which is a strain wave in the form of an annulus and does not reflect the shape of the afferent receptive field but rather the correlations present in the strain patterns ( Materials and methods ) . Importantly , however , the pattern overlapped the afferent’s receptive field and often peaked inside it . Furthermore , as expected , we did not observe such a clear pattern for the afferents having their receptive field outside the contact area ( Figure 5 , middle ) . Strikingly , clear patterns did not emerge for the SA-I afferents , suggesting that those afferents are less sensitive to the local surface strain changes ( Figure 5A , bottom ) . Note that we repeated the same analysis using the total ( cumulative ) strain instead of the strain changes , and that again , we did not observe any clear pattern . The peak values of the STA are shown in Figure 5B in orange and show strong values for FA-I units inside the contact and much weaker values for other afferents , confirming the previous observations . The same STA analysis was repeated using the two principal strain components , one compressive and one tensile , separately to build two STA maps ( see Materials and method ) . The results obtained are consistent with the STA obtained with strain rate norm , that is , that a clear pattern emerges only for FA-I afferents and that the STA peaks in the afferent receptive field ( Figure 5—figure supplement 1 ) . Moreover , we found that the compressive STA peaks were generally larger and more often found in the afferent receptive field than their tensile counterpart , suggesting that FA-I afferents are more sensitive to compression . This finding will be further supported in the next section . In the second , model-based approach , we aimed to predict the afferent discharge rate from the skin strain measured in the contact area . Inspection of the data led us to assume that , first , strains of opposite signs might not contribute in the same manner to the discharge as skin stretch seemed to evoke weaker responses than skin compression ( Figure 4 ) . Second , multiple components might be needed to explain responses in all directions . Therefore , we first set out to test if the afferents’ firing patterns could be reliably predicted for the skin strains using a model with six distinct predictors , that is , the three strain components each half-wave rectified , using both the positive ( stretch ) and negative part ( compression ) . The simplest model possible , a multiple linear regression including an intercept , was used . Our method was cross-validated , such that the regression models were fit on one friction condition and tested on the other ( Materials and methods ) . In the subsequent validations , we used data from different directions and forward vs . backward movements and observed quantitatively similar results ( see Figure 5—figure supplement 2 ) . The results obtained using the model-based approach revealed similar trends as the first model-free method . The linear model could predict the discharge pattern of the FA-I afferents , but not the SA-I afferents . Heatmaps built from the cross-validated R2 ( Figure 5—figure supplement 2 ) were qualitatively similar to those built from the STA . An example unit is shown in Figure 5C for an afferent with a substantial R2 ( 0 . 66 ) . This unit maximally responded in the proximal direction , when a compressive wave was observed in its receptive field . The maximal values of the R2 found within the receptive field ( or anywhere for the afferents outside the contact area ) are shown in Figure 5B in green . As with the model-free approach , only the firing rates of FA-I inside the contact area could be predicted from the strain . The high R2 value for the SA-I afferent outside the contact area is because this particular afferent was either active at a constant firing rate or silent , generating two separate clouds of data points and thus driving up the R2 ( Figure 2—figure supplement 1 ) . In summary , we observed that FA-I afferents respond to local strain patterns generated during partial slips . Having demonstrated that the recorded FA-I afferents respond to local strain patterns , we then sought to uncover what aspects of the strains were responsible for these responses . To that end , we aimed to predict FA-I afferent discharge rates with a subset of strain predictors and to compare their performance to the full model with six predictors . The analysis was performed only on the FA-I afferents for which we had optical measurements of the strains , that is , those having their receptive fields inside the contact area ( n = 6 , all shown in the top line of Figure 5A ) . Since all models are cross-validated , they can be compared irrespective of the number of predictors . First , we selected three single predictors that were invariant to the choice of a particular reference frame . The strain norm , informative about the intensity but neither the orientation nor the sign of the deformation , and the two principal strain components separately , the compressive ( e1 ) and the tensile ( e2 ) , obtained from single-value decomposition of the strain tensor ( Materials and methods ) , informative about the intensity and the sign of the deformation , irrespective of the orientation of the deformation . All those three single predictor models performed worse than the full model , as could be expected ( Figure 6A , left ) . However , we found that the compressive principal component always outperformed the tensile one , suggesting that the afferents are more sensitive to compression than to stretch . Next , we used each of the predictors of the full model separately . That is , the half-wave rectified positive and negative value of the three strain tensor components ( exx , eyy , and exy ) . Given that those components are dependent on the choice of a particular reference frame orientation , we repeated the fitting procedure for multiple rotations of the reference frame equally spaced from 0 to 90° ( Materials and methods ) . The results are shown in Figure 6B for an exemplar afferent , with the shear component ignored . In this figure , the performance of the prediction ( R2 ) based on a single strain component ( compressive in red and tensile in blue ) is shown as a function of the reference frame rotation . This afferent seemed to have a preferred strain orientation close to 90° with respect to the radial-ulnar axis ( i . e . , along the proximal-distal axis ) , where the compressive component peaks . That is , the afferent seemed to encode preferentially 'local' compressive strain rate along a particular orientation . Indeed , as already described in Figure 4 , this unit was responding strongly in the proximal condition , where a compressive strain wave along the proximal-distal orientation passed through its receptive field . Perpendicular to that orientation , the tensile component peaked as well but with a lower R2 . This is expected since compression in one orientation generates stretch in the other at the same time because the volume is mostly conserved . The same plot as in Figure 6B is provided for all FA-I afferents and the three cross-validation methods in Figure 6—figure supplement 1 . It is important to avoid the confusion between the units’ preferred direction described in Figure 3 , which relates to the robot movement direction and the maximal firing rate of the afferent , and the strain orientation preference described here , which is related to the orientation of the local deformation in the afferent receptive field . From these analyses , we draw two important conclusions . First , compressive strain change is a more effective stimulus than tensile strain change ( Figure 6A , right ) , confirming the observation in the principal component analyses , namely , that the FA-I responses are mostly driven by compression . Second , FA-I afferents did not respond to compression in any orientation but rather to compression along a certain preferred strain orientation . The argument for this is twofold: ( 1 ) the performance of a single compressive component in a particular orientation was always higher than the performance of the compressive principal component e1 ( Figure 6A ) and ( 2 ) models with one single component along its preferred strain orientation performed as well as the full model comprising all the components , suggesting that this component is mainly responsible for driving the afferent response . Note that the FA-I afferents’ preferred strain orientation seemed to coincide with the local fingerprint orientation , but more data is needed to confirm this trend ( Figure 2—figure supplement 2 , correlation r = 0 . 74 , p=0 . 10 , n = 6 ) . In sum , our analyses strongly suggest that the FA-I are sensitive to local skin strains , more so to changes in compressive strain than tensile strain and that they respond maximally along a preferred strain orientation . Finally , we asked how much the response of all types of afferents was related to the ‘global’ external 3D force vector . We computed the correlation between the afferent firing rates and each force components during the tangential loading ( all three phases ) . We also computed the same correlations with the force rates . We found low correlation values for both force and force rates for the fast-adapting afferents ( FA-I and FA-II ) . However , the correlations with tangential ( horizontal ) forces were high for slowly-adapting afferents ( SA-I and SA-II ) , especially for those not inside the contact area ( Figure 2—figure supplement 2 ) . Their firing rates were however not related to the small vertical ( normal ) force fluctuations . Our methods generate high-resolution representations of the distribution of surface strain in the contact area of the fingertip ( i . e . , 2D , x and y axes ) but do not allow measurements of the distribution of the deformations perpendicular to the fingertip surface inside the contact area ( i . e . , along the z-axis ) , nor outside the contact area . Given that the SA-I afferents did not respond to the surface deformation measured in this study , they could be sensitive to two other non-measured aspects of the local deformations . First , since SA-I afferents are known to be exquisitely sensitive to contact and contact pressure ( Goodwin and Wheat , 1999; Wheat et al . , 2010; Khamis et al . , 2015 ) , they may be more sensitive to the strains perpendicular to the surface , rather than the tangential ( i . e . , planar ) strains . Indeed , the tangential loading can cause a progressive redistribution of the pressure inside the contact area ( even if the global normal force is maintained constant ) that will consequently generate ‘local’ vertical strains ( not measured , which would be denoted ‘ezz’ ) . The SA-I shows slow firing rate variations that might be well attributed to a relatively slow movement of the center of pressure during tangential loading that results in local pressure change . Another possibility is that the SA-I afferents are sensitive to local shear ( which would be denoted ‘exz’ and ‘eyz’ ) . The fact that their firing rates correlate with the tangential forces supports this idea , and more measurements are needed to disentangle the two possibilities . Importantly , these two other aspects of skin deformations are likely much less informative about the partial slip state of the finger; therefore , SA-I afferents are less likely to contribute to the detection of the slip events . When considering the subdermal location of SA-II and FA-II afferent terminals , it is not surprising to observe a poor coupling between their responses and the measured surface strains . The tangential speed chosen for practical reasons in our experiments is certainly lower than that during normal manipulation actions , during which much higher strain rates are probably generated . In fact , it is unknown what the typical amplitudes of surface skin strain during manipulation are . Moreover , during active manipulation , humans typically also vary the amount of grip ( normal ) force according to the tangential force , while our experiment kept the normal force constant . The development of a manipulandum equipped with embedded imaging of skin deformation in the contact area will make it possible to quantify deformation in an ideal environment with both normal and tangential force fluctuations ( Barrea et al . , 2017 ) . Nevertheless , it is reasonable to expect that FA-I afferents respond even more vigorously during actual manipulation . It remains also to be demonstrated that in a typical manipulation task the timing of this signal enables online control of grip , that is , that there is enough time to react . In this study , the partial slip phase lasts several hundreds of milliseconds , which would definitely leave enough time for cutaneous signals to contribute to online control . Most surfaces encountered during natural object manipulation have frictional properties that cause stick-and-slip when the tangential load is sufficiently large , that is , local breaking of the bond between the surface and even a single fingerprint ridge . However , stick-and-slip events were largely absent in our experiment because we had to use a flat and transparent material to be able to measure skin surface deformations . With natural material , partial slip might therefore occur in a stepwise manner and result in readily measurable accelerations ( Johansson and Westling , 1984; Johansson and Westling , 1987 ) , as opposed to a very smooth progressive wave observed in our experiments . Nevertheless , while the timing might be affected , the patterns of deformation observed in their purest form in this study will generalize in some form to other surfaces and therefore the observations made in our study are likely to hold with natural materials . There are several reasons why the models' R2 are relatively low . First , the models are linear , whereas the response of tactile afferents to skin deformations is known to be far more complex than accounted for by simple linear relationships ( Dong et al . , 2013; Saal et al . , 2017 ) . Second , the inputs of the model were obtained from a single point inside the receptive field of the afferent . Yet , we know that human tactile afferents have complex receptive fields with multiple zones of sensitivity that they owe to the branching of the afferent terminals in the skin ( Johansson , 1976; Phillips et al . , 1992; Pruszynski and Johansson , 2014 ) . A richer stimulus using more stimulation conditions ( tangential speeds , normal forces , more orientations ) to reduce the effect of the correlations inherently present in the strain pattern might provide sufficient data to identify multiple zones of sensitivity , using the STA approach developed in this study . However , combining our stimulus with another one dedicated to identifying such receptive field topography ( Jarocka et al . , 2021 ) would be needed to establish a causal relationship between the two . Finally , our setup does not provide the ability to measure all aspects of local skin deformation as discussed above; we were unable to measure deformations along the axis normal to the glass surface . Analyses of strain normal to the contact surface and deformations outside the contact area would require other approaches , for example , finite element analysis . Our sense of touch enables us to grasp and manipulate objects with great dexterity . The ability to capture and extract very subtle mechanical phenomena arising continuously during fine manipulation is probably a key to its success . We have demonstrated in this study that subtle skin compressions taking place in the contact area with the object’s surface before the slippage provide essential feedback for grasp stability . Given that estimating friction is a complex problem influenced by many factors of the object itself ( texture , adhesion , hydrophilic or hydrophobic properties ) and by properties of the skin ( for instance , stiffness ) , some of which are constantly changing ( e . g . , humidity ) , the nature of this feedback and its independence from friction makes it particularly well suited for grasp stability . Sixteen healthy human subjects ( seven females; ages 19–24 years ) participated in the experiment . Each subject provided written informed consent to the procedures , and the study was approved by the local ethics committee at the host institution ( Université catholique de Louvain , Brussels , Belgium ) . Subjects were seated in a comfortable dentist chair with the forearm slightly pronated and abducted . The forearm was resting on a horizontal cushioned support . The right hand , with the volar side facing the ground , was fixed to a custom-made support ( Figure 1A ) , which enabled to lock one finger while keeping the rest of the fingers away in a safe position . The participant’s fingers were stabilized by gluing the nail of digit II and III to aluminum bars using cyanoacrylate and connecting the bars rigidly with the custom-made support . Such fixation hindered finger movement during the stimulation but allowed fingertip deformation due to the compliance of the fingerpad . The stimulations were applied by a robotic platform based on an industrial robot , as already described in previous studies ( Delhaye et al . , 2014; Delhaye et al . , 2016; Barrea et al . , 2018 ) . Briefly , a transparent plate was mounted horizontally on the end effector of an industrial robot ( Denso Robotics , Japan ) and its movement in three orthogonal directions could be controlled precisely . Two force transducers were mounted , one on each side of the plate , and measured the forces applied to the subject’s fingertip ( ATI force sensors , ATI Industrial Automation , USA ) . The measured RMS error of the force measurement was low ( ranging 0 . 01–0 . 02 N along the tangential axes and 0 . 03–0 . 06 N along the vertical axis ) . The stimulus was applied to the finger where the receptive field of the recorded afferent was located . The relative angle between the stimulus and the finger was around 30° . The transparent plate consisted of smooth glass and was either plain or covered with a hydrophobic coating , RainOff ( Arexons , Italy ) reducing friction . The robot followed a programmed tangential ( horizontal ) trajectory with the help of the manufactory position controller . A custom PID controller was developed to servo-control the normal ( vertical ) force applied to the fingerpad by feeding back the force measurements ( average RMS error is 5% during the whole tangential loading , and average peak error is 12% or 0 . 5 N ) . The robot was combined with a custom-made fingerprint imaging apparatus composed of a high-speed ( 50 fps ) and high-resolution ( 1200 dpi ) camera ( Mikrotron MC1362 , 1280 × 1024 pixels , Mikrotron GmbH , Germany ) and a coaxial light source ( White LED Backlight , Phlox , France ) . This optical apparatus enabled imaging of fingerprints in contact with the transparent stimuli to compute strains in the fingertip contact , as described in Delhaye et al . , 2016 . We used the microneurographic technique ( Vallbo and Hagbarth , 1968 ) to record single skin afferent activity ( ‘unit’ ) . An insulated tungsten needle electrode was percutaneously inserted into the right median nerve ~10 cm proximal to the elbow joint ( Figure 1A ) . Once the recording electrode was in an intraneural position , it was manipulated in minute steps until single-unit activity clearly stood up from the background noise . To evoke the responses of the afferents , the relevant skin areas of the fingertips were stimulated . Once a single afferent activity was identified and a corresponding receptive field located , the threshold force for the receptor was determined using von Frey hairs . According to well-known criteria ( Vallbo and Johansson , 1984 ) , receptors were identified as slowly-adapting ( SA ) if the discharge was sustained while pressing with a glass probe for at least 2 s; otherwise , they were considered fast-adapting ( FA ) . The type I units have small , easily located receptive fields , whereas type II units are characterized by large and poorly defined receptive fields . Furthermore , SA-II could be distinguished from SA-I by the regularity of their inter-spike intervals , and FA-II units from FA-I by their response to remote light taps . Once the spot with the highest sensitivity within the receptive field had been marked with a permanent marker , the experimental procedure was initiated . The afferent search was voluntarily biased toward type I afferents because of their relatively small receptive field , which makes them more likely to respond precisely to the local strain patterns recorded within the fingertip contact . While recording impulses from single tactile afferents innervating the right index or middle finger of subjects , the stimulus was moved vertically toward the fingerpad in single trials , made contact , and reached a preset normal force of 4 N ( ‘contact’ phase ) . That normal force was kept constant during the entire trial . After a delay of 1 . 2 s , which was necessary to induce occlusion and make the contact visible on the camera ( Dzidek et al . , 2017 ) , the surface was moved tangentially ( horizontally ) with a constant speed ( 5 . 5 mm/s ) first in one direction ( ulnar , radial , distal , or proximal ) for 8 mm ( ‘forward’ movement ) and then in the opposite direction for 12 mm ( ‘backward’ movement ) ( Figure 1B ) . The relatively slow tangential speed is explained by two reasons . First , since our optical system has a limited frame rate ( 50 Hz ) , we wanted to have a relatively slow stick-to-slip transition such that it was accurately measured by the imaging system . Second , the normal force servo-control was more effective at low speed . The movement amplitudes were thus large enough to ensure eventually full sliding between the fingertip and the surface in both directions . The surface was then retracted from the finger , and the trial was ended . The procedure was the same for all trials . Each protocol consisted of five repetitions of the four stimulation directions for each of the two friction conditions , totalizing 40 trials ( 2 frictions × 4 directions × 5 repetitions ) . For practical reasons , the same sequential order was used for all protocols: all high friction trials ( on plain glass ) were run first , followed by all low friction trials ( on coated glass ) . Moreover , the direction sequence was always the following: ulnar , distal , radial , and proximal , with all five repetitions made in a row . Plate position and forces exerted on the finger were sampled at 1 kHz along the three axes ( PCIe-1433 , National Instruments , and LabVIEW ) . Fingerprint deformations were monitored through the transparent plate during the tangential movement of the platform , allowing the derivation of surface strains at the fingertip contact . The neural signals were sampled at 12 . 8 kHz after amplification close to the recording site . The identification of single action potentials was made semiautomatically under visual control ( Edin et al . , 1988 ) . WINSC/WINZOOM software ( Umeå University , Sweden ) was used for recording and analyzing the neural data . The instantaneous firing rate ( as shown in Figures 2 and 4 ) was defined as the inverse of the time interval between two consecutive spikes for the interval duration . This rate was resampled at 1 kHz to obtain convenient time series for data analysis . For each trial , strains in the contact area were computed from the images as described in Delhaye et al . , 2016 ( Figure 1C ) . Briefly , the contact region was first extracted semiautomatically from each image of the sequence . Second , equally spaced features were sampled in the contact area of the initial frame and then tracked from frame to frame to measure the displacement field in the contact using the optical flow technique ( Lucas and Kanade , 1981 ) implemented in the OpenCV online computer vision toolbox ( Bradski , 2008 ) . Third , the tracked features were triangulated ( Delaunay triangulation ) , and Green-Lagrange strains were computed for each triangle by calculating the gradient of the displacement field . This operation yielded a 2-by-2 symmetric strain matrix for each triangle and each pair of consecutive frames . Axial strains , that is , the diagonal elements of the strain matrix , were denoted exx and eyy and shear strain ( off-diagonal ) was denoted exy . The x-axis was aligned to the radial-ulnar orientation ( Figure 1B ) . Strains were filtered , first spatially ( using Smooth Triangulated mesh from Matlab FEX , https://mathworks . com/matlabcentral/fileexchange/26710-smooth-triangulated-mesh ) , and then temporally ( median filtering over three strain values ) . By using the term 'strain' throughout the article , we refer to the elements of the strain tensor computed between two consecutive images and expressed in percent per second , that is , strain rates . The principal strain components denoted e1 and e2 were then obtained by an eigenvalue/eigenvector decomposition of the strain matrix . The principal strains e1 and e2 correspond to the maximum compressive and tensile strains at the location where the strain tensor is measured . The eigenvalue decomposition is equivalent to a rotation of the reference frame so that the shear strain is canceled and only axial strains remain , which thus corresponds to the maximum local compression and/or dilation . Therefore , the principal strains are not necessarily aligned with the axes of the x-y reference frame defined above and their orientation is potentially different for each local strain tensor . Therefore , the matrix components of the local strain tensor e are solely given by the axial and shear strains along the x-y plane ( exx , eyy , and exy ) . Besides , given the high stiffness of the outer layer of the skin ( Wang and Hayward , 2007 ) , it is reasonable to assume that the area of local patches of skin is conserved after deformation . Under this hypothesis , it can be shown that e1 and e2 have opposite signs . This means that if one of the principal strain is compressive , the other one is dilative and vice versa . Moreover , we follow the convention to sort the eigenvalues by ascending order . Therefore , e1 will always be smaller than e2 . Under the hypothesis of area conservation , this implies that e1 will be negative and thus compressive , whereas e2 will be positive and thus tensile . This hypothesis has been verified in practice on the data . To assess the strains taking place at each given point of the contact area across different trials , we interpolated the strains on an arbitrary rectangular grid , the ‘strain grid’ . First , an easily identifiable feature of the fingerprint located near the center of contact was manually spotted on the first image of each trial and used to precisely align all trials . Given the precise repeatability of the trials , the spotted feature only moved by a few pixels from trial to trial . Then , the contact area was centered on the grid using its geometric center on the first image . The feature coordinates on the first image were then used to interpolate the strains over the entire trial on the given grid using MATLAB’s scatteredInterpolant function . The ‘strain grid’ spacing was set to 10 pixels ( ~200 um ) and was composed of 120-by-90 elements ( 1200 × 900 pixels ) . The local fingerprint orientation was also estimated as described previously ( Delhaye et al . , 2016 ) by computing the image gray-level gradients ( Bazen and Gerez , 2000 ) over a 60-by-60 pixels window .	Each fingertip hosts thousands of nerve fibers that allow us to handle objects with great dexterity . These fibers relay the amount of friction between the skin and the item , and the brain uses this sensory feedback to adjust the grip as necessary . Yet , exactly how tactile nerve fibers encode information about friction remains largely unknown . Previous research has suggested that friction might not be recorded per se in nerve signals to the brain . Instead , fibers in the finger pad might be responding to localized ‘partial slips’ that indicate an impending loss of grip . Indeed , when lifting an object , fingertips are loaded with a tangential force that puts strain on the skin , resulting in subtle local deformations . Nerve fibers might be able to detect these skin changes , prompting the brain to adjust an insecure grip before entirely losing grasp of an object . However , technical challenges have made studying the way tactile nerve fibers respond to slippage and skin strain difficult . For the first time , Delhaye et al . have now investigated how these fibers respond to and encode information about the strain placed on fingertips as they are loaded tangentially . A custom-made imaging apparatus was paired with standard electrodes to record the activity of four different kinds of tactile nerve fibers in participants who had a fingertip placed against a plate of glass . The imaging focused on revealing changes in skin surface as tangential force was applied; the electrodes measured impulses from individual nerve fibers from the fingertip . While all the fibers responded during partial slips , fast-adapting type 1 nerves generated strong responses that signal a local loss of grip . Recordings showed that these fibers consistently encoded changes in the skin strain patterns , and were more sensitive to skin compressions related to slippage than to stretch . These results show how tactile nerve fibers encode the subtle skin compressions created when fingers handle objects . The methods developed by Delhaye et al . could further be used to explore the response properties of tactile nerve fibers , sensory feedback and grip .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2021	High-resolution imaging of skin deformation shows that afferents from human fingertips signal slip onset
The resection of DNA strand with a 5´ end at double-strand breaks is an essential step in recombinational DNA repair . RecJ , a member of DHH family proteins , is the only 5´ nuclease involved in the RecF recombination pathway . Here , we report the crystal structures of Deinococcus radiodurans RecJ in complex with deoxythymidine monophosphate ( dTMP ) , ssDNA , the C-terminal region of single-stranded DNA-binding protein ( SSB-Ct ) and a mechanistic insight into the RecF pathway . A terminal 5´-phosphate-binding pocket above the active site determines the 5´-3´ polarity of the deoxy-exonuclease of RecJ; a helical gateway at the entrance to the active site admits ssDNA only; and the continuous stacking interactions between protein and nine nucleotides ensure the processive end resection . The active site of RecJ in the N-terminal domain contains two divalent cations that coordinate the nucleophilic water . The ssDNA makes a 180° turn at the scissile phosphate . The C-terminal domain of RecJ binds the SSB-Ct , which explains how RecJ and SSB work together to efficiently process broken DNA ends for homologous recombination . DNA double-strand breaks ( DSBs ) are the most lethal form of DNA damage due to the free DNA ends . DSBs can be repaired by homologous recombination ( HR ) , which requires a pre-aligned homologous sequence prior to ligation . DNA end resection degrades the 5´ strand of a DSB end and is one of the earliest and most important processes in HR repair ( Symington and Gautier , 2011 ) . In bacteria DSBs are predominantly repaired by either the RecBCD or the RecF pathways . Compared with the RecBCD pathway , the RecF pathway is highly conserved across eubacteria and shows functional homology in the eukaryotic HR process . For example , RecA resembles yeast Rad51 protein with regard to the catalysis of DNA strand invasion and change reactions ( Cox and Lehman , 1982 ) . RecO is capable of annealing single-stranded DNA-binding ( SSB ) protein-coated ssDNA and facilitates RecA loading ( Kantake et al . , 2002 ) , a property that is shared with the yeast Rad52 protein . The RecBC complex contains both helicase and nuclease activities . In contrast , the RecF pathway requires RecJ , a 5´-3´ exonuclease , together with RecQ helicase and SSB protein to initiate DSB end resection ( Courcelle and Hanawalt , 1999; Han et al . , 2006; Handa et al . , 2009; Morimatsu and Kowalczykowski , 2014 ) . Deinococcus radiodurans is extremely resistant to DNA-damaging agents , such as ionizing radiation , ultraviolet radiation and mitomycin C ( MMC ) . This robustness correlates with its extraordinary DNA repair capabilities , especially DSB repair , which can rebuild a shattered genome within several hours ( Krisko and Radman , 2013; Slade et al . , 2009 ) . In contrast to Escherichia coli , RecBC is naturally absent from D . radiodurans , whereas the key components of the RecF pathway are present . Disruption of recF , recO , or recR significantly sensitized the cells to radiation similar as recA mutant ( Bentchikou et al . , 2010; Ithurbide et al . , 2015; Satoh et al . , 2012; Xu et al . , 2008 ) , which suggests that the RecF pathway is the dominant HR pathway in D . radiodurans . An extended synthesis-dependent strand annealing process prior to DNA recombination contributes substantially to the rapid restoration of an intact genome ( Bentchikou et al . , 2010; Slade et al . , 2009 ) . It was proposed that RecJ nuclease , which is associated with helicase , rapidly digests the DSB end in a 5´-3´ polarity after DNA damage . The resultant 3′-ssDNA overhang subsequently anneals to the complementary DNA strand via RecA- and RadA-mediated strand invasion , followed by extensive DNA synthesis . RecJ orthologs have been found widespread in most eubacteria and archaea ( Aravind and Koonin , 1998; Sanchez-Pulido and Ponting , 2011 ) . RecJ was firstly identified in E . coli by its effect on recombination that cells lacking both RecBCD and RecJ showed extreme recombination deficiency ( Lovett and Clark , 1984 ) . In addition to HR , RecJ is involved in ssDNA gap repair , base excision repair and methyl-directed mismatch repair ( Burdett et al . , 2001; Dianov et al . , 1994; Lovett and Kolodner , 1989 ) . As a processive nuclease , RecJ only degrades ssDNA in a 5´-3´ direction but not capable of DNA end digestion with a blunt end or 3´-ssDNA overhang ( Cheng et al . , 2015b; Han et al . , 2006; Jiao et al . , 2012 ) . Notably , RecJ also showed 'dRPase' activity to incise abasic sites ( Dianov et al . , 1994 ) . SSB can interact with RecJ and stimulate its DNA binding and nuclease activities ( Han et al . , 2006; Morimatsu and Kowalczykowski , 2014; Sharma and Rao , 2009 ) . Recently , it has been shown that the initiation of DSB end resection by coordinated activities of RecJ and RecQ depends on the nature of the DNA ends ( Morimatsu and Kowalczykowski , 2014 ) . In E . coli , RecJ nuclease alone is capable of digesting DNA with 5´-ssDNA overhang , whereas RecQ , a 3´-5´ helicase , is required to initiate DNA end resection with a blunt end or 3´-ssDNA overhangs ( Morimatsu and Kowalczykowski , 2014 ) . The crystal structures of RecJ orthologue from Thermus thermophilus ( ttRecJ ) have been solved in the absence of substrate DNA ( Wakamatsu et al . , 2010; Yamagata et al . , 2002 ) . The structure of ttRecJ reveals an O-shaped molecule , which comprises a novel oligonucleotide/oligosaccharide-binding ( OB ) fold domain . The hole close to the active site is too narrow to accommodate double-strand DNA ( dsDNA ) , suggesting the ssDNA preference of ttRecJ . The nuclease core of ttRecJ consists of a central parallel β-sheet encircled by α-helices . Two catalytic metal ions coordinated by conserved histidine and aspartate residues were observed in the active site . However , in the absence of DNA substrate , the conserved loop region of the nuclease core was disordered . Here , we report three crystal structures of D . radiodurans RecJ ( drRecJ ) in complex with deoxythymidine monophosphate ( dTMP ) , ssDNA , and the C-terminal region of drSSB protein ( SSB-Ct ) . These structures , together with mutagenesis and biochemical studies , provide mechanistic insights into DNA resection by drRecJ . DHH family proteins can be divided into three subfamilies based on their sequence similarity and domain arrangement ( Figure 1A ) . These subfamilies share a conserved N-terminal DHH domain that consists of consecutive DHH residues , which give rise to the name this family of proteins . The subfamily 1 group includes RecJ and nanoRNase ( RecJ-like protein ) , a nuclease that degrades short RNA , and has a distinct DHHA1 domain following the DHH domain ( Srivastav et al . , 2014; Wakamatsu et al . , 2010 ) . In contrast , the DHHA2 domain is present in the subfamily 2 group ( e . g . , exopolyphosphatase PPX1 and Drosophila Prune protein ) ( Ugochukwu et al . , 2007 ) . The subfamily 3 group is defined by the RecJ eukaryotic orthologue CDC45 , which has large insertions between the DHH domain and the C-lobe ( Krastanova et al . , 2012 ) . Notably , bacterial RecJ nucleases have an additional OB fold domain next to the nuclease core ( Figure 1A ) . 10 . 7554/eLife . 14294 . 003Figure 1 . Domain arrangement and substrate specificity of drRecJ . ( A ) Schematic of the domain arrangements of three DHH subfamilies . ( B ) Denaturing PAGE gel showing that drRecJ degrades different substrates , as shown at the top of the panel . 3′-Fluorescent labeled DNA or RNA ( 100 nM ) were incubated with drRecJ ( 0 , 5 and 20 nM ) in the presence of 100 nM Mn2+ ( see methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 00310 . 7554/eLife . 14294 . 004Figure 1—figure supplement 1 . Sequence alignments , secondary structure , and functional residues of RecJ and CDC45 . Names of species are dra , Deinococcus radiodurans; ttj , Thermus thermophilus; eco , Escherichia Coli; Homo , Homo sapiens; Drosophila , Drosophila melanogaster . Structural elements of RecJ are shown in distinct colors . Predicted secondary structure of human Cdc45 is shown at bottom ( grey ) . Conserved motifs ( I-VII ) are labeled . Conserved key residues are highlighted in distinct colors . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 00410 . 7554/eLife . 14294 . 005Figure 1—figure supplement 2 . Metal preference of drRecJ . 3′-fluorescent labeled 20 nt ssDNA ( KY08 , 100 nM ) was incubated with drRecJ ( 10 nM ) in the presence of Mn2+ and Mg2+ . For the metal competition assays , drRecJ was pre-incubated with 0 . 1 mM Mn2+ before addition of Mg2+ . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 005 drRecJ is conserved within bacteria , which shares 42% and 32% amino acid identity with ttRecJ and ecRecJ , respectively ( Figure 1—figure supplement 1 ) . The sequence alignment revealed that drRecJ contains all the seven signature motifs of the RecJ family nucleases , which belongs to the DHH ( motifs I-V ) domain and the DHHA1 domain ( motifs VI and VII ) ( Figure 1—figure supplement 1 ) . Compared with ecRecJ , drRecJ has an additional C-terminal domain that is conserved in the Deinococcus-Thermus phylum ( Figure 1A and Figure 1—figure supplement 1 ) . To determine the metal preference , various concentrations of Mn2+ and Mg2+were applied to reactions ( Figure 1—figure supplement 2 ) . In contrast to the degradation of DNA by ecRecJ in a reaction that requires Mg2+ , both Mn2+ and Mg2+ can activate the nuclease activity of drRecJ . However , the drRecJ activity in a reaction buffer containing 10 µM Mn2+ is substantially higher than that contains 10 mM Mg2+ ( Figure 1—figure supplement 2 ) . Overwhelming Mg2+ ( 10 mM ) does not affect the nuclease activity when drRecJ is pre-incubated with Mn2+ ( 0 . 1 mM ) , which suggests that Mn2+ ions are employed for drRecJ catalysis . To test the substrate specificity , drRecJ was incubated with different types of synthetic DNA fluorescence-labeled at the 3′ end in the presence of Mn2+ ( Figure 1B ) . drRecJ is able to processively digest ssDNA and DNA with a 5´-ssDNA overhang ( 14 nt overhang ) to a single nucleotide ( Figure 1B ) . In contrast , drRecJ cannot resect ssRNA or DNA with a blunt end or 3´-ssDNA overhang ( 6 nt overhang ) , which indicates that a free 5´-ssDNA is essential for drRecJ nuclease activity . To characterize the RecJ recognition and incision of the DNA substrate , we crystallized three types of drRecJ complexes: wild-type drRecJ complexed with dTMP ( complex I ) , a catalytic inactive mutant ( H160A ) RecJ complexed with DNA bearing a 5´-ssDNA overhang ( complex II ) , and the ternary complex of RecJ-ssDNA and the SSB-Ct ( complex III ) . The crystals were grown in the presence of Mn2+ ions and diffracted X-rays to 2 . 3–2 . 7 Å resolution . The structures are validated by the appearance of a well-formed active site with two catalytic metal ions and DNA or dTMP at the active site ( Figure 2A , B ) . The crystal data , together with the data collection and refinement statistics , are summarized in Table 1 . 10 . 7554/eLife . 14294 . 006Figure 2 . Structure of drRecJ complex . ( A ) Overall structure of drRecJ complex viewed from the side . Protein domains of drRecJ are labeled and shown in distinct colors . The DNA and SSB-Ct are colored orange and yellow , respectively . Two Mn2+ in the active site are shown as magenta spheres . Two regions that are disordered in the ttRecJ structures ( PDB code: 2ZXP ) are highlighted in cyan . Three helices that form a helical gateway are also labeled . ( B ) Overall structure of the drRecJ complex viewed from the top of the DNA . The downstream nucleotides stack well to mimic the double-stranded DNA . ( C ) Denaturing PAGE gel showing the nuclease activities of different truncations of drRecJ . 3′-Fluorescence-labeled 20 nt ssDNA ( 100 nM ) was incubated with various concentrations of different truncations of drRecJ proteins ( see methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 00610 . 7554/eLife . 14294 . 007Figure 2—figure supplement 1 . Structure of RecJ:DNA complex . ( A ) Comparison of complex II and ttRecJ structures . The ttRecJ ( PDB code: 2ZXP ) is colored white . Protein domains of drRecJ are shown in distinct colors and labeled . The DNA and ordered β3-α6 loop are colored orange and cyan , respectively . Loops and helices showing noticeable deviations are labeled . The relative domain movements are shown by the arrowheads . ( B ) Overall structure of complex II viewed from the top of the DNA entrance . Protein domains of drRecJ are shown in distinct colors and labeled . The DNA and SSB-Ct are colored orange and yellow , respectively . Two Mn2+ in the active site are shown as magenta spheres . Three helices ( α13 , α15 and α16 ) that form a helical gateway are also labeled . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 00710 . 7554/eLife . 14294 . 008Figure 2—figure supplement 2 . Comparison of the nuclease fold and DHHA1/DHHA2 domain . ( A ) Structural comparison among the nuclease fold from drRecJ , NrnA ( PDB code: 4LS9 ) , PpxI ( PDB code: 2QB6 ) and hFen1 ( PDB code: 3Q8K ) . The nuclease fold of each enzyme is shown in rainbow colors from the blue N- to the red C-terminus . The residues involved in catalysis are shown as sticks . The topologies of β-strands are also labeled . ( B ) Structural comparison between the DHHA1 domain ( drRecJ , NrnA ) and the DHHA2 domain ( PpxI , Pyrophosphatase ( PDB code: 1WPM ) ) . The topologies of β-strands and two inserted α-helices ( A and B ) are labeled . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 00810 . 7554/eLife . 14294 . 009Figure 2—figure supplement 3 . Relative position of DHH ( pink ) and DHHA1/DHHA2 ( blue ) domain . The active site of subfamily 1 group is formed between DHH domain and β3 to β5 of DHHA1 domain ( A ) . While the β1 of DHHA2 domain and DHH domain form the active site of subfamily 2 group proteins ( B ) . DNA and phosphate group are colored orange and DHH motif is shown as red stick . The helical gateway in DHHA1 domain and C-terminal α-helix in DHHA2 are also labeled . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 00910 . 7554/eLife . 14294 . 010Table 1 . Statistics from crystallographic analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 010RecJ-dTMPRecJd-DNARecJd-DNA-SSBctComplex IComplex IIComplex IIIData collectionSpace groupP 3221P 3221P 3221Cell dimensions a , b , c ( Å ) 106 . 53105 . 83102 . 22106 . 53105 . 82102 . 22161 . 90165 . 40166 . 12Wavelength ( Å ) 0 . 97920 . 97920 . 9792Resolution ( Å ) 30–2 . 7 ( 2 . 77–2 . 70 ) 30–2 . 6 ( 2 . 66–2 . 60 ) 30–2 . 3 ( 2 . 35–2 . 30 ) R-meas5 . 5 ( 64 . 1 ) 6 . 5 ( 79 . 1 ) 7 . 4 ( 63 . 7 ) I/σI27 . 0 ( 4 . 0 ) 22 . 6 ( 3 . 0 ) 17 . 0 ( 2 . 9 ) Completeness ( % ) 98 . 9 ( 99 . 5 ) 99 . 6 ( 99 . 4 ) 99 . 4 ( 94 . 1 ) Redundancy8 . 98 . 37 . 2RefinementResolution ( Å ) 30–2 . 730–2 . 630–2 . 3No . reflections297473359445296Rwork/Rfree20 . 61/25 . 2018 . 47/22 . 8422 . 56/23 . 89No . atomsProtein/DNA5262/-5373/2865342/182Ligand/Ion21/235/255/2Waters12-142B factorsProtein/DNA79 . 8/-70 . 4/103 . 958 . 2/93 . 7Ligand/Ion66 . 7/76 . 699 . 8/68 . 274 . 5/61 . 9Water62 . 7-48 . 9RmsdBond length ( Å ) 0 . 0120 . 0050 . 012Bond Angle ( ° ) 1 . 0850 . 8491 . 464Ramachandran statiticsFavored ( % ) 98 . 0 099 . 198 . 9Allowed ( % ) 2 . 00 . 91 . 1Outliers ( % ) 000Values in parentheses refer to the highest resolution shell . R factor = Σ\|\|F ( obs ) - F ( calc ) \|\|/Σ\|F ( obs ) \| . Rfree = R factor calculated using 5 . 0% of the reflection data randomly chosen and omitted from the start of refinement . RecJd denotes catalytic inactive drRecJ ( H160A ) drRecJ contains four domains: an N-terminal DHH domain ( residues 49–295 ) , a DHHA1 domain ( residues 328–424 ) , an OB fold domain ( residues 1–48 and residues 425–532 ) , and an extended C-terminal domain ( residues 533–705; Figure 2A–B and Figure 1—figure supplement 1 ) . The DHH domain and DHHA1 domain are interconnected by a long linker α-helix ( α15; residues 296–327 ) to form the nuclease core , as previously observed for the structure of ttRecJ ( Yamagata et al . , 2002 ) . The overall structure and conformation of the drRecJ complexes can be virtually superimposed on the ttRecJ ( PDB code: 2ZXP ) with the rmsd value of 1 . 937–2 . 307 Å over 476 Cα atoms ( Figure 2—figure supplement 1A ) . Although the DHHA1 domain and OB fold domain can be adequately superimposed , the DHH domain and linker α-helix further shift towards the DHHA1 domain , resulting in a much narrower cleft that accommodates the substrate DNA ( Figure 2—figure supplement 1A ) . The C-terminal domain shows large movement relative to the nuclease core , which is most likely due to the crystal lattice contacts . Three loops , which are located at the OB fold domain , also show noticeable deviations due to interactions with the substrate DNA binding ( Figure 2—figure supplement 1A ) . Above the active site , the β3-α6 loop , which is disordered in the ttRecJ structures , caps the DNA ( Figure 2A ) . The C-terminal-most α-helix ( α-CT ) also becomes ordered in all the drRecJ structures ( Figure 2A ) . Recognition of the ssDNA ( complex II and III ) by the drRecJ is mediated by nuclease core and OB fold domain ( Figure 2A–B and Figure 2—figure supplement 1B ) . The DHHA1 domain and the DHH domain bind the 5´-upstream region of the ssDNA . The two-metal ion active site in the DHH domain stabilizes the scissile phosphate . Three helices α13 , α15 and α16 form a helical gateway , which only permits ssDNA passage . The OB fold domain , located directly at the side of the DNA entrance , interacts with the downstream region of the ssDNA . The C-terminal domain , in contrast , is involved in SSB-Ct binding . These DNA binding properties were confirmed by mutagenesis coupled with nuclease assays ( Figure 2C ) . Compared with wild-type protein , the nuclease core of drRecJ exhibits a much decreased nuclease activity and processivity on ssDNA , which indicates that the OB fold domain is critical for substrate DNA binding . However , drRecJ lacking the C-terminal domain ( ΔCTD ) also exhibits reduced nuclease activity , which suggests that the C-terminal domain is possibly involved in OB fold domain orientation . As the representative member of the DHH family proteins , the DHH domain of drRecJ consists of α/β repeats , in which five central parallel β-strands ( β2 to β6 in drRecJ ) are surrounded by α-helices ( Figure 2—figure supplement 2A ) . The order of the parallel β-strands is 21345 , which is shared by all the DHH family proteins ( e . g . , NrnA in Figure 2—figure supplement 2A ) ( Uemura et al . , 2013; Ugochukwu et al . , 2007; Yamagata et al . , 2002 ) . The signature DHH residues ( motif III ) are located at the end of the fourth strand . The residues at the end of the first and third β-strands and the DHH motif coordinate two catalytic metal ions to form the active site ( Figure 2—figure supplement 2A ) . The arrangement of the catalytic residues appears to be employed by many other nucleases ( Figure 2—figure supplement 2A ) . For example , despite the different topology of the β-strands ( 32145 ) , the catalytic residues of human flap endonuclease 1 ( hFen1 ) are also located between the end of three central parallel β-strands ( Tsutakawa et al . , 2011 ) . As noted above , subfamily 1 and subfamily 2 DHH family proteins have distinct domains next to the DHH domain , which are denoted DHHA1 and DHHA2 , respectively ( Figure 1A ) . Both the DHHA1 domain and the DHHA2 domain are structurally similar with a mixed five-stranded central β-sheet surrounded by α-helices ( Figure 2—figure supplement 2B ) . The topologies of the β-strands , however , are different from DHHA1 ( 12354 , ↑↑↓↑↓ ) and DHHA2 ( 12345 , ↑↓↓↑↓ ) . In addition , the β1-β2 and β3-β4 strands in the DHHA1 domain have two inserted α-helices ( A and B in RecJ and NrnA , respectively , in Figure 2—figure supplement 2B ) ; in the DHHA2 domain , the α-helices are located between the β2-β3 and β4-β5 strands ( bottom two panels in Figure 2—figure supplement 2B ) . In drRecJ , the first ( A ) α-helix ( α16 ) of DHHA1 is critical to the composition of the helical gateway ( Figure 2A ) , which may explain the varying substrate specificity between these two subfamilies . drRecJ primarily contacts DNA with helix and loop elements ( Figure 3 and Figure 3—figure supplement 1A ) . In both complex II and complex III structures , a positively charged groove between the DHH domain and the DHHA1 domain was observed to bind 5 nt 5´-upstream ssDNA in the same manner ( Figure 3A and Figure 3—figure supplement 1B ) . Multiple loops located at the interface between the DHH domain and the DHHA1 domain bind the first two nucleotides ( Figure 3—figure supplement 1A ) . In contrast , the helical gateway , which is composed of three α-helices , primarily binds the +3 to +5 nucleotides ( Figure 3C , D ) . Ala substitutions of key residues in the helical gateway ( Arg280 in α13; Arg313/Arg314 in α15 , and Lys353 in α16 ) reduce the DNA-binding and nuclease activity with multi-stops at 5–7 nt ( Figure 3E and Figure 3—figure supplement 2 ) , which suggests that the helical gateway is essential for DNA binding and translocation . The interactions between the nuclease core and DNA phosphate groups are primarily formed by a number of Arg , His , and Asn residues ( Figure 3A , D ) . Notably , four side chain residues form stacking interactions to the 5´-upstream DNA bases ( Figure 3D ) . Tyr114 and Tyr80 ( motif I ) interact with the +1 base at the active site . The substitution of deoxyribose sugar with ribose sugar clashes with the Tyr114 residue , which is consistent with the notion that drRecJ nuclease only acts on DNA ( Figure 1B ) . Val224 ( motif IV ) and Phe269 ( motif V ) insert between the +2/+3 base and +4/+5 base . Ala substitutions of these two residues also impair the DNA-binding , nuclease activity and processivity of drRecJ ( Figure 3E and Figure 3—figure supplement 2 ) . 10 . 7554/eLife . 14294 . 011Figure 3 . The DNA binding in the RecJ-DNA complex . ( A ) Schematic of the numbered DNA substrate used for complex II crystallization . Interactions between nucleotides and DHH domain , DHHA1 domain and OB fold domain are colored pink , blue and red , respectively . Hydrogen bonds are defined as within 3 . 2 Å and van der Waals contacts within 4 . 2 Å ( dashed lines ) . Solid lines indicate residues that stack with DNA bases . ( B ) drRecJ surface and 2Fo−Fc electron density of DNA contoured at 1σ . The C7-G12 base pair is labeled . ( C ) The helical gateway is labeled and shown in cyan . Key residues interacting with DNA are labeled and shown as sticks . Nucleotides are labeled as in ( A ) . ( D ) Interactions between drRecJ and DNA in complex II structure . Protein side chains involved in the protein-DNA interactions are shown as sticks , and the key residues that form stacking interactions are highlighted in yellow . Nucleotides are labeled as in ( A ) . C7 and G12 ( orange ) form Watson-Crick base pair , as indicated by the dark dashed line . ( E ) Denaturing PAGE gel showing the reduced nuclease activity and processivity of mutant drRecJ proteins ( alanine substitutions of key residues involved in DNA binding ) . 3′-Fluorescence-labeled 46 nt ssDNA ( 100 nM ) was incubated with drRecJ proteins ( 0 , 5 and 20 nM ) in the presence of 100 nM Mn2+ ( see methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 01110 . 7554/eLife . 14294 . 012Figure 3—figure supplement 1 . The DNA binding of drRecJ . ( A ) Key structural elements in complex II . Structural elements of DNA binding and metal ions are indicated and colored according to the scheme in Figure 2A . Loops involved in nucleotide binding are also indicated . ( B ) The distribution of the electrostatic surface of drRecJ . Blue and red represent the positive and negative charge potential at the + and −10 kTe−1 scale , respectively . Electropositive patch were observed as DNA and SSB-Ct binding sites . ( C ) Denaturing PAGE gel showing that mutations of key residues involved in nuclease core-DNA interactions impaired the nuclease activity , processivity and the digestion of DNA with 5´-ssDNA overhang ( stops at ss-dsDNA junction ) . For the reaction , 3′-fluorescent labeled DNA containing 5′-ssDNA overhang ( KY04 , 100 nM ) were incubated with wild-type or mutant drRecJ proteins ( 10 and 20 nM ) in the presence of 100 nM Mn2+ . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 01210 . 7554/eLife . 14294 . 013Figure 3—figure supplement 2 . DNA binding activity of drRecJ mutant proteins as in the Figure 3E . ( A ) Electrophoretic mobility shift assays were performed with 100 nM 3′-fluorescent labeled 20 nt poly ( dA ) and different concentrations of RecJ ( 31 . 5 , 62 . 5 , 125 , 250 , 500 , 1000 and 2000 nM ) . ( B ) A plot of quantified relative band intensities of the RecJ-DNA complex bands from ( A ) . Data are represented as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 013 The OB fold domain of drRecJ at the side of the DNA entrance interacts with three downstream nucleotides ( +8 to +10; Figure 3A , D ) . This domain comprises an N-terminal region and a typical OB fold located next to the DHHA1 domain ( Figure 2A ) . Five β-strands ( β13 to β17 ) in the OB fold are orthogonally located to form a mixed β-barrel ( Figure 4A ) , which has a Greek topology ( 12354 ) that is identical to the topology of other proteins ( Figure 4—figure supplement 1 ) ( Bernstein et al . , 2004; Bochkareva et al . , 2002; Leiros et al . , 2005; Raghunathan et al . , 2000 ) . Compared with the canonical OB fold involved in ssDNA binding , the OB fold of drRecJ exhibits certain discrepancies . The L2 loop ( β14-β15 loop ) is short and the top of the β-barrel is capped by an additional α helix ( Figure 4A and Figure 4—figure supplement 1 ) . Conversely , the bottom α helix that connects the L3 loop ( β15-β16 loop ) is absent in drRecJ ( Figure 4A and Figure 4—figure supplement 1 ) . Interestingly , a similar structural features were observed in the previously solved drRecO structure ( Leiros et al . , 2005 ) ( Figure 4—figure supplement 1E ) , suggesting the possible coevolution of the RecF pathway . Surprisingly , no direct interaction between the side chain residues and the DNA phosphate group was observed ( Figure 3A , D ) . Two conserved aromatic residues , Tyr496 and Trp517 from the L3 and the L4 loop ( β16-β17 loop ) , with residues ( Met494 , Arg475 and Val477 ) between the end of the first three β-strands form the downstream DNA-binding surface , which interacts with the nucleotide bases ( Figure 4A , B ) . Ala substitutions of these residues reduce the nuclease activity of drRecJ on both ssDNA and DNA with 5´-ssDNA overhangs ( Figure 4C ) . Notably , Y496A mutant protein exhibits more severely impaired nuclease activity on DNA bearing 5´-ssDNA overhang compared with ssDNA ( Figure 4C ) , suggesting that the OB fold is critical for the drRecJ resection of DNA with 5´-ssDNA overhang . 10 . 7554/eLife . 14294 . 014Figure 4 . The OB fold domain is critical for drRecJ resection . ( A ) The OB fold domain is shown in rainbow-colored diagrams . Residues involved in DNA binding are labeled and shown as sticks . ( B ) A comparison of DNA in the complex II ( DNA with 5´-ssDNA overhang; orange ) with complex III ( ssDNA; white ) . Two conserved aromatic residues Tyr496 and Trp517 are shown as sticks ( yellow ) . The black arrowhead indicates the position of stacking interaction between Tyr496 and the guanine base . ( C ) Quantification and plot of ssDNA and DNA with 5´-ssDNA overhang , which are processed by wild-type and mutant drRecJ proteins ( alanine substitutions of key residues in the OB fold domain ) , using the same DNA substrate and reaction conditions as in Figure 1B . Data are represented as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 01410 . 7554/eLife . 14294 . 015Figure 4—figure supplement 1 . Structural comparison of the OB fold from E . coli SSB ( ecSSB ) , D . radiodurans SSB subunits ( drSSB_NTD and drSSB_CTD ) , D . radiodurans RecO ( drRecO ) and human replication protein A subunits ( RPA14 , RPA32 and RPA70 ) . Their PDB accession codes are shown in parentheses and references can be found in the main text . The mixed five β-strands are labeled and shown in distinct colors . Four loops are labeled as L1-L4 . Conserved aromatic residues in loop3 and loop4 are shown as sticks . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 015 Despite the disorder of the last four nucleotides , the G12 , G13 and C14 in complex II structure stack well , which mimic the double stranded DNA ( Figure 2B and Figure 3D ) . Unexpectedly , while drRecJ continuously interacts with the ssDNA ( complex III structure , Figure 4B ) , the C7 base in the 5´-ssDNA overhang of complex II is flipped out into the solvent ( Figure 3D and Figure 4B ) . The flipped out cytosine forms the Watson-Crick base pair with G12 , which causes severe bending at approximately 100° at the ss-dsDNA junction ( Figure 2A , Figure 3D and Figure 4B ) . This base flipping is most likely attributed to the stacking interaction between Tyr496 and the purine base ( G8 in complex II and G7 in complex III; Figure 4B ) . Poly ( dA ) and poly ( dT ) oligomers were synthesized to confirm the interaction between Tyr496 and purine base . Indeed , wild-type drRecJ exhibits approximate 2-fold increase in the catalytic efficiency ( kcat/Km ) for the poly ( dT ) substrate compared to the poly ( dA ) substrate ( Table 2 ) . When Tyr496 is replaced by Ala , the mutant drRecJ has a reduced Km and kcat but is no longer sensitive to the substrate sequence context , confirming the preferred stacking interaction between Tyr496 and purine base . 10 . 7554/eLife . 14294 . 016Table 2 . Kinetic parameters of wild-type and mutant drRecJ proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 016Protein-substrateKm ( nM ) kcat ( min-1 ) kcat/Km ( µM-1 min-1 ) WT-KY09 ( poly ( dT ) ) 74 . 9 ± 7 . 61 . 61 ± 0 . 0321 . 5WT-KY08 ( poly ( dA ) ) 100 . 3 ± 5 . 51 . 15 ± 0 . 0211 . 5Y496A -KY09 ( poly ( dT ) ) 102 . 2 ± 10 . 90 . 32 ± 0 . 013 . 1Y496A -KY08 ( poly ( dA ) ) 109 . 1 ± 8 . 40 . 27 ± 0 . 012 . 5WT-KY0390 . 3 ± 4 . 51 . 55 ± 0 . 0117 . 1Y80A-KY03223 . 1 ± 12 . 20 . 25 ± 0 . 021 . 1Y114A-KY03349 . 4 ± 23 . 80 . 77 ± 0 . 052 . 2R109A-KY03158 . 3 ± 19 . 90 . 25 ± 0 . 021 . 6S371A-KY03104 . 4 ± 8 . 80 . 78 ± 0 . 037 . 5R373A-KY03349 . 6 ± 26 . 20 . 61 ± 0 . 031 . 7R393A-KY03105 . 8 ± 6 . 91 . 51 ± 0 . 0214 . 2H397A-KY03321 . 3 ± 22 . 00 . 77 ± 0 . 032 . 4 In complex I ( drRecJ-dTMP ) structure , the dTMP is located at the interface between the DHH domain and the DHHA1 domain ( Figure 5A ) . The thymine base and the deoxyribose are coordinated by Tyr80 , Tyr114 and two β-strands ( β10 and β11 ) from the DHHA1 domain ( Figure 5A , B ) . The phosphate group of the dTMP is held in place by Arg109 , Arg280 ( motif V ) , Ser371 and Arg373 ( motif VI; Figure 5B ) . This mononucleotide-binding pocket is situated immediately above the active site , which is consistent with the translocation of the DNA by one nucleotide required for the exonuclease activity to proceed . Mutations of the binding pocket dramatically impair the enzymatic activity ( Figure 5C and Table 2 ) . The catalytic center of the complex I consists of two metal ions and two bridging ligands , a water molecule and the side chain of Asp135 ( motif II; Figure 5B ) . These two metal ions are bound by five aspartate residues and one histidine: one metal ion ( A ) is coordinated by Asp83 ( motif I ) , Asp135 , Asp223 ( motif IV ) and His159 ( motif III ) , whereas Asp79 , Asp81 ( motif I ) and Asp135 are the ligands to the other metal ion ( B; Figure 5C ) . Alanine substitutions of these metal-binding residues caused an almost complete inactivation of drRecJ ( Figure 5—figure supplement 1 ) , which is consistent with the results of equivalent mutations in ecRecJ ( Sutera et al . , 1999 ) . In addition , two conserved positively charged residues ( His397 and Arg393 ) from motif VII are situated at the entrance and exit to the active site ( Figure 5A , B ) . Alanine substitution of Arg393 shows a modest effect on drRecJ catalytic efficiency . 10 . 7554/eLife . 14294 . 017Figure 5 . Nucleotide binding site and the catalysis . ( A ) dTMP binding pocket between the DHH domain and DHHA1 domain . dTMP is shown as stick ( orange ) . Two conserved residues His397 and Arg393 at the entrance and exit to the active site are labeled and shown as sticks . ( B ) Close-up of the active site of drRecJ-dTMP ( complex I ) . Two Mn2+ ions ( A and B; magenta ) are coordinated by a water molecule ( red sphere ) and conserved Asp and His residues ( magenta dashed lines ) . Asp158 forms a hydrogen bond with His160 , as indicated by the pink dashed line . The phosphate of the dTMP is held by Arg109 , Arg280 , Ser371 and Arg373 ( blue dashed lines ) . ( C ) Catalytic efficiency of wild-type and mutant drRecJ as a bar graph showing the relative severity of the mutations . The Michaelis-Menten kinetics data are from Table 2 . Data are represented as mean ± SEM . ( D ) Close-up of the active of site of complex III . The scissile phosphate centered on two catalytic Mn2+ is highlighted in cyan . The red arrowhead indicates the position of P-O bond breakage . Interaction between His397 and the oxygen atom of the scissile phosphate group is indicated by blue dashed line . ( E ) A comparison of the active sites in the complex I ( white ) with complex III ( same color as in panel D ) . The nucleophilic water molecule observed in complex III occupies the position close to the His160 in drRecJ-dTMP structure ( complex I ) . The red arrowhead indicates the position of P-O bond breakage and the black arrowhead indicates the direction of nucleophilic attack . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 01710 . 7554/eLife . 14294 . 018Figure 5—figure supplement 1 . Denaturing PAGE gel showing the inactivation of mutant drRecJ proteins ( alanine substitutions of residues involved in metal-ion-chelation ) . KY08 was incubated with drRecJ proteins ( 0 , 5 and 20 nM ) in the presence of 100 nM Mn2+ . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 018 The active site of the drRecJ-DNA can be virtually superimposed well with that of the drRecJ-dTMP ( Figure 5D , E ) . The DNA scissile phosphate in drRecJ-DNA is centered on two Mn2+ ions and the A site metal ion seems to be in a position on the nucleophile side ( Figure 5D ) . The deoxyribonucleioside in complex I occupies almost the same place as the +1 nucleotide observed in drRecJ-DNA ( Figure 5E ) . Despite of the slightly movements , the coordination of the two catalytic metal ions is almost identical , except for an additional coordination between the A site metal ion and a water molecule ( Figure 5D , E ) . This water molecule is located directly below the scissile phosphate at a distance of 3 . 3 Å and is a probable nucleophile candidate for the nucleophilic attack . However , the angle between the water , scissile phosphate and 3´-O leaving group is ~115 degree ( Figure 5E ) , which explains the inactivity of the drRecJ-DNA complex . This DNA conformation is possibly attributed to the Ala substitution of His160 , which most likely serves as the general base essential for catalysis ( Figure 5E ) . Alanine substitutions of the conserved His160 and Asp158 , which is hydrogen bonded to His160 ( Figure 5B , E ) , inactivate drRecJ nuclease activity ( Figure 5—figure supplement 1 ) , which suggests that the DHH motif is critical for the catalysis . In addition , His397 from β10 undergoes a substantial rotamer change during DNA binding ( Figure 5E ) . This residue , which is located opposite to the direction of the nucleophilic attack in drRecJ-DNA complex , is hydrogen bonded to the oxygen atom of the scissile phosphate group ( Figure 5D , E ) . Alanine substitution of His397 drastically reduced the rate of the drRecJ-catalyzed reaction ( Figure 5C ) , suggesting that this residue may serve as a general acid to protonate the 3´-O leaving group . Inspection of the complex III electron density map reveals an additional electron density corresponding to the C-termini of the SSB-Ct peptide associated with the drRecJ C-terminal domain ( Figure 6A ) . Clear electron density was observed as the last four residues of SSB-Ct peptide ( residues 298–301 , EDDLPF ) used for co-crystallization . In contrast , the first two residues ( Glu296 and Asp297 ) of the SSB-Ct appear to be disordered . Three helices α20 , α21 and α23 form the SSB-Ct peptide-binding site , which anchors the last SSB-Ct residue ( Phe301 ) in a deep hydrophobic pocket ( Figure 6A ) . The aromatic Phe side chain is packed against Pro553 , Pro628 and the side chains of Met557 , Val572 , Tyr575 and Tyr600 from drRecJ ( Figure 6B ) . Moreover , Asp298 of SSB-Ct forms an apparent ionic bond with the Lys629 side chain , which composes an electropositive patch near the N-terminus of SSB-Ct ( Figure 6B and Figure 3—figure supplement 1B ) . These SSB-Ct binding site features are similar to those found in the ribonuclease HI SSB-Ct binding site and appears to be shared by other SSB partner proteins ( Figure 6—figure supplement 2 ) ( Marceau et al . , 2011; Petzold et al . , 2015; Ryzhikov et al . , 2011 ) . Y575A mutant drRecJ ( RecJ575 ) , wild-type drSSB ( SSBWT ) and drSSB , which lacks eight C-terminal residues ( SSBΔC ) , were purified to confirm the interactions between the SSB-Ct domain and the C-terminal domain ( Figure 6C and Figure 6—figure supplement 1 ) . drRecJ or RecJ575 alone can not process DNA with a 3´-ssDNA overhang ( Figure 6C , lanes 2–3 and 8–9 ) , which is consistent with previous biochemical observations of ecRecJ ( Morimatsu and Kowalczykowski , 2014 ) . For wild-type drRecJ , this nuclease activity is enhanced by SSBWT ( lanes 4–5 ) but not by SSBΔC ( lanes 6–7 ) . In contrast , neither SSBWT ( lanes 10–11 ) nor SSBΔC ( lanes 12–13 ) stimulate the RecJ575 resection of DNA with a 3´-ssDNA overhang substantially , indicating that the interactions between drRecJ and drSSB are required for the stimulation of drRecJ resection of DNA with a 3´-ssDNA overhang . For ssDNA , the stimulation is similar to that of the resection of the DNA with a 3´-ssDNA overhang , suggesting that the SSB-RecJ interaction is required for RecJ to resect SSB-covered ssDNA ( Figure 6—figure supplement 1 ) . 10 . 7554/eLife . 14294 . 019Figure 6 . The C-terminal domain interacts with the SSB-Ct . ( A ) SSB-Ct binding pocket . SSB-Ct is shown as stick . The electron density of SSB-Ct is shown in blue with the refined 2Fo-Fc map contoured at 1σ . ( B ) Interactions of the SSB-Ct and C-terminal domain . Both SSB-Ct ( yellow ) and residues involved in SSB-Ct interactions ( green ) are shown as sticks . The ionic bond between the Asp298 of SSB-Ct and Lys629 of drRecJ is indicated by a yellow dashed line . ( C ) SSB enhances wild-type drRecJ degradation . Y575A mutant drRecJ ( RecJ575 ) , wild-type drSSB ( SSBWT ) and drSSB lacking eight C-terminal residues ( SSBΔC ) were purified to perform the nuclease assays . For the reaction , DNA with a 3´-ssDNA overhang was pre-incubated with 200 nM SSBWT or SSBΔC and treated with drRecJ or RecJ575 ( 5 and 20 nM ) ( see Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 01910 . 7554/eLife . 14294 . 020Figure 6—figure supplement 1 . SSB enhances wild-type drRecJ degradation on ssDNA . Y575A mutant drRecJ ( RecJ575 ) , wild-type drSSB ( SSBWT ) and drSSB lacking eight C-terminal residues ( SSBΔC ) were purified to perform the nuclease assays . For the reaction , 100 nM ssDNA was pre-incubated with 200 nM SSBWT or SSBΔC and treated with drRecJ or RecJ575 ( 2 . 5 and 5 nM , or 10 and 20 nM ) ( see Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 02010 . 7554/eLife . 14294 . 021Figure 6—figure supplement 2 . Structural comparison of the SSB-Ct binding pockets of RecJ , Rnase HI ( 4Z0U ) , RecO ( 3Q8D ) and Pol III ( 3SXU ) . The SSB-Ct are shown as sticks and colored in yellow . Residues interact with SSB-Ct are labeled , shown as sticks and colored in green . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 021 The α-CT is located outside the C-terminal domain and exposed to solvent , which has no effect on drRecJ nuclease activity ( Figure 2C ) . To test whether the α-CT is required for drRecJ function in vivo , we performed the phenotypic assays ( Figure 7A ) . Cells that lack recJ ( ΔrecJ ) are thermosensitive and sensitive to MMC treatment ( Figure 7A ) . Interestingly , while complementation of the entire RecJ in ΔrecJ strain ( ΔrecJ/pk-recJ ) completely recovered the WT phenotype , overexpression of RecJ that lacks α-CT ( ΔrecJ/pk-recJΔCα ) , only partially recovered the phenotype , suggesting the essential role of the α-CT for drRecJ function in vivo ( Figure 7A ) . It has been shown that RecQ and RecJ biochemically cooperate to process DNA with 3´-ssDNA overhang in E . coli ( Morimatsu and Kowalczykowski , 2014 ) . However , drRecQ is able to activate both wild-type drRecJ and drRecJ lacking α-CT resection of DNA with 3´-ssDNA overhang ( unpublished data ) similarly , suggesting that the α-CT is possibly involved in the interaction between drRecJ and other helicases in vivo . Moreover , although both drSSB and drRecQ can enhance drRecJ nuclease activity on DNA with 3´-ssDNA overhang , the resection is substantially stimulated when both drRecQ and drSSB were present ( Figure 7B ) , which suggests that the DNA with 3´-ssDNA overhang is most likely processed by the combined activities of RecJ , RecQ and SSB in vivo ( Figure 7C ) . 10 . 7554/eLife . 14294 . 022Figure 7 . DSB end resection requires the coordinate activities of RecJ , RecQ and SSB proteins . ( A ) Functional analysis of the drRecJ α-CT in vivo . Wild-type ( WT ) , recJ mutant ( ΔrecJ ) and recJ complementary ( ΔrecJ/pk-recJ for the entire RecJ and ΔrecJ/pk-recJΔCα for RecJ lacking α-CT ) strains were spotted on TGY medium following high-temperature ( 37 degrees ) and MMC treatments . ( B ) drRecJ processes DNA bearing 3´-ssDNA overhang together with drRecQ and drSSB . The reaction contained drRecJ ( 5 , 10 and 20 nM ) , drRecQ ( 100 nM ) and drSSB ( 100 nM ) . ( C ) A model for DSB end resection by RecJ , RecQ and SSB proteins in D . radiodurans . RecQ ( yellow ) is bound to the ss-dsDNA junction , which unwinds them to generate 5´-tailed ssDNA . Following RecJ digestion , the SSB ( homodimer , white and grey ) is recruited to the resultant 3´-ssDNA overhang , which facilitates further strand exchange reaction . DOI: http://dx . doi . org/10 . 7554/eLife . 14294 . 022 DHH family proteins have been found widespread in bacteria , archaea and eukaryotes , which contain five conserved motifs . Analyses of the conserved DHH domain and the nonconserved DHHA1 domain in drRecJ structures provide insight into the conserved catalysis mechanism but different substrate specificity of these proteins . As a family of proteins , they share the conserved topology of β-strands in the DHH domain , the signature DHH motif at the end of the fourth strand and the catalytic residues ( Figure 2—figure supplement 2A ) . The core of the DHH domain , which consists of a central five parallel β-strands sandwiched by surrounding α helices , appears to be shared by many other nucleases . The similar Rossmann-like arrangement of the active site suggests that these nucleases may employ similar two-metal-ion catalysis ( Figure 2—figure supplement 2A ) . The DHHA1 domain in the subfamily 1 group , however , shows distinct features from the DHHA2 domain in the subfamily 2 group . In addition to the different topology of the β-strands , the subfamily 2 group lacks the helical gateway element ( α16 inserted between the first two β-strands in DHHA1 ) as observed in drRecJ . More importantly , the DHH domain of the subfamily 1 group faces the C-terminal three β-strands of DHHA1 to form the DNA binding site ( Figure 2—figure supplement 3A ) , whereas the active site of the subfamily 2 group is formed between the DHH domain and the first β-strand of DHHA2 ( Figure 2—figure supplement 3B ) . In addition , the C-terminal α-helix of DHHA2 caps the active site , resulting in a much narrower cleft that only accommodates the small-size pyrophosphate or polyphosphate . Thus , the relative positions of the DHH and the DHHA1/DHHA2 domain may determine the substrate specificity between these two subfamily groups . CDC45 , which is in the subfamily 3 group , is a eukaryotic orthologue of RecJ with no nucleolytic cleavage activity but retaining ssDNA binding capability ( Krastanova et al . , 2012; Petojevic et al . , 2015; Szambowska et al . , 2014 ) . In eukaryotes , CDC45 , the Mcm2-7 helicase and GINS proteins form the CMG complex , which is involved in DNA replication ( Miller et al . , 2014; Pacek et al . , 2006 ) . Recent studies revealed that CDC45 may guard the gate of Mcm2/5 helicase and capture the errant leading strand ( Petojevic et al . , 2015 ) . Based on sequence alignment , secondary structure prediction and our drRecJ structures ( Figure 1—figure supplement 1 ) , we provide a possible mechanism of CDC45 binding ssDNA . CDC45 exhibits sequence homology to the drRecJ DHH domain , which possesses five conserved motifs . However , three key residues ( Asp79 , Asp135 and His159 in drRecJ ) that coordinate catalytic metal ions are not conserved in CDC45 , which is consistent with the inactivation and the absence of metal ion in purified CDC45 proteins ( Krastanova et al . , 2012 ) . Second , residues that compose the terminal 5´-phosphate-binding pocket ( Arg109 , Ser371 and Arg373 in drRecJ ) are missing in CDC45 , which may explain the opposite 3´-5´ DNA binding polarity of CDC45 compared with RecJ ( Szambowska et al . , 2014 ) . Moreover , DNA binding for RecJ is metal ion independent ( Figure 3—figure supplement 2 ) ( Han et al . , 2006 ) . The residues involved in protein-DNA stacking interactions ( Tyr114 , Val224 and Phe269 in drRecJ ) and the helical gateway ( Arg280 , R313 , R314 and Lys353 in drRecJ ) are conserved in CDC45 , which suggests that the DNA binding of CDC45 is possibly similar to that of RecJ . The structures presented in this study reveal an elegant mechanism for how RecJ binding to DNA initiates DSB end resection in a 5´-3´ direction . First , the terminal of ssDNA is anchored to the 5´-phosphate binding pocket above the active site , which explains the RecJ hydrolyzing DNA with 5´-3´ polarity . This terminal 5´-phosphate-binding pocket appears to be shared by other 5´-3´ exonucleases such as the λ exonuclease ( Zhang et al . , 2011 ) , XrnI ( Jinek et al . , 2011 ) and RNaseJ ( Zhao et al . , 2015 ) . Second , the active site encloses a single deoxyribonucleioside , which renders RecJ an exonuclease discriminating DNA over RNA . Third , the entrance to the RecJ active site is guarded by the helical gateway , which prevents dsDNA from entering the active site . Fourth , a number of residues form stacking interactions with every two DNA bases . Biochemically , the end resection by the RecJ nuclease is processive . Alanine substitution of these residues resulted in multiple intermediate digestion bands , which indicates that the protein-DNA staking interactions are required for the mechanism of processivity of the RecJ nuclease . Fifth , the OB fold domain is located on the side of the DNA entrance , which positions the downstream DNA towards the active site . It has recently been shown that RecJ alone is able to process DNA with a 5´-ssDNA overhang ( Morimatsu and Kowalczykowski , 2014 ) . Mechanistically , RecJ is required to locally melt dsDNA and bind to the consequential 5´-ssDNA for the resection . Our analyses suggest that both the nuclease core and the OB fold domain contribute to the RecJ resection of DNA that bears a 5´-ssDNA overhang . In addition to the reduced nuclease activity and processivity , the mutation of the key residues involved in nuclease core-DNA interactions showed strong reaction-stops at the ss-dsDNA junction ( Figure 3—figure supplement 1C ) , which suggests that the complete processivity is essential for the RecJ resection of DNA with a 5´-ssDNA overhang . On the other hand , the OB fold domain is located at the ss-dsDNA junction . Ala substitution of Tyr496 impaired the nuclease activity against DNA with 5´-ssDNA overhang and ssDNA , but to a lesser extent , which indicates that the OB fold domain is also critical for this type of end resection . Interestingly , stacking interaction between Tyr496 and a purine base appears to be preferred , which may also be important for the dsDNA melting . Our drRecJ:DNA and drRecJ:dTMP structures suggest that RecJ nucleases employ the two-metal-ion catalysis observed for many nucleic acid processing enzymes ( Nakamura et al . , 2012; Yang , 2011; Zhao et al . , 2013; 2015 ) . In summary , after deprotonation , a water molecule serves as the nucleophile poised on the 5´-side to attack the scissile phosphate . After bond breakage , His397 , which is located at the opposite side of the nucleophilic attack , stabilizes the 3´-O leaving group . The resultant mononucleotide and one or both metal ions can be directly exported through the exit tunnel ( Arg393 ) . The newly exposed 5´-phosphate , which is generated in each cycle of the exonuclease reaction , further translocates into the terminal 5´-phosphate-binding pocket , making the next nucleotide ready for the cleavage . RecJ is able to digest DNA that contains abasic sites ( Dianov et al . , 1994 ) ; this binding pocket is located right above the active site , which indicates that the attraction of 5´-phosphate to this binding pocket helps to accurately position the scissile phosphodiester bond of the DNA substrate in the active site . D . radiodurans is capable of surviving high doses of ionizing radiation , which shatters its genome into several hundred fragments . The resulting numerous DSBs can be rapidly repaired by its super-efficient RecA-mediated DNA repair system ( Bentchikou et al . , 2010; Slade et al . , 2009 ) . The functional RecF pathway is required for the RecA-mediated DSB repair in vivo , as D . radiodurans naturally lacks RecB and RecC enzymes . E . coli has a minimum of three 5´-3´ exonucleases ( RecJ , RecBCD and ExoVII ) , whereas RecJ is the only 5´-3´ exonuclease in D . radiodurans , which plays an essential role in DSB end resection ( Lovett , 2011 ) . In contrast to the use of Mg2+ for DNA digestion by ecRecJ , Mn2+ appears to be preferred for drRecJ catalysis . This metal-ion preference may be attributed to the following two properties: First , the ionic radius of Mn2+ is similar to that of Mg2+ but with less-rigid coordination requirements ( Harding , 2000 ) , making it suitable for His159 coordination . In fact , only one Mg2+ was observed in crystal form that contains Mg2+ instead of Mn2+ ( unpublished data ) , which is consistent with the extremely low drRecJ nuclease activity in the presence of Mg2+ ( Figure 1—figure supplement 2 ) . Second , Daly and co-workers noted that D . radiodurans contained about 300 times more intracellular Mn2+ than that in E . coli , which plays a critical role in reactive oxygen species detoxification and protein protection ( Daly et al . , 2004; Daly , 2009 ) . Thus , the Mn2+ preference of drRecJ may also be an example of the evolution adaptation . Indeed , in D . radiodurans , many enzymes involved in DNA repair ( e . g . , PolI , PolX ) prefer to use Mn2+ for catalysis ( Cheng et al . , 2015a; Heinz and Marx , 2007; Lecointe et al . , 2004; Zhang et al . , 2014; Zhao et al . , 2015 ) , which indicates that the accumulation of Mn2+ may also contribute to the robust DSB repair . Importantly for the efficient HR , the DSB with different end structures must be promptly processed . Our structures and biochemical data are consistent with the helicase-nuclease coordination end resection mechanism ( Figure 7C ) and suggest that the extended C-terminal domain of drRecJ , which is absent in E . coli , is critical for its efficient end resection in D . radiodurans . The C-terminal domain can enhance the nuclease activity of drRecJ in vitro ( Figure 2C ) , which suggests that drRecJ may exhibit a more effective nuclease activity in D . radiodurans . In addition to ssDNA , drRecJ alone is able to initiate DNA end resection with 5´-ssDNA overhangs processively . Interestingly , drSSB stimulates drRecJ resection of DNA that bears a 3´-ssDNA overhang to a certain extent in vitro , possibly via the direct interaction between the drRecJ C-terminal domain and SSB-Ct , which suggests that drRecJ together with drSSB might be capable of processing this type of DSB end structure in vivo . On the other hand , SSB in E . coli recruits RecO protein to ssDNA through SSB-Ct . drRecO does not interact with SSB-Ct ( Ryzhikov et al . , 2011 ) , which suggests the existence of alternative pathway of HR initiation in D . radiodurans . drRecQ further stimulates the reaction , which indicates that the 3´-5´ helicase activity is also required for the efficient end resection . However , we found no evidence for the direct interaction between drRecJ and drRecQ , which suggests that other redundant helicases ( e . g . , UvrD ) may also assume the unwinding responsibilities in vivo ( Bentchikou et al . , 2010; Ithurbide et al . , 2015 ) . Collectively , we report the crystal structure of drRecJ in complex with SSB-Ct peptide and a DNA substrate representing a DSB end substrate , with 5´-ssDNA overhang placed within the active site , where DSB end resection occurs . These observations reveal the shared and distinctive features of DHH family proteins and enable us to propose a unified mechanism for substrate recognition and the exonucleolytic cleavage activity for RecJ family nucleases . The novel C-terminal domain involved in protein-protein interaction suggests a more effective DSB end resection in D . radiodurans . Full-length ( residues 1-705aa ) , nuclease-core ( drRecJcore , residues 48-431aa ) , C-terminal domain truncation ( drRecJΔC , residues 1-531aa ) , and C-terminal α helix truncation ( drRecJΔα-CT , residues 1-690aa ) of drRecJ were amplified by PCR and cloned to the modified pET28a expression vector , pET28-HMT , which contains a fused N-terminal 6×His-tag , a MBP-tag and a TEV protease recognition sequence ( His-MBP-TEV ) as described ( Austin et al . , 2009 ) . Full-length drSSB ( residues 1-301aa ) , C-terminal truncated drSSB ( drSSBΔC , residues 1-293aa ) , and full-length drRecQ ( residues 1-824aa ) were also cloned to pET28-HMT expression vector . All the constructed expression vectors were transformed into Escherichia coli Rossetta ( DE3 ) strain . For phenotypic assays , drrecJΔCα was also cloned to the shuttle vector pRADK , named as pk-recJΔCα . drrecJ complemented strain ( ΔrecJ/pk-recJΔCα ) was constructed by transforming pk-recJΔCα into drrecJ knockout strain ( ΔrecJ ) as described previously ( Jiao et al . , 2012 ) , followed by sequencing identification . Site directed mutagenesis was performed with a QuikChangeTM Site-Directed Mutagenesis Kit from Stratagene ( La Jolla , CA ) as described ( Cheng et al . , 2015a ) . Primers used for cloning and mutageneses are listed in Supplementary file 1 . All bacterial strains and plasmids used in this study are listed in Supplementary file 2 . All the drRecJ proteins were expressed and purified in a similar way as reported previously ( Cheng et al . , 2015b; Jiao et al . , 2012 ) . In brief , transformed Escherichia coli Rossetta ( DE3 ) clones were grown at 37°C in LB medium containing 50 μg/ml Kanamycin to an optical density at 600 nm of 0 . 6–0 . 8 . Protein expression was induced at 30°C for 5 hr by adding isopropyl-β-D-thioga-lactopyranoside ( IPTG ) with a final concentration of 0 . 4 mM . After harvesting , cells were resuspended in lysis buffer ( 20 mM Tris ( pH 7 . 5 ) , 1 M NaCl , 5% ( w/v ) glycerol , 3 mM β-ME and 10 mM imidazole ) , lysed by sonication and centrifuged at 15 , 000 × g for 30 min at 4°C . The supernatant was purified by HisTrap HP column ( GE Healthcare ) , equilibrated with buffer A ( 20 mM Tris ( pH 7 . 5 ) , 1 M NaCl , 5% ( w/v ) glycerol and 10 mM imidazole ) , washed with 30 mM imidazole and finally eluted with 300 mM imidazole . After TEV-tag-removal using TEV protease , the protein was loaded onto the MBPTrap HP column ( GE Healthcare , Fairfield , CT ) to remove the uncleaved protein . The flow-through fractions were collected and loaded onto a Heparin HP column ( GE Healthcare , Fairfield , CT ) pre-equilibrated with buffer B ( 20 mM Tris ( pH 7 . 5 ) , 100 mM NaCl , 1 mM DTT , 5% ( w/v ) glycerol ) . Fractions containing drRecJ proteins were eluted with a linear gradient from 100 mM to 500 mM NaCl . The protein was finally purified by Superdex 200 10/300 GL column ( GE Healthcare , Fairfield , CT ) with buffer C ( 20 mM Tris ( pH 7 . 5 ) , 100 mM NaCl , 0 . 1 mM EDTA and 1 mM DTT ) and stored at −80°C . drSSB and drSSBΔC were expressed similar as drRecJ , and purified as reported previously ( Cheng et al . , 2014 ) . drRecQ was induced at 16°C for 16 hr by adding isopropyl-β-D-thioga-lactopyranoside ( IPTG ) with a final concentration of 0 . 4 mM . The purification procedure of drRecQ was similar as drSSB , expect that Superdex 200 column was used instead of Superdex 75 in the final purification step . Crystallization trials were carried out by the sitting drop vapor diffusion method at 293 K . Fresh purified drRecJ was concentrated to ~18 mg/ml and centrifuged to remove insoluble fraction before crystallization . After a series of screening tests and optimizations , crystals of RecJ-dTMP ( complex I ) were obtained in 1 . 3 M Li2SO4 , 100 mM MES ( pH 6 . 5 ) , 2 . 5 mM MnCl2 and 10 mM dTMP . For RecJ:DNA complex , protein was mixed with DNA ( Supplementary file 3 ) at a 1:1 . 2 molar ratio and concentrated to ~18 mg/ml . Complex II crystals were grown in 1 . 5 M Li2SO4 , 100 mM MES ( pH 6 . 5 ) , and 2 . 5 mM MnCl2 . Complex III crystals were grown in similar condition , followed by soaking with 0 . 2 mg/mlSSB-Ct ( EDDLPF ) peptide for 24 hr . Cryocooling was achieved by stepwise soaking the crystals in reservoir solution containing 10 , 20 , and 30% ( w/v ) glycerol for 3 min and flash freezing in liquid nitrogen . Diffraction intensities were recorded on beamline BL17U at Shanghai Synchrotron Radiation Facility ( Shanghai , China ) and were integrated and scaled with the XDS suite ( Kabsch , 2010 ) . The structures were determined by molecular replacement using ttRecJ ( 2ZXP ) as the search model ( McCoy et al . , 2007 ) . Structures were refined using PHENIX ( Adams et al . , 2010 ) and interspersed with manual model building using COOT ( Emsley et al . , 2010 ) . Later stages of refinement utilized TLS group anisotropic B-factor refinement . The refined model contained one drRecJ molecule in the asymmetric unit . Two catalytic Mn2+ ions and dTMP ( complex I ) , DNA ( complex II ) or DNA/SSBct ( complex III ) were observed . The statistics for data collection and refinement are listed in Table 1 . All the residues are in the most favorable and allowed regions of the Ramachandran plot . All structural figures were rendered in PyMOL ( www . pymol . org ) . All the oligo DNA and RNA were purchased from Sangon ( Shanghai , China ) and Takara ( Dalian , China ) with 3′-end labeled by 6-carboxfluorescein ( 6-FAM ) . The sequences of oligos were listed in the Supplementary file 3 . DNA with 3′- ( KY06 , 6 nt overhang ) or 5′-overhang ( KY04 , 14 nt overhang ) ssDNA was obtained by annealing in annealing buffer ( 10 mM HEPES ( pH 8 . 0 ) , 50 mM NaCl , 0 . 1 mM EDTA ) by heating ( 95°C ) for 5 min and slow cooling to 4°C . KY03-KY07 were used to determine the drRecJ substrate preference ( Figure 1B ) . KY08 was used to test the nuclease activities of different truncations of drRecJ ( Figure 2C ) . KY03 was used to test the nuclease activities of mutant drRecJ proteins ( Figure 3E and 5C ) . KY03 and KY04 were used to test the nuclease activity of drRecJ on ssDNA and DNA with 5´-ssDNA overhangs ( Figure 4C and Figure 3—figure supplement 1C ) . KY08 and KY09 were used to test the different digestion efficiency of drRecJ on poly ( dA ) and poly ( dT ) . KY06 was used to test the stimulation of drRecJ nuclease activities by drSSB and drRecQ ( Figure 6C and 7B ) . For a typical nuclease assay , 100 nM DNA was incubated with various concentrations ( 2 . 5–500 nM ) of freshly prepared full-length , truncated or mutant drRecJ proteins in a 10 μl reaction volume containing 50 mM Tris ( pH 7 . 5 ) , 60 mM KCl , 0 . 1 mg/ml BSA , 1 mM DTT , 5% ( v/v ) glycerol and 0 . 1 mM MnCl2 at 30°C for 20–30 min , in the presence or absence of 100–200 nM drSSB or drRecQ proteins . The reactions were stopped with 2×stop buffer ( 10 mM EDTA , 98% formamide ) , incubating at 95°C for 10 min and flash cooled on ice . Reaction products were resolved on 12–20% polyacrylamide gels containing 7 M urea . The gels were imaged at fluorescence mode ( FAM ) on Typhoon FLA 9500 ( GE ) , and bands were analyzed using Image J Software ( National Institutes of Health , USA ) . To determine the metal preference , KY08 was used and MnCl2 ( 0 . 01–10 mM ) , MgCl2 ( 0 . 01–10 mM ) or EDTA ( 10 mM ) was added in the reaction buffer . Additional Mg2+ ( 2 mM ) and ATP ( 2 mM ) were present in the reaction buffer when drRecQ was needed . For steady state measurements , typically 5–10 nM drRecJ were incubated with saturated substrate ( KY03 , 0–2 µM ) at 30°C for 20 min . All reactions were independently repeated at least three times . The kcat and KM were derived from generalized nonlinear least-squares using Michaelis-Menten equation , from which the apparent second order rate constant ( kcat/KM ) was determined from a plot of normalized initial rate ( v/[E] ) versus substrate concentration ( [S] ) . Kinetic parameters of wild-type and mutant drRecJ proteins can be found in Table 2 . 100 nM 3′-labeled ssDNA ( KY08 ) was incubated with different concentrations of RecJ ( 31 . 5 , 62 . 5 , 125 , 250 , 500 , 1000 and 2000 nM ) in a 20 μl reaction volume containing 50 mM Tris ( pH 7 . 5 ) , 60 mM KCl , 0 . 1 mg/ml BSA , 1 mM DTT and 5% ( v/v ) glycerol at 30°C for 20 min . Samples were separated on 8% native polyacrylamide gels in 1 × TBE buffer . The gels were imaged at fluorescence mode ( FAM ) on Typhoon FLA 9500 ( GE ) . The bands were analyzed using Image J Software ( National Institutes of Health , USA ) and the graph was created by Graphpad Prism 6 Software . D . radiodurans wild type strain R1 and its derivatives were grown at 30°C either in TGY broth ( 0 . 5% tryptone , 0 . 1% glucose , 0 . 3% yeast extract ) or on TGY agar plate ( with 1 . 25% agar ) . Phenotypic assays were performed as described previously ( Cheng et al . , 2015b ) . Cells were grown to early exponential phase ( OD600 = 0 . 6-0 . 8 ) and incubated with 20 mg/ml of MMC at 30°C for 20 min , diluted to various concentrations and dotted onto TGY plates . Plates were cultured at 30°C for 2–3 days . The cells without MMC treatment were set as control . For temperature-dependent assay , a plate was cultured at 37°C for 2–3 days .	DNA encodes information that cells need to create the molecules and proteins that are essential for life . It is therefore vital that damaged DNA is repaired rapidly and accurately . Some DNA-damaging agents , such as gamma radiation , break both strands of the DNA double helix , which can be fatal to cells if not repaired quickly and accurately . One important pathway in charge of repairing such double-strand breaks is called the homologous recombination repair pathway . The first stage of this repair involves cutting away part of one of the DNA strands at the break . This exposes a single-stranded stretch of the partner strand , which can be used for the repair . One organism that is highly resistant to having its DNA damaged by radiation is the bacterium Deinococcus radiodurans . In this bacterium , an enzyme called RecJ performs part of the first step in the repair of DNA double-strand breaks by progressively shortening one end of a DNA strand . Cheng et al . have now used crystallography to look at the structure that RecJ forms when it binds to DNA . This , together with the results from biochemical experiments , revealed how RecJ recognizes where to bind on a broken DNA strand and how it moves along the broken strand along with cutting that strand . Further investigations revealed that two other proteins enhance the ability of RecJ to process the ends of broken DNA strands . Cheng et al . also examined the structure that RecJ forms with one of these additional proteins , called SSB . A future goal is to determine how all three proteins co-ordinate with each other to efficiently and accurately repair double stranded breaks in the D . radiodurans bacteria .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics" ]	2016	Structural basis for DNA 5´-end resection by RecJ
To optimize fitness a plant should monitor its metabolism to appropriately control growth and defense . Primary metabolism can be measured by the universally conserved TOR ( Target of Rapamycin ) pathway to balance growth and development with the available energy and nutrients . Recent work suggests that plants may measure defense metabolites to potentially provide a strategy ensuring fast reallocation of resources to coordinate plant growth and defense . There is little understanding of mechanisms enabling defense metabolite signaling . To identify mechanisms of defense metabolite signaling , we used glucosinolates , an important class of plant defense metabolites . We report novel signaling properties specific to one distinct glucosinolate , 3-hydroxypropylglucosinolate across plants and fungi . This defense metabolite , or derived compounds , reversibly inhibits root growth and development . 3-hydroxypropylglucosinolate signaling functions via genes in the ancient TOR pathway . If this event is not unique , this raises the possibility that other evolutionarily new plant metabolites may link to ancient signaling pathways . Herbivory , pathogen attacks and weather fluctuations are just some of the factors that constantly fluctuate within a plants environment . To optimize fitness under this wide range of conditions , plants utilize numerous internal and external signals and associated signaling networks to plastically control metabolism and development ( Henriques et al . , 2014; Smith and Stitt , 2007; Rexin et al . , 2015 ) . This metabolic and developmental plasticity begins at seed germination , where early seedling growth is maintained by heterotrophic metabolism relying solely on nutrients and energy stored in the seed including the embryo . Upon reaching light , the seedling transitions to autotrophy by shifting metabolism to initiate photosynthesis and alters development to maximize photosynthetic capacity ( Rexin et al . , 2015; Galili et al . , 2014 ) . Until light is available , it is vital for the plant to prioritize usage from the maternal energy pool , to ensure the shoot will breach the soil before resources are depleted . Because the time to obtaining light is unpredictable , seedlings that had the ability to measure and accordingly adjust their own metabolism would likely enjoy a selective advantage . In this model , energy availability would an essential cue controlling growth throughout a plant’s life and not solely at early life-stages . On a nearly continuous basis , photo-assimilates , such as glucose and sucrose , are monitored and their internal levels used to determine the growth potential by partitioning just the right amount of sugars between immediate use and storage ( Smith and Stitt , 2007 ) . Illustrating the key nature of metabolite measurement within plants is that glucose , is measured by two separate kinase systems that are oppositely repressed and activated to determine the potential growth capacity , SnRKs1 ( sucrose non-fermenting 1 ( SNF1 ) -related protein kinases 1 ) and the Target of Rapamycin ( TOR ) kinase ( Sheen , 2014 ) . SnRKs1s are evolutionarily conserved kinases that are activated when sugars are limiting ( Lastdrager et al . , 2014 ) . Arabidopsis thaliana ( Arabidopsis ) has two catalytic SnRK1-subunits , KIN10 and KIN11 ( SNF kinase homolog 10 and 11 ) , that activate vast transcriptional responses to repress energy-consuming processes and promote catabolism ( Baena-González , 2010; Crozet et al . , 2014; Baena-González et al . , 2007 ) . This leads to enhanced survival during periods of energy starvation . Oppositely , the TOR kinase is a central developmental regulator , whose sugar-dependent activity controls a myriad of developmental processes including cell growth , cell-cycle , and cell-wall processes . The TOR pathway functions to modulate growth and metabolism by altering transcription , translation , primary and secondary metabolism , as well as autophagy ( Sheen , 2014; Sablowski et al . , 2014 ) . The TOR kinase primarily functions in meristematic regions where it promotes meristem proliferation . Within these cell types , TOR measures the sugar content and if the tissue is low in sugar , TOR halts growth , even overruling hormone signals that would otherwise stimulate growth ( Xiong et al . , 2013 ) . In plants , TOR functions within a conserved complex that includes RAPTOR ( regulatory-associated protein of TOR ) and LST8 ( lethal with sec-13 protein 8 ) ( Henriques et al . , 2014 ) . RAPTOR likely functions as an essential substrate-recruiting scaffold enabling TOR substrate phosphorylation ( Rexin et al . , 2015 ) , and LST8 is a seven WD40 repeats protein with unclear function ( Moreau et al . , 2012 ) . TOR complex ( TORC ) activity is positively linked with growth ( Rexin et al . , 2015 ) as mutants in any component lead to qualitative or quantitative defects in growth and development and even embryo arrest in strong loss-of-function alleles ( Menand et al . , 2002; Deprost et al . , 2005 ) . Although the energy sensory kinases KIN10/11 and TOR sense opposite energy levels , they govern partially overlapping transcriptional networks , which are intimately connected to glucose-derived energy and metabolite signaling ( Sheen , 2014; Baena-González et al . , 2007 ) . Having two systems to independently sense sugar shows the importance of measuring internal metabolism . A key pathway controlled by TOR in all eukaryotes is autophagy ( Liu and Bassham , 2010; Shibutani and Yoshimori , 2014 ) . In non-stressed conditions , continuous autophagy allows the removal of unwanted cell components like damaged , aggregated or misfolded proteins by vacuolar/lysosomal degradation ( Inoue et al . , 2006 ) . Under low energy conditions , TORC inhibition leads to an induction of autophagy to free up energy and building blocks , through degradation of cytosolic macromolecules and organelles ( Shibutani and Yoshimori , 2014 ) . Autophagy-mediated degradation is facilitated by formation of autophagosomes; double membrane structures that enclose cytoplasmic cargo , and delivers it to the vacuole ( Shibutani and Yoshimori , 2014; Feng et al . , 2014; Le Bars et al . , 2014; Zhuang et al . , 2016 ) . In nature , plant plasticity is not only limited to responding to the internal energy status , but to an array of external environmental inputs . The multitude of abiotic and biotic factors that plants continuously face often require choices between contradictory responses , that requires integrating numerous signals across an array of regulatory levels to create the proper answer . For example; plant defense against biotic organisms requires coordination of metabolic flux to defense and development while in continuous interaction with another organisms and the potential for interaction with other organisms . A proper defense response is vital for the plant as a metabolic defense response to one organism can impart an ecological cost by making the plant more sensitive to a different organism ( Züst et al . , 2012 ) . Therefore , a plant must choose the most appropriate defense response for each situation to optimize its fitness and properly coordinate its defense response with growth and development . A key defense mechanism intricately coordinated across development is the synthesis of specific bioactive metabolites that are often produced in discrete tissues at specific times . A current model is that developmental decisions hierarchically regulate defense metabolism with little to no feed-back from defense metabolism to development . However , work on the glucosinolate and phenolic pathways is beginning to suggest that defense metabolites can equally modulate development ( Katz et al . , 2015; Bonawitz et al . , 2012; Bonawitz et al . , 2014; Kim et al . , 2014; Francisco et al . , 2016a; Francisco et al . , 2016b ) , thus suggesting that development and defense metabolism can directly cross-talk . To asses if and how defense metabolites can signal developmental changes , we chose to investigate the glucosinolate ( GSL ) defense metabolites . The evolution of the core of GSL biosynthesis is relatively young , and specifically modified GSL structures are even more recent ( Sønderby et al . , 2010 ) . There are >120 known GSL structures limited to plants from the Brassicales order and some Euphorbiaceae family members , with Arabidopsis containing at least 40 structures ( Sønderby et al . , 2010 ) . GSLs are amino acid derived defense metabolites that , after conversion to an array of bioactive compounds , provide resistance against a broad suite of biotic attackers ( Sønderby et al . , 2010; Lambrix et al . , 2001; Kliebenstein et al . , 2002 ) . GSLs not only exhibit a wide structural diversity , but their composition varies depending on environmental stimuli , developmental stage and even across tissues . This , combined with the information-rich side chain , makes GSLs not only an adaptable defense system , but also prime candidates for having distinctive signaling functions . Previous work has suggested that there may be multiple signaling roles within the GSLs ( Francisco et al . , 2016a; Francisco et al . , 2016b; Clay et al . , 2009; Khokon et al . , 2011; Bednarek et al . , 2009 ) . Tryptophan-derived indole GSLs or their breakdown products alter defense responses to non-host pathogens illustrated by biosynthetic mutants devoid of indole GSLs being unable to deposit PAMP-induced callose in cell walls via an unknown signal and pathway ( Clay et al . , 2009; Bednarek et al . , 2009 ) . Similarly , indole GSL activation products have the ability to directly alter auxin perception by interacting with the TIR1 auxin receptor ( Katz et al . , 2015 ) . In contrast to indole GSLs , mutants in aliphatic GSL accumulation alter flowering time and circadian clock oscillations ( Kerwin et al . , 2011; Jensen et al . , 2015 ) . The aliphatic GSL activation product , allyl isothiocyanate can induce stomatal closure but it is unknown if this is specific to allylGSL or a broader GSL property ( Khokon et al . , 2011; Boller and Felix , 2009 ) . AllylGSL ( other names are 2-propenylGSL and sinigrin ) can also alter plant biomass and metabolism in Arabidopsis ( Francisco et al . , 2016a; Francisco et al . , 2016b ) . While these studies have provided hints that the GSL may have signaling potential , there is little understanding of the underlying mechanism or the structural specificity of the signal . To explore whether built-in signaling properties are a common attribute of GSLs , we screened for altered plant growth and development in the presence of specific purified aliphatic GSLs . In particular , we were interested in identifying candidate signals whose activity could ensure fast repartitioning of resources between development and defense . Here we present a novel signaling capacity specific to the aliphatic 3-hydroxypropylglucosinolate ( 3OHPGSL ) . Our results suggest that 3OHPGSL signaling involves the universally conserved TOR pathway for growth and development , as mutants in TORC and autophagy pathways alter responsiveness to 3OHPGSL application . We reasoned that if a GSL can prompt changes in plant growth it is an indication of an inherent signaling capacity . Using purified compounds , we screened for endogenous signaling properties among short-chain methionine-derived aliphatic GSLs by testing their ability to induce visual phenotypic responses in Arabidopsis seedlings . We found that 3OHPGSL causes root meristem inhibition , at concentrations down to 1 µM ( Figure 1A ) . The observed response is concentration-dependent ( Figure 1A–B ) . All Arabidopsis accessions accumulate 3OHPGSL in the seeds , and this pool is maintained at early seedling stages ( Chan et al . , 2011; Petersen et al . , 2002; Brown et al . , 2003 ) . Previous research has shown that GSLs in the seed are primarily deposited in the embryo , accumulating to about 3µmol/g suggesting that we are working with concentrations within the endogenous physiological range ( Fang et al . , 2012; Kliebenstein et al . , 2007 ) . We tested how exogenous 3OHPGSL exposure to the roots alters 3OHPGSL accumulation within the shoot , and how this compares to endogenously synthesized 3OHPGSL levels . We grew the Col-0 reference accession and the myb28-1 myb29-1 mutant that is devoid of endogenous aliphatic GSLs in the presence and absence of exogenous 3OHPGSL . At day 10 , the foliar 3OHPGSL levels were analyzed . Col-0 without treatment had average foliar levels of 3OHPGSL of 3 . 2 µmol/g and grown on media containing 5 µM , 3OHPGSL contributed an additional 2 . 2 µmol/g raising the total 3OHPGSL to 5 . 4 µmol/g ( Figure 1C ) . The myb28-1 myb29-1 mutant had no measurable 3OHPGSL on the control plates and accumulated ~1 . 4µmol/g upon treatment ( Figure 1D ) . In agreement with the lower foliar 3OHPGSL accumulation in the myb28-1 myb29-1 mutant background , this double mutant had a lower root growth response to exogenous 3OHPGSL ( Figure 1—figure supplement 1 ) . Importantly , this confirms that the level of 3OHPGSL application is within the physiological range . We then tested if 3OHPGSL or potential activation products inhibit root growth because of cell death or toxicity . The first evidence against toxicity came from the observation that even during prolonged exposures , up to 14 days of length , Col-0 seedlings continued being vital and green ( Figure 1E ) . If there was toxicity the seedlings would be expected to senesce and die . We next tested if the strong root growth inhibition by 50 µM 3OHPGSL is reversible . Importantly , root inhibition is reversible , as the 3OHPGSL-mediated root stunting could be switched on and off by transfer between control media and media containing 3OHPGSL ( Figure 1E ) . Based on the toxicity and GSL assays , we conclude that the 3OHPGSL treatments are at reasonable levels compared to normal Arabidopsis physiology , and that the phenotypic responses we observed were not caused by flooding the system with 3OHPGSL or toxicity . To evaluate whether 3OHPGSL mediated root inhibition is a general GSL effect or if it is structurally specific to 3OHPGSL , we tested if aliphatic GSLs with similar side-chain lengths , but different chain modifications , would induce similar root growth effects . First , we assessed 3-methylsulfinyl-propyl ( 3MSP ) GSL , the precursor of 3OHPGSL , and the alkenyl-modified three carbon glucosinolate allyl ( Figure 2A ) . In contrast to 3OHPGSL , neither of these structurally related GSLs possessed similar root-inhibiting activities within the tested concentration range ( Figure 2B–D ) . We also analyzed the potential root inhibition for the one carbon longer C4-GSLs 4-methylsulfinylbutyl ( 4MSB ) and but-3-enyl ( Figure 2E ) . Neither of these compounds could inhibit root growth at the tested concentrations ( Figure 2F–H ) . There is no viable commercial , synthetic or natural source for the 4-hydroxybutyl GSL which prevented us from testing this compound . To test if the presence of a hydroxyl is essential for this response , we tested if either the R or S form of 2-hydroxybut-3-enylGSL had similar effects . This showed that neither enantiomer was similar to 3OHPGSL and that both actually stimulated root growth ( Figure 2—figure supplement 1 ) The fact that only 3OHPGSL inhibits root elongation suggests that the core GSL structure ( comprised of a sulfate and thioglucose ) does not cause the effect . Importantly , this indicates that 3OHPGSL root inhibition is not a generic result of providing extra sulfur or glucose from the GSL core structure to the plant , as these compounds would be equally contributed by the other GSLs . Furthermore , the results confirm that there is no general toxic activity when applying GSLs to Arabidopsis . This evidence argues that the 3OHPGSL root inhibition effect links to the specific 3OHP side chain structure , indicating the presence of a specific molecular target mediating the root inhibition response . The evolution of the GSL defense system is a relatively young phylogenetic event that occurred within the last ~92 Ma and is largely limited to the Brassicales order ( Edger et al . , 2015 ) . The aliphatic GSL pathway is younger still ( ~60 Ma ) and is limited to the Brassicaceae family with the enzyme required for 3OHPGSL production , AOP3 , being limited to Arabidopsis thaliana and Arabidopsis lyrata within the Arabidopsis lineage ( Edger et al . , 2015; Kliebenstein et al . , 2001a ) . However , 3OHPGSL is also found in the vegetative tissue of the close relative Olimarabidopsis pumila ( dwarf rocket ) ( Windsor et al . , 2005 ) , and in seeds of more distant Brassicaceae family members such as the hawkweed-leaved treacle mustard ( Erysimum hieracifolium ) , virginia stock ( Malcolmia maritima ) , shepherd's cress ( Teesdalia nudicaulis ) , and alpine pennycress ( Thlaspi alpestre ) ( Daxenbichler et al . , 1980; Daxenbichler et al . , 1991 ) . These species are evolutionarily isolated from each other , suggesting that they may have independently evolved the ability to make 3OHPGSL ( Kliebenstein et al . , 2001a; Windsor et al . , 2005; Cni , 2016 ) . As such , 3OHPGSL is an evolutionarily very young compound and we wanted to determine if the molecular pathway affected by 3OHPGSL is equally young , or whether 3OHPGSL affects an evolutionarily older , more conserved pathway . First we tested for 3OHPGSL responsiveness in plant species belonging to the GSL-producing Brassicales order ( Figure 3A ) . We found that four of the five tested Brassicales species responded to 5 µM 3OHPGSL with root growth inhibition regardless of their ability to synthesize 3OHPGSL ( Figure 3B–F ) . This suggests that responsiveness to 3OHPGSL application does not link to the ability to make 3OHPGSL . We expanded the survey by including plants within the eudicot lineage that do not have the biosynthetic capacity to produce any GSLs ( Figure 3G ) and found that 5 µM 3OHPGSL can inhibit root growth in several of the non-Brassicales species tested ( Figure 3H–L ) . The ability of 3OHPGSL to alter growth extended to Saccharomyces cerevisae where 3OHPGSL led to slower log phase growth than the untreated control ( Figure 3—figure supplement 1 ) . Allyl GSL in the media had no effect on S . cerevisae growth showing that this was a 3OHPGSL mediated process ( Figure 3—figure supplement 1 ) . The observation that 3OHPGSL responsiveness is evolutionarily older than the ability to synthesize 3OHPGSL suggests that the molecular target of We hypothesized that 3OHPGSL application may alter root cellular development to create the altered root elongation phenotype . A reduction in root growth can be caused by inadequate cell division in the root meristematic zone or by limited cell elongation in the elongation zones ( Figure 4A ) ( Henriques et al . , 2014 ) . To investigate how 3OHPGSL affects the root cellular morphology , we used confocal microscopy of 4-d-old Arabidopsis seedlings grown vertically with or without 10 µM 3OHPGSL . We used propidium iodide stain to visualize the cell walls of individual cells , manually counted the meristematic cells , and measured the distance to the point of first root hair emergence . Root meristems of 3OHPGSL treated seedlings were significantly reduced in cell number compared to untreated controls ( Figure 4B–C ) . Moreover , we also observed a premature initiation of the differentiation zone , as the first root hairs were closer to the root tip upon 3OHPGSL treatment ( Figure 4D–E ) . In addition , we saw bulging and branching of the root hairs in 3OHPGSL treated roots ( Figure 4F ) . There was no morphological evidence of cell death in any root supporting the argument that 3OHPGSL is not a toxin . These results indicate that 3OHPGSL leads to root growth inhibition by reducing the size of the meristematic zone within the developing Arabidopsis root . The observed response to 3OHPGSL suggests that the target of this compound is evolutionarily conserved and alters root growth but does not affect the patterning of the root meristem . This indicates that key root development genes like SHR and SCR are not the targets as they affect meristem patterning ( Sabatini et al . , 2003; Hao and Cui , 2012 ) . Mutants in GSL biosynthetic genes can lead to auxin over-production phenotypes as indicated by the superroot ( SUR ) 1 and 2 loci ( Mikkelsen et al . , 2004; Boerjan et al . , 1995 ) . However , the SUR genes are not evolutionarily conserved and 3OHPGSL does not create a superroot phenotype , showing that the genes are not the targets . A remaining conserved root regulator that does not alter meristem formation , but still alters root growth , is the TOR pathway ( Xiong et al . , 2013 ) . Thus , we proceeded to test if mutants in the TOR pathway alter sensitivity to 3OHPGSL . Because TORC activity is sugar responsive , we investigated whether 3OHPGSL application may alter the response to sugar in genotypes with altered TORC activity . We first used the TOR kinase overexpression line GK548 ( TORox ) because it was the only one of several published TOR overexpression lines ( Deprost et al . , 2007 ) that behaved as a TOR overexpressor within our conditions ( Figure 5—figure supplement 1 ) . The GK548 TORox line exhibits accelerated TORC signaling and consequently grows longer roots on media containing sucrose ( Figure 5 and ( Deprost et al . , 2007 ) ) . In addition , GK548 TORox meristems are harder to arrest ( Figure 5B ) . Applying 3OHPGSL to the GK548 TORox line showed that this genotype had an elevated 3OHPGSL-mediated inhibition of meristem reactivation in comparison to the WT ( Figure 5C–D ) . This suggests that TORC activity influences the response to 3OHPGSL ( Figure 5E ) . We next investigated how genetically disrupting additional components of TORC affects 3OHPGSL responsiveness . In addition to the catalytic TOR kinase subunit , TORC consists of the substrate binding RAPTOR ( Henriques et al . , 2014; Rexin et al . , 2015 ) , and LST8 ( 13 ) ( Figure 5E ) . In Arabidopsis- there is one copy of TOR , and two copies of both RAPTOR and LST8 ( RAPTOR1/RAPTOR2 and LST8-1/LST8-2 ) ( Moreau et al . , 2012; Deprost et al . , 2005 ) . RAPTOR1 and TOR null mutants are lethal as homozygotes ( Menand et al . , 2002; Deprost et al . , 2005 ) , and heterozygous raptor1 mutants did not display a significant change in 3OHPGSL responsiveness ( Figure 5—figure supplement 2 ) . We therefore tested insertion mutants within the weaker homolog RAPTOR2 , whose null mutant is viable , and in our conditions shows mildly reduced root length on sucrose-containing media ( Figure 5—figure supplement 3A and C ) . We found that , for two independent insertion lines raptor2-1 ( Deprost et al . , 2005; Anderson et al . , 2005 ) and raptor2-2 ( Deprost et al . , 2005 ) ( Figure 5—figure supplement 3E ) , there was a statistically significant reduction in 3OHPGSL response ( Figure 5—figure supplement 3A–C ) . This supports the hypothesis that 3OHPGSL-associated signaling proceeds through TORC and that RAPTOR2 may play a stronger role in 3OHP perception than RAPTOR1 . A key function of TORC activity is to control meristem cell division and this can be measured by meristem reactivation assays ( Xiong et al . , 2013 ) . Thus , to further test if TORC dependent responses are altered by 3OHPGSL , seedlings were germinated in sugar-free media and photosynthesis-constrained under low light conditions to induce root meristem arrest when the maternal glucose is depleted ( three days after germination ) . The root meristems were reactivated by applying exogenous sucrose ( Figure 6A–C ) . By treating arrested root meristems with sucrose alone or in combination with 3OHPGSL we found that 3OHPGSL could inhibit meristem reactivation of sugar-depleted and photosynthesis-constrained seedlings ( Figure 6D–E ) . Further , this response was dependent upon the 3OHPGSL concentration utilized . A similar response was found when treating with a TOR inhibitor such as rapamycin ( Xiong et al . , 2013 ) , providing additional support to the hypothesis that 3OHPGSL may reduce root growth by altering TORC activity . To further examine the possibility that 3OHPGSL may be affecting the TOR pathway , we proceeded to compare the effect of 3OHPGSL to published chemical TOR inhibitors . The active site TOR inhibitors were originally developed for mammalian cells and inhibit root growth in various plant species ( Montané and Menand , 2013 ) . Similar to 3OHPGSL , the active-site TOR inhibitor AZD-8055 ( AZD ) induces a reversible concentration-dependent root meristem inhibition ( Montané and Menand , 2013 ) . By directly comparing 3OHPGSL treatment with known TOR chemical inhibitors in the same system , we can test for interactions between 3OHPGSL and the known TOR inhibitors . An interaction between 3OHPGSL application and a known TOR inhibitor , e . g . an antagonistic relationship , is an indication that the same target is affected . To assess whether interactions between 3OHPGSL and TOR signaling occur , we grew seedlings vertically on media with combinations of 3OHPGSL and AZD and root-phenotyped the plants to compare the effect on root morphology . This identified a significant antagonistic interaction between AZD and 3OHPGSL ( 3OHP x AZD ) , both in terms of root length response ( Figure 7A ) and in initiation of the differentiation zone ( Figure 7B ) . This antagonistic interaction is also supported by the appearance of first root hair ( Figure 7B ) , as the premature initiation of the differentiation zone in the presence of 10 µM 3OHPGSL did not change further upon co-treatment ( Figure 7B ) . Moreover , there was a vast overlap in the phenotypic response to both compounds ( Figure 7C–D ) ; notably the closer initiation of the root differentiation zone to the root tip ( Figure 7B–C ) and the decreased cell elongation ( Figure 7C , right panel ) . Together , this suggests that the TOR inhibitor AZD and 3OHPGSL have a target in the same signaling pathway as no additive effect is observed . Supporting this is the observation that 3OHPGSL treatment is phenotypically similar to a range of TOR active site inhibitors , as well as an inhibitor of S6K1 ( one of the direct targets of TOR ) ( Figure 7—figure supplement 1 ) . Interestingly , the short root hair phenotype induced by AZD showed a synergistic interaction between AZD and 3OHPGSL suggesting that they may target different components of the TORC pathway that interact ( Figure 7C ) . Further , while there is strong phenotypic overlap between AZD and 3OHPGSL , there are also specific activities . AZD induced a rounding of the root tip ( Dolan and Davies , 2004 ) , but co-treatment with 3OHPGSL restored a wildtype-like tip phenotype ( Figure 7D ) . The lack of root rounding and root hair inhibition suggest that AZD and 3OHPGSL both target the TOR pathway , but at different positions . Alternatively , the 3OHPGSL may be a more specific TOR inhibitor and the additional AZD phenotypes could be caused by the ATP-competitive inhibitor having alternative targets in plants . Together , these results suggest that 3OHPGSL directly or indirectly targets the same molecular pathway as known TOR inhibitors ( Jia et al . , 2009 ) . Activation or repression of the TOR pathway leads to regulatory shifts in numerous downstream pathways ( Figure 8E ) ( Sheen , 2014; Sablowski et al . , 2014 ) . For example , active TOR negatively regulates autophagy across eukaryotic species including Arabidopsis ( Liu and Bassham , 2010; Shibutani and Yoshimori , 2014 ) . To test if pathways downstream of TORC areaffected by , or involved in , 3OHPGSL signaling , we analyzed mutants of two key autophagic ( ATG ) components , atg2-1 ( 18 ) and atg5-1 ( 58 ) . ATG2 is part of the ATG9 cycling system that is essential for autophagosome formation ( Feng et al . , 2014; Velikkakath et al . , 2012; Ryabovol and Minibayeva , 2016 ) . ATG9-containing vesicles are a suggested membrane source for the autophagosome , and vesicles containing ATG9 are cycled to-and-from the phagophore via the ATG9 cycling system ( Feng et al . , 2014; Ryabovol and Minibayeva , 2016 ) . ATG5 , is part of the dual ubiquitin-like conjugation systems responsible for ATG8 lipidation ( Feng et al . , 2014; Ryabovol and Minibayeva , 2016; Fujioka et al . , 2008 ) . There are nine ATG8 paralogues in Arabidopsis ( Ryabovol and Minibayeva , 2016 ) and together with the single copy of ATG5 , they are essential for autphagosome initiation , expansion , closure , and vacuolar fusion ( Feng et al . , 2014; Ryabovol and Minibayeva , 2016 ) . After the first conjugation system has conjugated ATG8 to an E2-like enzyme , the E3 ligase-like activity of the second ATG5-containing system enables ATG8 lipidation at the autophagic membrane ( Feng et al . , 2014; Walczak and Martens , 2013; Kaufmann et al . , 2014 ) . We found that atg5-1 enhanced 3OHPGSL responsiveness ( Figure 8A–B ) while atg2-1 had a wild type response ( Figure 8C–D ) . One possible explanation for this difference between the two mutants is that , apart from macro-autophagy , plants also have micro-autophagy ( Ryabovol and Minibayeva , 2016 ) , a process that , in animal systems , has been shown to be negatively regulated by TOR ( Li et al . , 2012 ) . Micro-autophagy does not involve de novo assembly of autophagosomes , and ATG5 has been shown to be involved in several forms of micro-autophagy whereas the role of ATG2 is more elusive and may not be required ( Li et al . , 2012 ) . Thus , the elevated 3OHPGSL response in the atg5-1 mutant supports the hypothesis that 3OHPGSL signaling proceeds through the TOR pathway , but also suggests that this response requires parts of the autophagic machinery as it was not observed for atg2-1 . In this study we describe a novel signaling capacity associated with 3OHPGSL , a defense metabolite present in Arabidopsis , and provide evidence that the linked signal proceeds via the TOR pathway . Application of exogenous 3OHPGSL caused reversible root meristem inhibition by morphological reprogramming of the root zones , i . e . dramatically reduced the root meristem size and limited root cell elongation ( Figure 4 ) . This response occurred at levels within the endogenous range and there was no evidence of cell death in any treated root , suggesting that this is not a toxicity response ( Figure 1 ) . Additionally , these morphological responses were specific to 3OHPGSL and not caused by any structurally or biosynthetically related GSL , suggesting that these responses were not because of generic properties shared by GSLs ( Figure 2 ) . Exposing a wide phylogenetic array of plants , including lineages that have never produced GSLs , to 3OHPGSL showed that application of this compound can inhibit growth broadly across the plant kingdom as well as in yeast ( Figure 3 , Figure 3—figure supplement 1 ) . This suggests conservation of the downstream signaling pathway across these diverse plant lineages . Equally , if the signaling compound is not 3OHPGSL itself , but a derivative , then the required biosynthetic processes must be conserved . This conservation largely rules out the specific GSL activation pathway controlled by Brassicales specific thioglucosidases , myrosinases ( Barth and Jander , 2006; Nakano et al . , 2017; Bones and Rossiter , 2006 ) . The phylogenetic conservation of the 3OHPGSL response led us to search for a target pathway controlling growth and development that would be evolutionary well conserved between the tested species . By comparing the root phenotype identified with 3OHPGSL application to the published literature , we hypothesized that 3OHPGSL treatment may affect TORC , a key primary metabolic sensor that controls growth and development , and is conserved back to the last common eukaryotic ancestor ( Henriques et al . , 2014 ) . Active site TOR inhibitors inhibit root growth in numerous plant species similar to 3OHPGSL application ( Montané and Menand , 2013 ) , supporting the hypothesis that 3OHPGSL may function via TORC . A model with 3OHPGSL affecting TORC would explain how 3OHPGSL can alter root development across the plant kingdom ( Figure 3 ) . Mechanistic support for this hypothesis came from a number of avenues . First , 3OHPGSL can block the TOR-mediated sugar activation of arrested meristems ( Figure 6 ) . Second , the TORox mutant intensifies 3OHPGSL linked signaling ( Figure 5 ) , and correspondingly loss-of-function mutants of the substrate binding TORC component raptor2 diminish the 3OHPGSL effect ( Figure 5—figure supplement 3 ) . Additionally , there are clear phenotypic overlaps between the root phenotypes induced by known TOR inhibitors and 3OHPGSL , e . g . root inhibition , inhibition of cell elongation , and notably the dramatic reduction of the meristem sizes ( Figure 7 ) . Critically , 3OHPGSL and known small-molecule inhibitors of TOR were mutually antagonistic for a number of phenotypes . In pharmacology , the outcomes of a drug combination can either be antagonistic , additive or synergistic , depending on whether the effect is less than , equal to , or greater than the sum of the effects of the two drugs ( Jia et al . , 2009 ) . Antagonistic interactions , as observed with 3OHPGSL and AZD , can occur if two drugs exhibit mutual interference against the same target site , or if their targets converge on the same regulatory hub ( Jia et al . , 2009 ) . Together , these lines of evidence suggest that 3OHPGSL or a derived metabolite targets the TOR pathway to alter root meristem development within Arabidopsis and potentially other plant species . Extending the analysis to pathways downstream of TORC , showed that loss of ATG5 , a vital component of the autophagic machinery ( Liu and Bassham , 2010; Shibutani and Yoshimori , 2014 ) , intensifies the 3OHPGSL response ( Figure 8A-DB ) . This supports the hypothesis that 3OHPGSL signaling proceeds through the TOR complex , but also suggest that this signal requires parts of the autophagic machinery . Loss of another autophagic component , ATG2 , did not influence the 3OHP response ( Figure 8C–D ) . Together , this raises the possibility that 3OHPGSL influenced responses involve predominantly a micro-autophagy pathway , which is ATG5- but may not be ATG2-dependent , rather than the macro-autophagy pathway that depends upon both genes ( Ryabovol and Minibayeva , 2016; Li et al . , 2012 ) . Micro-autophagy removes captured cytoplasmic components directly at the site of the vacuole via tonoplast invagination . The cargo to be degraded ranges from non-selective fractions of the cytoplasm to entire organelles , dependent on the type of micro-autophagy . The two ubiquitin-like conjugation systems , and thereby ATG5 , have been shown to be involved in several forms of micro-autophagy , such as starvation-induced , non-selective , and glucose-induced selective autophagy ( Li et al . , 2012 ) . Interestingly , micro-autophagy involves vacuolar movement of cargo , and the vacuole is considered the main storage site for glucosinolates ( Mikkelsen et al . , 2004 ) . Thus , ATG5 may be responsible for enabling the movement of exogenously applied 3OHPGSL out of the cytoplasm where it or a derivative metabolite could interact with the TORC pathway and into the vacuole . This would decrease the concentration of the 3OHPGSL associated signal and could explain why the atg5-1 mutant is more sensitive to 3OHPGSL application . Further work is required to test if ATG5 is functioning to attenuate the 3OHPGSL associated signal . A conundrum for defense signaling compounds to affect growth is the evolutionary age discrepancy; defense metabolites are typically evolutionarily very young , as they are often species or taxa specific , while growth regulatory pathways are highly conserved across broad sets of plant taxa . This raises the question of which mechanism ( s ) may allow this connection between young metabolites and old regulatory pathways . This suggests that plants may sense young metabolites using evolutionarily old signaling pathways . Similar evidence is coming from other secondary metabolite systems suggesting that this may be a general phenomenon . For example , an indolic GSL activation product can interact with the conserved TIR1 auxin receptor to alter auxin sensitivity within Arabidopsis ( Katz et al . , 2015 ) . Similarly , an unknown phenolic metabolite appears to affect regulation of growth and development by influencing the Mediator complex that is conserved across all eukaryotes ( Bonawitz et al . , 2012; Bonawitz et al . , 2014; Kim et al . , 2014 ) ; and the plant polyphenol resveratrol directly inhibits the mammalian TOR to induce autophagy ( Park et al . , 2016 ) . Thus , young plant metabolites can influence evolutionarily conserved pathways . Interestingly , this strongly resembles the action of virulence-associated metabolites within plant pathogens . Pseudomonad bacteria produce the evolutionarily young coronatine that alters the plant defense response by interacting with the conserved JA-Ile receptor COI1 ( Xie et al . , 1998 ) . In plant/pathogen interactions , this ability of pathogen-derived metabolites to alter plant defense signaling is evolutionarily beneficial because it boosts the pathogen’s virulence in planta . It is less clear if this selective pressure model also applies to plant defense compounds that interact with endogenous signaling pathways . Following the plant/pathogen derived model , it is tempting to assume that such plant defense metabolites have been co-selected on their ability to affect the biotic attacker and simultaneously provide information to the plant . However , an alternative hypothesis that is that these examples may simply be serendipitous cases , where the defense metabolites happened to interact with a pathway and are potentially of no evolutionary benefit . In this model , the plant might still be adapting to the evolution of this new regulatory linkage . In the particular case of 3OHPGSL , the AOP3 enzyme that makes this compound evolved prior to the split between A . thaliana and A . lyrata suggesting that these species have had at least several million years/generations of potential to adapt . However , the two hypotheses need to be empirically tested . Central to testing between the two hypotheses is to assess if the observed signaling effects have any fitness benefit for the plant suggesting that even if the connections arose by serendipity that they have been maintained by a selective benefit . This will require field testing the fitness of plants that contrast for the presence of these connections . An alternate way to test between these hypotheses would be to conduct a broad survey of plant metabolites to test how many can affect signaling within the plant . If a large fraction of metabolites have potential signaling function , it is unlikely that all of these are simply serendipitous cases that have not had sufficient time to be removed by natural selection . However; earlier studies have provided evidence that both allyl GSL and the GSL breakdown product indole-3-carbinol affect plant signaling and growth ( Katz et al . , 2015; Francisco et al . , 2016a; Francisco et al . , 2016b ) . In addition , R- and S-2-hydroxybut-3-enyl GSL promoted root growth ( Figure 2—figure supplement 1 ) , suggesting that dual effects of defense metabolites such as 3OHPGSL are possibly more general . Within this report , we provided evidence that 3OHPGSL , or derived compounds , appears to function as a natural endogenous TORC inhibitor that can work across plant lineages . This creates a link whereby the plant’s endogenous defense metabolism can simultaneously coordinate with growth . Such a built-in signaling capacity would allow coordination between development and defense , as the plant could use the defense compound itself as a measure of the local progress of any defense response and readjust development and defense to optimize against the preeminent threat . Future work is required to identify the specific molecular interaction that allows this communication to occur , this will help to illuminate how and why plants measure their own defense metabolism to coordinate available resources more broadly with growth . Future work might also ascertain whether there is a broader class of plant produced TOR inhibitors . If this is true , they might be highly useful in understanding TOR function across kingdoms of life and possibly to reveal significant aspects of this universally conserved pathway that may have gone unnoticed in other eukaryotic models . The genetic background for the Arabidopsis ( Arabidopsis thaliana ) mutants and transgenic lines described in this study is the Col-0 accession . The following lines were described previously: myb28-1 myb29-1 ( SonderbySønderby et al . , 2007 ) , atg2-1 ( Inoue et al . , 2006 ) , atg5-1 ( 58Thompson et al . , 2005 ) , raptor1-1 ( Deprost et al . , 2005; Anderson et al . , 2005 ) , raptor1-2 ( Deprost et al . , 2005 ) , raptor2-2 ( Deprost et al . , 2005 ) , raptor2-1 ( Deprost et al . , 2005; Anderson et al . , 2005 ) , and the TORox lines G548 , G166 , S784 , and S7817 ( Deprost et al . , 2007 ) . All genotypes were obtained and validated both genetically and phenotypically as homozygous for the correct allele . Seeds were vapour sterilized for 2–3 hr , by exposure to a solution of 100 mL household bleach ( Klorin Original , Colgate-Palmolive A/S ) mixed with 5 mL hydrochloric acid ( 12M ) , and ventilated for 30 min to one hour . After plating , on ½ strength Murashige and Skoog ( MS ) medium ( 2 . 2 g/l MS +vitamins ) ( Duchefa ) with 1% ( w/v ) sucrose ( Nordic Sugar ) , and 0 . 8% ( w/v ) micro agar ( Duchefa ) , pH adjusted to 5 . 8 ) , the seeds were stratified for two days in the dark at 4°C . For root length assays at normal light ( 115-130 µE ) Arabidopsis seedlings were grown vertically at 22 °C day 20°C night under a 16 hr photoperiod and 80% humidity ( long day ) . Concentrated 3OHPGSL in water was added to the agar post-sterilization to create media with the described concentration for each assay . The same method and water was used to create the media for the testing of the other specific GSLs . For meristem reactivation assays plants were grown as described in ( Xiong et al . , 2013 ) , except that in our conditions we needed to go to 25µE to obtain meristem inhibition . Daily root lengths were manually marked ( from day 3 ) with a permanent marker pen on the backside of the plate . After photography of 7-d-old seedlings the root growth was quantified using the ImageJ software ( Schneider et al . , 2012 ) . The least square means ( lsmeans ) for the genotypes in response to different treatments were calculated across experiments ( in R , see statistics ) , and plotted in excel . To test 3OHPGSL perception in other plant orders , seeds were obtained as listed in Supplementary File 1 . Except for Solanum lycopesicum , here San Marzano tomatoes were bought in a local supermarket and the seeds were harvested , fermented and dried . All seeds were vapour sterilized for three hours ( as above ) . Before plating , and Lotus japonicus MG20 ( Lotus ) were emerged in water and kept at 4°C for 1–2 weeks . Seeds were plated on vertical ½MS plates as specified in Supplementary File 1 , stratified for four days in the dark at 4°C before being transferred to a long day growth chamber ) . Root growth was measured approximately every 24 hr ( as described above ) . The yeast strain , NMY51 with pOST1-NubI and pDHB1-LargeT ( ( Stagljar et al . , 1998; Möckli et al . , 2007 ) ; DUALsystem Biotech ) , was grown in liquid YPD media ( 2% w/v bactopeptone ( Duchefa Biochemie ) , 1% w/v yeast extract ( Becton , Dickinson and Company ) , 2% w/v glucose ) with or without added GSLs , at 30°C and 150 rpm shaking . On day one; a 5 ml overnight culture was started from cryostock . Day two; four new 4 ml cultures were inoculated with 1 ml overnight culture , and grown overnight . On day three; an OD600 0 . 4 and a 0 . 04 dilution was prepared from each of the four cultures . 500 µl of each of the four cultures , at both dilutions , were transferred to a 96-well culture plate containing 500 µl YPD liquid media with 3OHPGSL or Allyl GSL , to final OD600 0 . 2 and GSL concentrations of 50 , 10 , 5 , 1 and 0 µM . The yeast growth was measured at 0 , 4 , 6 , 8 , 24 and 48 hr . For each growth measurement 100 µl culture was transferred to a 96-well Elisa-plate together with three wells of YPD liquid media for standardization . Growth was measured with a SpectraMAX 190 ( Molecular Devices ) and SoftMax Pro 6 . 2 . 2 software . Growth rates and statistical analysis was calculated using the R software . The linear growth range was determined , and a linear regression using the lm ( ) function in R was carried out to determine OD600 increase per hour ( slope ) and the yeast doubling time was calculated . Glucosinolates were extracted from whole plant tissue of adult plants ( for 3OHPGSL extraction ) , or from or 10-d-old seedlings ( 3OHPGSL uptake ) ( Kliebenstein et al . , 2001a; Kliebenstein et al . , 2001b; Kliebenstein et al . , 2001c ) , and desulfo-glucosinolates were analysed by LC-MS/TQ as desulfo-GSLs as described in ( Crocoll et al . , 2016 ) . The R software with the R studio interface was used for statistical analysis ( R Core Team , 2017; RStudio Team , 2015 ) . Significance was tested using the Anova function ( aov ) , lsmeans were obtained using the ‘lsmeans’ package ( version 2 . 17 ) ( Lenth , 2016 ) . The letter groupings ( Tukey’s HSD Test ) were obtained using the ‘agricolae’ package ( version 1 . 2–3 ) ( de Mendiburu , 2010 ) . To examine the root tip zones , we used confocal laser-scanning microscopy of 4-d-old seedlings grown vertically with or without treatment ( with 3OHPGSL and/or various inhibitors ) . Samples were mounted on microscopy slides in propidium iodide solution ( 40 µM , Sigma ) and incubated for 15 min . Confocal laser scanning microscopy was carried out on a Leica SP5-X confocal microscope equipped with a HC PL FLUOTAR 10 DRY ( 0 . 3 numerical aperture , 10X magnification ) or a HCX lambda blue PL APO 320 objective ( 0 . 7 numerical aperture , 20X magnification ) for close-up pictures of the meristem . To visualize the cell walls of individual cells the propidium iodide stain was excited at 514 nm and emission was collected at 600 nm to 680 nm . To determine the size of the meristems the confocal pictures we manually inspected and the meristematic cells marked and counted ( the meristem region is defined as in ( Dolan and Davies , 2004; Perilli and Sabatini , 2010 ) ) . To measure the distance from the root tip to the point of first root hair emergence we used ImageJ ( Schneider et al . , 2012 ) . The AZD8055 ( Chresta et al . , 2010 ) , Torin2 ( Liu et al . , 2011 ) , KU-63794 ( García-Martínez et al . , 2009 ) , and WYE-132 ( Yu et al . , 2010 ) were purchased from Selleckchem . PF-4708671 ( Pearce et al . , 2010 ) and allyl/sinigrin were purchased from Sigma-Aldrich . 4MSB and 3MSP GSLs were purchased at C2 Bioengineering . But-3-enyl GSL was purified from Brassica rapa seeds while 3OHPGSL was purified from the aerial parts of 4–5 weeks old greenhouse-grown plants of the Arabidopsis accession Landsberg erecta ( Kliebenstein et al . , 2001c; Crocoll et al . , 2016 ) . The concentration of 3OHP and but-3-enyl GSL was determined by LC-MS/TQ as desulfo-GSLs . All inhibitors were dissolved in DMSO and stored as 10 mM stocks at –20°C . For allyl , 3MSP , and 4MSB ~ 100 mM GSL stocks were made with H2O and the concentration of GSLs within these stocks was determined by LC-MS/TQ ( see above ) .	Plants , like all organisms , must invest their resources carefully . Growing new roots or shoots may allow a plant to better exploit its environment . But a plant should never leave itself vulnerable to disease . As such , there must be a balance between allocating resources to growth or to defense . Brassicas like cabbage , Brussels sprouts and wasabi use unique compounds called glucosinolates to protect themselves against pests and disease-causing microbes . These same compounds give these vegetables their distinctive flavors , and they are the source of many of the health benefits linked to eating these vegetables . Yet it was not known if glucosinolates could also affect a plant’s growth and development . Malinovsky et al . tested a number of purified glucosinolates with the model plant Arabidopsis thaliana , and found that one ( called 3-hydroxypropylglucosinolate ) caused the plants to grow with stunted roots . When 10 other species of plant were grown with this glucosinolate , almost all had shorter-than-normal roots . The effect was not limited to plants; baker’s yeast also grew less when its liquid media contained the plant-derived compound . The reason glucosinolates can protect plants against insect pests , provide us with health benefits , and widely inhibit growth is most likely because they have evolved to interact with proteins that are found in many different organisms . Indeed , through experiments with mutant Arabidopsis plants , Malinovsky et al . revealed that their glucosinolate influences the TOR complex . This complex of proteins works in an ancient and widespread signaling pathway that balances growth and development with the available energy and nutrients in organisms ranging from humans to yeast to plants . The TOR complex plays such a vital role in living cells that problems with this complex have been linked to diseases such as cancer and heart disease . Importantly , the chemical structure of this glucosinolate is unlike other compounds that have already been tested against the TOR complex . As such , it is possible that this glucosinolate might lead to new drugs for a range of human diseases . Further , as this compound affects plant growth , it could also act as a starting point for new herbicides . Together these findings show how studying molecules made in model organisms and understanding how they function can lead to the identification of new compounds and targets with an unexpectedly wide range of potential uses .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology" ]	2017	An evolutionarily young defense metabolite influences the root growth of plants via the ancient TOR signaling pathway
Adult stem cells continuously undergo self-renewal and generate differentiated cells . In the Drosophila ovary , two separate niches control germ line stem cell ( GSC ) self-renewal and differentiation processes . Compared to the self-renewing niche , relatively little is known about the maintenance and function of the differentiation niche . In this study , we show that the cellular redox state regulated by Wnt signaling is critical for the maintenance and function of the differentiation niche to promote GSC progeny differentiation . Defective Wnt signaling causes the loss of the differentiation niche and the upregulated BMP signaling in differentiated GSC progeny , thereby disrupting germ cell differentiation . Mechanistically , Wnt signaling controls the expression of multiple glutathione-S-transferase family genes and the cellular redox state . Finally , Wnt2 and Wnt4 function redundantly to maintain active Wnt signaling in the differentiation niche . Therefore , this study has revealed a novel strategy for Wnt signaling in regulating the cellular redox state and maintaining the differentiation niche . Stem cells have two important properties , self-renewal and differentiation , which are critical for continuously generating new functional cells to maintain tissue homeostasis . The self-renewal property is controlled in various stem cell systems by interplays between signals from the niche and intrinsic factors ( Li and Xie , 2005; Morrison and Spradling , 2008; Losick et al . , 2011 ) . Germ line stem cells ( GSCs ) in the Drosophila ovary and testis are attractive systems for studying stem cell self-renewal at the molecular and cellular level ( Fuller and Spradling , 2007; Xie , 2013 ) . Although stem cell differentiation was widely thought to be a developmentally default state , we have recently proposed that GSC lineage differentiation is also controlled extrinsically by a differentiation niche formed by inner germarial sheath cells ( ISCs , also known as escort cells ) . However , it remains unclear how the maintenance and function of the differentiation niche are regulated at the molecular level . In this study , we show that autocrine Wnt2/4 signaling maintains the differentiation niche by regulating ISC proliferation and survival via redox regulation . In the Drosophila ovary , two or three GSCs at the tip of the germarium , the most anterior region of the Drosophila ovary , continuously self-renew and generate differentiated GSC daughters , cystoblasts ( CBs ) . The CBs further divide four times synchronously with incomplete cytokinesis to form 2-cell , 4-cell , 8-cell , or 16-cell cysts ( de Cuevas et al . , 1997 ) . GSCs and their differentiated progeny can be reliably identified by their unique morphology of germ line-specific intracellular organelles known as fusomes: GSCs and CBs contain a spherical fusome known as the spectrosome , whereas differentiated germ cell cysts contain a branched fusome ( Lin et al . , 1994 ) . GSCs can be reliably distinguished from CBs by their direct contact with cap cells ( Figure 1A ) . Cap cells function as the self-renewing niche to maintain GSCs by activating BMP signaling and maintaining E-cadherin-mediated cell adhesion ( Song et al . , 2002; Xie and Spradling , 1998 , 2000 ) . In addition , various classes of intrinsic factors work with BMP signaling and E-cadherin to control GSC self-renewal ( Xie , 2013 ) . Therefore , GSC self-renewal is controlled by coordinated functions of niche-initiated signaling pathways and intrinsic factors . 10 . 7554/eLife . 08174 . 003Figure 1 . Canonical Wnt signaling in ISCs promotes germ cell differentiation . ( A ) The Drosophila germarium dividing into three regions 1 , 2a , 2b and 3 . Abbreviations: TF-terminal filament; CPC-cap cell; ISC-inner germarial sheath cell; FC-follicle cell; GSC-germ line stem cell; CB-cystoblast; DC-developing cyst; SS-spectrosome; FS-fusome . In B–L , cap cells are highlighted by broken ovals , whereas CBs and cysts are indicated by arrowheads and arrows , respectively . ( B ) In the c587>>UAS-GFP germarium containing two GSCs ( spectrosomes indicated by arrowheads ) close to cap cells , one CB and a few differentiated cysts are surrounded by GFP-positive ISCs . ( C–E ) In armKD1 ( C ) dshKD1 ( D ) germaria , many spectrosome-containing CBs accumulate far away from cap cells . ( E ) Quantification results on the percentages of the germaria exhibiting the germ cell differentiation defect ( ≥4 CBs ) . ( F–H ) fz fz2 double knockdown ( F ) , axn- ( G ) , and sgg-overexpressing ( H ) germaria contain excess CBs . ( I–L ) In arm-overexpressing control ( I ) and dshKD ( K , L ) germaria , GSC progeny differentiate into cysts containing a branched fusome ( arrow ) . J: Quantification results . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 00310 . 7554/eLife . 08174 . 004Figure 1—figure supplement 1 . Wnt receptors FZ and FZ2 function redundantly in ISCs to promote germ cell differentiation . Broken ovals highlight cap cells and GSCs , while arrowheads denote spectrosomes in CBs ( A–C , E ) . ( A–D ) fzKD ( B ) and fz2KD ( C ) germaria contain 0 and 1 CB , respectively , in comparison with the gfpKD germarium carrying one CB ( A ) . ( D ) The quantification results on CB numbers in one-week-old ( 1w ) and two-week-old ( 2w ) control and single knockdown germaria . ( E , F ) fzKD fz2KD ( E ) germarium contains significantly more CBs . ( F ) The quantification results on CB numbers . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 004 Following GSC division , differentiating GSC daughters , CBs , are always positioned away from the self-renewal niche . ISCs sit on the surface of the germarium to send their cellular processes to wrap up underneath CBs , mitotic cysts , and early 16-cell cysts , which move posteriorly ( Decotto and Spradling , 2005; Kirilly et al . , 2011; Morris and Spradling , 2011 ) . Our recent study suggests ISCs and their associate long cellular processes act as the differentiation niche to promote GSC progeny differentiation in the Drosophila ovary because disrupting long ISC processes leads to an accumulation of CB-like cells , indicative of a germ cell differentiation defect ( Kirilly et al . , 2011 ) . A series of genetic studies have further supported the existence of the differentiation niche . The epidermal growth factor ( EGF ) signaling pathway is active in ISCs to promote GSC lineage differentiation partly by repressing dally expression ( Schultz et al . , 2002; Liu et al . , 2010 ) . In addition , Rho signaling is also required in ISCs to promote GSC differentiation partly by repressing dally and dpp expression . dally encodes a proteoglycan protein , which is capable of promoting Dpp/BMP diffusion to the differentiation niche ( Guo and Wang , 2009; Hayashi et al . , 2009 ) . Ecdysteroid signaling also operates in ISCs to promote germ cell differentiation because inactivating ecdysteroid receptors EcR and Usp in ISCs disrupts cyst formation ( Morris and Spradling , 2012 ) . One potential mechanism is that ecdysteroid signaling controls the formation of ISC cellular processes , thereby promoting the interaction between ISCs and germ cells ( Konig and Shcherbata , 2015 ) . Gap junction protein Inx2 functions in ISCs to promote germ cell differentiation , but its transmitted substances between ISCs and germ cells remain identified ( Mukai et al . , 2011 ) . The importance of gap junctions between ISCs and germ cells could also explain why ISC cellular processes are important for germ cell differentiation . Therefore , physical interactions and signaling-mediated communications between ISC cellular processes and GSC progeny likely contribute to GSC progeny differentiation collectively . In addition , chromatin regulators are also important in ISCs to promote GSC differentiation . Eggless , a Drosophila H3K9 trimethyltransferase , maintains ISCs and represses dally and dpp expression in ISCs , thereby promoting germ cell differentiation ( Wang et al . , 2011 ) . Similarly , Piwi functions in ISCs likely as a chromatin regulator to control germ cell differentiation partly by repressing dpp expression ( Jin et al . , 2013; Ma et al . , 2014 ) . dBre1 ( a E3 ubiquitin ligase ) and dSet1 ( a H3K4 trimethylase ) together control H3K4 trimethylation in ISCs and promote germ cell differentiation partly by limiting BMP signaling from the differentiation niche ( Xuan et al . , 2013 ) . The potential chromatin factor without children ( Woc ) maintains the ISC-germ cell physical interaction via regulation of Stat-Zfh1 ( Maimon et al . , 2014 ) . The histone demethylase Lsd1 is required in ISCs to promote germ cell differentiation by maintaining ISC survival , maintaining ISC morphology , and preventing BMP signaling from the differentiation niche ( Eliazer et al . , 2011 , 2014 ) . Therefore , ISCs function as the differentiation niche by preventing BMP signaling and direct communication . A recent study showed that tyrosine kinase Btk29A maintains Wnt signaling in ISCs by phosphorylating Drosophila β-catenin homolog Armadillo ( Arm ) ( Hamada-Kawaguchi et al . , 2014 ) . It also argues that Wnt4 activates Wnt signaling to maintain Piwi expression and repress E-cadherin expression in ISCs , thereby promoting germ cell differentiation . In contrast , this study has demonstrated that both ISC-expressing Wnt2 and Wnt4 , but not Wnt4 alone , act through known Wnt pathway components to maintain active Wnt signaling , promoting germ cell differentiation . More importantly , we show that Wnt signaling is required to maintain ISCs by promoting ISC survival and proliferation . Surprisingly , Wnt signaling is dispensable for Piwi expression . Instead , we demonstrate that Wnt signaling controls the expression of four Gst genes to maintain the reduced redox , thereby promoting ISC maintenance and germ cell differentiation . Finally , knocking down one of the Gst gene , GstD2 , in ISCs leads to the germ cell differentiation defect . Therefore , our study has revealed a novel mechanism which autocrine Wnt signaling utilizes to maintain ISCs and promote germ cell differentiation . To identify the genes that function in ISCs to promote germ cell differentiation , we used c587-driven UAS-RNAi expression to knockdown individual genes in ISCs . c587 is a GAL4 line that is specifically expressed in ISCs and early follicle cell progenitors based on UAS-GFP expression ( Song et al . , 2004 ) ( Figure 1B ) . To facilitate the identification of GSCs and differentiated germ cells , spectrosomes and fusomes are labeled by Hts staining ( Lin et al . , 1994 ) , and germ cells are visualized by Vasa staining ( Lasko and Ashburner , 1988 ) . In contrast to the control germarium containing 0 to 2 CBs , knocking down Wnt downstream genes armadillo ( arm ) and disheveled ( dsh ) in ISCs leads to an accumulation of many spectrosome-containing CBs ( collectively referred to single germ cells at least one cell diameter away from cap cells ) ( Figure 1C , D ) . Based on the fact that control germaria rarely contain three CBs , the germaria containing four or more CBs are considered to exhibit a germ cell differentiation defect . Over 90% of arm and dsh knockdown germaria ( armKD1 and dshKD1 ) exhibit the germ cell differentiation defect ( Figure 1E ) . Similarly , c587-driven expression of the RNAi lines against different arm or dsh sequences ( armKD2 and dshKD2 ) also generates the similar germ cell differentiation defect ( Figure 1E ) . In Drosophila , Wnt ligands bind the receptor complex composed of Frizzled ( Fz ) , Frizzled 2 ( Fz2 ) and Arrow to activate Dsh and stabilize Arm , which forms a protein complex with a TCF ( T Cell Factor ) -like Pangolin in the nucleus to activate target gene expression , whereas Axin ( Axn ) and Shaggy ( Sgg ) negatively modulate Wnt signaling by promoting Arm degradation ( Logan and Nusse , 2004 ) . Simultaneous knockdown down of both fz and fz2 ( fzKD+fz2KD ) in ISCs recapitulate the germ cell differentiation defect caused by either armKD or dshKD , although either fzKD or fz2KD yields a much weaker germ cell differentiation defect ( Figure 1E , F; Figure 1—figure supplement 1 ) . Consistently , overexpression of axn and sgg in ISCs also leads to the germ cell differentiation defect similar to that caused by armKD and dshKD ( Figure 1E , G , H ) . These results indicate that Wnt signaling is required in ISCs to promote germ cell differentiation . The expression of a constitutively active mutant armS10 ( arm ) leads to hyperactive Wnt signaling independently of Wnt ligands ( Pai et al . , 1997 ) . c587-driven arm* expression alone does not cause any obvious GSC maintenance and germ cell differentiation defects , but results in severe egg chamber budding defects likely due to defective follicle cell development ( Figure 1I ) . Interestingly , c587-mediated arm* expression can fully rescue the germ cell differentiation defect caused by the two independent dshRNAi knockdowns , indicating that Arm functions downstream of Dsh in ISCs to promote germ cell differentiation ( Figure 1J–L ) . These results further suggest that the two dshRNAi lines are highly specific . Taken together , our findings demonstrate that canonical Wnt signaling works in ISCs to promote GSC lineage differentiation . The c587-gal4 driver is known to be expressed in somatic precursor cells in the developing female gonad , which give rise to terminal filament , cap cells , adult ISCs , and follicular stem cells ( Zhu and Xie , 2003 ) . To determine if adult ISCs require active Wnt signaling for GSC lineage differentiation , we took advantage of c587-deriven expression of a temperature-sensitive gal4 repressor gal80 ( c587>>UAS-gal80ts ) to allow the expression of UAS-RNAi lines against dsh and arm only in adult ISCs using temperature shift . At 25°C , gal80ts is functional to prevent gal4-driven expression of a UAS transgene , but at 29°C , gal80ts is inactivated to allow gal4-driven gene expression ( McGuire et al . , 2003 ) . After the c587>>gal80ts control and c587>>gal80ts UAS-dshRNAi or UAS-armRNAi females eclosed at 25°C ( RNAi expression is extremely low or not expressed ) , they either continued to be kept at 25°C for one week ( keeping RNAi expression repressed ) or were moved to 29°C for one week ( turning on the RNAi expression due to gal80 inactivation ) . For the c587>>gal80ts control females , which were cultured at 25°C or 29°C as adults for one week , their germaria contain the normal number of CBs ( Figure 2A , B ) . Interestingly , for the c587>>gal80ts UAS-dshRNAi or UAS-armRNAi females , which were cultured at constant 25°C , their germaria contain the normal or close to normal numbers of CBs ( Figure 2B–F ) . In contrast , for the c587>>gal80ts UAS-dshRNAi or UAS-armRNAi females , which were shifted from 25°C to 29°C for one week , their germaria contain excess CBs , indicative of the germ cell differentiation defect ( Figure 2B , C′–F′ ) . These results demonstrate that Wnt signaling is required in adult ISCs to promote germ cell differentiation . 10 . 7554/eLife . 08174 . 005Figure 2 . Wnt signaling is required in adult ISCs to promote germ cell differentiation . Broken ovals highlight germ line stem cells ( GSCs ) , whereas arrowheads indicate cystoblasts ( CBs ) . Two experimental regimens 25–25 and 25–29 mean the females cultured under 25°C or 29°C for one additional week after reaching adulthood at 25°C , respectively . ( A , A′ ) Control c587>>gal80ts germaria contain one or two CBs under the 25–25 ( A ) or 25–29 ( A′ ) condition . ( B ) Quantification results on the percentages of the germaria carrying 4 or more CBs show that adult stage-specific arm or dsh knockdown causes an accumulation of more CBs in comparison with the control . ( C–F ) armRANi- ( C , D ) or dshRNAi- ( E , F ) carrying germaria contain one or two CBs under 25–25 , indicating that RNAi-mediated arm or dsh knockdown in adult inner germarial sheath cells ( ISCs ) under 25°C is very limited based on the germ cell differentiation phenotype . ( C′–F′ ) armRANi- ( C′ , D′ ) or dshRNAi- ( E′ , F′ ) carrying germaria contain many more CBs under 25–29 , indicating that RNAi-mediated arm or dsh knockdown under 29°C in adult ISCs leads to the severe germ cell differentiation defects . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 005 Our previous studies have shown that a severe ISC loss also causes the similar germ cell differentiation defect ( Kirilly et al . , 2011; Wang et al . , 2011 ) . Then , we determined if Wnt signaling controls ISC maintenance by using the PZ1444 reporter to quantify ISC numbers in control and Wnt signaling-defective germaria . PZ1444 expresses nuclear LacZ in ISCs and cap cells , which can be distinguished based on nucleus size and location ( Xie and Spradling , 2000 ) . In the one-week-old control germaria , PZ1444 labels 30–35 ISCs in addition to cap cells ( Figure 3A , C ) . In the one-week-old arm and dsh knockdown germaria , only fewer than 5 ISCs remain ( Figure 3B–D ) . Some knockdown germaria have completely lost ISCs ( Figure 3B ) , whereas others retain one or a few ISCs ( Figure 3D ) . Consistently , overexpression of axn and sgg also leads to a severe ISC loss ( Figure 3E , F ) . These results demonstrate that Wnt signaling is required to maintain ISCs . 10 . 7554/eLife . 08174 . 006Figure 3 . Wnt signaling maintains the differentiation niche by promoting cell proliferation and survival . Cap cells are highlighted by broken ovals . ( A ) The control germarium contains PZ1444-labeled cap cells and ISCs ( one by arrow ) . ( B–D ) The armKD ( B ) and dshKD ( D ) germaria maintain zero and one ISC ( arrow ) , respectively . ( C ) Quantification results on ISC numbers ( mean ± standard deviation; Student's t-test is used to calculate p values ) . ( E , F ) c587-directed axn- ( E ) and sgg- ( F ) overexpressing germaria contain no ISCs ( E ) and a few ISCs ( F , arrows ) , respectively . ( G ) c587-directed arm* expression drastically increases the ISC number ( one by arrow ) . ( H–L ) c587-mediated knockdown of arm ( I ) and dsh significantly decreases BrdU-positive PZ1444-labeled ISCs ( arrows ) , whereas c587-directed arm* expression ( K ) significantly increases BrdU-labeled ISCs in comparison with the control ( H ) . The arrowhead in I indicates a BrdU-positive germ cell cyst . J and L show quantification results . ( M–Q ) c587-mediated knockdown of arm and dsh ( N ) significantly increases TUNEL-positive PZ1444-labeled ISCs ( arrows ) , whereas c587-directed arm* expression ( P ) significantly decreases TUNEL-positive ISCs in comparison with the control ( M ) . ( O , Q ) Quantification results . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 00610 . 7554/eLife . 08174 . 007Figure 3—figure supplement 1 . Hyperactive Wnt signaling increases the ISC population . Cap cells are highlighted by ovals , while some of the ISCs are denoted by arrowheads . ( A ) In the control germarium , PZ1444 labels cap cells and ISCs . ( B–D ) The axnKD ( B ) and sggKD ( C , D ) germaria accumulate more ISCs . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 007 Wnt signaling maintains ISCs possibly by promoting cell proliferation , survival , or both . Interestingly , arm-expressing germaria contain significantly more ISCs ( Figure 3G ) . In contrast with the one-week-old control germaria containing 32 ISCs , the one-week-old arm-expressing germaria carry 130 ISCs ( Figure 3C ) . Consistently , c587-mediated axn and sgg knockdown , which increases Wnt-signaling activity ( Heslip et al . , 1997; Willert et al . , 1999 ) , also leads to more ISCs ( Figure 3—figure supplement 1 ) . Interestingly , germ cell differentiation proceeds normally in the germaria carrying extra ISCs , suggesting that ISCs provide a permissive environment for germ cell differentiation ( Figure 3G; Figure 3—figure supplement 1 ) . To further investigate how Wnt signaling maintains the ISC population , we examined ISC proliferation and apoptosis in the control , armKD , dshKD , and arm-expressing germaria using BrdU incorporation and TUNEL ( Terminal deoxynucleotidyl transferase dUTP nick end labeling ) -labeling assays , respectively . In order to avoid severe ISC loss in these experiments , we purposely cultured the control , armKD , dshKD , and arm-expressing females at 29°C for shorter time than earlier experiments . BrdU incorporation labels replicating ISCs in the S-phase of the cell cycle , whereas TUNEL labeling detects fragmented DNA in dying cells . Based on BrdU-labeling results , defective Wnt signaling significantly decreases ISC proliferation , whereas hyperactive Wnt signaling significantly increases ISC proliferation ( Figure 3H–L ) . Based on TUNEL-labeling results , defective Wnt signaling significantly increases ISC apoptosis , whereas hyperactive Wnt signaling significantly decreases ISC apoptosis ( Figure 3M–Q ) . These results suggest that Wnt signaling maintains the ISC population by promoting proliferation and increasing survival . Our previous studies have also shown that ISC loss disrupts germ cell differentiation by increasing BMP signaling ( Kirilly et al . , 2011; Wang et al . , 2011; Ma et al . , 2014 ) . In the Drosophila germarium , BMP signaling leads to production of phosphorylated Mad ( pMad ) and activation of Dad-lacZ reporter expression in GSCs ( Chen and McKearin , 2003; Kai and Spradling , 2003; Casanueva and Ferguson , 2004; Song et al . , 2004 ) . In the control germaria , pMad and Dad-lacZ expression is restricted to GSCs ( Figure 4A , A′ , D , D′ ) . In the armKD and dshKD germaria , pMad and Dad-lacZ expression is activated in the accumulated CBs locating one to a few cells away from cap cells in addition to GSCs ( Figure 4B , B′ , C , C′ , E , E′ , F , F′ ) . These results indicate that BMP signaling is spread to the differentiation zone in the Wnt signaling-defective germarium . 10 . 7554/eLife . 08174 . 008Figure 4 . Wnt signaling in ISCs prevents BMP signaling in the differentiated germ cell zone and maintains long ISC cellular processes . Ovals indicate GSCs , whereas arrows in A-F′ denote CBs or CB-like single germ cells . ( A′–F′ ) Images only show green fluorescence of ( A–F ) . ( A , A′ ) In the control germarium , GSCs are positive for pMad , but one CB is negative . ( B–C′ ) In the armKD ( B , B′ ) and dshKD ( C , C′ ) germaria , GSCs are pMad-positive as in the control . Although most of the accumulated CBs are negative for pMad , some CBs are pMad-positive ( arrows ) . ( D , D′ ) The control germarium shows that GSCs are positive for Dad-lacZ expression , but one CB is negative . ( E–F′ ) In the armKD ( E , E′ ) and dshKD ( F , F′ ) germaria , GSCs are Dad-lacZ-positive as in the control . Some of the accumulated CBs are negative for Dad-lacZ , but the other ones exhibit low Dad-lacZ expression ( arrows ) . ( G ) RNA sequencing ( RNA-seq ) results on the purified ISCs show that mRNA levels for the known BMP pathway components , including dpp , gbb ( also encoding a BMP ligand ) and dally , remain largely unchanged in axn-overexpressing ISCs or dskKD ISCs in comparison with the control ( FKPM = reads per kilobase per million mapped reads ) . ( H , H′ ) Our CB and GSC quantification results indicate that the heterozygous dpp mutations have partial suppression on the germ cell differentiation defects caused by dshKD , and the suppression is not due to the changes in GSC numbers ( 50 germaria examined for each genotype ) . ( I–K′ ) armKD ( J , J′ ) and dshKD ( K , K′ ) ISCs lack their CD8GFP-positive cellular processes extending into the accumulated CBs ( arrowheads ) in contrast with the control ISCs extending the cellular processes wrapping up underneath germ cells ( arrowheads , I′ ) . ( I′–K′ ) A higher magnification of highlighted areas in ( I–K ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 008 To investigate how Wnt signaling regulates germ cell differentiation at the molecular level , we compared the gene expression changes in fluorescence-activated cell sorting ( FACS ) -purified GFP-labeled control , Wnt signaling-defective dsh knockdown , and axn-overexpressing ( AxnOE ) ISCs using deep RNA sequencing ( RNA-seq ) . Our RNA-seq results show that known Drosophila BMP-signaling components and regulators , including dally and dpp , which are often upregulated in defective ISCs ( Liu et al . , 2010; Kirilly et al . , 2011; Wang et al . , 2011; Jin et al . , 2013; Ma et al . , 2014 ) , remain unchanged in the dshKD and AxnOE ISCs in comparison with the control ISCs ( Figure 4G ) . To further determine if upregulated BMP signaling is responsible for the germ cell differentiation defect resulted from defective Wnt signaling , we determined the effect of heterozygous dpp mutations on the differentiation defect caused by dshKD because they have been shown to suppress the differentiation defect caused by BMP-signaling upregulation ( Kirilly et al . , 2011; Wang et al . , 2011 ) . Our results indicate that the dpp heterozygous mutation dpphr4 significantly rescues the germ cell differentiation defect caused by dshKD1 and dshKD2 , whereas the dpp heterozygous mutation dpphr56 significantly rescues that caused by dshKD1 but not dshKD2 ( Figure 4H ) . This rescue effect by the dpp heterozygous mutations is unlikely caused by decreasing GSCs ( Figure 4H′ ) . These results suggest that Wnt signaling in ISCs regulates germ cell differentiation partly by preventing BMP signaling in the differentiation niche . Long ISC cellular processes are also required to promote germ cell differentiation ( Kirilly et al . , 2011 ) . They can be easily visualized by c587-driven expression of membrane-tethered GFP , CD8GFP . In the control germaria , ISC cellular processes wrap up differentiated germ cells ( Figure 4I , I′ ) . In contrast , cellular processes in the remaining ISCs of the armKD and dshKD germaria fail to extend into the accumulated CBs ( Figure 4J–K′ ) . These results indicate that Wnt signaling is required to maintain ISC cellular processes . A recent study proposes that Wnt signaling upregulates piwi expression in ISCs , thereby promoting germ cell differentiation ( Hamada-Kawaguchi et al . , 2014 ) . To verify if Piwi protein is indeed downregulated in Wnt signaling-defective ISCs , we quantified the expression of Piwi protein in the PZ1444-labeled dshKD or armKD ISCs . Piwi protein shows comparable expression levels among the examined 203 control ISCs , 48 armKD1 ISCs , 58 armKD2 ISCs , and 46 dshKD2 ISCs ( Figure 5A–D ) . Surprisingly , Piwi protein levels are significantly elevated in the dshKD1 ISCs ( 93 examined ) in comparison with the control ISCs ( Figure 5D ) . In addition , we also used FACS to purify the GFP-labeled control ISCs , axnOE ISCs , and dshKD1 ISCs for RNA-seq . The RNA-seq results show that the piwi mRNA levels are slightly upregulated in the axnOE ISCs and are significantly upregulated in the dshKD ISCs in comparison to the control ISCs ( Figure 5E ) . Therefore , our results indicate that Wnt signaling promotes germ cell differentiation unlikely by maintaining Piwi expression in ISCs . However , Wnt signaling possibly works with Piwi in an unknown means to promote germ cell differentiation because the loss of their functions in ISCs leads to similar germ cell differentiation defects ( Jin et al . , 2013; Hamada-Kawaguchi et al . , 2014; Ma et al . , 2014 ) . 10 . 7554/eLife . 08174 . 009Figure 5 . Piwi mRNA and protein expression remain normally expressed in the Wnt signaling-defective ISCs . PZ1444 is used to highlight ISCs . ( A ) The control germarium shows that ISCs ( arrow ) express more Piwi proteins than cap cells ( oval ) and underneath germ cells . ( B , C ) The remaining armKD ( B ) and dshKD ( C ) ISCs ( arrows ) still retain high Piwi protein expression . ( D ) Quantification results show that Piwi protein levels remain largely unchanged in the remaining armKD1 , armKD2 , and dshKD2 ISCs in comparison with the wild-type control ISCs , but it appears to increase its expression in dshKD1 ISCs . ( E ) RNA-seq results on the purified ISCs show that piwi mRNA expression levels are not changed in axn-overexpressing ISCs , but increase in dskKD ISCs , in comparison with the wild-type control ISCs . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 009 Our RNA-seq results also show that four Gst genes , GstD2 , GstD4 , GstD10 , and GstE3 , decrease their mRNA expression levels significantly in both dshKD- and axn-overexpressing ISCs in comparison with the control ( Figure 6A ) . GST proteins comprise a family of eukaryotic metabolic enzymes for eliminating hydrogen peroxide ( H2O2 ) and catalyzing the conjugation of the reduced form of glutathione ( GSH ) to oxidized substrates for the purpose of detoxification . Thus , downregulated expression of the Gst genes in the Wnt signaling-defective ISCs might cause the increase in cellular reactive oxygen species ( ROS ) , which can be detected by dihydroethidium ( DHE ) fluorescence . In the anterior half of control germaria , ISCs and underneath differentiating germ cells show low DHE fluorescence , indicating that these cells have low cellular ROS levels , including ISCs ( Figure 6B; Figure 6—figure supplement 1A , B′ ) . In contrast , dshKD and armKD ISCs elevate DHE fluorescence , and germ cells underneath also increase DHE fluorescence , further supporting the idea that defective Wnt signaling results in the increased cellular ROS in ISCs and their interacting germ cells ( Figure 6B′; Figure 6—figure supplement 1C–G ) . As a distinct GST family member from the GST-D family members , GST2 has similar function in regulating the cellular redox state ( Singh et al . , 2001 ) . Consistently , GST2 overexpression in the dshKD ISCs can also restore low DHE fluorescence in the anterior half of the germaria , suggesting that Wnt signaling controls the cellular redox state in ISCs , GSCs , and early GSC progeny by regulating Gst gene expression ( Figure 6B″; Figure 6—figure supplement 1H ) . Catalase ( CAT ) can also help eliminate cellular H2O2 by converting it to free oxygen and water . As expected , CAT overexpression in the dshKD ISCs also restores low cellular ROS in the anterior half of the germaria ( Figure 6B′″; Figure 6—figure supplement 1I ) . These results suggest that Wnt signaling regulates the cellular redox state in ISCs by controlling Gst gene expression . 10 . 7554/eLife . 08174 . 010Figure 6 . Wnt signaling maintains the reduced redox state in ISCs , promoting germ cell differentiation . ( A ) RNA-seq results of the purified ISCs show that GstD2 , GstD4 , GstD10 , and GstE3 mRNA expression levels are significantly lower in axn-overexpressing or dskKD ISCs than the control ISCs . In B–B′″ , D–I , L–N , and P–R , cap cells and GSCs are highlighted by broken ovals , some CBs are indicated by arrowheads and some PZ1444-positive ISCs are denoted by arrows . ( B–B′″ ) The dshKD germarium ( B′ ) exhibits a drastic increase of dihydroethidium ( DHE ) fluorescence in the anterior region , including ISCs , GSCs , and early GSC progeny , in comparison with the control germarium ( B ) . GST2 ( B″ ) or catalase ( CAT ) ( B′″ ) overexpression in ISCs restores low DHE fluorescence in the dshKD germariun . ( C–J ) GST2 ( D , G ) , CAT ( E , H ) , or superoxide dismutase1 ( SOD1 ) ( F , I ) overexpression in dshKD ISCs significantly decreases CBs ( arrowheads; D–F ) and significantly increase ISCs ( arrows; G–I ) in comparison with dshKD . ( C , J ) Quantification results on CB and ISC numbers , respectively ( for each genotype , 50 or more germaria examined ) . ( K ) Quantitative RT-PCR results on the purified ISCs show that GstD2 knockdown by RNAi Line 1 ( GstD2KD1 ) , but not the line 2 ( GstD2KD2 ) , significantly decreases GstD2 mRNA levels in comparison to the control ( c587 ) . ( L–O ) GstD2KD1 ( L ) and CatKD germaria contain 4 CBs and 0 CB , respectively , but the GstD2KD1 CatKD germarium ( N ) contains 5 CBs ( O: CB quantitative results ) . ( P–S ) GstD2KD1 CatKD germarium ( R ) contains fewer ISCs ( arrows ) than GstD2KD1 ( P ) and CatKD ( Q ) germaria ( S: ISC quantitative results ) . Note: PZ1444 expression appears to be downregulated in GstD2KD1 CatKD ISCs . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 01010 . 7554/eLife . 08174 . 011Figure 6—figure supplement 1 . Wnt signaling maintains the reduced redox state in ISCs . ( A ) The control ovariole exhibit low DHE staining in the germarial region and gradually upregulates DHE staining in differentiated germ cells in egg chambers . B′–F′ only show DHE staining in B–F . In B–F′ , broken ovals highlight cap cells , whereas arrowheads point to PZ1444-labeled ISCs . ( B , B′ ) Control germarium shows low DHE staining in ISCs and cap cells ( broken oval ) . ( B–F′ ) armKD ( C–D′ ) and dshKD ( E–F′ ) ISCs as well as their underneath germ cells increase DHE staining . ( G–I ) c587-directed GST2 ( H ) or CAT ( I ) overexpression in ISCs restores low DHE fluorescence in the anterior region of the dshKD2 germariun ( G ) . Bars in A and B ( B–I in the same scale ) represent 20 μm and 10 μm , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 01110 . 7554/eLife . 08174 . 012Figure 6—figure supplement 2 . Reduced redox state is important for ISC maintenance . In A–F , broken ovals highlight cap cells and GSCs , whereas arrowheads point to CBs . ( A–G ) c587-directed overexpressing GST2 ( A ) , CAT ( B ) , or SOD1 ( C ) in ISCs does not affect GSC maintenance and germ cell differentiation , but drastically reduces CBs in the dshKD2 germaria ( D–F ) . ( G ) GSC quantification results . In H–M , broken ovals highlight cap cells and GSCs , whereas arrows point to PZ1444-labeled ISCs . ( H–M ) c587-directed overexpressing GST2 ( H ) , CAT ( I ) , or SOD1 ( J ) in ISCs does not affect ISC maintenance , but drastically rescues the ISC loss phenotype in the dshKD2 germaria ( K–M ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 012 Then , we determined if the restoration of the reduced redox state in dshKD ISCs could rescue the germ cell differentiation defect . Cytoplasmic superoxide dismutase1 ( SOD1 ) is also important for the clearance of cellular ROS . The germaria overexpressing GST2 , CAT , and SOD1 in ISCs contain normal CB numbers ( Figure 6C; Figure 6—figure supplement 2A–C ) . GST2 , CAT , and SOD1 overexpression in the dshKD ISCs drastically and significantly reduces CB numbers in the germaria , but does not change GSC numbers significantly , indicating that increased cellular ROS levels in dshKD ISCs are largely responsible for the germ cell differentiation defect ( Figure 6C–F; Figure 6—figure supplement 2D–G ) . Similarly , GST2 , CAT , and SOD1 overexpression in the dshKD ISCs significantly and drastically rescue the ISC number , but not to the wild-type ISC number , indicating that increased ROS is at least partly responsible for the loss of dshKD ISCs ( Figure 6G–J; Figure 6—figure supplement 2H–M ) . Taken together , these results demonstrate that Wnt signaling in ISCs maintains the reduced redox state , which is partly responsible for ISC maintenance and germ cell differentiation . Our RNA-seq results indicate that GstD2 is the most abundantly expressed Gst genes in ISCs . In addition , it is also the most severely downregulated Gst gene in the Wnt signaling-defective ISCs , which prompted us to further investigate its function in promoting germ cell differentiation ( Figure 6A ) . We generated two independent RNAi lines against GstD2 , among which Line 1 ( GstD2KD1 ) efficiently knocks down GstD2 mRNA but Line 2 ( GstD2KD2 ) does not based on quantitative RT-PCR results on the purified ISCs ( Figure 6K ) . In contrast with the control germaria containing one CB on average , the GstD2KD1 germaria contain two CBs , which are significantly more than the control ( Figure 6L , O ) . As expected , the GstD2KD2 germaria behave like the control , containing one CB on average ( Figure 6O ) . Interestingly , c587-mediated expression of Gst2 and Cat can also rescue the moderate germ cell differentiation defect caused by GstD2KD1 , indicating that the differentiation defect is caused by GstD2 knockdown in ISCs ( Figure 6O ) . Considering the fact that additional three Gst genes are downregulated in the Wnt signaling-defective ISCs , these results indicate that Gst genes are required in ISCs to promote germ cell differentiation . Both Cat and Gst genes work together to remove cellular hydrogen peroxide ( H2O2 ) . The c587-mediated Cat knockdown ( CatKD ) germaria behave like the control germaria , containing one CB on average ( Figure 6M , O ) . Interestingly , c587-mediated CatKD significantly enhances the germ cell differentiation defect caused by GstD2KD1 , and consequently the double knockdown germaria accumulate significantly more CBs than the single knockdown germaria ( Figure 6N , O ) . In addition , the CatKD germaria carry a normal number of ISCs , whereas the GstD2KD germaria carry slightly fewer ISCs ( Figure 6P , Q , S ) . Consistently , the GstD2KD1 CatKD germaria carry significantly fewer ISCs than the GstD1KD1 or CatKD germaria ( Figure 6R , S ) . These results indicate that Cat and Gst genes work together in ISCs to maintain ISCs and promote germ cell differentiation , and further suggest that redox control is important for ISC maintenance . Our RNA-seq results indicate that four Wnt-like genes are expressed in the purified ISCs , but wingless shows little expression ( Figure 7A ) . Wnt2 and Wnt4 are expressed at high levels , whereas Wnt5 and Wnt6 are present in much lower levels ( Figure 7A ) . A recent study has shown that wnt4 mRNAs are indeed restricted to ISCs , whereas wnt2 mRNAs are also present in ISCs ( Luo et al . , 2015 ) . Knocking down Wnt2 or Wnt4 alone in ISCs by two or three independent RNAi lines results in no or slight germ cell differentiation defect based on CB numbers ( Figure 7B–D ) . This is in contrast with the recent study claiming that Wnt4 alone in ISCs is responsible for germ cell differentiation ( Hamada-Kawaguchi et al . , 2014 ) . Consistent with the idea that Wnt2 and Wnt4 function redundantly in ISCs , simultaneous knockdown of Wnt2 and Wnt4 via different combinations of RNAi lines leads to a severe germ cell differentiation defect , which is similar to that caused by dshKD or armKD ( Figure 7D–H ) . These results demonstrate that Wnt2 and Wnt4 in ISCs serve as redundant autocrine signals for promoting germ cell differentiation . 10 . 7554/eLife . 08174 . 013Figure 7 . Wnt2 and Wnt4 function redundantly in ISCs to promote germ cell differentiation . ( A ) RNA-seq results show that Wnt2 and Wnt4 are highly expressed in the purified ISCs in comparison with Wnt5 and Wnt6 . In B–C′ and E–H , cap cells and GSCs are highlighted by ovals , CBs are indicated by arrowheads . ( B–D ) Wnt2 ( B–B″ ) or Wnt4 ( C , C′ ) knockdown by independent RNAi lines causes a slight accumulation of CBs . ( D ) Quantification results on CB numbers ( the numbers on the top of bars represent the p values in comparison with the wild-type control , whereas those on the lines indicate the p values between single and double knockdown ) . ( E–H ) Knocking down both Wnt2 and Wnt4 significantly increases CBs . ( I ) A schematic diagram showing that autocrine Wnt2 and Wnt4 signals control ISC maintenance and promote germ cell differentiation at least in part by maintaining the reduced redox state and preventing BMP signaling . DOI: http://dx . doi . org/10 . 7554/eLife . 08174 . 013 Although the differentiation niche is critical for promoting GSC progeny differentiation , little is known about its regulation . Here , we have identified Wnt2 and Wnt4 as autocrine signals to maintain the GSC differentiation niche partly through redox regulation ( Figure 7I ) . First , Wnt signaling is required in ISCs for their maintenance by promoting cell proliferation and survival . Defective Wnt signaling causes a severe ISC loss , thereby preventing germ cell differentiation . Second , Wnt signaling is required to maintain the reduced redox state by sustaining the expression of Gst genes . This represents a novel connection between Wnt signaling and redox control . In addition , this study has also revealed that the reduced redox state is critical for ISC survival and thus for promoting germ cell differentiation . Third , Wnt signaling in ECs promotes GSC progeny differentiation partly by repressing BMP signaling in differentiated GSC progeny . Fourth , Wnt2 and Wnt4 represent redundant autocrine signals for maintaining Wnt signaling in the germ cell differentiation niche . Wnt signaling has been shown to control stem cell self-renewal directly ( Holland et al . , 2013 ) , whereas ROS has shown to prime hematopoietic progenitor differentiation in Drosophila ( Owusu-Ansah and Banerjee , 2009 ) . This study has demonstrated the importance of Wnt signaling in maintaining the GSC differentiation niche by reducing ROS and thus promoting GSC lineage differentiation . Therefore , this study not only has identified critical signals for maintaining the GSC differentiation niche but also has revealed a novel function of Wnt signaling in the regulation of cellular redox . This study has demonstrated that autocrine Wnt signaling controls ISC maintenance , thereby promoting germ cell differentiation . First , canonical Wnt signaling is required in ISCs to promote germ cell differentiation . c587-mediated knockdown of Wnt signal transducers Fz/fz2 , dsh , and arm as well as c587-directed overexpression of Wnt signaling inhibitors sgg and axn leads to similar germ cell differentiation defects . Moreover , c587-directed overexpression of a constitutively active form of arm* can fully rescue the germ cell differentiation defect caused by dsh knockdown . Second , Wnt signaling maintains ISCs by promoting proliferation and survival . c587-mediated knockdown of dsh and arm leads to a severe ISC loss , whereas c587-mediated knockdown of sgg and axn or c587-directed arm* overexpression expands the ISC population . In addition , hyperactive Wnt signaling increases ISC proliferation and decreases apoptosis , whereas Wnt signaling downregulation increases ISC apoptosis and decreases proliferation . Thus , canonical Wnt signaling maintains the differentiation niche by promoting ISC proliferation and survival . However , we could not completely rule out the possibility that defective Wnt signaling leads to the loss of adult ISCs due to early developmental defects . Third , ISC-expressing Wnt2 and Wnt4 function redundantly in the differentiation niche to promote germ cell differentiation . Our RNA-seq results show that wnt2 and wnt4 mRNAs are present in the purified ISCs at high levels , while other wnt genes are expressed at much lower levels . c587-mediated wnt2 and wnt4 double knockdown results in more severe germ cell differentiation defects than wnt2 or wnt4 single knockdown . Piwi has recently been shown to be required in the differentiation niche for promoting germ cell differentiation partly by repressing dpp expression ( Jin et al . , 2013; Ma et al . , 2014 ) . Although a recent study proposes that Wnt signaling controls germ cell differentiation by regulating piwi expression ( Hamada-Kawaguchi et al . , 2014 ) , this study shows that Piwi protein and mRNAs are not downregulated in Wnt signaling-defective ISCs , suggesting that Wnt signaling does not sustain piwi expression in the differentiation niche to promote GSC progeny differentiation . Instead , this study has further revealed that Wnt signaling maintains the differentiation niche by controlling the cellular redox . First , our RNA-seq results show that GstD2 , GstD4 , GstD10 , and GstE3 mRNA expression levels are significantly downregulated in the purified Wnt signaling-defective ISCs in comparison with the control ISCs . Second , defective Wnt signaling in ISCs elevates ROS levels in themselves and underneath germ cells , indicating that Wnt signaling is required in the differentiation niche to maintain low ROS levels . It remains unclear if increased ROS levels in early germ cells contribute to their differentiation defects . Third , c587-directed GST2 , SOD1 , and CAT overexpression can dramatically suppress the ROS elevation , the germ cell differentiation retardation , and the ISC loss caused by defective Wnt signaling , indicating that ROS elevation is responsible for the germ cell differentiation defect and the ISC loss . Finally , c587-mediated knockdown of GstD2 and Cat results in the ISC loss and the germ cell differentiation defect , indicating that ROS elevation in ISCs is sufficient to cause ISC loss and retard germ cell differentiation . This study has , for the first time , demonstrated that autocrine Wnt signaling controls cellular redox state in the differentiation niche and thus promotes germ cell differentiation . So far , various studies have revealed a number of important signaling pathways and factors in ISCs to promote germ cell differentiation by maintaining ISC cellular process-mediated germ cell–soma interaction and preventing BMP signaling ( Xie , 2013 ) . This study shows that Wnt signaling is required for preventing BMP signaling in differentiated germ cells and for maintaining long ISC cellular processes . Rho , Eggless , Woc , Lsd1 , Piwi , EGFR signaling , and ecdysone signaling have been shown to be required in ISCs to maintain long ISC cellular processes encasing germ cells ( Schultz et al . , 2002; Kirilly et al . , 2011; Wang et al . , 2011; Eliazer et al . , 2014; Ma et al . , 2014; Maimon et al . , 2014; Konig and Shcherbata , 2015 ) . Since properly differentiated germ cells are also required to maintain long ISC cellular processes ( Kirilly et al . , 2011 ) , it is difficult to distinguish the cause and effect of the germ cell differentiation defect and the ISC cellular process loss . In contrast , three known strategies operate in the differentiation niche to prevent BMP signaling , thereby producing a permissive environment for germ cell differentiation . First , Lsd1 , Rho , and Piwi are required in ISCs to repress dpp mRNA expression , thereby directly preventing BMP signaling in the differentiation niche ( Eliazer et al . , 2011; Kirilly et al . , 2011; Jin et al . , 2013; Ma et al . , 2014 ) . dpp encodes a BMP ligand , which activates BMP signaling important for GSC self-renewal ( Xie and Spradling , 1998 ) . Second , Rho , Eggless , and EGFR signaling function in ISCs to repress the expression of dally , preventing BMP diffusion from the self-renewal niche to the differentiation niche ( Liu et al . , 2010; Kirilly et al . , 2011; Wang et al . , 2011 ) . Dally , which is a proteoglycan protein facilitating BMP differentiation and signaling , is expressed in cap cells , the GSC self-renewal niche , to restrict BMP signaling to GSCs ( Akiyama et al . , 2008; Guo and Wang , 2009 ) . Third , a recent study has proposed that cap cell-initiated Wnt signaling maintains the expression of tkv encoding a type I BMP receptor in ISCs , thereby preventing BMP signaling in differentiated germ cells ( Luo et al . , 2015 ) . Our findings from this study have also supported the idea that autocrine Wnt signaling in ISCs promotes GSC progeny differentiation partly by repressing BMP signaling . However , our findings have also argued that inactivating Wnt signaling does not regulate the mRNA expression levels of tkv and other BMP pathway components in ISCs . Therefore , future research will be needed to investigate the molecular mechanisms for Wnt signaling in the differentiation niche to prevent BMP signaling and maintain long ISC cellular processes . The Drosophila stocks used in this study include: c587 ( Kirilly et al . , 2011 ) , PZ1444 ( Kirilly et al . , 2011 ) , UAS-SOD1 ( Pan et al . , 2007 ) , UAS-axn ( BL7225 ) , UAS-sgg ( BL6818 ) , armRNAi ( BL31304; BL31305 ) , dshRNAi ( BL31306; BL31307 ) , axnRNAi ( BL31705 ) , sggRNAi ( BL35364 ) , Wnt2RNAi ( BL36722; TH00483; TH00484 ) , and Wnt4RNAi ( BL29442; TH00485 ) . TH lines for Wnt2 and Wnt4 were constructed for this study according to the published procedure ( Ni et al . , 2011 ) ; the coding region of Gst2 and Cat were cloned into a UASp vector to generate UAS-Gst2 and UAS-Cat using standard molecular biology techniques . Drosophila strains were maintained and crossed at room temperature on standard cornmeal/molasses/agar media unless specified . To maximize the RNAi-mediated knockdown effect , newly eclosed flies were cultured at 29°C for a week before the analysis of ovarian phenotypes . BrdU incorporation , TUNEL labeling , and DHE staining were performed according to the published procedures ( Xie and Spradling , 1998; Owusu-Ansah and Banerjee , 2009; Wang et al . , 2011 ) . Immunohistochemistry was performed according to our previously published procedures ( Xie and Spradling , 1998 ) . The following antibodies were used in this study: rabbit polyclonal anti-β-galactosidase antibody ( 1:100 , Cappel ) , mouse monoclonal anti-Hts antibody ( 1:50 , DSHB ) , Guinea pig polyclonal anti-Piwi antibodies ( 1:100; provided by H . Lin ) , rabbit monoclonal anti-pS423/425 Smad3 antibody ( 1:100 , Epitomics ) , and rat monoclonal anti-Vasa antibody ( 1:50 , DSHB ) . All images were taken with a Leica TCS SP5 confocal microscope . The GFP-positive dshKD or axn-overexpressing ISCs were sorted by FACS , and mRNA isolation was carried out according to the published procedure ( Ma et al . , 2014 ) . Following manufacturer's directions and using the Illumina TruSeq Stranded mRNA LT Kit ( Illumina , MA , United States; Cat . No . RS-122-2101/2 ) , short fragment libraries were constructed . The resulting libraries were purified using Agencourt AMPure XP system ( Backman Coulter , CA , United States; Cat . No . A63880 ) and were then quantified using a LabChip® GX ( PerkinElmer , MA , United States ) and a Qubit Fluorometer ( Life Technologies , CA , United States ) . All libraries were pooled , re-quantified , and run as 50 bp single-end lanes on an Illumina HiSeq 2500 instrument , using HiSeq Control Software 2 . 0 . 5 and Real-Time Analysis ( RTA ) version 1 . 17 . 20 . 0 . Secondary Analysis version CASAVA-1 . 8 . 2 was run to demultiplex reads and generate FASTQ files .	An animal or plant has many different types of cells that have specific roles in the life of the organism . These cells are organized into tissues . In most tissues in adult animals , small groups of cells called stem cells are responsible for replacing the other cells that have been lost due to disease , injury , or as part of normal body maintenance . The ‘germ line’ stem cells of female fruit flies—which produce female sex cells ( or eggs ) —are an effective system for studying how stem cells are regulated . These cells live in an area of the ovary called a stem cell niche . Each time a stem cell divides , it produces one stem cell and one other daughter cell . This daughter cell then moves into another niche called the ‘differentiation’ niche and undergoes a series of divisions that produce the egg cells . The differentiation niche is formed by escort cells and is crucial for producing the egg cells , but it is not clear how the escort cells promote this process , or how the niche is maintained . Wang et al . have now studied the differentiation niche in more detail . The experiments show that a cell communication system called Wnt signaling maintains the differentiation niche by controlling the ability of the escort cells to grow and divide . If Wnt signaling is defective , the differentiation niche is lost , which disrupts the formation of egg cells . Further experiments show that two proteins called Wnt2 and Wnt4 in the differentiation niche—which activate Wnt signaling—act as signals to regulate the niche , mainly by controlling the expression of four particular genes . These four genes encode enzymes that remove ‘reactive oxygen species’ from cells . Wang et al . 's findings have revealed an important role for Wnt signaling in maintaining the differentiation niche . The next step is to figure out the details of how this works .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine" ]	2015	Wnt signaling-mediated redox regulation maintains the germ line stem cell differentiation niche
Physiological adaptation to proteotoxic stress in the endoplasmic reticulum ( ER ) requires retrotranslocation of misfolded proteins into the cytoplasm for ubiquitination and elimination by ER-associated degradation ( ERAD ) . A surprising paradox emerging from recent studies is that ubiquitin ligases ( E3s ) and deubiquitinases ( DUBs ) , enzymes with opposing activities , can both promote ERAD . Here we demonstrate that the ERAD E3 gp78 can ubiquitinate not only ERAD substrates , but also the machinery protein Ubl4A , a key component of the Bag6 chaperone complex . Remarkably , instead of targeting Ubl4A for degradation , polyubiquitination is associated with irreversible proteolytic processing and inactivation of Bag6 . Importantly , we identify USP13 as a gp78-associated DUB that eliminates ubiquitin conjugates from Ubl4A to maintain the functionality of Bag6 . Our study reveals an unexpected paradigm in which a DUB prevents undesired ubiquitination to sharpen substrate specificity for an associated ubiquitin ligase partner and to promote ER quality control . In eukaryotic cells , secretory and membrane proteins are processed in the endoplasmic reticulum ( ER ) . In this protein maturation process , nascent polypeptides need to pass stringent checkpoints monitored by a sophisticated quality control program . Only correctly folded proteins can exit the ER and reach their final destinations , whereas terminally misfolded or unassembled polypeptides are retained and eliminated by the ER-associated degradation ( ERAD ) system ( Vembar and Brodsky , 2008; Smith et al . , 2011 ) . This mechanism not only adapts cells to proteotoxic stress ( Balch et al . , 2008 ) , but also fine-tunes signaling processes essential for the development of multicellular organisms ( Ron and Walter , 2007 ) . ERAD requires retrotranslocation of misfolded polypeptides across the membrane for ubiquitination by one of a few ER-associated ubiquitin conjugating systems , each comprising a ubiquitin ligase ( E3 ) and a cognate conjugating enzyme ( E2 ) ( Hirsch et al . , 2009; Ye and Rape , 2009 ) . Ubiquitin ligases are a key component of the cellular ubiquitination machinery that acts at the final step of ubiquitin transfer to ensure only certain substrates are selectively ubiquitinated in both temporally and spatially regulated manners ( Deshaies and Joazeiro , 2009 ) . Substrate specificities of E3s are thought to be an intrinsic property of these enzymes , but whether it can be modulated by additional factors is an open question . In ERAD , substrate ubiquitination occurs on the cytosolic side of the ER membranes , immediately after substrates have emerged from the ER lumen . This modification serves as a signal to recruit the p97/VCP ATPase , leading to the extraction of substrates from the ER membrane ( Bays et al . , 2001; Ye et al . , 2001; Jarosch et al . , 2002; Flierman et al . , 2003 ) . Additionally , it facilitates subsequent targeting of retrotranslocated substrates to the proteasome ( Vembar and Brodsky , 2008 ) . In mammalian cells , the multi-spanning membrane protein gp78 is one of the most extensively studied E3s in ERAD ( Fang et al . , 2001; Song et al . , 2005; Jo et al . , 2011; Chen et al . , 2012 ) . gp78 is homologous to Hrd1p , an ERAD-specific E3 in yeast proposed to form a ‘retrotranslocon’ that channels misfolded proteins out of the ER ( Carvalho et al . , 2010 ) . Similar to Hrd1p , gp78 also forms a large membrane complex that communicates with machinery proteins on both sides of the membrane ( Zhong et al . , 2004; Christianson et al . , 2011 ) , establishing it as a master regulator of retrotranslocation . On the luminal side , ER chaperones and lectins recognize and deliver misfolded proteins to a gp78-containing complex for retrotranslocation . In the cytosol , gp78 uses a VIM ( VCP-interacting motif ) segment to bind p97/VCP ( Ballar et al . , 2006 ) . In addition , it has a CUE ( coupling ubiquitin to ER degradation ) domain that recruits a multiprotein complex comprising Bag6 and its cofactors Ubl4A and Trc35 ( Chen et al . , 2006; Wang et al . , 2011 ) . Bag6 contains an unusual chaperone ‘holdase’ activity , capable of maintaining retrotranslocated polypeptides in a soluble state to enhance their turnover ( Wang et al . , 2011 ) . Moreover , Bag6 was reported to interact transiently with the proteasome ( Minami et al . , 2010 ) , raising the possibility that it may facilitate substrate hand over from the gp78–p97 complex to the proteasome . Importantly , this ‘holdase’ activity can also act in conjunction with other ubiquitin ligases to help degrade defective cytosolic and mislocalized proteins ( Minami et al . , 2010; Hessa et al . , 2011 ) , establishing Bag6 as a major E3-associated chaperone in protein quality control ( Lee and Ye , 2013 ) . In addition to ubiquitin ligases and associated factors , recent studies have implicated deubiquitinases ( DUBs ) , enzymes that cleave ubiquitin conjugates , as important regulators of ERAD . However , unlike ubiquitin ligases , the functions of DUBs in ERAD are obscure . Many DUBs involved in ERAD associate with p97/VCP either directly or indirectly . These include YOD1 ( Ernst et al . , 2009 ) , ataxin-3 ( Wang et al . , 2006; Zhong and Pittman , 2006 ) , USP25 ( Blount et al . , 2012 ) , USP13 , USP50 , and VCPIP1 ( Sowa et al . , 2009 ) . Interestingly , although deubiquitination has often been reported to cause removal of the proteasome targeting signals and therefore inhibition of protein turnover , many p97-associated DUBs apparently serve as positive regulators in ERAD ( Wang et al . , 2006; Ernst et al . , 2009; Sowa et al . , 2009 ) . Several models have been proposed to explain these surprising findings , but experimental data in support of any of these models are lacking ( Liu and Ye , 2012 ) . In this study , we characterize a p97-associated deubiquitinase termed USP13 . Our study reveals an unexpected interaction between USP13 and the ERAD-specific ubiquitin ligase gp78 . Despite having opposing activities , USP13 can cooperate with gp78 to promote ERAD . Mechanistically , USP13 plays a crucial role in fine-tuning the substrate specificity of gp78 as in its absence ubiquitin chains assembled by gp78 can accumulate on an ERAD machinery component—the Bag6-Ubl4A-Trc35 complex that is associated with proteolytic processing of Bag6 and altered interaction between Bag6 and SGTA . Together , these results established that the activity of an E3 ubiquitin ligase can be safeguarded by an accompanying deubiquitinase to enhance its specificity . USP13 shares ∼80% sequence similarity to USP5 ( Figure 1—figure supplement 1A ) , an isopeptidase that preferentially cleaves unanchored ubiquitin chains ( Reyes-Turcu et al . , 2006 ) . However , only USP13 , but not USP5 has been implicated in ERAD . To understand how these homologous enzymes can have distinct functions , we first purified recombinant USP13 and USP5 from both mammalian cells and Escherichia coli . We tested their ability to bind free ubiquitin and to cleave ubiquitin-AFC ( Ub-AFC ) in vitro . Interestingly , USP5 readily bound to free ubiquitin and cleaved Ub-AFC with high efficiency , whereas USP13 was almost completely inactive ( Figure 1A , Figure 1—figure supplement 1B , C ) . USP13 could not efficiently cleave K48- or K63-linked di-ubiquitin either ( data not shown ) . We next generated and purified USP13 mutants in each of which a fragment in USP13 was replaced by a corresponding segment in USP5 ( Figure 1B ) . Ub-AFC assay showed that the chimera ( USP13/5-4 ) in which the first 722 amino acids of USP13 were replaced with a corresponding segment in USP5 still had no deubiquitinating activity . However , further incorporation of a USP5 segment encompassing the second ubiquitin-associated ( UBA ) domain ( amino acids 717–781 ) to USP13 ( USP13/5-5 ) led to an activity similar to that of USP5 ( Figure 1B , C ) . These data suggest that USP13 and USP5 are distinct deubiquitinases and the second UBA domain accounts for the differential activities of these enzymes . 10 . 7554/eLife . 01369 . 003Figure 1 . USP13 and USP5 have distinct activities despite sequence homology . ( A ) Purified USP13 is inactive whereas USP5 is active . The activities of the purified USP5 and USP13 were measured by the Ub-AFC assay . The Coomassie blue-stained gels show the purified proteins . ( B ) A schematic illustration of the USP13-USP5 chimeras and a summary of their deubiquitinating activities . ( C ) The deubiquitinating activity of the indicated USP13-USP5 chimeras as measured by the Ub-AFC assay . The USP13 catalytic inactive mutant C345S was used as a negative control . The inset shows the purified proteins . ( D ) USP13 does not bind p97 directly . The indicated GST-tagged proteins were immobilized and incubated with recombinant p97 . The precipitated proteins were analyzed by immunoblotting ( IB ) . ( E ) USP13 , but not USP5 , binds p97 through an adaptor . As in D , except that a whole cell extract was used in replace of p97 . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 00310 . 7554/eLife . 01369 . 004Figure 1—figure supplement 1 . USP13 and USP5 have distinct deubiquitinating activities . ( A ) Sequence alignment of USP13 and USP5 . ( B ) Deubiquitinating activity of recombinant USP13 and USP5 purified from E . coli . Purified USP5 and USP13 ( 200 nM ) were incubated with Ub-AFC ( 0 . 5 µM ) and the fluorescence intensity was measured . ( C ) USP13 does not bind free ubiquitin directly . The indicated GST-tagged proteins were immobilized and incubated with ubiquitin . The precipitated proteins were analyzed by SDS-PAGE and Coomassie blue staining . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 004 USP13 , but not USP5 was reported to bind p97 ( Sowa et al . , 2009 ) , suggesting that these enzymes also differ in binding partners . To understand the role of USP13 in ERAD , we characterized its interaction with p97 using an in vitro binding assay . We immobilized purified GST-USP13 or GST-USP5 on glutathione beads and incubated the beads with recombinant p97 . As a positive control , we used GST-VIMP , a known p97 interactor ( Ye et al . , 2004 ) . As expected , the glutathione beads containing GST-VIMP pulled down p97 efficiently . By contrast , neither GST-USP13 nor GST-USP5 bound p97 ( Figure 1D ) . However , when glutathione beads containing the same set of proteins were incubated with whole cell extracts , both VIMP and USP13 co-precipitated with endogenous p97 , but USP5 still failed to bind p97 ( Figure 1E ) . These results indicate that USP13 , but not USP5 , interacts with p97 through an adaptor ( s ) in cells , which explains why USP13 , but not USP5 can function in ERAD . We used the GST pull-down assay to screen a collection of known p97 interactors in order to identify factor ( s ) that promotes the USP13-p97 interaction . We expressed these proteins in cells individually and examined the binding of endogenous p97 to GST-USP13 . Compared to the control ( cells transfected with an empty vector ) , expression of several UBX domain-containing p97 cofactors including ASPL , SAK1 , and UbxD1 reduced the p97-USP13 interaction ( Figure 2—figure supplement 1A , lanes 3 , 5 , 7 vs lane 1 ) , suggesting that these factors may compete with USP13 for p97 binding . By contrast , several other factors enhanced the p97-USP13 interaction . Among them , the most significant enhancer was gp78 ( lane 16 ) , a ubiquitin ligase known to mediate retrotranslocation and ubiquitination of many ERAD substrates ( Fang et al . , 2001; Song et al . , 2005; Christianson et al . , 2011; Jo et al . , 2011; Chen et al . , 2012 ) . Several lines of evidence confirmed that gp78 could bind to USP13 directly and promote its interaction with p97 . Firstly , GST-USP13 co-precipitated with both endogenous and overexpressed gp78 in addition to p97 ( Figure 2A , lanes 4 , 7 ) . By contrast , USP5 did not bind endogenous gp78 , and only a negligible amount of overexpressed gp78 could be pulled down by GST-USP5 ( lanes 5 , 8 ) . Secondly , depletion of gp78 by shRNA-mediated gene knockdown or gene deletion reduced the USP13-p97 interaction by threefold to fourfold ( Figure 2B , Figure 2—figure supplement 1B ) . An endogenous interaction between USP13 and gp78 was also detected ( Figure 2C ) . Moreover , GST-tagged gp78 cytosolic domain ( gp78c ) could bind recombinant USP13 in vitro ( Figure 2D ) . These results demonstrated a specific and direct interaction between USP13 and gp78 that facilitates USP13-p97 association . The data also suggest that the interaction of USP13 with p97 is mediated by several cellular factors with gp78 being a major one . 10 . 7554/eLife . 01369 . 005Figure 2 . USP13 forms a complex with the ERAD E3 gp78 . ( A ) gp78 enhances the interaction of USP13 with p97 . Cell extracts prepared from control or gp78-expressing cells were incubated with glutathione beads containing the indicated proteins . The asterisk indicates a fraction of endogenous gp78 that is co-precipitated with GST-USP13 . The numbers indicate the levels of p97 . ( B ) Depletion of gp78 reduces the interaction of USP13 with p97 . Whole cell extracts ( WCE ) from control or gp78 knockdown cells were incubated with glutathione beads containing either GST or GST-USP13 . ( C ) An interaction between endogenous USP13 and gp78 . Whole cell extract was subject to immunoprecipitation ( IP ) with either control or anti-USP13 antibodies . ( D ) USP13 binds the gp78 cytosolic domain ( gp78c ) directly . The indicated GST-tagged proteins were immobilized and incubated with recombinant USP13 . ( E and F ) Interactions of the indicated USP13-USP5 chimeras with gp78 . WCEs from cells expressing the indicated DUB chimeras and gp78 were subject to IP with FLAG beads . ( G ) ERAD inhibition by overexpressed USP13 is dependent on gp78 interaction . TCRα-YFP cells transfected with the indicated USP13-expressing plasmids were treated with cycloheximide for the indicated time points . Cell extracts were analyzed by immunoblotting . The graph represents quantification of three independent experiments . Error bars , SD ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 00510 . 7554/eLife . 01369 . 006Figure 2—figure supplement 1 . Identification of factors that influence the USP13-p97 interaction . ( A ) Cells expressing the indicated proteins were lysed . Cell extracts were incubated with glutathione beads containing GST-USP13 . The precipitated materials and a fraction of the input were analyzed by immunoblotting with the indicated antibodies . ( B ) Depletion of gp78 reduces the interaction of USP13 with p97 . Whole cell extracts ( WCE ) from control or gp78 knockout cells ( −/− ) were incubated with glutathione beads containing either GST or GST-USP13 . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 006 We next tested the interaction of gp78 with the various USP13/5 chimeras to map the domain in USP13 responsible for gp78 interaction . Co-immunoprecipitation showed that the chimeras defective in cleaving Ub-AFC also behaved similarly to USP13 in binding gp78 ( Figure 2E , lanes 3–7 ) . By contrast , USP13/5-5 had similar DUB activity as USP5 , and both the proteins only co-precipitated with marginal amount of gp78 ( lanes 2 , 8 ) . It appears that the second UBA domain is crucial for USP13 to bind gp78 . To further test this idea , we created an additional construct ( USP13/5-6 ) that had the amino acid sequence downstream of the first UBA domain of USP13 replaced with the corresponding segment from USP5 ( Figure 1B ) . Co-immunoprecipitation showed that like USP5 , this mutant also bound gp78 with significantly reduced affinity when compared to USP13 ( Figure 2F ) . Collectively , these results suggest that the second UBA in USP13 inhibits its DUB activity , but at the same time allows it to gain interaction with gp78 . When the degradation of the model ERAD substrate TCRα was examined in cells overexpressing either USP13 or the USP13 variants , we found that wild-type USP13 significantly increased the half-life of TCRα ( from 1 . 6 hr to 4 . 6 hr ) . Because this phenotype was similarly observed in cells expressing the catalytically inactive USP13 mutant ( USP13 C345S ) ( Figure 2G ) , as well as in USP13 knockdown cells ( see below , Figure 3C ) , we concluded that overexpressed USP13 had a dominant negative effect on ERAD . This was not entirely surprising as several reports had shown similar dominant negative activities with overexpression of other wild-type ERAD factors , presumably because the overexpressed proteins disrupt the normal stoichiometry of endogenous complexes . Consistent with this interpretation , we found that the ERAD inhibitory activity of USP13 was strongly dependent on gp78 binding because USP13/5-4 , a chimera bound gp78 similarly to wild-type USP13 , inhibited TCRα degradation ( t1/2 = 3 . 6 hr ) , whereas the chimera ( USP13/5-5 ) that had reduced affinity to gp78 only slightly increased the half-life of TCRα ( t1/2 = 2 . 3 hr ) . The requirement of gp78 interaction for ERAD inhibition by ectopic USP13 indicates that the interaction of USP13 with gp78 is functionally relevant to ERAD . 10 . 7554/eLife . 01369 . 007Figure 3 . USP13 is required to maintain the solubility of a retrotranslocation substrate . ( A–C ) USP13 knockdown stabilizes the ERAD substrate TCRα . ( A ) HEK293 cells stably expressing TCRα-YFP were transfected with the indicated shRNA-expressing plasmids . Proteins in the whole cell extracts ( WCE ) were analyzed by immunoblotting . ( B ) Control or USP13 knockdown cells expressing TCRα-YFP were analyzed by flow cytometry . The proteasome inhibitor MG132 was used as a positive control . The numbers indicate the calculated X-mean values . ( C ) The stability of TCRα-YFP in control and USP13 knockdown cells was analyzed by CHX chase . The number indicated the relative amount of TCRα-YFP averaged from two independent experiments . ( D ) USP13 knockdown does not stabilize a non-ERAD substrate . As in B , except that cells stably expressing ΔsTCRα-YFP was used . ( E ) USP13 functions at a step downstream of p97-mediated dislocation . The TCRα-YFP-expressing cells treated with the indicated shRNA-expressing plasmids were exposed to the proteasome inhibitor MG132 ( 10 µM , 15 hr ) . TCRα+CHO , glycosylated TCRα; TCRα-CHO , deglycosylated TCRα . ( F ) TCRα-YFP-containing aggregates accumulate in USP13- and Bag6-depleted HEK293 cells . Shown are representative cells transfected with a TCRα-YFP-expressing plasmid together with the indicated shRNA constructs . Arrowheads indicate TCRα-YFP punctae formed upon depletion of USP13 or Bag6 . ( G ) TCRα-YFP-containing aggregates in USP13-depleted cells are resistant to extraction by NP40 . As in F , except that cells were subject to sequential extraction , first by an NP40-containing lysis buffer , then by the Laemmli buffer . UBE1 and H2A serve as the markers for the NP40 soluble ( S ) and insoluble ( P ) fractions , respectively . Note that the NP40-insoluble fractions were loaded two times of the soluble fractions . The numbers indicate the ratio of TCRα in the pellet vs the soluble fraction . ( H and I ) As in G , except that cells expressing the indicated ERAD substrates were used . The numbers in I indicate the levels of TTR D18G . Note that a Bag6 antibody against the C-terminus of Bag6 was used in G and H . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 00710 . 7554/eLife . 01369 . 008Figure 3—figure supplement 1 . USP13 loss-of-function causes TCRα to accumulate in aggregates in cells . ( A ) Cells co-expressing TCRα-YFP together with the indicated shRNAs were subject to sequential extraction , first by an NP40-containing lysis buffer , then by the Laemmli buffer . The NP40 soluble ( S ) and insoluble ( P ) fractions were analyzed by immunoblotting . UBE1 and H2A serve as markers for the NP40 soluble ( S ) and insoluble ( P ) fractions , respectively . ( B ) Expression of USP13 inhibits TCRα degradation in a dominant negative manner . COS7 cells transfected with TCRα-YFP together with either control or the indicated USP13-expressing constructs were stained by a PDI antibody in red . Cells were imaged using a Zeiss Axiovert fluorescence microscope with the exposure time set to auto . Note that in USP13 and USP13 C345S transfected cells , TCRα-YFP often accumulates in aggregates that are not localized to the same focal plane as the ER marker PDI . Scale bars , 10 µm . ( C ) USP13 knockdown stabilizes an ERAD substrate in both glycosylated and deglycosylated forms . Cells co-expressing TCRα-ΔTMD together with the indicated shRNAs were subject to sequential extraction , first by an NP40-containing lysis buffer , then by the Laemmli buffer . Where indicated , the extracts were treated with EndoH before immunoblotting . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 008 To understand how USP13 regulates ERAD , we characterized the ERAD defects in USP13 knockdown cells . First , we used two USP13 specific shRNAs to knock down USP13 in a cell line stably expressing TCRα-YFP . USP13 depletion significantly increased the levels of TCRα-YFP , as revealed by both immunoblotting and flow cytometry analyses ( Figure 3A , B ) . This phenotype could be attributed to increased TCRα stability as demonstrated by cycloheximide ( CHX ) chase experiments ( Figure 3C , lanes 5–8 vs lanes 1–4 ) . Because depletion of DUBs may cause ubiquitin deficiency that can indirectly inhibit proteasomal degradation ( Reyes-Turcu et al . , 2009 ) , we generated a cell line expressing a TCRα variant lacking the signal peptide ( ΔsTCRα-YFP ) . This TCRα mutant is not targeted to the ER . Consequently , it is degraded by an ERAD-independent mechanism ( Fiebiger et al . , 2004 ) . Notably , USP13 knockdown had no effect on the steady state level of ΔsTCRα-YFP ( Figure 3D ) . We conclude that USP13 is specifically required for efficient turnover of ERAD substrates . We next determined which step in ERAD is blocked by USP13 knockdown . To this end , we analyzed the glycosylation status of TCRα in USP13-depleted cells that had been exposed to a proteasome inhibitor . TCRα is a glycosylated type I membrane ERAD substrate that undergoes deglycosylation upon retrotranslocation ( Yu et al . , 1997 ) . Accordingly , in cells treated with the proteasome inhibitor MG132 , TCRα is accumulated in two forms , an ER-localized , glycosylated precursor and a deglycosylated intermediate that has emerged into the cytosol . The level of deglycosylated TCRα therefore can be used to gauge the retrotranslocation activity . As anticipated , knockdown of p97 , the retrotranslocation-driving ATPase , reduced deglycosylated TCRα in MG132-treated cells due to inhibition of retrotranslocation ( Figure 3E , lane 3 vs 1 ) . By contrast , USP13 depletion did not affect deglycosylation of TCRα ( lane 2 vs 1 ) , suggesting that USP13 promotes ERAD downstream of p97 . Because we recently established the Bag6-Ubl4A-Trc35 complex as a key mediator that channels substrates from the sites of retrotranslocation to the proteasome , we asked whether USP13 may cooperate with Bag6 to promote substrate delivery to the proteasome . As demonstrated previously ( Wang et al . , 2011 ) , depletion of Bag6 caused accumulation of TCRα-YFP in aggregates in a fraction of cells ( ∼30% ) due to lack of chaperoning during the proteasome targeting steps . TCRα in these cells was resistant to extraction by the non-ionic detergent NP40 ( Figure 3F , G ) . Intriguingly , we observed accumulation of TCRα in aggresome-like punctae in similar number of the TCRα-YFP-expressing cells under USP13 knockdown conditions ( Figure 3F ) , whereas in control cells , TCRα-containing aggregates was detected only in ∼8% of the cells . Immunoblotting confirmed that USP13 depletion caused more TCRα to partition into the NP40-insoluble fractions ( Figure 3G ) compared to control cells . For comparison , knockdown of p97 resulted in greater levels of TCRα stabilization , but only a small fraction of TCRα was present in the NP40-insoluble fraction ( Figure 3—figure supplement 1A , lane 8 vs 4 ) . Because p97 depletion caused TCRα to accumulate primarily inside the ER , our data suggest that TCRα forms aggregates in USP13-depleted cells most likely after it has entered the cytosol . To further characterize the localization of the TCRα-containing aggregates , we examined the localization of stabilized TCRα in cells expressing either USP13 or USP13 C345S . Compared to control cells in which TCRα was expressed at low levels and displayed a reticulum-like pattern co-localized with the ER marker PDI ( Figure 3—figure supplement 1B , left column ) , ∼50% of the USP13- and USP13 C345S-expressing cells contained TCRα-positive aggregates ( Figure 3—figure supplement 1B , middle and right columns ) . These punctae were not co-localized with PDI , suggesting that they had left the ER . The fact that similar ERAD phenotypes were detected under both USP13 overexpression and depletion conditions further confirmed that USP13 overexpression inhibits ERAD by a dominant negative manner . We also tested the effect of USP13 knockdown on the solubility of several other ERAD substrates including a transmembrane domain ( TMD ) -deleted TCRα variant predicted to be an ERAD-C substrate ( Soetandyo et al . , 2010 ) , and the non-glycosylated ERAD substrate TTR D18G ( Christianson et al . , 2011 ) . Cells expressing these ERAD substrates together with USP13 shRNA were subjected to extraction using the NP40 lysis buffer followed by the Laemmli buffer . Immunoblotting showed that USP13 depletion caused accumulation of these substrates in cells . Importantly , in both cases , more substrates were detected in the NP40-insoluble fractions in USP13 knockdown cells than in control cells ( Figure 3H , I ) . Moreover , consistent with the model that USP13 acts together with Bag6 at a post-dislocation step , a significant fraction of TCRα-ΔTMD accumulated in USP13- and Bag6-knockdown cells in a deglycosylated form ( Figure 3H , Figure 3—figure supplement 1C ) . These results demonstrate a function of USP13 in ERAD downstream of retrotranslocation and suggest a potential functional interplay between USP13 and Bag6 that is required to maintain retrotranslocated substrates in a soluble and degradable state . We next tested whether USP13 could physically interact with Bag6 and/or its cofactors . Immunoprecipitation of USP13 from a whole cell extract resulted in co-precipitation of a fraction of Bag6 and its cofactor Ubl4A in addition to gp78 and p97 . By contrast , the cytosolic chaperone Hsp90 was not co-precipitated ( Figure 4A ) , demonstrating a specific interaction between USP13 and the Bag6 complex . The interaction of USP13 with the Bag6 complex and gp78 did not require its deubiquitinating activity ( Figure 4B ) . This interaction could be further demonstrated by pull-down experiments using GST-tagged USP13 and whole cell extracts ( Figure 4C ) . Notably , USP13 not only bound Bag6 , but also interacted with MMS1 and α6 , subunits of the proteasome ( lane 5 ) , further implicating USP13 in delivery pathways that targets substrates to the proteasome . Because depletion of gp78 did not affect the interaction of USP13 with Bag6 ( Figure 4C , lane 7 vs 5 ) , USP13 might bind Bag6 directly . Indeed , GST pull-down experiments using purified proteins showed that GST-USP13 bound Bag6 , and the N-terminal ubiquitin-like ( UBL ) domain in Bag6 was both necessary and sufficient for interaction with USP13 ( Figure 4D , E ) . Taken together , these results suggest a direct interaction between USP13 and Bag6 that is mediated by the Bag6 UBL domain . 10 . 7554/eLife . 01369 . 009Figure 4 . USP13 interacts with the Bag6 complex . ( A ) Endogenous interaction of USP13 with the Bag6 complex . Extracts from cells were subject to IP by the indicated antibodies followed by immunoblotting . ( B ) USP13 binds gp78 and Bag6 independent of its DUB activity . Extracts from HEK293 cells-transfected with the indicated plasmids were subject to IP with anti-FLAG beads . Asterisk indicates a non-specific band . ( C ) USP13 interacts with both the proteasome and Bag6 in a gp78 independent manner . Control ( ctrl . ) or gp78-depleted cell extracts were incubated with glutathione beads containing either GST or GST-USP13 . The precipitated materials were analyzed by immunoblotting ( Upper two panels ) or by Ponceau S staining ( the lower panel ) . ( D ) The UBL domain in Bag6 is required for interaction with USP13 . A GST pull-down experiment was performed using the indicated Bag6 proteins purified from mammalian cells . ( E ) Bag6 UBL binds USP13 more tightly than Ubl4A UBL . Recombinant proteins purified from E . coli were used . ( F ) USP13 , gp78 and Bag6 form a multi-protein complex . Sequential IP first by FLAG beads then by gp78 antibody using extracts from either control ( − ) or FLAG-USP13 ( F-USP13 ) -expressing cells ( indicated by ‘+’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 009 Since Bag6 can interact with both gp78 and USP13 , we asked whether these proteins could form a multiprotein complex . We performed sequential immunoprecipitation using cells expressing FLAG-USP13 and untagged gp78 . As expected , proteins eluted from FLAG beads contained gp78 and each component of the Bag6 complex in addition to FLAG-USP13 ( Figure 4F , lane 2 ) . When this eluate was subject to another round of immunoprecipitation using gp78 antibody , a significant fraction of the Bag6 complex remained bound to gp78 together with USP13 ( lane 8 ) , demonstrating that these proteins can form a multiprotein complex . Given that USP13 and gp78 are enzymes modulating ubiquitin dynamics , we reasoned that they might regulate Bag6 by controlling the ubiquitination status of a component in the Bag6 complex . We first asked whether Bag6 itself could be a substrate of USP13 . Immunoblotting analyses of immunoprecipitated Bag6 failed to detect significant ubiquitin conjugates on Bag6 either in control or USP13 knockdown cells ( Figure 5—figure supplement 1A ) , suggesting that Bag6 is unlikely regulated by USP13-mediated deubiquitination . We next used the same approach to determine the ubiquitination status of Ubl4A and Trc35 , the two Bag6 partners . Our results showed that Ubl4A , but not Trc35 , could be ubiquitinated in cells ( Figure 5A , lane 3 vs 7 ) . Biochemical fractionation experiments showed that ubiquitination of Ubl4A preferentially occurred on the membrane-associated pool of Ubl4A ( Figure 5B , lane 2 vs 1 ) , consistent with the notion that it may be functionally relevant to ERAD regulation . Ubl4A appeared to be conjugated with Lys48-linked ubiquitin chains because expression of the ubiquitin K48R mutant , but not the K63R mutant , significantly reduced ubiquitinated Ubl4A ( Figure 5C , lane 2 vs 1 , 3 ) . Intriguingly , despite carrying Lys48-linked ubiquitin chains , Ubl4A was a stable protein ( Figure 5—figure supplement 1B , Figure 3C ) ( ‘Discussion’ ) . 10 . 7554/eLife . 01369 . 010Figure 5 . USP13 regulates ubiquitination of Ubl4A A , Ubl4A but not Trc35 is ubiquitinated in cells . ( A ) Ubl4A or Trc35 was immunoprecipitated from an extract of HEK293 cells under denaturing condition and blotted with the indicated antibodies . ( B ) Membrane-associated Ubl4A is preferentially ubiquitinated . HEK293 cells were fractionated into an ER-containing membrane and cytosol fractions . Ubl4A immunoprecipitated from these fractions was analyzed by immunoblotting . ( C ) Ubl4A is conjugated with Lys48-linked ubiquitin chains . Ubl4A immunoprecipitated from extracts of cells expressing the indicated ubiquitin variants was analyzed by immunoblotting . ( D and E ) USP13 depletion causes accumulation of ubiquitinated Ubl4A in cells . ( D ) Ubiquitinated Ubl4A was analyzed in cells expressing the indicated shRNAs and HA-ubiquitin . ( E ) Ubiquitinated Ubl4A was analyzed in control wild-type cells and USP13 knockout ( −/− ) cells expressing HA-ubiquitin . ( F ) USP19 depletion does not affect Ubl4A ubiquitination . Asterisk indicates a non-specific band . ( G ) In vitro deubiquitination of Ubl4A by USP13 . Ubl4A purified from USP13 knockout cells was treated with either buffer or increased concentration of FLAG-USP13 ( 0 . 15 µM–0 . 6 µM ) at 37°C for 2 hr . Where indicated , Ub-aldehyde ( Ub-Al , 1 µM ) was included . ( H ) TCRα purified from USP13 knockout cells was incubated with either buffer or FLAG-USP13 ( 0 . 15 µM–0 . 6 µM ) at 37°C for 2 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 01010 . 7554/eLife . 01369 . 011Figure 5—figure supplement 1 . Regulation of Ubl4A by ubiquitination . ( A ) USP13 does not regulate ubiquitination of Bag6 . HEK293 cells were transfected with the indicated shRNA constructs . Endogenous Bag6 was immunoprecipitated under denaturing condition and was analyzed by immunoblotting . Where indicated , a fraction of the lysate ( input ) was analyzed directly by immunoblotting . ( B ) The proteasome inhibitor MG132 does not increase the steady state level of Ubl4A . Immunoblotting analysis of whole cell extracts from HEK293 cells either untreated or treated with MG132 ( 20 µM , 6 hr ) . The numbers show the quantification of Ubl4A levels . Note that MG132 stabilizes the transcription factor ATF4 , but does not affect Ubl4A protein level . ( C ) Ubiquitinated TCRα accumulates in USP13 knockdown cells . ( D ) The USP13 activity is increased at pH 8 . 0 . Purified FLAG-USP13 ( 120 nM ) was incubated with Ub-AFC ( 2 . 5 µM ) under the indicated pH condition . ( E ) Ubl4A purified from USP13 knockout cells was incubated with increased amount of GST-USP13 at 37°C for 2 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 011 If USP13 regulated Ubl4A ubiquitination , depletion of USP13 was expected to cause accumulation of ubiquitinated Ubl4A . Immunoblotting showed that Ubl4A purified from either USP13 knockdown cells or USP13 deficient cells indeed carried more ubiquitin conjugates than that from control cells ( Figure 5D , E ) . Given that knockdown of the ER-localized deubiquitinase USP19 ( Hassink et al . , 2009 ) did not result in accumulation of ubiquitinated Ubl4A ( Figure 5F ) , we concluded that ubiquitination of Ubl4A is specifically regulated by USP13 . Because USP13 depletion also led to accumulation of ubiquitinated ERAD substrate TCRα ( Figure 5—figure supplement 1C , lane 2 ) , it is important to distinguish whether USP13-mediated deubiquitination acts on Ubl4A , the ERAD substrates , or both . To this end , we screened several pH conditions in order to find a way to activate USP13 because previous studies showed that increased pH can facilitate deprotonation of the catalytic cysteine in some USP DUBs to promote their activities ( Lee et al . , 2013 ) . Indeed , at pH 8 . 0 , the activity of USP13 was significantly elevated compared to pH 7 . 4 ( Figure 5—figure supplement 1D ) . We then treated endogenous ubiquitinated Ubl4A purified from USP13-depleted cells with recombinant USP13 under this condition . This treatment did significantly reduce ubiquitinated Ubl4A , as demonstrated by immunoblotting ( Figure 5G , lane 4 vs 1 , Figure 5—figure supplement 1E ) , which was blocked by the specific DUB inhibitor ubiquitin aldehyde ( Figure 5G , lanes 5–8 vs 1–4 ) . As a control , we treated purified ERAD substrate TCRα under the same condition , but no obvious deubiquitination was observed ( Figure 5H ) . Together , these results suggest that USP13 preferentially removes ubiquitin conjugates from Ubl4A , probably because it forms a specific interaction with the Bag6 complex via the Bag6 UBL domain ( Figure 4 ) . Although our results have not ruled out the possibility that USP13 may contribute to deubiquitination of TCRα with the assistance of a substrate recruiting adaptor in cells , the inability to deubiquitinate TCRα by purified USP13 suggests that the accumulation of ubiquitinated TCRα in USP13 knockdown cells is probably due to ERAD inhibition at a step downstream of ubiquitination , as shown in Bag6-depleted cells ( Wang et al . , 2011 ) . Given that membrane-associated Ubl4A is preferentially ubiquitinated , we hypothesized that gp78 might contribute to Ubl4A ubiquitination . Incubating purified Ubl4A , E1 , Ube2g2 , ubiquitin with wild-type gp78c , but not a catalytically inactive gp78c mutant resulted in accumulation of high molecular weight products corresponding to ubiquitinated Ubl4A ( Figure 6A , lane 2 vs 3 ) . Likewise , overexpression of wild-type gp78 , but not the catalytically inactive RING mutant ( gp78 Rm ) in cells significantly increased ubiquitination of endogenous Ubl4A ( Figure 6B ) . These data suggest that gp78 can ubiquitinate Ubl4A directly in vitro and in vivo . Surprisingly , knockdown of gp78 did not significantly reduce ubiquitinated Ubl4A ( Figure 6C ) . One possible way to explain these seemingly contradictory results is that Ubl4A might be ubiquitinated by several Bag6-associated ligases , but only those conjugates assembled by gp78 were preferentially removed by USP13 . If this was correct , ubiquitinated Ubl4A accumulated upon USP13 depletion should be reduced when gp78 was depleted together with USP13 , which was indeed observed in USP13 knockdown ( Figure 6D , lane 9 vs 8 ) or deficient cells expressing gp78 specific shRNA ( data not shown ) . These data suggest that USP13 and gp78 play antagonizing roles in regulation of Ubl4A ubiquitination: While gp78 assembles ubiquitin chains on Ubl4A , USP13 antagonizes this activity to limit Ubl4A ubiquitination . 10 . 7554/eLife . 01369 . 012Figure 6 . USP13 preferentially removes gp78-assembled ubiquitin chain from Ubl4A . ( A ) In vitro ubiquitination of Ubl4A by gp78 . Purified Ubl4A was incubated with E1 , Ube2g2 , ubiquitin , ATP in the absence ( no E3 ) or presence of the wild-type gp78 cytosolic domain ( gp78c ) or a catalytically inactive gp78c RING mutant ( Rm ) . The reactions were analyzed by immunoblotting with anti-Ubl4A antibody . ( B ) Overexpression of wild-type gp78 promotes ubiquitination of Ubl4A . HEK293 cells were transfected with control or the indicated gp78-expressing plasmid . Ubiquitination of endogenous Ubl4A was analyzed by immunoprecipitation followed by immunoblotting . WCE , whole cell extracts . ( C ) gp78 knockdown in cells does not significantly reduce basal ubiquitination of Ubl4A . Ubl4A immunoprecipitated under denaturing conditions from extracts of cells transfected with the indicated shRNA constructs was analyzed by immunoblotting . Where indicated , a fraction of the whole cell extract was directly analyzed by immunoblotting to verify the knockdown efficiency . Asterisk indicates a non-specific band , which serves as a loading control . ( D ) gp78 antagonizes USP13 in regulation of Ubl4A . Lane 1–6 , A fraction of the cells expressing the indicated shRNAs and HA-Ubiquitin were subject to sequential extraction by NP40- and SDS-containing buffers . Lane 7–9 , the remaining cells were used to measure the levels of ubiquitinated Ubl4A as in B . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 012 To see how ubiquitination of Ubl4A influenced the function of the Bag6 complex , we analyzed the Bag6 protein level in USP13-depleted cells by immunoblotting . Intriguingly , we observed that a fraction of Bag6 ( designated as Bag6* ) had increased mobility on SDS-PAGE gel ( Figure 6D , lane 5 ) . This fast migrating Bag6 species was probably caused by proteolytic processing of Bag6 that removed a small C-terminal segment because Bag6* could not be detected by an antibody against the C-terminus of Bag6 ( Figure 3G ) . Bag6* was primarily detected in NP40-insoluble fractions ( Figure 6D , lane 5 vs 1 ) , and its generation appears to be associated with hyper-ubiquitination of Ubl4A because co-depletion of gp78 together with USP13 reduced ubiquitinated Ubl4A and also abolished Bag6* accumulation ( Lane 6 vs 5 ) . Moreover , Bag6* was preferentially detected in the membrane bound pool of Bag6 in USP13-depleted cells ( Figure 7A ) . Coincidentally , membrane-associated Ubl4A was more prone to ubiquitination ( Figure 5B ) . Furthermore , overexpression of gp78 , a condition that significantly increased ubiquitinated Ubl4A , also induced Bag6 cleavage ( Figure 7B , C ) . Expression of a ubiquitin-Ubl4A fusion protein ( Ub-Ubl4A ) but not wild-type Ubl4A also caused accumulation of Bag6* in cells ( Figure 7D ) . Importantly , even though overexpression of gp78 was previously shown to promote turnover of an ERAD substrate ( CD3δ ) ( Fang et al . , 2001 ) whose degradation is independent of Bag6 ( data not shown ) , the degradation of the Bag6 substrate TCRα was significantly inhibited in cells expressing gp78 ( Figure 7E ) . Moreover , expression of Ub-Ubl4A also significantly reduced the rate of TCRα degradation ( Figure 7F ) . Together , these results suggest that hyper-ubiquitination of Ubl4A can cause ERAD inhibition in a dominant negative manner , probably because it is associated with the cleavage and subsequent inactivation of a fraction of membrane bound Bag6 . 10 . 7554/eLife . 01369 . 013Figure 7 . Hyper-ubiquitination of Ubl4A is associated with Bag6 clipping and ERAD inhibition . ( A ) Membrane-associated Bag6 is preferentially cleaved in USP13 knockdown cells . C , cytosol fraction; M , membrane fraction . ( B and C ) Overexpression of gp78 causes cleavage of Bag6 . ( B ) A fraction of the cells transfected as indicated were extracted sequentially with NP40- and SDS-containing buffers . The corresponding extracts were analyzed by immunoblotting . ( C ) Protein extracts from HEK293 cells transfected with control or gp78-expressing plasmid were subject to immunoprecipitation with Ubl4A antibodies or with IgG as a negative control . ( D ) Overexpression of a Ub-Ubl4A fusion protein induces Bag6 cleavage . As in B , except that cells expressing the indicated proteins were analyzed . ( E ) Overexpression of wild-type gp78 inhibits TCRα degradation . The number indicates the relative levels of TCRα averaged from two independent experiments . ( F ) Overexpression of Ub-Ubl4A inhibits TCRα degradation . The graph shows the quantification results from three independent experiments . Error bars , SD ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 013 To further dissect the mechanism by which ubiquitination of Ubl4A inhibits Bag6 function , we used mass spectrometry to map the ubiquitination site in Ubl4A . The result identified Lys48 in Ubl4A as the major ubiquitination site ( Figure 8—figure supplement 1A ) . To confirm that this Lys residue was ubiquitinated in cells , we initially attempted to overexpress a Ubl4A mutant bearing the Lys48 to Arg substitution . However , we found that most transiently expressed Ubl4A was not incorporated into the endogenous Bag6 complex because the complex had a slow turnover rate . As a result , overexpressed Ubl4A was not regulated by the same manner as endogenous Ubl4A . To overcome this problem , we established stable cell lines expressing either FLAG-tagged wild-type Ubl4A or Ubl4A K48R mutant under an inducible promoter . A careful titration experiment showed that when cells were grown in the presence of a low concentration of the inducer tetracycline ( ‘Materials and method’ ) , Ubl4A and Ubl4A K48R could be expressed at a level comparable to that of endogenous Ubl4A , and a significant fraction of ectopic Ubl4A interacted with Bag6 . Under this condition , ubiquitination still preferentially occurred on endogenous Ubl4A ( data not shown ) . However , significant amount of ubiquitinated Ubl4A-FLAG could be detected when gp78 was expressed ( Figure 8A , lane 7 ) . By contrast , overexpression of gp78 did not lead to ubiquitination of the Ubl4A K48R mutant ( lane 8 ) , suggesting that gp78-mediated ubiquitination primarily occurs at Lys48 in Ubl4A in cells . Moreover , when HA-tagged ubiquitin was expressed in these cells , gp78-independent basal ubiquitination of Ubl4A could be detected on wild-type Ubl4A , but not on the Ubl4A K48R mutant ( Figure 8B , lane 3 vs 4 ) . These results establish Lys48 as the major site in Ubl4A for both gp78-dependent and gp78-independent ubiquitination . 10 . 7554/eLife . 01369 . 014Figure 8 . USP13 maintains a functional interaction between Bag6 and SGTA . ( A ) Lys48 of Ubl4A is required for gp78-mediated ubiquitination in cells . Cells stably expressing wild type and K48R mutant Ubl4A were transfected with either an empty vector ( control ) or a gp78-expressing vector . Whole cell extracts were either directly analyzed by immunoblotting ( lanes 1–4 ) or subject to immunoprecipitation with anti-FLAG antibodies prior to immunoblotting ( lane 5–8 ) . ( B ) Mutating Lys48 in Ubl4A reduces basal ubiquitination of Ubl4A . Ubl4A ubiquitination was analyzed using wild type and the K48R mutant Ubl4A cells transfected with an HA-Ub expressing plasmid . ( C ) A model of SGTA-N-Ubl4A UBL complex . The structure of SGTA-N domain ( PDB: 4GOD ) ( Chartron et al . , 2012 ) and Ubl4A UBL domain ( PDB: 2DZI ) was superimposed on the complex of the corresponding yeast complex ( PDB: 2LXC ) ( Chartron et al . , 2012 ) . Note that Lys48 in Ubl4A interacts with two negatively-charged residues in SGTA . ( D ) A Bag6- and SGTA-containing complex can be crosslinked in cells . HEK293 cells were treated with the indicated concentrations of BMH prior to lysis . Cell extracts were subject to IP with either IgG as a control or anti-SGTA antibody followed by immunoblotting . The arrow indicates a crosslinking species that is detected by both anti-Bag6 and anti-SGTA blotting . ( E ) As in D , except that cells expressing the indicated shRNA constructs were used and that IP was only conducted with the anti-SGTA antibody . SE , short exposure; LE , long exposure . ( F ) As in E , except that cells transfected with either control or USP13 shRNA were used . Asterisks indicate SGTA-containing crosslinking products that are not affected by USP13 knockdown . The numbers show the relative levels for the crosslinking products . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 01410 . 7554/eLife . 01369 . 015Figure 8—figure supplement 1 . Bag6 and SGTA form a transient interaction dependent on Ubl4A . ( A ) Mass spectrometry analysis of purified ubiquitinated Ubl4A . Purified Ubl4A was incubated with E2 , Ube2g2 , gp78 , Ub , and ATP . After incubation , ubiquitinated Ubl4A was re-purified and analyzed by mass spectrometry . ( B ) Purified SGTA and the Bag6 ternary complex were analyzed by SDS-PAGE and Coomassie blue staining . ( C ) In vitro crosslinking demonstrates that SGTA can be crosslinked with the Bag6 complex . In vitro crosslinking was performed with the indicated proteins in the presence ( + ) or absence of 50 µM BMH . The reactions quenched by DTT were subject to immunoprecipitation with the indicated antibodies . The precipitated samples were analyzed by immunoblotting . ( D ) The formation of the 250 kD crosslinking product in cells requires Ubl4A . Control or Ubl4A knockdown cells were lysed under a denaturing condition . A fraction of the whole cell extracts ( WCE ) , was analyzed by immunoblotting directly . The remaining samples were subject to immunoprecipitation with anti-SGTA antibody followed by immunoblotting analyses . The arrowhead indicates the 250 kD Bag6-SGTA crosslinking product , which is reduced under Ubl4A knockdown conditions . Note that a fraction of Bag6 also accumulated as the fast migrating Bag6* species in Ubl4A knockdown cells . ( E ) Ubl4A depletion also induces Bag6 cleavage in cells . Cells were directly lysed in the Laemmli buffer and the extracts were analyzed by immunoblotting . ( F ) USP13 helps sharpening the substrate specificity for gp78 . gp78 can act not only on ERAD substrates but also on ERAD machinery proteins that interact with gp78 such as Ubl4A . USP13 antagonizes gp78 to remove ubiquitin conjugates from Ubl4A , thus prevent undesired ubiquitination by gp78 . DOI: http://dx . doi . org/10 . 7554/eLife . 01369 . 015 Because our structural and biochemical studies showed previously that Ubl4A promotes Bag6 binding to a co-chaperone named SGTA via positively-charged residues including Lys48 on Ubl4A ( Figure 8C ) ( Chartron et al . , 2012; Xu et al . , 2012 ) , and because Lys48 is the major site of ubiquitination in Ubl4A , hyper-ubiquitination of Ubl4A under USP13 knockdown conditions might inhibit directly the interaction of Ubl4A with SGTA , and therefore disrupts this functional link between Bag6 and SGTA . To follow the transient interaction between SGTA and the Bag6 complex , we employed a crosslinking approach . We treated HEK293 cells with a cysteine reactive crosslinker to stabilize protein–protein interactions . Cell extracts were subject to immunoprecipitation under denaturing conditions with SGTA antibody . Immunoblotting with SGTA antibodies detected a major crosslinking product whose molecular weight was consistent with a SGTA dimer . In addition , several SGTA-containing crosslinking products were detected only in anti-SGTA precipitated samples , but not in control samples ( IgG pulldown ) . Among them , a crosslinking product of 250 kD was detected by both anti-SGTA and anti-Bag6 blotting ( Figure 8D ) . Two pieces of evidence further suggested that this 250 kD crosslinking product was a crosslinked complex containing both SGTA and Bag6 . First , in vitro crosslinking using purified SGTA and the Bag6 complex generated a similar 250 kD complex ( Figure 8—figure supplement 1B , C ) . This complex was not detected when the Bag6 complex or SGTA was omitted from the reaction . Second , knockdown of either Bag6 or SGTA by shRNA significantly reduced this crosslinking product ( Figure 8E , lanes 2 , 4 vs 6 ) . Consistent with the notion that Ubl4A facilitates the association of SGTA with Bag6 , knockdown of Ubl4A also decreased this crosslinking product ( Figure 8—figure supplement 1D ) . Importantly , USP13 knockdown specifically reduced the 250 kD Bag6-SGTA crosslinking product without affecting either the SGTA dimer or other SGTA-containing crosslinking products ( Figure 8F ) . Collectively , these results indicate that USP13 facilitates the association between SGTA and Bag6 . We report a novel functional link between the deubiquitinase USP13 and gp78 , an ER-associated E3 postulated to be part of a ‘retrotranslocon’ for some misfolded proteins ( Fang et al . , 2001 ) . Despite having opposing activities , these enzymes interact with each other to form a complex that coordinately promotes ERAD . We identify the Bag6 cofactor Ubl4A as a shared substrate of gp78 and USP13 . USP13 depletion is associated with hyper-ubiquitination of Ubl4A and altered interaction between the Bag6 complex and its co-chaperone SGTA . SGTA is an ortholog of Sgt2p , a chaperone implicated in biogenesis of tail-anchored ER proteins in S . cerevisiae ( Wang et al . , 2010 ) . Like Bag6 , SGTA also contains a chaperone-like activity that preferentially binds hydrophobic segments . We recently demonstrated that this activity facilitates substrate binding by Bag6 in ERAD ( Xu et al . , 2012 ) . We also showed that the interaction of Bag6 with SGTA is facilitated by Ubl4A as the latter binds directly to SGTA in a highly dynamic manner ( Chartron et al . , 2012 ) . Because the interaction of Ubl4A with SGTA is mediated by positively-charged residues in Ubl4A including Lys48 ( Chartron et al . , 2012; Xu et al . , 2012 ) , which happens to be the major ubiquitination site , the simplest model to explain reduced Bag6-SGTA interaction in USP13 knockdown cells is that ubiquitin conjugates on Ubl4A sterically hinder SGTA binding . However , given that USP13 can also interact physically with Bag6 , it is also possible that USP13 may serve an adaptor function to promote Bag6–SGTA interaction . Hyper-ubiquitination of Ubl4A in USP13-depleted cells is also associated with increased proteolysis of Bag6 , resulting in a truncated variant ( Bag6* ) . Based on the molecular weight , the cleavage seems to occur within or near the C-terminal BAG domain . It has been shown that the Bag6 co-factor Ubl4A binds to a site near the BAG domain ( Xu et al . , 2013 ) . Thus , it is possible that Ubl4A normally covers a protease site in Bag6 either by itself or by recruiting a Bag6 cofactor . Hyper-ubiquitination of Ubl4A in USP13 knockdown cells may alter its function , leading to increased cleavage of Bag6 . In support of this idea , Bag6* also accumulates in cells depleted of Ubl4A ( Figure 8—figure supplement 1E ) or cells expressing the Ub-Ubl4A fusion protein ( Figure 7D ) . Our results suggest that depletion of USP13 can impair Bag6 function via at least two ways . One is to cause its cleavage and the other is to inhibit its interaction with SGTA ( Figure 8—figure supplement 1F ) . The two events may be linked as increased cleavage of Bag6 may also contribute to reduced interaction with SGTA . Although hyper-ubiquitination of Ubl4A is clearly detrimental to ERAD , it is noteworthy that our results do not exclude the possibility that transient ubiquitination of Ubl4A by gp78 at the site of retrotranslocation may also serve a positive role in ERAD . Because the accumulation of Bag6* upon USP13 depletion is significantly reduced in cells co-depleted of USP13 and gp78 ( Figure 6D ) , we propose that USP13 is required to antagonize a promiscuous activity of gp78 towards Ubl4A , which would otherwise impair the function of the Bag6 complex by altering its interaction pattern and/or increasing its cleavage by a cellular protease . In this model , a DUB can cooperate with an E3 ligase to enhance its substrate specificity . The specificity of a ubiquitination reaction has generally been thought to be controlled at the E3 ligase level . In proteasomal degradation pathways , many E3s appear to directly recognize substrates bearing degradation signals , leading to their ubiquitination . However , in the complex cellular environment , ubiquitin ligases often function in large protein complexes , meaning that in addition to substrates , many cellular proteins containing ubiquitin acceptor lysine residues are also in proximity to these enzymes . How these E3 cofactors evade ubiquitination is completely unknown . Our study suggests that cooperation between an E3 ligase and an associated DUB may provide a simple solution that sharpens substrate specificity for the ligase . It is conceivable that while acting on substrates , E3 ligases may also ubiquitinate other factors that function in proximity . Such undesired ubiquitination , even occurring at a low frequency , could cause significant damage overtime , particularly if it leads to irreversible inactivation of the modified proteins . Removal of these unwanted ubiquitination products by DUBs ensures that only desired ubiquitination signals are maintained in cells . This concept is in line with a recent study , showing that even non-specific DUB activities can amplify the specificity of a quality control E3 , allowing discrimination between two misfolded substrates bearing subtle structural differences ( Zhang et al . , 2013 ) . Our study also suggests that Lys48-linked ubiquitination can have a non-destructive function , which is to regulate protein–protein interactions . Previous studies showed that cells can use ubiquitin to regulate protein–protein interaction in many ways . The most common strategy is to have ubiquitin serve as a ‘signal’ that recruits a downstream effector containing a ubiquitin-binding domain . By contrast , the regulation of Bag6 by the gp78-USP13-mediated ubiquitination cycle seems to represent another paradigm; Ubiquitination inhibits the recruitment of a co-factor . It is noteworthy that a similar ubiquitin-dependent disassembly of a protein complex involved in TGFβ signaling has been reported ( Dupont et al . , 2012; Eichhorn et al . , 2012 ) . How Ubl4A escape proteasomal degradation despite carrying a bone fide proteasome degradation signal remains to be investigated . It is possible that ubiquitin conjugates on Ubl4A is bound by one of the many ERAD factors that carry ubiquitin-binding motifs ( Xu et al . , 2012 ) . These interactions may prevent substrate recognition by the proteasome . Such a mechanism has been reported to maintain the stability of the ubiquitinated transcription factor Met4 in yeast ( Flick et al . , 2006 ) . Our model predicts that the USP13 activity needs to be tightly regulated both spatially and temporally in cells , a theme generally applicable to DUBs ( Reyes-Turcu et al . , 2006 ) . Indeed , we found that purified USP13 is inactive despite sequence homology to USP5 , a constitutively active isopeptidase . Our data suggest that sequence variations in a UBA domain allow USP13 to acquire new interactors , which may activate it at the ER membrane . Since purified gp78 is insufficient to activate USP13 ( Liu , Y unpublished results ) , further experiments using gp78-containing retrotranslocation complexes reconstituted in proteoliposome are required to elucidate how USP13 is regulated in the context of retrotranslocation . In summary , our study demonstrates that a ubiquitin ligase and a DUB can act in concert to coordinate protein turnover at the ER by regulating the ubiquitination status of a degradation machinery factor . In addition , our findings reveal a mechanism by which an E3-DUB ‘tag team’ can enhance E3 specificity , a paradigm that may be applicable to other E3-DUB complexes . The HEK293 cell line stably expressing TCRα-YFP was described previously ( Soetandyo et al . , 2010 ) . Mammalian expression constructs for Ufd1 , Npl4 , Derlin-1 , Ube2g2 , VIMP , Hrd1 , and gp78 were described previously ( Ye et al . , 2001 , 2003 , 2004; Li et al . , 2007 ) . Constructs for expression of UbxD8 , ASPL , UbxD5 , SAK1 , FAF1 , UbxD1 , UbxD7 , p47 were kindly provided by R Deshaies ( Caltech , Pasadena , CA ) ( Alexandru et al . , 2008 ) . FLAG-USP13 and FLAG-USP5 expression plasmids were constructed by cloning the corresponding cDNA into the SalI and NotI sites of the pRK5-FLAG vector ( Li et al . , 2008 ) . The USP5/USP13 chimera constructs were made by fusing USP5 and USP13 DNA fragments ( as indicated in Figure 1B ) using a ‘sewing’ PCR method . GST-USP13 and GST-USP5 were made by cloning the coding regions of USP13 or USP5 to the SalI and NotI sites in the pET42c ( + ) vector ( Novagen/Merck , Germany ) . The pET42-Bag6 UBL and pET42-Ubl4A UBL constructs were described previously ( Xu et al . , 2012 ) . Constructs for expression of Erasin and gp78 shRNA were provided by M Monteiro , and S Fang , respectively ( University of Maryland , Baltimore , MD ) . Plasmids for expression of Bag6 and its truncated variants were generated by cloning the coding DNA fragments as indicated in Figure 4 into the EcoRV and NotI sites in the pRK5-FLAG vector . Plasmid expressing FLAG-tagged Ubl4A and FLAG-tagged SGTA were purchased from Origene Technologies ( Rockville , MD ) . To generate the gp78 RING mutant for mammalian expression , Cys356 and His361 in the gp78 RING domain were converted to Glycine and Alanine , respectively . The pGEX-gp78c Rm construct was described previously ( Li et al . , 2007 ) . The constructs for expression of the various GST-gp78c truncation mutants were generated by a PCR-based mutagenesis approach using the pGEX-gp78c as the template and the TagMaster Site-Directed Mutagenesis kit from GM Biosciences ( Frederick , MD ) . Other point mutations were generated by PCR based site-directed mutagenesis following a standard protocol . All mutations were confirmed by DNA sequencing . To construct pCMV- Ub-Ubl4A , ubiquitin-coding sequence bearing a G76V mutation was first cloned into pET28a using the NdeI and SalI sites . Ubl4A coding sequence was then inserted downstream of Ub-G76V using HindIII and XhoI sites . The whole Ub-G76V-Ubl4A fragment was then amplified by PCR and subcloned into pCMV6-ENTRY vector using the SgfI and MluI sites . For USP13 knockdown experiments , the targeting sequences are as follows:USP13-1 shRNA: cctgaatacttggtagtgcagataaagaaUSP13-2 shRNA: gcgcatgtttaaggcctttgt The knockdown constructs were purchased from Origene ( Rockville , MD ) . shRNA constructs for depletion of Bag6 and Ubl4A were described previously ( Wang et al . , 2011 ) . TransIT-293 ( Mirus ) was used for DNA plasmid transfection except for the gene knockdown experiments in which transfections were carried out using lipofectamine2000 ( Invitrogen ) . The HEK293 cell lines stably expressing Ubl4A-FLAG or Ubl4AK48R-FLAG were made according to the Invitrogen Fln-In T-Rex system protocol . Briefly , Ubl4A-FLAG or Ubl4AK48R-FLAG fragments were sub-cloned into the HindIII and BamHI sites in the pCDNAa5/FRT/TO vector . The construct , together with a help vector pOG44 , was then transfected into the Flp-In T-Rex expression cells . Positive cell clones resistant to hygromycin ( 125 µg/ml ) and Blasticidin ( 15 µg/ml ) were pooled together and maintained in the presence of a low dose of tetracycline ( 3 . 3 ng/ml ) to induce the expression of Ubl4A . gp78 knock out cells were generated using the CRISPR/Cas9 technology ( Wiedenheft et al . , 2012 ) . The vector pX330 ( Cong et al . , 2013 ) for inserting the guide sequence was purchased from Addgene . The following two oligoes ( Forward primer , 5′-CACCgcccagcctccgcacctaca; Reverse primer , 3′-GcgggtcggaggcgtggatgtCAAA-5′ ) were annealed and the resulting double strand DNA fragment was inserted into pX330 at the BbsI sites . The construct was then transfected into 293T cells following the standard protocol . 48 hr post-transfection , 50% of the cells were used to prepare genomic DNA for SURVOR assay to validate the cleavage of the target DNA . The remaining cells were cloned by infinite dilution . Immunoblotting was used to validate the positive knockout clones . USP13 knockout cells were generated by the same way . The two oligos are 5′-CACCgccgcccggcatgccgaaca and; 3′-GcggcgggccgtacggcttgtCAAA-5′ . The antibody for USP13 was raised against recombinant GST-USP13 protein . Rabbit polyclonal antibodies to Ubl4A , Trc35 , gp78 , Ube2g2 , Histone H2A , ATF4 , and GFP were described previously ( Li et al . , 2009; Wang et al . , 2009 , 2011 ) . The anti-SGTA antibody was raised against the following peptide ( CRSRTPSASNDDQQE-COOH ) and affinity-purified using the same peptide conjugated to Sulfolink coupling resin ( Thermo Scientific ) . Unless specified in the figure legends , Bag6 antibodies raised against a recombinant protein containing the N-terminal 249 amino acids were used for immunoblotting to detect the full length Bag6 and the cleaved Bag6 . Other antibodies used are FLAG ( M2 ) ( Sigma , St . Louis , MO ) , p97 ( Fitzgerald , Acton , MA ) , UBE1 ( Sigma ) , UbxD8 ( Proteintech Group , Chicago , IL ) , Myc ( Santa Cruz , Dallas , TX ) , Trc35 ( Novus Biologicals , Littleton , CO ) , ubiquitin ( P4D1 , Santa Cruz ) , Bag6 ( Creative BioMart , Shirley , NY ) , PDI ( 1D3 ) , Hsp90 ( AC88 ) ( Enzo Life Sciences , Farmingdale , NY ) , and MMS1 ( BIOMOL/Enzo , Farmingdale , NY ) . Trc35 antibody from Novus Biological was used for immunoprecipitation . HRP-linked or fluorescence-labeled secondary antibodies ( Rockland , Boyertown , PA ) were used for immunoblotting detection . GST-E1 , ubiquitin , ubiquitin-aldehyde , methylated ubiquitin , and ubiquitin K48R were purchased from Boston Biochem . MG132 was purchased from EMD Bioscience . EndoH was purchased from New England Biolab . The purification of Bag6 and the Bag6-Ubl4A-Trc35 complex were described previously ( Wang et al . , 2011 ) . The truncated Bag6 variants , SGTA , FLAG-USP13 and the USP13/5 chimeras were all purified using the same FLAG affinity chromatography procedure . Ube2g2 and GST-gp78c were purified from E . coli according to a previously described method ( Ye et al . , 2003 ) . GST-USP13 and GST-USP5 were purified from E . coli according to a published protocol ( Russell and Wilkinson , 2005 ) . Proteins eluted from glutathione column ( GE Healthscience , Piscataway , NY ) or Ni-NTA beads ( Qiagen , Valencia , CA ) were further fractionated on a Superdex 200 HR ( 10/30 ) column in a buffer containing 50 mM Tris-HCl , pH 8 . 0 , 150 mM potassium chloride , 5% glycerol , 1 mM DTT , and 2 mM magnesium chloride . Cell lysates were prepared using the NP40 lysis buffer ( 50 mM Tris–HCl pH 7 . 4 , 150 mM sodium chloride , 2 mM magnesium chloride , 0 . 5% NP40 , and a protease inhibitor cocktail ) . For most experiments , NP40 soluble fractions were used for immunoprecipitation or immunoblotting . Where indicated , the NP40-insoluble pellets were re-solubilized by the Laemmli buffer for immunoblotting . For immunoprecipitation under denaturing condition , cells were first lysed in a buffer containing 1%SDS and 5 mM DTT . The lysates were heated at 65°C for 15 min and then diluted 10-fold by the NP40 lysis buffer . The samples were subject to centrifugation at 20 , 000×g for 10 min to remove insoluble materials . The soluble supernatant fractions were used for immunoprecipitation by antibodies as indicated in the figure legends . For sequential immunoprecipitation , cell extract was first incubated with FLAG beads . The bound materials were eluted by a buffer containing FLAG peptide ( 0 . 2 mg/ml ) . The eluate was subject to another round of immunoprecipitation by anti-gp78 antibody or a control antibody . Immunoblotting was performed following the standard protocol . To measure the levels of YFP-tagged proteasomal substrates , cells were analyzed by the Cytomics FC 500MPL Flow Cytometry System ( Beckman Coulter , Brea , CA ) following a previously described procedure ( Wang et al . , 2011 ) . For the immunofluorescence imaging studies , cells transiently expressing YFP-tagged TCRα were grown in a Lab-Tek slide chamber ( Nalge Nunc International , Rochester , NY ) were fixed for 20 min in phosphate-buffered saline ( PBS ) containing 4% paraformaldehyde . Images were taken using an Axiovert 200 inverted microscope with a 63× oil objective lens ( Zeiss , Germany ) . For immunostaining experiments , fixed cells were permeabilized in PBS containing 0 . 1% NP40 and 5% normal donkey serum and then stained in the same buffer with the indicated antibodies . Ubiquitination experiments were performed as described previously ( Ye et al . , 2003 ) . Briefly , E1 ( 60 nM ) , Ube2g2 ( 200 nM ) , gp78c ( 400 nM ) , and ubiquitin ( 10 µM ) were incubated with the substrate Ubl4A at 37°C for 1 hr in a buffer containing 25 mM Tris-HCl , pH 7 . 4 , 2 mM magnesium/ATP , and 0 . 1 mM DTT . The DUB activities were measured by detecting the increase in fluorescence upon cleavage of Ubiquitin-AFC ( Boston Biochem , MA ) . Purified DUBs ( ∼50 nM ) were added individually to 200 μl deubiquitinating assay buffer ( 50 mM Tris-HCl pH 7 . 4 , 20 mM potassium chloride , 5 mM magnesium chloride , 1 mM DTT , including 0 . 5 μM Ub-AFC ) , and incubated at 37°C . The fluorescence intensity was measured using an Aminco Bowman Luminescence spectrometer with excitation and emission wavelengths set at 400 and 505 nm , respectively . In vitro deubiquitination of Ubl4A was performed as follows . USP13-depleted cells were transfected with HA-Ub and TCRα-YFP and lysed in the NP40 lysis buffer . The whole cell extract was incubated with protein A beads pre-bound with GFP or Ubl4A antibodies to purify TCRα and endogenous Ubl4A , respectively . After immunoprecipitation , the beads were washed twice with the NP40 washing buffer and once with 2× deubiquitination buffer . The beads were then divided into equal portions , each containing 20 µl 2× DUB buffer ( 100 mM Tris pH 8 . 0 , 40 mM KCl , 10 mM MgCl2 , 2 mM DTT ) . The reactions were started by adding recombinant FLAG-USP13 ( purified from mammalian cells ) or GST-USP13 ( purified from E . coli ) as indicated in the figure legends . The samples were incubated at 37°C for 2 hr with shaking and then stopped by addition of 40 µl Laemmli buffer . The samples were then analyzed by immunoblotting with anti-HA ( ubiquitin ) , anti-Ubl4A , or anti-GFP antibodies . In vivo crosslinking experiments were performed by incubating 2 × 106 cells in a labeling buffer ( 95 mM NaCl , 50 mM KCl , 10 mM glucose , 10 mM HEPES pH 7 . 2 , 1 . 2 mM CaCl2 , 1 . 2 mM MgCl2 ) containing the indicated concentration of 1 , 6-bis-Maleimidohexane ( BMH ) ( Pierce ) at room temperature for 20 min . The cells were then treated with 5 mM DTT for 10 min on ice , followed by addition of 1% SDS . Cell extracts were heated at 65°C for 15 min . The soluble extracts were diluted in the NP40 lysis buffer prior to immunoprecipitation using 2 . 5 µg affinity purified anti-SGTA antibody . In vitro crosslinking was performed similarly except that purified proteins were used . Cycloheximide chase experiments were performed by incubating cells in DMEM medium containing 50 µg/ml cycloheximide at 37°C . At the time points indicated in the figures , ∼0 . 6 × 106 cells were harvested . The cell extracts were analyzed by immunoblotting .	Cells make proteins inside a structure called the endoplasmic reticulum . However , some of these proteins cannot fold into the correct shape , so cells rely on a process called the ERAD pathway to degrade and eliminate these faulty proteins . First , however , the misfolded proteins must be moved from the endoplasmic reticulum to the main body of the cell ( the cytosol ) . The process by which the misfolded proteins are moved through the membrane that encloses the endoplasmic reticulum is complex , with ‘ERAD machinery proteins’ playing an important role . Among them , a series of enzymes called E3 ligases ‘tag’ the faulty proteins with a small protein called ubiquitin , and a complex called the proteasome then recognizes and degrades those proteins that have been tagged with ubiquitin . However , it is not clear why the E3 ligases that tag the misfolded proteins with ubiquitin don’t also tag the machinery proteins that from complexes with the faulty proteins . Now Liu et al . have used a combination of biochemical and genetic tools to shed light on this puzzle by studying the interaction of gp78—which is an E3 ligase—and USP13 , an enzyme that opposes the actions of the E3 ligases by removing ubiquitin . Liu et al . showed that gp78 can indeed tag certain machinery proteins with ubiquitin , which would stop the removal of misfolded proteins from the endoplasmic reticulum . However , USP13 opposed the action of gp78 , thus allowing the removal to continue . It has been known for some time that enzymes with opposing roles—the addition and removal of ubiquitin—can work together , but the biological significance of this phenomenon was not fully understood . The work of Liu et al . suggests that USP13 makes the elimination of misfolded proteins more efficient by ensuring that gp78 only tags those proteins that are misfolded: it does this by removing ubiquitin from proteins that should not have been tagged . A similar phenomenon is known to occur in genetics during DNA replication , with the enzyme complex that replicates the DNA including an enzyme that performs a proofreading role .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2014	USP13 antagonizes gp78 to maintain functionality of a chaperone in ER-associated degradation
Type 1 diabetes ( T1D ) is an autoimmune disease caused by loss of pancreatic β cells via apoptosis while neighboring α cells are preserved . Viral infections by coxsackieviruses ( CVB ) may contribute to trigger autoimmunity in T1D . Cellular permissiveness to viral infection is modulated by innate antiviral responses , which vary among different cell types . We presently describe that global gene expression is similar in cytokine-treated and virus-infected human islet cells , with up-regulation of gene networks involved in cell autonomous immune responses . Comparison between the responses of rat pancreatic α and β cells to infection by CVB5 and 4 indicate that α cells trigger a more efficient antiviral response than β cells , including higher basal and induced expression of STAT1-regulated genes , and are thus better able to clear viral infections than β cells . These differences may explain why pancreatic β cells , but not α cells , are targeted by an autoimmune response during T1D . Type 1 diabetes ( T1D ) is an autoimmune disease in which pancreatic β cells are targeted by a protracted attack by the immune system . This leads to death of most of the β cells while neighboring α cells survive . The disease is characterized by pancreatic islet inflammation ( insulitis ) and progressive β cell loss by apoptosis ( Ziegler and Nepom , 2010; Morgan et al . , 2014 ) . The triggering of T1D probably depends on environmental factors that interact with predisposing genes to induce autoimmunity against the β cells ( Eizirik et al . , 2009; Santin and Eizirik , 2013; Morgan and Richardson , 2014; Morgan et al . , 2014 ) . Among the potential environmental factors , epidemiological , clinical , and pathological studies in humans support the implication of viral infections , particularly enteroviruses ( e . g . , coxsackievirus; CVB ) , as triggers for the development of T1D ( Helfand et al . , 1995; Dotta et al . , 2007; Yeung et al . , 2011; Morgan and Richardson , 2014; Richardson et al . , 2014 ) . CVB-specific antibodies and enteroviral RNA are more frequently observed in serum samples from T1D patients than in healthy individuals ( Helfand et al . , 1995; Lonnrot et al . , 2000 ) , and staining of human pancreatic islets revealed that the enteroviral capsid protein VP1 is present at a higher frequency in insulin-containing islets from patients with recent-onset T1D when compared to healthy controls ( Dotta et al . , 2007; Richardson et al . , 2013 ) . A meta-analysis of 33 prevalence studies involving 1931 T1D cases and 2517 controls confirmed a clinically significant association between enteroviral infections and islet autoimmunity and T1D in humans ( Yeung et al . , 2011 ) . Additionally , it has been shown that the presence of enterovirus RNA ( Oikarinen et al . , 2011 ) or antibodies anti-CVB1 ( Laitinen et al . , 2014 ) in blood can predict the development of T1D . Insulitis is established and exacerbated in the context of a ‘dialog’ between pancreatic β cells and the immune system , regulated by the local production and release of chemokines and cytokines . These proteins attract and stimulate cells of the immune system , such as macrophages and cytotoxic T lymphocytes ( Eizirik et al . , 2009; Willcox et al . , 2009; Arif et al . , 2014 ) . The immune cells cause selective β cell destruction both directly and via the production of pro-inflammatory cytokines , such as interleukin-1β ( IL-1β ) , type II interferon ( IFNγ ) , and tumor necrosis factor α . These cytokines are released closely to the target cells and modulate the expression of complex gene networks in β cells , leading to the release of chemokines and eventually to the activation of the intrinsic pathway of cell death in β cells ( Eizirik et al . , 2009; Gurzov and Eizirik , 2011 ) . A key unanswered question in the field is why highly differentiated and specialized pancreatic β cells express these immune-related pathways . Host defense in vertebrates is usually viewed as the task of specialized immune cells . This perception underestimates the capacity of many non-immune cells to trigger ‘self-defense’ or ‘cell autonomous immune responses’ against infection . These mechanisms are pre-existing in many cell types and can be up-regulated upon virus infection ( Yan and Chen , 2012; Randow et al . , 2013 ) . This is an ancient form of cellular protection , present both in bacteria and metazoans . In vertebrates , for instance , cellular self-defense synergizes with innate and adaptive immunity to fight infections ( Randow et al . , 2013 ) . The cell susceptibility/resistance of highly differentiated and poorly proliferating cells , such as neurons and pancreatic β cells , to microbial infection is a major determinant of clinical outcome . It has been shown that cerebellum granule cell neurons and cortical neurons ( CNs ) have unique self-defense programs that confer differential resistance to infection by West Nile virus . Specifically , higher basal expression and faster up-regulation of IFN-induced genes improves the survival of granule cell neurons infected by West Nile virus ( Cho et al . , 2013 ) . These responses rely on detection of microbial signatures by pattern recognition receptors ( PRRs ) ( Randow et al . , 2013 ) . Several candidate genes for T1D expressed in human islets , such as the RIG-like receptor MDA5 ( Colli et al . , 2010 ) and the regulators of type I IFNs PTPN2 and USP18 ( Moore et al . , 2009; Colli et al . , 2010; Santin et al . , 2012 ) , modulate viral detection , antiviral activity , and innate immunity . The candidate genes described above ( Moore et al . , 2009; Colli et al . , 2010; Santin et al . , 2012 ) and CVB5 infection ( Colli et al . , 2011 ) regulate β cell apoptosis via activation of the BH3-only protein Bim . These observations support the concept that genetically modulated self-defense responses in β cells might play an important role in determining the outbreak of insulitis and the progression to T1D in face of viral infection or other stimuli ( Santin and Eizirik , 2013 ) . Against this background , we have presently evaluated the global gene expression of cytokine-treated and virus-infected human islet cells , observing that these two treatments lead to similar up-regulation of a large number of genes , gene networks , and transcription factors involved in cell autonomous immune responses . This conclusion generated two additional questions , namely whether this self-defense response is islet cell specific and , if yes , whether these putative cellular differences may explain the preferential β cell targeting by the autoimmune assault . To answer these questions , we next compared the responses of FACS-purified rat pancreatic α and β cells to infection by potentially diabetogenic CVB5 and CVB4 . The results obtained indicate that α cells trigger a more effective antiviral response than β cells , including higher basal and induced expression of STAT1-regulated genes , and are thus able to better clear viral infections as compared to β cells . We used previous microarray and RNA sequencing ( RNAseq ) analysis made by our group to compare the global gene expression of CVB5-infected human islets , evaluated by microarray analysis 48 hr after viral infection ( HV ) ( Ylipaasto et al . , 2005 ) , against the gene expression of human islets exposed to the pro-inflammatory cytokines IL-1β + IFNγ , evaluated either by microarray analysis at 24 , 36 , or 48 hr ( HC1 ) ( Lopes et al . , 2014 ) or by RNAseq at 48 hr ( HC2 ) ( Eizirik et al . , 2012 ) , focusing the analysis on over-expressed genes ( Figure 1 ) . Comparison of human islets exposed to cytokines and analyzed by either microarray or RNAseq showed a strong similarity in the top 20% ranked genes ( 50% common genes; Figure 1 ) . Comparison between CBV5-infected human islets against cytokine-treated human islets indicated a large number of common genes , in particular among the top 20% genes ( 30–50% common genes ) . Interestingly , the area under the curve ( AUC ) for a comparison between different batches of human islets exposed to cytokines and analyzed either by microarray or RNAseq analysis was 0 . 209 ( subtracted by a null area of 0 . 5 ) , while the AUC for the comparisons virus vs cytokines ( microarray vs microarray or microarray vs RNAseq ) was , respectively , 0 . 154 and 0 . 127 , that is , 74% and 61% of the cytokines vs cytokines comparison , indicating a close similarity between human islet cell responses to virus or cytokines . To exclude that these similarities were the result of non-specific cell stress responses , we compared the viral-induced gene expression ( Ylipaasto et al . , 2005 ) against genes modified by palmitate ( HP ) ( Cnop et al . , 2014 ) , a metabolic stress unrelated to the immune response . There was limited similarity between virus- and palmitate-induced genes , with a curve close to random ( Figure 1 ) and an AUC of 0 . 027 , that is , <20% of the area observed when comparing virus- against cytokine-induced genes . 10 . 7554/eLife . 06990 . 003Figure 1 . Ranking similarity between gene expression of human islets after cytokine exposure ( HC1 and HC2 ) or after virus exposure ( HV ) . The similarity between HC1 and HC2 and between HV and palmitate exposure ( HP ) is also presented . The area under the curve ( subtracted by a null threshold of 0 . 5 ) is indicated , as well as similarity curves corresponding to a p-value of 0 . 001 . The figure represents the ranking similarity ordered by up-regulation . ( For a detailed explanation of the calculations done , see ‘Materials and methods’ , ‘gene expression ranking similarity’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 00310 . 7554/eLife . 06990 . 004Figure 1—figure supplement 1 . Venn diagram of the up-regulated genes in human islets exposed to cytokines ( HC1 and HC2 ) or infected with CVB5 ( HV ) . In each data set , a group of up-regulated genes was identified . The criterion for differential expression was p-value < 0 . 05 ( paired t-test; BH-adjusted in the HC1 case ) . Up-regulation was identified through mean fold change . 894 genes were identified to be up-regulated in HC1 , 1100 in HC2 , and 954 in HV . The number of genes in each separated area is indicated . The intersection between HC1 and HV has a p-value of 2 . 46e-151 , between HC2 and HV 2 . 78e-180 , and between HC1 and HC2 approximately e-897 ( hypergeometric distribution , 9504 genes in total ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 00410 . 7554/eLife . 06990 . 005Figure 1—figure supplement 2 . IPA analysis of the up-regulated genes in human islets exposed to cytokines HC1 or infected with CVB5 ( HV ) . The up-regulated genes in HC1 and HV ( 348 in total ) were analyzed in QIAGEN's Ingenuity Pathway Analysis ( IPA ) to identify enriched pathways in both cytokine-treated and virus-infected human islets . The top 30 pathways ( ordered by p-value ) are shown . Enrichment p-value ( Fisher exact test ) and ratio to the number of pathway elements are also indicated . QIAGEN's Ingenuity Pathway Analysis ( IPA , QIAGEN , Redwood City , www . qiagen . com/ingenuity ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 00510 . 7554/eLife . 06990 . 006Figure 1—figure supplement 3 . IPA analysis of the up-regulated genes in human islets exposed to cytokines HC2 or infected with CVB5 ( HV ) . The up-regulated genes in HC2 and HV ( 413 in total ) were analyzed in QIAGEN's IPA to identify enriched pathways in both cytokine-treated and virus-infected human islets . The top 30 pathways ( ordered by p-value ) are shown . Enrichment p-value ( Fisher exact test ) and ratio to the number of pathway elements are also indicated . QIAGEN's Ingenuity Pathway Analysis ( IPA , QIAGEN , Redwood City , www . qiagen . com/ingenuity ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 006 The superposition of up-regulated genes between virus-infected and cytokine-treated human islets confirmed the similarity between virus- and cytokine-induced genes , with >40% of the genes up-regulated in common between both islet treatments ( Figure 1—figure supplement 1 ) . By further comparing the RNAseq data obtained in human islets ( Eizirik et al . , 2012; Lopes et al . , 2014 ) against known PRRs and other antiviral/antibacterial factors ( Randow et al . , 2013; Carty et al . , 2014 ) , we noticed that cytokine-exposed human islets highly express and up-regulate antiviral ( e . g . , TLR3 , MDA5 , RIG-1 , APOBEC36 , SAMHD1 , TRIM22 , CNP , Tetherin , Viperin etc ) factors , while antibacterial factors ( e . g . , TLR4 , NLRP1 , CLEC6A , CLEC7A . ) are lowly or not expressed , suggesting that these cells are under evolutionary pressure to counteract viral but not bacterial infections , probably because they are seldom confronted by bacteria . Pathway analysis of genes induced in common by cytokines and viral infection ( Figure 1—figure supplement 2 and Figure 1—figure supplement 3; for clarity only the top 30 correlations are shown ) indicated presence of large groups of genes involved in interferon signaling ( top correlation for both cytokine data sets compared to the virus data set ) , activation of IRFs ( family of interferon regulatory transcription factors ) , PRRs , T1D signaling , role of PKR in IFN induction , Jak/Stat signaling ( activated downstream of IFNs and STATs ) , and so on . These observations indicate that both a viral infection by CVB5 and exposure to pro-inflammatory cytokines trigger a cell autonomous immune response in human islet cells . The experiments described above were performed with whole-human islets , composed to a large extent by β and α cells . Since it has been described that different brain cells present diverse innate immune response programs to viral infection ( Cho et al . , 2013 ) , and since β but not α cells are killed in T1D , we next examined whether pancreatic β and α cells present different susceptibility to viral infections or to pro-inflammatory cytokines . It is presently not technically feasible to FACS-purify human β and α cells for long-term in vitro experiments due to the high-background fluorescence of human β cells caused by marked lipofuscin accumulation ( Cnop et al . , 2010 ) and the putative impact of antibody-mediated techniques on the long-term survival and function of the sorted cells . These follow-up experiments were thus performed in FACS-purified rat β and α cells ( >90% pure cells and with >90% viability after 4 days in culture; there was also similar expression of the housekeeping gene GAPDH between α and β cells under different experimental conditions; Figure 2 ) . Both β and α cells were killed to a same extent after a 24 hr exposure to different concentrations of IL-1β + IFNγ ( Figure 3A and Figure 3—figure supplement 1 ) , indicating that under the present experimental conditions FACS-purified rat α and β cells are similarly affected by cytokine-induced apoptosis . One of the mechanisms involved in cytokine-induced cell death is NO production ( Eizirik and Mandrup-Poulsen , 2001 ) , and β and α cells showed roughly similar iNOS expression and medium nitrite accumulation , with slightly higher nitrite production by β cells , but basal iNOS expression was higher in α than β cells ( Figure 3—figure supplement 2A , B ) . Similarly , cytokines induced expression of the chemokines CXCL10 and CCL2 in both cell types , but CCL2 expression was higher in α cells basally and following cytokine-exposure ( Figure 3—figure supplement 2C , D ) . In line with the cytokine data , the synthetic dsRNA polyinosinic-polycitidilic acid ( PIC ) induced a similar percentage of cell death in β and α cells ( Figure 3B ) , but a slightly higher expression of the downstream genes IFNα ( Figure 3—figure supplement 3B ) and CCL2 ( Figure 3—figure supplement 3F ) in α compared to β cells . Furthermore , basal expression of the transcription factor STAT1 , iNOS and of the chemokines CXCL10 and CCL2 was also higher in α than in β cells ( Figure 3—figure supplement 3C , F ) . 10 . 7554/eLife . 06990 . 007Figure 2 . Purity , viability , and GAPDH mRNA expression of the β- and α-cell fractions after single-step FACS purification . ( A ) Immunostaining for insulin or glucagon of the rat islet cell preparations used in this study . Percentage of insulin- and glucagon-positive cells in the β and α cell preparations . ( B ) Cell viability was evaluated by staining the β and α cells with the nuclear dyes Hoechst 33342 and propidium iodide after 4 days in culture . Results are plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points of 32 independent preparations . ( C ) GAPDH values were measured by RT-PCR and compared with a standard curve . Results are plotted as scatter dot plot of each single measurement ( n = 70 for each cell type ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 00710 . 7554/eLife . 06990 . 008Figure 3 . Pancreatic α cells are more resistant than β cells against CVB5- but not against cytokine- or PIC-induced cell death . FACS-purified rat β and α cells ( >90% purity for both cell types ) were treated with interleukin-1β ( IL-1β ) + type II interferon ( IFNγ ) ( 50 or 500 U/ml , respectively ) ( A ) or PIC ( 1 μg/ml ) for 24 hr or infected with CVB5 ( multiplicity of infection—M . O . I . 5 ) for 36 hr ( C–G ) . ( A–C ) Apoptosis was evaluated by staining with the nuclear dies Hoechst 33342 and PI . ( D ) VP1 mRNA expression was assayed by RT-PCR and normalized by the housekeeping gene GAPDH . ( E and F ) The figures show representative Western blots of VP1 protein expression after CVB5 infection and α-tubulin for loading control . ( F ) The Western blot ( E ) was overexposed to allow visualization of VP1 expression in α cells . ( G ) Titration of the supernatants from β and α cells infected with CVB5 for 36 hr . Results of 4–6 experiments are plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points; p < 0 . 05 , p < 0 . 01 , and p < 0 . 001 treated vs untreated ( A and B ) or CVB5 vs mock infection ( C , D , and G ) ; ###p < 0 . 001 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 00810 . 7554/eLife . 06990 . 009Figure 3—figure supplement 1 . Dose-response of cytokine-induced apoptosis in pancreatic α and β cells . FACS-purified rat β and α cells ( >90% purity for both cell types ) were treated with different concentrations of IL-1β + IFNγ , as indicated in the figure , for 24 hr . Apoptosis was evaluated by staining with the nuclear dies Hoechst 33342 and PI . Results are mean and SEM of 3 independent experiments; p < 0 . 01 and **p < 0 . 001 treated vs untreated; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 00910 . 7554/eLife . 06990 . 010Figure 3—figure supplement 2 . Pancreatic rat α and β cells have similar NO production , cytokine , and chemokine expression following exposure to cytokines . FACS-purified β and α cells ( >90% purity for both cell types ) were treated with IL-1β + IFNγ ( 50 or 500 U/ml , respectively ) ( A–D ) . iNOS ( A ) , CXCL10 ( C ) , and CCL2 ( D ) mRNA expression were assayed by RT-PCR and normalized by the housekeeping gene GAPDH . ( B ) Cytokine-induced NO production by primary rat β or α cells was evaluated by medium nitrite accumulation . Results of 4–5 experiments are plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points; p < 0 . 05 , p < 0 . 01 , and p < 0 . 001 treated vs untreated; #p < 0 . 05 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 01010 . 7554/eLife . 06990 . 011Figure 3—figure supplement 3 . Pancreatic rat α and β cells have similar cytokine and chemokine expression following exposure to PIC . FACS-purified β and α cells ( >90% purity for both cell types ) were treated with intracellular PIC ( 1 μg/ml ) for 24 hr ( A–F ) . IFNβ ( A ) , IFNα ( B ) , STAT1 ( C ) , iNOS ( D ) , CXCL10 ( E ) , and CCL2 ( F ) mRNA expression were assayed by RT-PCR and normalized by the housekeeping gene GAPDH . Results of 4–5 experiments are plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points; p < 0 . 05 , p < 0 . 01 , and p < 0 . 001 treated vs untreated; #p < 0 . 05 and ##p < 0 . 01 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 01110 . 7554/eLife . 06990 . 012Figure 3—figure supplement 4 . Prolonged time-course of CVB5-induced apoptosis in pancreatic α and β cells . FACS-purified rat β and α cells ( >90% purity for both cell types ) were infected with CVB5 ( multiplicity of infection—M . O . I . 5 ) for 48 , 72 , or 96 hr . Apoptosis was evaluated by staining with the nuclear dies Hoechst 33342 and PI . Results are mean and SEM of 3 independent experiments; *p < 0 . 001 CVB5 vs mock infection; ###p < 0 . 001 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction; h . p . i . , hours post-infection . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 01210 . 7554/eLife . 06990 . 013Figure 3—figure supplement 5 . UV-inactivated CVB5 does not induce cell death in pancreatic α and β cells . FACS-purified rat β and α cells ( >90% purity for both cell types ) were infected with CVB5 ( multiplicity of infection—M . O . I . 5 ) or UV-inactivated CVB5 ( 1000 J/m2 ) for 36 hr . Apoptosis was evaluated by staining with the nuclear dies Hoechst 33342 and PI . Results are mean and SEM of 3 independent experiments; #p < 0 . 05 CVB5 vs . UV-inactivated CVB5 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 01310 . 7554/eLife . 06990 . 014Figure 3—figure supplement 6 . Cell counting after CVB5 infection of pancreatic α cells . FACS-purified rat α cells ( >90% purity ) were infected with CVB5 ( M . O . I . 5 ) for 36 hr . Supernatants were collected and the cells attached to the well trypsinized and also collected for quantification . The counting of supernatant cells ( A ) or attached cells ( B ) was performed in Neubauer chambers , and each point was measured in triplicate by two observers , one of them unaware of sample identity . Results are mean and SEM of 4 independent experiments; Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 01410 . 7554/eLife . 06990 . 015Figure 3—figure supplement 7 . Pancreatic α cells infected with CVB5 under different medium conditions . FACS-purified rat α cells ( >90% purity ) were infected in parallel with CVB5 ( multiplicity of infection—M . O . I . 5 ) for 36 hr in two different media , that is , medium used for β cell culture , with 10 mM glucose and 5% fetal bovine serum , ( gray ) and the usual medium used in α cells culture ( white ) . Apoptosis was evaluated by staining with the nuclear dies Hoechst 33342 and PI . Results are mean and SEM of 3 independent experiments; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 015 In contrast with the observations made with cytokines and dsRNA ( see above ) , infection of these cells with CVB5 at a multiplicity of infection ( M . O . I . ) 5 induced death of nearly 50% of β cells after 36 hr , but it did not increase α cell death at this time point ( Figure 3C ) or after a more prolonged follow-up ( up to 96 hr ) post-infection ( Figure 3—figure supplement 4 ) . UV-inactivated virus did not kill β cells ( Figure 3—figures supplement 5 ) , indicating that cell death is the consequence of actual viral infection and proliferation . Additionally , there were no differences in the number of α cells attached to the well before and after the infection ( Figure 3—figure supplement 6 ) , supporting the assumption that the α cells remain alive after CVB5 infection . α cells resistance was independent of the glucose/fetal bovine serum ( FBS ) concentration used in the medium ( Figure 3—figure supplement 7 ) , and it was paralleled by a markedly lower expression of the CVB5 marker VP1 in α than in β cells , as evaluated by mRNA ( Figure 3D ) and protein ( Figure 3E , F ) expression , and by titration of the production of infective virus ( Figure 3G ) . Dose-response experiments in α cells indicated that even at a M . O . I . of 100 , CVB5 still induced markedly less cell death in α than in β cells infected at an M . O . I . 20-fold lower , that is , M . O . I . 5 ( Figure 4A ) . This is not a phenomenon restricted to CVB5 , since α cells were also much more resistant than β cells to CVB4-induced cell death ( Figure 4B ) . This cannot be explained by different expression of receptors for the virus , since the CVB receptors coxsackievirus and adenovirus receptor ( CAR ) and decay accelerating factor for complement have similar or higher mRNA expression in rat α as compared to β cells ( Figure 4C , D ) . The similar expression of CAR was confirmed at the protein level in α and β cells ( Figure 4—figures supplement 1 ) . Consistently , the same quantity of CVB5 remained associated with α and β cells after adsorption of the virus as measured by titration 2 hr after virus exposure ( data not shown ) . In line with these findings , the percentage of infected α and β cells by a non-replicating adenoviral vector encoding GFP , which also enters the cells via CAR , was similar in α and β cells ( Figure 5A–C ) . On the other hand , the intensity of GFP fluorescence was several-fold lower in α than in β cells ( Figure 5A , B , D ) , indicating that the virus enters α cells but cannot properly translate its cargo protein . These observations may explain a common knowledge in the field that α cells are ‘difficult to transduce’ with adenoviral vectors . 10 . 7554/eLife . 06990 . 016Figure 4 . The higher susceptibility of β cells to virus-induced cell death , as compared to α cells , is not due to higher expression of virus receptors . FACS-purified rat α and β cells ( >90% purity ) were infected with CVB5 ( M . O . I . 5 , 50 or 100 for α cells; M . O . I . 5 for β cells ) ( A ) , CVB4 ( M . O . I . 5 ) ( B ) , or CVB5 ( M . O . I . 5 ) ( C and D ) for 36 hr . ( A and B ) Apoptosis was evaluated by staining with the nuclear dies Hoechst 33342 and PI . Coxsackievirus and adenovirus receptor ( CAR ) ( C ) and DAF ( D ) mRNA expression were assayed by RT-PCR and normalized by the housekeeping gene GAPDH . Results are from 4–8 experiments , plotted as box plots indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points; p < 0 . 05 and **p < 0 . 001 CVB5 or CVB4 vs mock infection; #p < 0 . 05 and ###p < 0 . 001 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction . DAF , decay accelerating factor . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 01610 . 7554/eLife . 06990 . 017Figure 4—figure supplement 1 . CAR protein expression in pancreatic rat β and α cells . FACS-purified β and α cells ( >90% purity for both cell types ) were collected after 3 days in culture . Basal expression of CAR protein level was assayed by Western blot ( 2 out of 4 similar blots are shown ) ( A ) ; Densitometry quantification of 4 independent samples of each cell type is shown in ( B ) . Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 01710 . 7554/eLife . 06990 . 018Figure 5 . β and α cells are infected with similar efficiency by an adenoviral vector encoding GFP , but the translation of GFP protein is lower in α cells . ( A–D ) FACS-purified β and α cells ( >90% purity for both cell types ) were infected with adeno-GFP ( M . O . I . 1 , 5 or 10 ) for 48 hr . Presence of GFP protein was evaluated by fluorescence microscopy ( A ) and flow cytometry ( B–D ) . ( A ) Pictures show nucleus ( blue ) and GFP fluorescence ( green ) . β and α cells were infected with adeno-GFP M . O . I . 1 or 5 ( B and D ) for 48 hr and then sorted based on green fluorescence and forward-scattered light . ( B ) Representative 2-D plot of 4 independent experiments . ( C ) Quantification of GFP positive cells . ( D ) Average of green fluorescence intensity in cells infected at M . O . I . 5 . Results of 4 experiments are plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points; p < 0 . 05 , p < 0 . 01 and p < 0 . 001 adeno-GFP vs mock infection; #p < 0 . 05 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 018 We next compared the expression of known components of the cell-autonomous immune response in α and β cells under basal conditions ( Figures 6 , 7 ) and following infection with CVB5 ( Figure 8 ) . The cell preparations used were highly pure ( >90%; Figure 2A ) , as confirmed by the 21-fold higher expression of glucagon ( GCG ) and the α cell transcription factor ARX in α cells as compared to β cells , and the lower expression of the β cell markers insulin ( decrease of 185-fold ) and PDX-1 ( decrease of 12-fold ) in these cells as compared to β cells ( Figure 6A ) . In comparison to β cells , α cells have higher basal expression of the chemokines CXCL10 and CCL2 , of the cytokines IFNα and IL-1β ( Figure 6B ) , of the transcription factor STAT1 ( which mediates IFN signal transduction ) and of several IFN-related downstream genes previously described to have a role in cell autonomous immune response in the brain ( Cho et al . , 2013 ) , including Viperin , Tetherin , PKR , Mx1 , USP18 , Oas1 , and so on ( Figure 6C ) . Several of these genes have higher expression in granule cell neurons of the cerebellum , as compared to CNs , explaining why these cells are more resistant to infection by positive-stranded RNA viruses ( Cho et al . , 2013 ) . There is remarkable similarity between gene expression in neurons and β cells ( Atouf et al . , 1997; Villate et al . , 2014 ) , and we next used available microarray and RNAseq data of , respectively , mouse granule neurons compared to CNs ( Cho et al . , 2013 ) and mouse α cells compared to β cells ( Benner et al . , 2014 ) to determine whether cells that have higher resistance to viral infection ( i . e . , granule neurons and α cells ) have also increased basal expression of similar genes of the cell autonomous immune responses . The data shown in Figure 7 and Supplementary file 1 indicate a clear similarity between granule neurons and α cells for the genes present in both data sets; indeed , 64% of these genes were similarly increased in both cell types , with only 4% showing opposite direction of expression . 10 . 7554/eLife . 06990 . 019Figure 6 . Basal expression of cell-autonomous immune response genes is higher in α cells than in β cells . mRNA expression of genes related to identity of β and α cells ( A ) or cell-autonomous immune response ( B and C ) was assayed by RT-PCR and normalized by the housekeeping gene GAPDH . Graphs represent relative expression of mRNAs in α cells vs β cells ( dotted line indicates 1 , i . e . , no change ) . Results from 4–9 experiments are plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points; p < 0 . 05 , p < 0 . 01 , and p < 0 . 001 α cells vs β cells; Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 01910 . 7554/eLife . 06990 . 020Figure 7 . Similarity between up-regulated genes in granule neurons and pancreatic α cells . ( A ) Correlation between up-regulated genes in brain ( cerebellum granule cell neurons ( CGNs ) vs cortical neurons [CNs] ) and islet cells ( alpha vs beta ) . ( B ) Venn diagram of the up-regulated genes in brain ( CGN vs CN ) and islets cells ( alpha vs beta ) . The absolute values are shown in Supplementary file 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 02010 . 7554/eLife . 06990 . 021Figure 8 . Differential expression of virus recognition and antiviral response genes in β and α cells exposed to CVB5 . ( A–I ) FACS-purified β and α cells ( >90% purity for both cell types ) were infected with CVB5 ( M . O . I . 5 ) for 36 hr . mRNA expression of genes related to virus recognition ( A–C ) and antiviral responses ( D–I ) was assayed by RT-PCR and normalized by the housekeeping gene GAPDH . Results for 4–8 experiments are plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points; p < 0 . 05 and **p < 0 . 001 CVB5 vs mock infection; #p < 0 . 05 , ##p < 0 . 01 , and ###p < 0 . 001 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 021 Following a 36 hr infection of α and β cells with CVB5 ( Figure 8 ) , α cells presented higher levels of the viral sensors MDA5 and PKR , of iNOS and CCL2 , but lower induction of Viperin . A time-course analysis of α and β cells infected with CVB5 showed a progressive increase in the viral capsid protein VP1 in β cells ( Figure 9A ) , while there was an early increase in VP1 expression in α cells ( see inset in Figure 9A ) , with peak at 8 hr and subsequent decrease by 24 hr , suggesting that these cells are indeed infected by CVB5 but manage to eradicate the virus without dying . In line with this , α cells have both a higher basal expression and induction of two mRNAs encoding key antiviral proteins , namely STAT1 ( Figure 9B ) and MX1 ( Figure 9C ) as compared to β cells . 10 . 7554/eLife . 06990 . 022Figure 9 . Time-course analysis of gene expression in β and α cells infected with CVB5 . ( A–C ) FACS-purified β ( squares and solid lines ) and α cells ( circles and dotted lines ) were infected with CVB5 ( M . O . I . 5 ) for 1 , 2 , 4 , 6 , 8 , 24 hr . VP1 ( A ) , STAT1 ( B ) , and MX1 ( C ) mRNA expression were assayed by RT-PCR and normalized by the housekeeping gene GAPDH . Inset in 6A show details of VP1 expression in α cells . Results are mean values ± SEM of 3–4 independent experiments; p < 0 . 05 , p < 0 . 01 , and p < 0 . 001 α vs β; ANOVA followed by Student's t-test with Bonferroni correction . Inset; p < 0 . 01 and *p < 0 . 001 CVB5 vs mock infection; One-way ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 022 Additional experiments confirmed higher basal STAT1 mRNA ( Figure 10A ) and protein ( Figure 10B , C ) expression in α cells as compared to β cells . Knockdown of STAT1 in α cells by a previously validated ( Moore et al . , 2011 ) siRNA ( Figure 10D ) prevented MX1 up-regulation in response to CVB5 infection ( Figure 10E ) and enabled a more intense CVB5 infection , as indicated by increased VP1 expression ( Figure 10F ) . These observations confirm that STAT1 plays a key role in α cell resistance to viral infection . 10 . 7554/eLife . 06990 . 023Figure 10 . Knockdown of STAT1 decreases MX1 and increases VP1 expression in pancreatic rat α cells after CVB5 infection . ( A–C ) FACS-purified β and α cells ( >90% purity for both cell types ) were collected after 3 days in culture . Basal expression of STAT1 mRNA level ( A ) was assayed by RT-PCR and normalized by the housekeeping gene GAPDH . STAT1 protein expression was measured by Western blot ( B and C ) . Densitometry quantification of 4 independent samples of each cell type is shown in ( C ) . ( D–F ) FACS-purified α cells ( >90% purity ) were transfected with siCTRL or siSTAT1 ( D–F ) . After 48 hr of recovery , cells were infected with CVB5 ( M . O . I . 5 ) for 36 hr . STAT1 ( D ) , MX1 ( E ) , or VP1 ( F ) mRNA expression was assayed by RT-PCR and normalized by the housekeeping gene GAPDH . Results from 4–6 experiments are plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points; p < 0 . 05 and **p < 0 . 01 and CVB5 vs mock infection; #p < 0 . 05 and ###p < 0 . 001 as indicated by bars; ANOVA followed by Student's t-test with Bonferroni correction . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 023 It has been previously shown that human β cells are more sensitive than α cells to CVB-induced infection and functional impairment both in vivo and in vitro ( Dotta et al . , 2007; Richardson et al . , 2009; Anagandula et al . , 2014 ) , leading to the suggestion that CVB does not infect human α cells during the progression of T1D . Our present findings suggest an alternative hypothesis , namely that α cells may become infected but are able to eradicate the virus more effectively than β cells ( see Figure 9 ) . If this hypothesis is correct , CVB infection should not be detectable in α cells of T1D patients when their islets are examined months/years after the putative initial infection . To test whether CVB may indeed infect human α cells , we first infected dispersed human islets from two donors ( Figure 11—figure supplement 1 ) in vitro with CVB5 ( M . O . I . 10 ) for 8 hr , the time point when maximum expression of VP1 was observed in a time-course experiment on α cells ( Figure 9 ) . We detected by immunofluorescence the presence of the enteroviral capsid protein VP1 in both α cells ( glucagon-positive ) and β cells ( insulin-positive ) ( Figure 11 ) . Thus , 52% ( human islet sample 1 ) and 33% ( human islet sample 2 ) insulin-positive cells were also positive for VP1 , while 28% ( human islet sample 1 ) or 27% ( human islet sample 2 ) of cells were double positive for glucagon and VP1 ( Figure 11 ) . Similar results were observed with CVB4 ( data not shown ) . To confirm these results in clinically relevant samples , we examined the pancreas of three children who died from myocarditis during the course of an acute and severe CVB infection . In these samples , we detected the presence of the enteroviral capsid protein , VP1 , in β cells in all three cases and in α cells in two of three cases ( Figure 12 and Figure 12—figure supplement 1 ) . In these latter patients , the number of infected α cells was clearly lower than the number of infected β cells . 10 . 7554/eLife . 06990 . 024Figure 11 . Infection of both α and β cells in dispersed human islets exposed to high titers of CVB5 for 8 hr . Dispersed human islets were mock infected or infected with CVB5 ( M . O . I . 10 ) for 8 hr . After infection , cells were fixed and used for histological studies . Fluorescent microscopy analysis of insulin ( A and C , in green ) , glucagon ( B and D , in cyan ) , and VP-1 ( E–H , in red ) shows the presence of double-positive cells for insulin and VP-1 ( O , merged panels , in yellow ) and glucagon and VP-1 ( P , merged panels , in yellow/white ) after CVB5 infection . No VP-1 positive cells ( E and F ) were observed in mock-infected cells . Nuclear staining was performed with Hoechst ( I–L , in blue ) . Double-positive cells for insulin and VP-1 and for glucagon and VP-1 are indicated by the arrows ( C , D , G , H , K , L , O , and P panels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 02410 . 7554/eLife . 06990 . 025Figure 11—figure supplement 1 . Characteristics of the 2 human donors used in the present study . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 02510 . 7554/eLife . 06990 . 026Figure 12 . Fluorescence photomicrographs of an islet from a neonate with an acute coxsackievirus infection . Viral VP1 ( green; A , D ) co-localizes with glucagon ( red; B , D ) in certain cells ( orange arrows ) and with insulin ( light blue; C , D ) in another ( white arrow ) . Nuclei were stained with DAPI ( dark blue ) in the merged image ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 02610 . 7554/eLife . 06990 . 027Figure 12—figure supplement 1 . List of human samples used . Cases were randomly selected from a previously described collection ( Richardson et al . , 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06990 . 027 A long-term puzzle in the pancreatic islet field has been why the highly specialized pancreatic β cells express a large number of immune-related genes upon exposure to pro-inflammatory cytokines ( Cardozo et al . , 2001; Eizirik et al . , 2009 , 2012; Ortis et al . , 2010; Lopes et al . , 2014 ) . From an evolutionary point of view , it would not make sense that these glucose-sensing and insulin-producing cells express these complex gene networks with the sole purpose to commit suicide during insulitis . The present observations , indicating a close similarity between cytokine- and virus-induced gene expression , suggest that these gene networks are actually part of a complex cell autonomous immune response , regulated at least in part by candidate genes for T1D ( Santin and Eizirik , 2013 ) , aiming to eradicate putative viral infections without excessive cell loss . Indeed , β cells , like neurons , have a very limited replication potential ( Cnop et al . , 2010 ) , and an excessive loss of β cells would be disastrous for the host . In some genetically susceptible individuals , however , the putative initial viral infection might not be resolved leading to a persistent low-level infection , which in these individuals could trigger a specific autoimmune assault against β cells ( Dotta et al . , 2007; Santin and Eizirik , 2013; Richardson et al . , 2014 ) . Pancreatic α and β cells are neighboring endocrine cells with a common embryonic origin ( Teitelman , 1993 ) . If an islet viral infection , which in theory should affect all islet cells , contributes to trigger T1D , an important question arises; namely , why are β cells killed while α cells survive ? It has been previously suggested that α cells are more resistant to cytokine-induced apoptosis than β cells ( Mandrup-Poulsen et al . , 1987 ) . The present observations , however , based on highly purified and viable rat α and β cells , cultured under similar conditions , indicate that both cell types are killed roughly to the same extent by the pro-inflammatory cytokines IL-1β + IFNγ . α and β cells are also equally susceptible to dsRNA-induced apoptosis , but α cells are several-fold more resistant to CVB-induced infection and consequent cell death than β cells . This suggests that the main difference between these two cell types is not their actual ability to detect signals of the viral infection ( e . g . , dsRNA ) but how they respond to the virus once it is actively translating proteins inside the cells . In line with this hypothesis , both α and β cells were infected with an adenoviral vector encoding GFP , but α cells efficiently blocked ( by 90% ) the translation of virally-encoded GFP . A limitation of the present study is that the experiments described above were performed in FACS-purifed rat α and β cells , which limits extrapolation to the human disease . Analysis of the expression of cell autonomous immunity-related genes in α and β cells ( present data ) and comparison between global expression of these genes in virus-resistant mouse granule cell neurons ( Cho et al . , 2013 ) and α cells indicate that a large number of IFN- and STAT1-dependent genes are expressed at higher level in α cells as compared to β cells . This suggests that this IFN-STAT1 gene network contributes to the resistance of α cells to viral infection . In support of this hypothesis , KD of STAT1 prevents CVB-induced MX1 expression in α cells and enables a more active infection with CVB5 , as evaluated by higher expression of the enteroviral capsid protein VP1 . Of note , unphosphorylated STAT1 supports the induction of antiviral genes in other cell types even without IFN-dependent stimulation ( Yang and Stark , 2008; Cheon and Stark , 2009 ) and increased basal expression of STAT1 amplifies cell-intrinsic immune responses ( Hu et al . , 2008; Amit et al . , 2009 ) . Previous histological studies in pancreas from T1D individuals have shown that CVB is detected in β but not α cells ( Ylipaasto et al . , 2004; Richardson et al . , 2009 , 2013; Anagandula et al . , 2014; Morgan et al . , 2014 ) , leading to the suggestion that CVB does not infect human α cells . Our present findings , obtained in highly purified α cells in vitro and by the analysis of dispersed human islets and pancreas from children with an acute and severe CVB infection , suggest an alternative hypothesis , namely that α cells indeed get infected but they rapidly eradicate the virus , probably due to their enhanced cell autonomous immunity responses . Our proposal that a strong cell immune response protects α cells against viral infection and subsequent death in T1D seems at odds with the association of polymorphic variants of MDA5 that reduce helicase activity with a lower risk to develop T1D ( Nejentsev et al . , 2009; Winkler et al . , 2011; Lincez et al . , 2015 ) . The long-lasting detection of enteroviral protein expression and IFN-induced markers in β cells in pre-diabetic or diabetic donors ( Dotta et al . , 2007; Richardson et al . , 2013 , 2014 ) ; however , most probably reflects a chronic non-cytolytic infection . In this context , the continuous stimulation of vigorous antiviral responses , in an attempt to control the infection , may lead to protracted presentation of β cell autoantigens , local release of chemokines and cytokines and eventually autoimmunity ( Santin and Eizirik , 2013 ) . This process , however , would not take place in α cells , which , as presently shown , have in place adequate mechanisms to swiftly eliminate the virus in the early stages of infection . As mentioned above , pancreatic α and β cells have a common embryonic origin and are localized in the same microorgan ( Teitelman , 1993 ) . They thus provide a unique model to understand how molecular regulation of self-defense against viral infection in β cells , as compared to α cells , may determine cell death , local inflammation , and eventual diabetes . The present observations suggest that pancreatic α and β cells have different cell autonomous signatures . This may explain their different ability to clear viral infections and potentially explain why putative chronically infected pancreatic β cells , but not α cells , are targeted by an autoimmune response and killed during T1D . Gene expression after cytokine exposure was analyzed based on two data sets . The first ( HC1 ) consists of 10 samples of human islets evaluated by RNAseq at 48 hr of IL-β + IFNγ exposure ( Eizirik et al . , 2012 ) . The second ( HC2 ) consists of 9 samples of human islets evaluated by microarray analysis at 24 hr , 36 hr , and 48 hr of IL-β + IFNγ exposure ( Lopes et al . , 2014 ) . Gene expression after coxsackievirus exposure ( HV ) was evaluated in 3 samples of human islets evaluated by microarray analysis at 48 hr of CVB5 infection ( Ylipaasto et al . , 2005 ) . The palmitate data set ( HP ) consists of 5 samples of human islets , evaluated by RNAseq at 48 hr of palmitate exposure ( Cnop et al . , 2014 ) . All the samples are paired with their respective non-treated controls , that is , human islets obtained from the same donor and control condition ( Ylipaasto et al . , 2005; Eizirik et al . , 2012; Lopes et al . , 2014 ) . A group of 9504 genes commonly identified in all data sets were considered for the analysis . In order to assess the similarity of gene expression in two different data sets , the following procedure was adopted . Gene expression fold change was calculated for each paired sample in each data set ( both composed of the same genes ) . In each data set , genes were then ranked by mean fold change . A plot was drawn associating to each number of genes n ( as a ratio to the total , x-axis ) the number of common genes in the first n of the two rankings ( divided by n ) . Lines corresponding to p-values thresholds ( in this case 0 . 001 ) and the AUC are also presented ( this statistic is known as the Sørensen–Dice index ) . AUC values and lines delimiting an area of null similarity ( i . e . , expected similarity of two random rankings ) corresponding to a p-value of 0 . 001 ( hypergeometric distribution ) are also shown . Note that this analysis concerns similarity of up-regulation ( genes are ranked from high- to low-fold change ) . Male Wistar rats ( Charles River Laboratories , L'Arbresle Cedex , France ) were housed and used according to the guidelines of the Belgian Regulations for Animal Care , with the approval by the local Ethical Committee ( protocol number 465N; period of validity 07/2013-07/2017 ) . Rat islets were isolated by collagenase digestion and hand picked . For β and α cells isolation , islets were dissociated into single cells by mechanical and enzymatic dispersion using trypsin ( 1 mg/ml ) ( Sigma , Bornem , Belgium ) and DNase I ( 1 mg/ml ) ( Roche Applied Science , Indianapolis , USA ) for 5 min at 31°C under agitation . Dissociated cells were re-suspended in HEPES-buffered Earle's medium containing 2 . 8 mM glucose and purified by FACS as described in ( Marroqui et al . , 2015 ) . After sorting , purified β cells were cultured in Ham's F-10 medium containing 10 mM glucose , 2 mM GlutaMAX , 0 . 5% bovine serum albumin ( BSA ) , 50 µM isobutylmethylxanthine , 50 units/ml penicillin and 50 µg/ml streptomycin and 5% heat-inactivated fetal bovine serum ( FBS , Gibco Life Technologies , Germany ) . α Cells were cultured in the same medium but with 6 . 1 mM glucose and 10% FBS . Purity of the β and α cell preparations was evaluated by immunofluorescence . α and β cells were immunostained with mouse monoclonal anti-insulin ( Sigma , Bornem , Belgium ) or mouse monoclonal anti-glucagon ( Sigma , Bornem , Belgium ) for 1 hr followed by rabbit anti-mouse secondary antibody conjugated with AlexaFluor 488 or AlexaFluor 567 . The purity was calculated as a % of positive cells in each cell type ( Figure 2A ) . Human islets were isolated from 2 non-diabetic organ donors ( Figure 11—figure supplement 1 ) with approval from the local Ethical Committee in Pisa , Italy . Organ and tissue donation in Italy is regulated by the art . 23 of the national law n . 91 , issued on 1 April 1999; in Tuscany the regional transplant organization ( OTT , Organizzazione Toscana Trapianti ) allows that organs not suitable for clinical transplantation are used for research purposes provided informed consent has been signed by the responsible relative . Prof Marchetti's group has access to donated pancreases for the preparation and study of isolated islets on the basis of approval by their local ethics committee , renewed in 2013 . Isolation of human islets was done by collagenase digestion and density-gradient purification ( Marchetti et al . , 2007 ) . Subsequently , isolated islets were cultured in M199 medium containing 5 . 5 mM glucose ( Marchetti et al . , 2007 ) . Within 1–5 days of isolation , the human islets were shipped to Brussels . After arrival in Brussels and overnight recovery , the human islets were dispersed and cultured in Ham's F-10 medium containing 6 . 1 mM glucose , 2 mM GlutaMAX , 50 μM 3-isobutyl-1-methylxanthine , 1% charcoal-absorbed bovine serum albumin , 10% FBS , 50 mg/ml streptomycin , and 50 units/ml penicillin . The proportion of β cells and α cells in the preparations was determined by immunocytochemistry for insulin and glucagon , respectively ( Eizirik et al . , 2012 ) . Cells were treated with recombinant human IL-1β ( R&D Systems , Abingdon , U . K . ) and recombinant rat IFNγ ( R&D Systems , Abingdon , U . K . ) for 24 hr . The concentrations of cytokines ( Marroqui et al . , 2014 ) used are indicated in the figures . The synthetic dsRNA analog PIC ( Invitrogen , San Diego , CA , USA ) was used at the final concentration of 1 μg/ml and its transfection into cells was performed under the same conditions as used for siRNA ( see below ) but using 0 . 15 ml of Lipofectamine 2000 ( Colli et al . , 2011 ) . Culture supernatants were collected for nitrite determination ( nitrite is a stable product of nitric oxide [NO] oxidation ) at OD540 nm using the Griess method ( Green et al . , 1982 ) . The prototype strains of enterovirus ( CVB5/Faulkner; CBV-4/J . V . B . ) were obtained from American Type Culture Collection ( Manassas , VA ) . This virus was passaged in Green Monkey Kidney cells . The identity of the enterovirus preparations used was confirmed using a plaque neutralization assay with type-specific antisera ( Roivainen et al . , 2000 ) . We choose to analyze the effect of CVB5 and CVB4 , both serotypes detected in T1D donors ( Yeung et al . , 2011 ) , in order to allow comparisons with our previous studies based on infection of β cells with these viral serotypes ( Ylipaasto et al . , 2004 , 2005 , 2012 ) . Importantly , both CVB4 and CVB5 have been detected in islets of T1D donors ( Yeung et al . , 2011 ) and CVB4 has been associated to diabetes onset ( Dotta et al . , 2007; Gallagher et al . , 2015 ) . We did not test the CVB1 serotype because it has been shown that CVB1 does not multiply in rat β cells ( Nair et al . , 2010 ) . Viral stocks were prepared in GMK cells and titrated by plaque assay as previously described ( Roivainen et al . , 2000 ) or by limit dilution assay; viral titers obtained in plaque forming unit/ml and in 50% tissue culture infectious dose/ml were similar . β cells , α cells , and dispersed human islets were infected with virus diluted in β and α cells , or human islets medium in the absence of serum at the indicated M . O . I . After adsorption for 2 hr at 37°C , the inoculum virus was removed and cells were washed 3 times with medium . Serum-containing medium was added to the plates and the virus was allowed to replicate for indicated time periods . Inactivated virus was prepared by UV irradiation with a 1000J/m2 dose and proper inactivation verified by titration on GMK cells . Infected cells were frozen in their medium and thawed three times to release the virus . Total infectivity was assayed using end-point dilutions in microwell cultures of GMK cells . Cytopathic effects were read on day 6 by microscopy , and 50% tissue culture infectious dose titers were calculated using the Kärber formula ( Lennette and Schemidt , 1969 ) . β and α cells were infected with an adenovirus encoding Green fluorescent protein ( adeno-GFP; [Heimberg et al . , 2001] ) diluted in β or α cell medium in the absence of serum at M . O . I . 1 , 5 , or 10 . After adsorption for 3 hr at 37°C , the inoculum was removed , serum-containing medium was added to the plates , and the cells were allowed to express GFP for 48 hr . Cells were detached with mild trypsin treatment and suspended in 2% paraformaldehyde-containing phosphate buffered saline ( PBS ) . Cells were then analyzed on a flow cytometer ( FacsCalibur , BD Biosciences , San Jose , CA ) . Analysis was performed using CellQuest Pro software version 6 . 0 ( BD Biosciences , San Jose , CA ) . The cellular populations were selected based on size and cell granularity and analyzed for green fluorescence . α Cells were transfected with 30 nM of the previously validated siRNA for STAT1 ( 5′-CCCUAGAAGACUUACAAGAUGAAUA-3 , Invitrogen , Carlsbad , CA , USA; [Moore et al . , 2011] ) or Allstars Negative Control siRNA ( siCTRL , Qiagen , Venlo , the Netherlands; used as a negative control ) using the Lipofectamine RNAiMAX lipid reagent ( Invitrogen , Carlsbad , CA , USA ) . siCTRL does not affect α cell gene expression , function , or viability ( data not shown ) . Cells were cultured for 48 hr after transfection and then infected with CVB5 . The percentage of viable , apoptotic , and necrotic cells was determined after incubation with the DNA-binding dyes propidium iodide ( 5 μg/ml; Sigma , Bornem , Belgium ) and Hoechst 33342 ( 5 μg/ml; Sigma , Bornem , Belgium ) ( Rasschaert et al . , 2005 ) . A minimum of 600 cells was counted in each experimental condition . Viability was evaluated by two independent observers , one of them unaware of sample identity . The agreement between observers was >90% . Cell counting of floating ( i . e . , in the supernatant ) or attached cells was performed in Neubauer chambers , and each point was measured in triplicate by two observers , one of them unaware of sample identity . Poly ( A ) +mRNA was isolated from primary rat β and α cells using the Dynabeads mRNA DIRECT kit ( Invitrogen , Carlsbad , CA , USA ) , reverse transcribed , and amplified by real-time PCR using SYBR Green as described ( Rasschaert et al . , 2005 ) . Quantitative real-time PCR was compared with a standard curve ( Overbergh et al . , 1999 ) . Expression values were corrected for the reference gene glyceraldehyde-3-phosphate dehydrogenase ( GAPDH ) , whose expression is not modified by the presently utilized experimental conditions ( Figure 2C and data not shown ) . Primers are detailed in Supplementary file 2 . Cells were washed with cold PBS and lysed in Laemmli buffer . Immunoblot analysis was performed with anti-STAT1 ( 1:1000; Santa Cruz , Dallas , USA ) , enterovirus-specific rabbit antiserum ( 1:1000; KTL-510 ) , anti-CAR ( 1:100; Santa Cruz Biotechnology ) , anti-α-tubulin ( 1:5000; Sigma , Bornem , Belgium ) , and anti-β-actin ( 1:2000; Cell signaling ) . Membranes were exposed to secondary peroxidase-conjugated antibody for 1 hr at room temperature . Immunoreactive bands were revealed using the SuperSignal West Femto chemiluminescent substrate ( Thermo Scientific , Rockford , USA ) and detected using a Bio-Rad chemi DocTM XRS+ ( Bio-Rad laboratories ) . The densitometry of the bands was evaluated using Image Laboratoty software ( Bio-Rad laboratories ) . Immunofluorescence was performed as described ( Gurzov et al . , 2010 ) . Briefly , cells were plated on polylysine-coated cover slips , infected with CVB5 or CVB4 M . O . I . 10 for 8 hr , and fixed with 4% paraformaldehyde . After permeabilization with 0 . 3% Triton X-100 , cells were incubated for 1 hr with an enterovirus-specific rabbit antiserum ( 1:1000; KLT-510 ) , mouse monoclonal anti-insulin ( 1:1000; Sigma , Bornem , Belgium ) , or mouse monoclonal anti-glucagon ( 1:1000; Sigma , Bornem , Belgium ) . Alexa Fluor 568 goat anti-rabbit IgG or rabbit anti-mouse IgG and Alexa Fluor 488 goat anti-mouse IgG were , respectively , applied for 1 hr ( 1:1000 ) . After nuclear staining with Hoechst , cover slips were mounted with fluorescent mounting medium ( DAKO , Carpintera , USA ) , and immunofluorescence was visualized on a Zeiss microscope equipped with a camera ( Zeiss-Vision , Munich , Germany ) . Images were acquired at 40× magnification and analyzed using AxiVision software . For the histological study of clinical samples , three human pancreases removed at autopsy from neonatal patients ( 3–14 days ) with fatal coxsackievirus infections were employed ( Figure 12—figure supplement 1 ) . The cases were selected randomly from within a previously described collection ( Foulis et al . , 1990; Richardson et al . , 2009 ) . Specimens had been fixed in buffered formalin or unbuffered formol saline , and they were all paraffin-embedded . The coxsackievirus infected human tissue samples were from a historical collection compiled in the 1980s , when fully informed consent was not required . They are held in the Glasgow Diabetes Biobank and were analyzed under authority of the UK Human Tissue Authority ( licence number 12276 ) . Ethical permission was granted by Greater Glasgow Clyde Research Ethics Committee ( Ref: 10/S0704/25 ) . Serial sections ( 4 μm ) were mounted on glass slides coated in ( 3-aminopropyl ) -triethoxysilane ( Sigma , Dorset , UK ) . Antigens were unmasked by heat-induced epitope retrieval in 10 mM citrate buffer pH 6 . 0 . To examine the islet cell subtypes expressing VP1 , triple immunofluorescence staining was performed . Sections were incubated with antisera with specificity for the enteroviral capsid protein , VP1 ( Dako; 5D8/1 ) overnight at 1:500 ( optimal for these specimens ) and detected using the Life Technologies TSA kit#2 ( AlexaFluor 488 ) as per the manufacturer's instructions . The sections were washed and stained with rabbit anti-glucagon ( Abcam; 1:4000 ) for 1 hr followed by goat anti-rabbit secondary antibody conjugated with AlexaFluor 555 . Finally , sections were incubated with a guinea pig anti-insulin serum ( DAKO; 1/600 ) , which was detected with a goat anti-guinea pig secondary antibody conjugated with AlexaFluor 647 . DAPI ( 1:1000 , Invitrogen ) was included in the final incubation to stain cell nuclei . Some slides were processed in the absence of primary antibody or with isotype control antisera to confirm the specificity of labeling . Sections were mounted in fluorescence mounting medium ( Dako ) under glass cover slips and examined on a Leica AF6000 Microscope . Multi-channel images were collected , processed , and analyzed using LAS AF software . Data are presented as mean values ± SEM or plotted as box plots , indicating lower quartile , median , and higher quartile , with whiskers representing the range of the remaining data points . Comparisons were performed by two-tailed paired Student's t-test or by analysis of variance ( ANOVA ) followed by Student's t-test with Bonferroni correction , as indicated . A p value <0 . 05 was considered as statistically significant .	Type 1 diabetes is caused by a person's immune system attacking the cells in their pancreas that produce insulin . This eventually kills off so many of these cells—known as beta cells—that the pancreas is unable to make enough insulin . As a result , individuals with type 1 diabetes must inject insulin to help their bodies process sugars . One of the mysteries of type 1 diabetes is why the beta cells in the pancreas are killed by the immune system while neighboring alpha cells , which produce the hormone glucagon , are spared . Scientists suspect a combination of genetic and environmental factors contributes to type 1 diabetes . Certain viruses , including one called Coxsackievirus , appear to trigger type 1 diabetes in susceptible individuals . Other factors may also make these individuals more likely to develop the disease . For example , they may ‘express’ genes that are thought to increase the risk of type 1 diabetes , many of which control how the immune system responds to viral infections . These genes may make susceptible individuals experience excessive inflammation , because inflammation is what ultimately kills off the beta cells . Now , Marroqui , Lopes , dos Santos et al . provide evidence that suggests why the alpha cells are spared the immune onslaught in type 1 diabetes . In initial experiments , clusters of cells—known as islets—from the human pancreas were either exposed to small proteins that cause inflammation or infected with the Coxsakievirus . Both events caused a similar increase in the expression of particular immune response genes in the islets . This indicates that these islet cells are able to react to the virus and trigger a first line of defense , which will be further boosted when the immune system is subsequently called into action . Islets contain both alpha and beta cells , and so further experiments on alpha and beta cells from rats investigated whether the two cell types respond differently when infected by the Coxsakievirus . The results revealed that alpha cells boost the expression of the genes needed to clear the virus to a greater extent than the beta cells , and so respond more efficiently to the virus . Therefore , an infection is more likely to establish itself in the beta cells and consequently trigger inflammation and the immune system's attack on the cells . These observations explain one of the puzzling questions in the diabetes field and reinforce the possibility that a long-standing viral infection in beta cells—which seem to have a limited capacity to clear viral infections—may be one of the mechanisms leading to progressive beta cell destruction in type 1 diabetes . This knowledge will help in the search for ways to protect beta cells against both viral infections and the consequent immune assault .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "immunology", "and", "inflammation" ]	2015	Differential cell autonomous responses determine the outcome of coxsackievirus infections in murine pancreatic α and β cells
Calcium in the flagellum controls sperm navigation . In sperm of marine invertebrates and mammals , Ca2+ signalling has been intensely studied , whereas for fish little is known . In sea urchin sperm , a cyclic nucleotide-gated K+ channel ( CNGK ) mediates a cGMP-induced hyperpolarization that evokes Ca2+ influx . Here , we identify in sperm of the freshwater fish Danio rerio a novel CNGK family member featuring non-canonical properties . It is located in the sperm head rather than the flagellum and is controlled by intracellular pH , but not cyclic nucleotides . Alkalization hyperpolarizes sperm and produces Ca2+ entry . Ca2+ induces spinning-like swimming , different from swimming of sperm from other species . The “spinning” mode probably guides sperm into the micropyle , a narrow entrance on the surface of fish eggs . A picture is emerging of sperm channel orthologues that employ different activation mechanisms and serve different functions . The channel inventories probably reflect adaptations to species-specific challenges during fertilization . Fertilization is a complex task that , for different species , happens in entirely different spatial compartments or ionic milieus . In aquatic habitats , gametes are released into the water where sperm acquire motility and navigate to the egg . By contrast , mammalian fertilization happens in confined compartments of the female oviduct . From invertebrates to mammals , sperm use various sensing mechanisms , including chemotaxis , rheotaxis , and thermotaxis , to gather physical or chemical cues to spot the egg . These sensory cues activate various cellular signalling pathways that ultimately control the intracellular Ca2+ concentration ( [Ca2+]i ) and , thereby , the flagellar beat and swimming behaviours ( Alvarez et al . , 2012; Darszon et al . , 2008; Eisenbach and Giojalas , 2006; Florman et al . , 2008; Guerrero et al . , 2010; Ho and Suarez , 2001; Kaupp et al . , 2008; Publicover et al . , 2008 ) . In species as phylogenetically distant as sea urchin and mammals , these pathways target a sperm-specific , voltage-dependent Ca2+ channel , called CatSper . Signalling events open CatSper by shifting its voltage-dependence to permissive , more negative Vm values . This shift is achieved by different means . In sea urchin sperm , opening of a K+-selective cyclic nucleotide-gated channel ( CNGK ) causes a transient hyperpolarization ( Bönigk et al . , 2009; Strünker et al . , 2006 ) ; the hyperpolarization activates a sperm-specific Na+/H+ exchanger ( sNHE ) ( Lee , 1984 , 1985; Lee and Garbers , 1986 ) resulting in a long-lasting alkalization that shifts the voltage dependence of CatSper and leads to a Ca2+ influx ( Kaupp et al . , 2003; Seifert et al . , 2015 ) . By contrast , in human sperm , the shift is achieved by direct stimulation of CatSper with prostaglandins and progesterone in the seminal fluid or the oviduct ( Brenker et al . , 2012; Lishko et al . , 2011; Smith et al . , 2013; Strünker et al . , 2011 ) . Navigation of fish sperm and the underlying signalling pathways must be arguably different . First , teleost fish are lacking CatSper channels ( Cai and Clapham , 2008 ) , although activation of sperm motility requires Ca2+ influx ( Alavi and Cosson , 2006; Billard , 1986; Cosson et al . , 2008; Morisawa , 2008; Takai and Morisawa , 1995 ) stimulated by hyper- or hypoosmotic shock after spawning into seawater or freshwater , respectively ( Alavi and Cosson , 2006; Cherr et al . , 2008; Krasznai et al . , 2000; Morisawa , 2008; Vines et al . , 2002 ) . Therefore , Ca2+ signalling in fish sperm must involve molecules different from those in marine invertebrates and mammals . Second , the ionic milieu seriously constrains ion channel function . Sperm of freshwater fish , marine invertebrates , and mammals are facing entirely different ionic milieus . K+ and Na+ concentrations in freshwater are extremely low ( 70 μM and 200 µM , respectively ) compared to the orders-of-magnitude higher concentrations in seawater or the oviduct ( Alavi and Cosson , 2006; Hugentobler et al . , 2007 ) . Furthermore , [Ca2+] in seawater is high ( 10 mM ) , whereas in freshwater it is low ( < 1 mM ) . The low salt concentrations in freshwater probably require distinctively different ion channels . In fact , none of the ion channels controlling electrical excitation and Ca2+ signalling of fish sperm are known . Finally , fish sperm are not actively attracted to the whole egg from afar by chemical or physical cues , i . e . chemotaxis , thermotaxis , or rheotaxis ( Cosson et al . , 2008; Morisawa , 2008; Yanagimachi et al . , 2013 ) . Instead , many fishes deposit sperm directly onto the eggs . For fertilization , sperm must search for the narrow entrance to a cone-shaped funnel in the egg coat – the micropyle – that provides access to the egg membrane . Sperm reach the micropyle probably by haptic interactions with tethered molecules that line the egg surface and the opening or interior of the micropyle ( Iwamatsu et al . , 1997; Ohta and Iwamatsu , 1983; Yanagimachi et al . , 2013 ) . At surfaces , sperm swim with their flagellum slightly inclined , which pushes the head against the wall and stabilizes sperm at the surface ( Denissenko et al . , 2012; Elgeti et al . , 2010 ) . Thus , fish sperm motility might be governed by specific hydrodynamic and haptic interactions with the egg surface and the micropyle . Although the principal targets of a CNGK-mediated hyperpolarization – the Na+/H+ exchanger and CatSper – are absent in fish , vertebrate orthologues of the sperm CNGK channel are present in various fish genomes ( Figure 1A ) . Here , we study the function of the CNGK channel in sperm of the freshwater fish Danio rerio ( DrCNGK ) . The DrCNGK channel constitutes the principal K+ channel in D . rerio sperm . Unexpectedly , cyclic nucleotides neither regulate CNGK channel activity nor sperm motility; instead , intracellular alkalization , a key mechanism to control sperm function in many species , strongly activates CNGK and , thereby , triggers a Ca2+ signal and a motility response . Although its mechanism of activation is entirely different compared to sea urchin sperm , the principal CNGK function , namely to provide a hyperpolarization that triggers a Ca2+ signal , is conserved . Our results show that sperm signalling among aquatic species shows unique variations that probably represent adaptations to vastly different ionic milieus and fertilization habits . 10 . 7554/eLife . 07624 . 003Figure 1 . Identification of DrCNGK channel homologues and of a K+ channel in D . rerio sperm . ( A ) Phylogenetic tree ( Page , 1996 ) of various ion channel families . The CNGK channel family exists in protozoa ( dark blue ) , marine invertebrates and fish ( medium blue ) , and freshwater fish ( light blue ) . The HCN , CNG , and KCNH channel families are highlighted in green; voltage-gated Nav and Cav channels are highlighted in yellow; and voltage-gated Kv channels are highlighted in red . The following ion channel sequences were used: CNGK channels from zebrafish ( DrCNGK ) , rainbow trout ( OmCNGK ) , spotted gar ( LoCNGK ) , West Indian Ocean coelacanth ( LcCNGK ) , sea urchin ( ApCNGK ) , acorn worm ( SkCNGK ) , amphioxus ( BfCNGK ) , starlet sea anemone ( NvCNGK ) , vasa tunicate ( CiCNGK ) , sponge ( AqCNGK ) , choanoflagellate ( SrCNGK ) ; murine HCN channel subunits 1 ( mHCN1 ) , 2 ( mHCN2 ) , 3 ( mHCN3 ) , 4 ( mHCN4 ) , and the HCN channel from sea urchin ( SpHCN1 ) ; rat CNGA subunits A1 ( rCNGA1 ) , A2 ( rCNGA2 ) , A3 ( rCNGA3 ) , and A4 ( rCNGA4 ) ; the KCNH channels from fruit fly ( DmEAG ) and human ( hERG ) ; murine voltage-gated Nav ( mNav 1 . 1 and mNav 1 . 6 ) and Cav channels ( mCav1 . 1 , mCav2 . 3 and mCav3 . 1 ) and voltage-gated Kv channels from fruit fly ( DmShaker ) and mouse ( mKv3 . 1 ) . Full-length Latin names and accession numbers are given in experimental procedures . Scale bar represents 0 . 1 substitutions per site . ( B ) Pseudo-tetrameric structure of CNGK channels . Numbers 1 to 4 , homologous repeats; S1 to S6 , transmembrane segments; yellow cylinders , cyclic nucleotide-binding domain CNBD; asterisks , epitopes recognized by antibodies anti-repeat1 of DrCNGK ( polyclonal ) and anti-repeat3 of DrCNGK ( YENT1E2 , monoclonal ) . ( C ) Whole-cell recordings from zebrafish sperm at low ( left upper panel ) and high ( middle panel ) extracellular K+ concentrations . Left lower panel: Voltage step protocol . Right panel: corresponding IV relations . ( D ) Whole-cell recordings from an isolated sperm head . Description see part C . ( E ) Whole-cell recording from zebrafish sperm ( upper panel ) and an isolated head ( lower panel ) . ( F ) IV relation of recordings from part E . ( G ) Pooled IV relations ( ± sd ) of currents from zebrafish sperm ( filled circle , n = 23 ) and sperm heads ( open squares , n = 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 00310 . 7554/eLife . 07624 . 004Figure 1—figure supplement 1 . Amino-acid sequence of the DrCNGK channel . Different colors indicate the transmembrane segments ( red shades , repeats 1–4 ) , the four pore regions ( green ) , the four CNBDs ( gray ) , and the unusual insert in the C-linker of the third repeat ( blue ) . Lines below the sequence indicate peptides that were identified by mass spectrometry in testis and sperm preparations . The position of the last amino-acid residue in a row is given on the right . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 00410 . 7554/eLife . 07624 . 005Figure 1—figure supplement 2 . Separation of heads and flagella from whole sperm . Dark-field micrographs of whole sperm ( left ) , purified heads ( middle ) , and purified flagella ( right ) . Bar represents 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 00510 . 7554/eLife . 07624 . 006Figure 1—figure supplement 3 . Electrophysiological characterization of currents recorded from zebrafish sperm . ( A ) Whole-cell recordings at different intracellular Cl- concentrations ( Cli ) . ( B ) IV relations from part A . ( C ) Reversal potentials ( Vrev ) from whole-cell recordings at different extracellular K+ ( Ko ) and intracellular Cl- concentrations ( Cli ) in whole sperm and isolated heads . Individual data ( symbols ) and mean ± sd ( gray bars ) , number of experiments in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 006 We recorded currents from whole D . rerio sperm ( Figure 1—figure supplement 2 , left panel ) in the whole-cell patch-clamp configuration . Voltage steps from a holding potential of -65 mV evoked slightly inwardly rectifying currents ( Figure 1C ) . Two pieces of evidence established that currents are carried by K+ channels and not by Cl- channels . The reversal potential ( Vrev ) shifted from -77 ± 3 mV ( n = 18 ) at 5 . 4 mM extracellular [K+]o to -7 ± 1 mV ( n = 7 ) at 140 mM [K+]o ( △Vrev = 51 ± 3 mV/log [K+] , n = 7 ) ( Figure 1C , Figure 1—figure supplement 3C ) . Changing the intracellular [Cl-] did not affect Vrev ( Figure 1—figure supplement 3A , B ) . These results demonstrate that the current is predominantly carried by K+ ions . To localize the underlying K+ channel , we recorded currents from isolated sperm heads ( Figure 1D-G , Figure 1—figure supplement 2 , middle panel ) . Head and whole-sperm currents displayed a similar K+ dependence ( △Vrev = 52 ± 2 mV/log [K+] , n = 6 ) ( Figure 1D ) , rectification ( Figure 1F , G ) , and amplitude ( Figure 1F , G ) , suggesting that the underlying K+ channel is primarily located in the head . This result is unexpected , as ion channels involved in sperm signalling are usually localized to the flagellum . To test whether the DrCNGK channel is also localized to the head , we used Western blot analysis and immunocytochemistry . To this end , the DrCNGK protein was first characterized by heterologous expression in mammalian cell lines . DrCNGK constructs with a C-terminal HA-tag or with two tags , a C-terminal HA-tag and an N-terminal flag-tag , were expressed in CHOK1 cells ( Figure 2A ) . In Western blots , the anti-HA-tag and the anti-flag-tag antibody labelled proteins of the same apparent molecular mass ( Mw ) ( 174 ± 4 kDa ( n = 13 ) and 175 ± 4 kDa ( n = 3 ) , respectively ) ( Figure 2A ) . The Mw is smaller than the predicted Mw of 244 . 4 kDa . Because flag-tag and HA-tag antibodies recognized the N- and C-terminal end of the CNGK protein , respectively , we conclude that the 175-kDa band represents the full-length protein that , however , displays an abnormal electrophoretic mobility similar to other CNG channels ( Körschen et al . , 1995; 1999 ) . 10 . 7554/eLife . 07624 . 007Figure 2 . Localization of the DrCNGK channel . ( A ) Western blot of membrane proteins ( 15 µg ) from CHOK1 cells transfected with cDNA encoding either DrCNGK with a C-terminal HA-tag alone ( lane C ) or with both , a C-terminal HA-tag and an N-terminal flag-tag ( N/C ) . Apparent molecular weight Mw is indicated on the left . ( B ) Characterization of anti-DrCNGK antibodies . Left: Western blot of membrane proteins ( 10 µg ) from HEK293 cells transfected with cDNA encoding DrCNGK ( Tr ) and wild-type cells ( wt ) . Right: Western blot of membrane proteins ( 15 µg ) from zebrafish testis . ( C ) Western blot of membrane proteins ( 15 µg ) from different zebrafish tissues . ( D ) Upper panel: Scheme of a testis cross-section . GC , germinal compartment; IC , intertubular compartment; SER , Sertoli cells; SGA , primary spermatogonia; SGB , secondary spermatogonia; SC , spermatocytes; ST , spermatids; scheme according to ( Nobrega et al . , 2009 ) . Lower panel: Staining with anti-repeat1 antibody ( red , left ) and superposition ( right ) of the immunohistochemical image with a bright-field image of an in situ hybridization using an anti-DrCNGK-specific RNA probe ( arrows ) . Bar represents 50 µm . ( E ) Staining of zebrafish sperm with anti-repeat1 ( upper left panel ) and anti-repeat3 antibody ( lower left panel ) . Bars represent 10 µm . The respective bright-field images are shown ( upper and lower right panels ) . ( F ) Western blot of equal amounts of total membrane proteins ( 15 µg ) from purified heads and purified flagella . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 00710 . 7554/eLife . 07624 . 008Figure 2—source data 1 . Indicators of merit for the mass spectrometric results of a testis preparation . a m indicates oxidized methionine , b△M [ppm] relative mass error , c XCorr , △Score , and △Cn indicate results based on searches with the sequest algorithm in Proteome Discoverer; d PEP ( Posterior Error Probability ) describes the probability that the observed hit is a chance event . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 00810 . 7554/eLife . 07624 . 009Figure 2—source data 2 . Indicators of merit for the mass spectrometric results of different sperm preparations: whole sperm , head and flagella . a m indicates oxidized methionine , b△M [ppm] for all peptides is below ± 2 . 5 , c XCorr indicates results based on searches with the sequest algorithm; Amanda indicates results based on searches with the MS Amanda algorithm , d PEP ( Posterior Error Probability ) describes the probability that the observed hit is a chance event; △Score and △Cn are not indicated separately since the scores for all peptides are 1 and 0 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 009 We raised two antibodies against epitopes in repeat 1 and 3 of the DrCNGK protein ( Figure 1B , asterisks ) . Both antibodies labelled membrane proteins of about 170 kDa in Western blots of DrCNGK-expressing CHOK1 cells , D . rerio testis , and sperm , but not of heart , brain , ovaries , and eyes ( Figure 2B , C ) . To scrutinize the antibody specificity , we analyzed by mass spectrometry the ~170-kDa protein band from testis , mature whole sperm , isolated heads , and isolated flagella; 7 , 23 , 18 , and 15 proteotypic DrCNGK peptides were identified , respectively ( Figure 1—figure supplement 1 , Figure 2—source data 1 , 2 ) . Peptides covered almost the entire polypeptide sequence ( Figure 1—figure supplement 1 ) . The presence of DrCNGK in testis was confirmed by immunohistochemistry and in situ hybridization of D . rerio testis slices . The anti-repeat1 antibody labelled structures , most likely sperm , in the lumen of testicular compartments ( Figure 2D , bottom left ) . An antisense RNA probe stained sperm precursor cells , in particular spermatocytes ( Figure 2D , bottom right ) , but almost no primary or secondary spermatogonia . Finally , the anti-repeat1 and anti-repeat3 antibodies intensely labelled the head and , to a lesser extent , the flagellum of single sperm cells ( Figure 2E ) . In Western blots of isolated heads and flagella , the DrCNGK was readily identified in head preparations , yet was barely detectable in flagella preparations ( Figure 2F , n = 4 ) . In summary , the CNGK channel is located primarily in the head of mature D . rerio sperm . The sea urchin ApCNGK channel is opened by cyclic nucleotides and mediates the chemoattractant-induced hyperpolarization ( Bönigk et al . , 2009; Strünker et al . , 2006 ) . Unexpectedly , patch-clamp recordings of K+ currents from D . rerio sperm required no cyclic nucleotides in the pipette ( Figure 1C-G ) . Therefore , we scrutinized the action of cyclic nucleotides on sperm K+ currents . Mean current amplitudes were similar in controls and in the presence of either cAMP or cGMP ( 100 µM ) in the pipette solution ( Figure 3A ) . We used also caged cyclic nucleotides to study the K+ current in the absence and presence of cyclic nucleotides in the same sperm cell ( Kaupp et al . , 2003 ) . Photo-release of cAMP or cGMP from caged precursors did not affect K+ currents , suggesting that cyclic nucleotides do not modulate DrCNGK ( Figure 3B ) . In contrast , the photo-release of cAMP or cGMP induced a rapid current increase in heterologously expressed cyclic nucleotide-gated channels ApCNGK from sea urchin ( Figure 3—figure supplement 3A ) . We also heterologously expressed the DrCNGK channel in X . laevis oocytes . The current-voltage ( IV ) relation ( Figure 3C–F ) , K+ dependence ( Figure 3—figure supplement 1A-D ) , and block by external TEA ( Figure 3—figure supplement 1E , F ) was similar to that of the K+ current recorded from D . rerio sperm . Moreover , currents in oocytes were also insensitive to the membrane-permeable analogs 8Br-cAMP and 8Br-cGMP ( Figure 3C–F ) , whereas perfusion with 8Br-cGMP increased currents in oocytes that express the ApCNGK channel from A . punctulata sperm ( Figure 3—figure supplement 3B ) . 10 . 7554/eLife . 07624 . 010Figure 3 . Cyclic nucleotides do not activate K+ channels in sperm . ( A ) Current amplitude of whole-cell recordings from zebrafish sperm at +25 mV in the absence or presence of 100 µM cAMP or cGMP in the pipette ( control: 91 ± 49 pA ( n = 23 ) ; cAMP: 73 ± 25 pA ( n = 6 ) ; cGMP: 109 ± 44 pA ( n = 5 ) ) . Individual data ( symbols ) and mean ± sd ( gray bars ) , number of experiments in parentheses . ( B ) Photo-release of cyclic nucleotides from caged precursors inside sperm . Left panel: Whole-cell recordings at +15 mV from sperm loaded with 100 µM BCMACM-caged cAMP ( upper panel ) or BCMACM-caged cGMP ( lower panel ) . Arrows indicate the delivery of the UV flash to release cyclic nucleotides by photolysis . Right panel: Mean current 3 s before ( - ) and 3 s after ( + ) the release of cAMP or cGMP . Statistics as in part A . Data points from individual sperm are indicated by identical colours . ( C-F ) Currents of heterologously expressed DrCNGK channels in the absence or presence of 8Br- analogs of cyclic nucleotides . ( C ) Left: Two-Electrode Voltage-Clamp recordings from DrCNGK-injected Xenopus oocytes . Currents shown are in the absence ( left traces ) and presence ( right traces ) of 10 mM 8Br-cAMP . Voltage steps as shown in Figure 3—figure supplement 1A . Right: IV relations of current recordings from the left panel . ( D ) Pooled IV curves from DrCNGK injected and control oocytes; recordings in the absence and presence of 10 mM 8Br-cAMP . ( E ) Left: Two-Electrode Voltage-Clamp recordings from DrCNGK injected Xenopus oocytes . Currents shown are in the absence ( left traces ) and presence ( right traces ) of 10 mM 8Br-cGMP . Right: IV relations of current recordings from the left panel . ( F ) Pooled IV curves from DrCNGK-injected and control oocytes; recordings in the absence and presence of 10 mM 8Br-cGMP . ( G ) Swimming path before ( green line ) and after ( red line ) photo-release ( black flash ) of cAMP ( left panel ) or cGMP ( right panel ) . The blue arrow indicates the swimming direction . Photo-release of cyclic nucleotides was verified by monitoring the increase of fluorescence of the caging group ( Figure 3—figure supplement 5 ) ( Hagen et al . , 2003 ) . ( H ) Path curvature before ( - ) and after ( + ) release of cAMP or cGMP . Sperm were loaded with 30 µM DEACM-caged cAMP or DEACM-caged cGMP . Statistics as in part A . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 01010 . 7554/eLife . 07624 . 011Figure 3—figure supplement 1 . K+ dependence of heterologously expressed DrCNGK channels in oocytes and channel block by tetraethylammonium ( TEA ) . ( A ) Two-Electrode Voltage-Clamp recordings of heterologously expressed DrCNGK channels in the presence of different K+ concentrations ( left panel: 7 mM , middle panel: 96 mM ) and corresponding IV relation ( right panel ) . ( B ) Two-Electrode Voltage-Clamp recordings of uninjected oocytes ( control ) in the presence of different K+ concentrations ( left panel: 7 mM , middle panel: 96 mM ) . ( C ) Pooled IV curves of Two-Electrode Voltage-Clamp recordings of DrCNGK-injected and uninjected oocytes ( control ) in the presence of different K+ concentrations ( left panel: 7 mM , middle panel: 96 mM ) . Number of experiments is given in parentheses . ( D ) Reversal potentials ( Vrev ) of DrCNGK-injected and control oocytes in the presence of different K+ concentrations . Number of experiments is given in parentheses . ( E ) Whole-cell recordings from zebrafish sperm and Two-Electrode Voltage-Clamp recordings from heterologously expressed DrCNGK channels in the absence and presence of different TEA concentrations ( 0 , 1 , and 100 mM TEA ) . ( F ) Normalized ( ( I-IminFit ) /ImaxFit ) dose dependence of TEA block . Mean current ± sd . Individual dose response curves were fitted with the Hill equation . Mean ( ± sd ) Ki value and Hill coefficient for sperm K+ current were 4 . 5 ± 1 . 1 mM and 1 . 0 ± 0 . 2 ( n = 5 ) and for DrCNGK in oocytes 1 . 6 ± 0 . 3 mM and 1 . 0 ± 0 . 1 ( n = 11 ) , respectively . The solid lines were calculated with the Hill equation using mean values for Ki and the Hill coefficient . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 01110 . 7554/eLife . 07624 . 012Figure 3—figure supplement 2 . Sequence alignment of the individual CNBDs from the DrCNGK and ApCNGK channels . The secondary structure elements of CNBDs are indicated above the sequences . A key Arg residue between β6 and β7 is indicated by an asterisk . An FGE motif important for interaction with cyclic nucleotides and highly conserved Gly residues that are important for the CNBD fold are highlighted in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 01210 . 7554/eLife . 07624 . 013Figure 3—figure supplement 3 . Photo-release of cyclic nucleotides in HEK cells expressing ApCNGK channels and use of 8Br-analogs in ApCNGK-injected oocytes . ( A ) Left and middle panel: Pooled IV curves of ApCNGK channels heterologously expressed in HEK cells before ( -cGMP or -cAMP ) and after the release of cGMP or cAMP . Cells were loaded with 100 µM BCMACM-GMP or BCMACM-cAMP . Right panel: Whole-cell recordings at +15 mV from HEK cells heterologously expressing ApCNGK channels loaded with 100 µM BCMACM-caged cGMP ( upper panel ) or BCMACM-caged cAMP ( lower panel ) . Arrows indicate the delivery of the UV flash to release cyclic nucleotides by photolysis . ( B ) Pooled IV curves of Two-Electrode Voltage-Clamp recordings of ApCNGK-injected oocytes in the presence and absence of 3 mM 8Br-cGMP . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 01310 . 7554/eLife . 07624 . 014Figure 3—figure supplement 4 . Photo-release of cyclic nucleotides ( A ) or Ca2+ ( B ) in sperm . ( A ) Mean velocity ( averaged-path velocity , VAP ) before ( - ) and after ( + ) release of cAMP or cGMP . Sperm were loaded with 30 µM DEACM-caged cAMP or DEACM-caged cGMP . Individual data ( symbols ) and mean ± sd ( gray bars ) , number of experiments in parentheses . ( B ) Mean velocity ( VAP ) before and after the 1st ( + ) and 2nd ( ++ ) UV flash . Statistics as in part A . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 01410 . 7554/eLife . 07624 . 015Figure 3—figure supplement 5 . Control of loading and release of DEACM-cAMP in zebrafish sperm . ( A ) Dark-field micrograph ( using red light ) of sperm loaded with DEACM-caged cAMP ( 30 µM ) . ( B ) Fluorescence image after 15 s of continuous illumination with 365 nm UV light ( 1 . 75 mW power ) . ( C ) Time course of the release for the cell marked with a red circle . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 015 Membrane-permeant caged cyclic nucleotides have successfully been used to study sperm motility in sea urchin ( Böhmer et al . , 2005; Kashikar et al . , 2012; Wood et al . , 2005 ) and humans ( Gakamsky et al . , 2009 ) . We studied D . rerio sperm motility before and after photo-release of cAMP ( Figure 3G , left ) or cGMP ( Figure 3G , right ) from caged precursors . The photo-release was followed by the increase of fluorescence of the free coumaryl cage ( Figure 3—figure supplement 5 ) ( Bönigk et al . , 2009 ) . Swimming behaviour , i . e . path curvature ( Figure 3H ) and swimming speed ( Figure 3—figure supplement 4A ) were not altered by photo-release , showing that neither cAMP nor cGMP play a major role in the control of sperm motility . In conclusion , we observe no action of cyclic nucleotides on the DrCNGK channel and on the swimming behaviour of D . rerio sperm . We noticed a much stronger rectification of currents carried by sea urchin ApCNGK compared to zebrafish DrCNGK ( Figure 4A ) and , therefore , investigated the origin of this pronounced difference . In classical K+ channels , block by intracellular Mg2+ ( Matsuda et al . , 1987 ) or spermine ( Fakler et al . , 1995 ) produces inward rectification . However , neither Mg2+ nor spermine affected ApCNGK rectification ( Figure 4B ) . Instead , intracellular Na+ blocked outward currents in a strong voltage- and dose-dependent fashion ( Figure 4B , D ) . In the absence of Na+ , the IV relation of ApCNGK and DrCNGK channels converged ( Figure 4A , B ) . We searched the pore regions of CNGK channels for clues regarding the molecular basis of the Na+ block . In three of the four ApCNGK pore motifs , we identified a Thr residue that in most K+ channels is replaced by a Val or Ile residue ( Figure 4C ) . When these Thr residues were changed to Val , the strong rectification of the mutant ApCNGK channel was lost and the IV relation became similar to that of the DrCNGK channel ( Figure 4E ) . We also tested the reverse construct , introducing Thr residues into the pore motif of DrCNGK channels . For unknown reasons , the mutants did not form functional channels . 10 . 7554/eLife . 07624 . 016Figure 4 . Comparison of sperm K+ current with current from heterologously expressed ApCNGK channels . ( A ) Normalized IV relations of whole-cell recordings from zebrafish sperm and ApCNGK channels expressed in HEK293 cells . Pipette solution: standard IS . Bath solution: standard ES . Currents were normalized to -1 at -115 mV . ( B ) Normalized IV relations ( mean current ± sd , n = 6 ) of inside-out recordings from ApCNGK channels expressed in HEK293 cells . Pipette solution: standard ES , bath solution: NMDG-based IS with the indicated concentrations of Na+ , Mg2+ , and spermine . Currents were normalized to -1 at -103 mV . ( C ) Alignment of pore regions from different CNGK channels . Freshwater fishes are highlighted in light blue and seawater species in dark blue . The position of the last amino-acid residue is given on the right . Asterisks indicate the G ( Y/F ) GD selectivity motif . A key threonine residue that is conserved in three repeats of the ApCNGK channel and other seawater species is highlighted in red ( arrow ) . Hydrophobic amino acids at this position are indicated in gray . ( D ) IV relations of inside-out recordings of ApCNGK channels expressed in HEK293 cells . Pipette solution: ES; bath solution: NMDG-based IS . Different Na+ concentrations were added to the bath solution . ( E ) Normalized IV relations ( mean current ± sd ) of whole-cell recordings from zebrafish sperm ( n = 18 ) and from ApCNGK-4V channels ( n = 7 ) expressed in HEK293 cells . Currents were normalized to -1 at -115 mV . Inset: amino-acid sequence of the pore region of the mutant ApCNGK-4V . ApCNGK channels were activated with 100 µM cGMP . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 016 Of note , the Thr residues are absent in CNGKs of freshwater organisms yet present in seawater organisms except for the sponge AqCNGK ( Figure 4C ) , suggesting that the different CNGK pores represent adaptations to vastly different ionic milieus . Intracellular pH ( pHi ) is an important factor controlling sperm motility in marine invertebrates and mammals ( Alavi and Cosson , 2005; Dziewulska and Domagala , 2013; Hirohashi et al . , 2013; Lishko et al . , 2010; Lishko and Kirichok , 2010; Nishigaki et al . , 2014; Santi et al . , 1998; Seifert et al . , 2015 ) . Moreover , in mouse sperm , the Slo3 channels and the CatSper channels are exquisitely pH-sensitive ( Kirichok et al . , 2006; Schreiber et al . , 1998; Zeng et al . , 2011; 2013; Zhang et al . , 2006a; 2006b ) . Therefore , we examined whether DrCNGK is controlled by pHi . At pHi 6 . 4 , almost no CNGK current was recorded from D . rerio sperm ( Figure 5A , left panel , B , C ) . To rule out that an increase of Ca2+ ( from 91 pM to 7 . 7 nM , see Materials and methods ) due to the reduced buffering capacity of EGTA at pH 6 . 4 is responsible for current inhibition , we recorded sperm currents at pH 7 . 4 with a free Ca2+ concentration of 1 µM ( Figure 5—figure supplement 2 ) . Under these conditions , the K+ current in sperm was still large , indicating that protons and not Ca2+ , at least for the concentrations tested , are responsible for current inhibition . Exposing sperm to 10 mM NH4Cl rapidly elevates pHi ( Figure 5G ) , because NH4Cl overcomes the buffer capacity of HEPES at pH 6 . 4 ( Boron and De Weer , 1976; Seifert et al . , 2015; Strünker et al . , 2011 ) . Alkaline pHi strongly enhanced CNGK currents ( Figure 5A , middle panel , B , C ) . Subsequent superfusion with 10 mM propionic acid , which lowers pHi , completely reversed the NH4Cl-induced CNGK currents ( Figure 5A , right panel , B and C ) . The NH4Cl action was very pronounced: 1 mM activated approximately 40% of the CNGK current; at 10 mM , the current was maximal ( Figure 5D , triangles ) . To quantitatively determine the pH dependence , we recorded sperm K+ currents at different intracellular pHi values ( Figure 5C , D circles ) . The current is half-maximally activated at pH 7 . 08 . The pH dependence allows to calibrate the NH4Cl action by superposing the data of the two different experimental conditions ( Figure 5D ) : For example , at an NH4Cl concentration of 1 . 5 mM , pHi in sperm will increase from 6 . 4 to 7 . 08 ( Figure 5D ) . Under current-clamp conditions , alkalization of D . rerio sperm with 10 mM NH4Cl evoked a rapid and reversible hyperpolarization from -49 ± 7 mV to -71 ± 4 mV ( Figure 5E , n = 10 ) . We also studied the pH regulation of the DrCNGK channel expressed in Xenopus oocytes . Oocytes were first perfused with a K+ bicarbonate solution , followed by a K+ gluconate-based solution including 1 mM NH4Cl ( Figure 5—figure supplement 1 ) . 1 mM NH4Cl reversibly increased DrCNGK currents by approximately 71% , demonstrating regulation by pHi ( Figure 5F ) . Higher concentrations of NH4Cl further increased the DrCNGK current; however , these conditions also elicited significant currents in control oocytes , thus precluding quantitative analysis . Furthermore , we tested , whether DrCNGK channels in oocytes are activated by hypoosmotic conditions . Reducing the osmolarity by ~50% does not significantly change DrCNGK currents in oocytes ( Figure 5—figure supplement 3 , n = 4 ) . In sea urchin sperm , CNGK-mediated hyperpolarization leads to an increase of intracellular Ca2+ . Therefore , we tested , whether the NH4Cl-induced hyperpolarization ( Figure 5E ) and alkalization ( Figure 5G ) evokes a Ca2+ response . In fact , mixing of D . rerio sperm with 10 or 30 mM NH4Cl gave rise to a rapid Ca2+ signal ( Figure 5H , n = 4 ) . The time course of the pHi- and Ca2+ signal was similar , suggesting that the CNGK-mediated hyperpolarization triggers a Ca2+ influx . We conclude that DrCNGK represents a pH-sensitive channel that is strongly activated at alkaline pHi; the ensuing hyperpolarization , like in sea urchin sperm , produces a Ca2+ signal . 10 . 7554/eLife . 07624 . 017Figure 5 . pH regulation of the DrCNGK channel . ( A ) Whole-cell recordings from zebrafish sperm after perfusion with NH4Cl or propionic acid . Voltage steps as shown in Figure 1C . Recordings at extracellular pH 7 . 4 and pipette pH 6 . 4 ( left ) . NH4Cl ( 10 mM , middle ) or propionic acid ( 10 mM , right ) was added to the bath . ( B ) Pooled IV curves for recordings from zebrafish sperm at a pipette pHi of 6 . 4 and in the presence of 10 mM NH4Cl or 10 mM propionic acid ( PA ) . ( C ) Pooled IV curves from recordings of zebrafish sperm at different intracellular pHi . ( D ) Dependence of mean current ( ± sd ) on intracellular pHi ( circles , bottom axis ) or in the presence of either 10 mM propionic acid ( PA ) or different NH4Cl concentrations ( triangles , top axis ) . ( E ) Recording of the voltage signal of zebrafish sperm in the current-clamp configuration . Pipette solution with an intracellular pHi of 6 . 4; recording in the presence of 10 mM NH4Cl or 10 mM propionic acid ( PA ) . Left panel: single recording . Right panel: individual data ( symbols ) and mean ± sd ( gray bars ) , n = 10 . ( F ) Pooled IV curves of Two-Electrode Voltage-Clamp recordings from heterologously expressed DrCNGK channels and uninjected wild-type oocytes in 96 mM K+ bicarbonate solution ( black and red symbols ) or 96 mM K+ gluconate , including 1 mM NH4Cl ( white symbols , see Figure 5—figure supplement 1 for recordings ) . ( G ) Changes in fluorescence of a zebrafish sperm population incubated with the pH indicator BCECF , recorded as the ratio of fluorescence at 549/15 nm and 494/20 nm ( excited at 452/28 nm ) , before ( black ) and after the addition of 10 mM ( red ) or 30 mM ( green ) NH4Cl . ( H ) Stimulation of sperm with NH4Cl as in panel G using the Ca2+ indicator Cal-520 . Fluorescence was excited at 494/20 nm and recorded at 536/40 nm . Fluorescence F was normalized to the control value F0 before stimulation . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 01710 . 7554/eLife . 07624 . 018Figure 5—figure supplement 1 . pH dependence of heterologously expressed DrCNGK channels in oocytes . Two-Electrode Voltage-Clamp recordings of DrCNGK-injected ( A ) and uninjected ( B ) oocytes . Cells were reversibly perfused with 96 mM K+ bicarbonate followed by 1 mM NH4Cl added to K+ gluconate ( 10 min each ) . Voltage steps from -100 mV to + 30 mV from a holding potential of -60 mV . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 01810 . 7554/eLife . 07624 . 019Figure 5—figure supplement 2 . High intracellular Ca2+ does not suppress DrCNGK currents . Pooled IV relations of whole-cell recordings from whole sperm and isolated heads under standard conditions ( ES/IS , pH 7 . 4 ) ( Figure 1G ) and from whole sperm , when [Ca2+]i in the pipette solution was adjusted to 1 µM . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 01910 . 7554/eLife . 07624 . 020Figure 5—figure supplement 3 . Hypoosmotic conditions do not stimulate or diminish DrCNGK currents in Xenopus oocytes . DrCNGK-injected oocytes were recorded in the TEVC mode in ND96-7K and ND48-7K olutions ( n = 4 ) . No significant differences were observed . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 020 We studied the role of Ca2+ for motility of D . rerio sperm using photo-release of Ca2+ from caged Ca2+ ( NP-EGTA ) . Sperm motility was activated by hypoosmotic dilution ( 1:20 into 70 mM Na+ ES , 167 mOsm x L-1 ) and was followed under a dark-field microscope ( Figure 6 , Video 1 ) . Unstimulated sperm swam on curvilinear trajectories of low curvature ( Figure 6A-F , green segment , Video 1 ) . A UV flash almost instantaneously increased path curvature in NP-EGTA-loaded sperm ( Figure 6D-F , I ) , but not in control sperm ( Figure 6A-C , I ) ; sperm swam on much narrower arcs ( Figure 6D-F , red segment , Video 1 ) . The increase of path curvature was even more pronounced after a second UV flash ( Figure 6 D-F , cyan segment , I , Video 1 ) . Many cells were pushing against the wall of the observation chamber and performed a ‘spinning’ or ‘drilling’ behaviour , as if to penetrate the wall ( Figure 6G ) . The asymmetry of the flagellar beat increased with each consecutive photo-release of Ca2+ and eventually the flagellum pointed away from the glass surface ( Figure 6H , Video 1 ) . This swimming behaviour likely represents a strategy followed by sperm on its search for the micropyle , a small opening ( outer diameter about 8 µm in diameter ) ( Hart and Danovan , 1983 ) on the surface of the much larger egg ( about 0 . 75 mm in diameter ) ( Selman et al . , 1993 ) . A similar swimming behaviour during fertilization has been reported for sperm of herring and black flounder ( Cherr et al . , 2008; Yanagimachi et al . , 1992; 2013 ) . We propose that the spinning or drilling movements observed after Ca2+ release reflect the swimming behaviour in vivo down the narrow micropyle . 10 . 7554/eLife . 07624 . 021Figure 6 . Sperm swimming behaviour upon Ca2+ release . ( A ) , ( B ) , and ( C ) representative swimming paths of three different DMSO loaded sperm before and after application of UV light . ( D ) , ( E ) , and ( F ) representative averaged swimming paths of three different sperm before ( green ) and after Ca2+ release by one ( red ) or two ( cyan ) consecutive UV flashes ( black arrows ) . Curved blue arrows indicate the swimming direction of sperm . ( G ) Same swimming path shown in ( F ) including a temporal axis to facilitate the visualization of the changes in swimming path after consecutive flashes . Upon release ( black arrows ) , the curvature of the swimming path progressively increases and the cell finally spins around the same position . ( H ) Representative flagellar shapes before ( - ) , after Ca2+ release by one ( + ) or two consecutive flashes ( ++ ) , and during cell spinning against the wall ( bottom right ) . Consecutive frames every 100 ms are shown in different colours . Sequence order: red , green , blue , and yellow . ( I ) Mean curvature before ( - ) and after one ( + ) or two ( ++ ) UV flashes . Individual data ( symbols ) and mean ± sd ( gray bars ) , number of experiments in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 02110 . 7554/eLife . 07624 . 022Video 1 . Behavioural response of zebrafish sperm to successive Ca2+ release . Representative recording of zebrafish sperm loaded with NP-EGTA ( 40 µM ) . Upon release of Ca2+ , the swimming path curvature increases and , eventually , sperm spin against the wall of the recording chamber with their flagellum pointing away from the wall . Video recorded using dark-field microscopy at 30 frames per second using a 20x magnification objective . The field of view corresponds to 410 µm . The Video is shown in real time . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 022 A growing body of evidence reveals unexpected commonalities , but also notable differences among sperm from different species ( for review ( Darszon et al . , 2006; Kaupp et al . , 2008; Yoshida and Yoshida , 2011 ) . Organisms , as phylogenetically distant as sea urchins and humans , share the CatSper channel as a common site of Ca2+ entry into the sperm flagellum . By the same token , Slo3 K+ channel orthologues in mouse and human sperm evolved different selectivity for intracellular ligands and might serve different functions ( Brenker et al . , 2014; Chavez et al . , 2013; Lishko et al . , 2012; Santi et al . , 2009; Zeng et al . , 2013 ) . Here , we characterize a novel variant of CNGK channels in zebrafish sperm , whose key features depart from those of CNGK channels of marine invertebrates ( Figure 7 ) . 10 . 7554/eLife . 07624 . 023Figure 7 . Models of signalling pathways in sea urchin and zebrafish sperm . Sea urchin ( upper panel ) : Binding of the chemoattractant resact to a receptor guanylyl cyclase ( GC ) activates cGMP synthesis . Cyclic GMP opens K+-selective CNG channels ( CNGK ) , thereby , causing a hyperpolarization , which in turn activates a sperm-specific Na+/H+ exchanger ( sNHE ) that alkalizes the cell . Alkalization and subsequent depolarization by hyperpolarization-activated and cyclic nucleotide-gated ( HCN ) channels lead to the opening of sperm-specific CatSper channels . Zebrafish ( lower panel ) : Upon spawning , K+ efflux through CNGK hyperpolarizes sperm . An unknown mechanism of alkalization ( dashed lines ) modulates the open probability of CNGK channels; the ensuing hyperpolarization opens voltage-gated Ca2+ channels ( Cav ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07624 . 023 First , although the D . rerio CNGK carries four canonical CNBDs , it is gated by pHi rather than cyclic nucleotides , indicating that the CNBDs have lost their genuine ligand selectivity . The related ApCNGK channel from sea urchin sperm is also unique in that it displays an unusual cGMP dependence: Unlike “classic” cooperative CNG channels , it is gated by binding of a single cGMP molecule to the third CNBD , implying that the other CNBDs are non-functional ( Bönigk et al . , 2009 ) . Because the ApCNGK channel is activated through binding of cGMP to the third repeat , we searched for sequence alterations in the third repeat of the DrCNGK channel . Strikingly , in the C-linker region of the third repeat , we identified an insert of 42 amino-acid residues ( Figure 1—figure supplement 1 , blue ) that is absent in other cyclic nucleotide-regulated channels . The insert shows no sequence similarity to any known functional domain of ion channels . One hypothesis is that this insert prevents the transmission of the binding signal to the channel pore . Another sequence peculiarity is identified in the second repeat ( Figure 3—figure supplement 2 ) . At amino-acid position 934 , an Ala residue replaces a highly conserved Arg residue that is crucial for cyclic-nucleotide binding ( Kaupp and Seifert , 2002 ) . The CNBDs of repeat 1 and 4 do not show obvious sequence abnormalities and could represent bona fide CNBDs ( Figure 3—figure supplement 2 ) . In recent years , many structures of CNBDs have been solved ( Clayton et al . , 2004; Kesters et al . , 2015; Kim et al . , 2007; Rehmann et al . , 2003; Schünke et al . , 2011; Schünke et al . , 2009; Zagotta et al . , 2003 ) . Can we learn from these structures something about the DrCNGK channel ? Most of the CNBD structures feature a similar fold and are able to bind both cAMP and cGMP . However , for some CNG channels , ligands are full agonists , like cAMP and cGMP in the CNGA2 channel ( Dhallan et al . , 1990 ) , only partial agonists , like cAMP in the CNGA1 channel ( Altenhofen et al . , 1991 ) , or competitive antagonists , like cGMP in the bacterial SthK channel ( Brams et al . , 2014 ) . Finally , in HCN channels , CNBDs interact and form a so-called gating ring ( Zagotta et al . , 2003 ) , whereas in MloK1 channel , CNBDs do not interact at all ( Cukkemane et al . , 2007; Schünke et al . , 2009; 2011 ) . In conclusion , at present , no sequence features can be identified that unequivocally explain the lack of cyclic-nucleotide regulation of the DrCNGK channel . The insensitivity of the DrCNGK channel to cyclic nucleotides is , however , reminiscent of EAG and hERG channels that carry classic CNBDs , yet are not gated by cyclic nucleotides ( Brelidze et al . , 2009; 2010; 2012 ) . Instead , small molecules such as flavonoids have been suggested as ligands that bind to the CNBD and modulate channel activity ( Brelidze et al . , 2010; Carlson et al . , 2013 ) . Moreover , in the C-terminus of these CNBDs , a conserved segment of residues was identified that occupies the CNBD and serves as an intrinsic “ligand” ( Brelidze et al . , 2012; Carlson et al . , 2013 ) . We can only speculate , that , in addition to protons , as yet unidentified ligands might bind to and regulate the DrCNGK channel . The apparent pKa value for channel activation by pH was approximately 7 , suggesting that a His residue controls channel opening . There are a number of His residues in the C-linker of the four repeats that might serve as candidate sites . Future work is necessary to identify the site of pH regulation of the DrCNGK channel . To take on a new ligand selectivity or activation mechanism is also reminiscent of orthologues of the sperm-specific K+ channel Slo3 . Whereas the mouse Slo3 channel is exclusively controlled by pHi ( Brenker et al . , 2014; Schreiber et al . , 1998; Yang et al . , 2011; Zeng et al . , 2011; Zhang et al . , 2006a ) , the human Slo3 is primarily regulated by Ca2+ ( Brenker et al . , 2014 ) . In conclusion , the zebrafish CNGK is a striking example for a channel featuring a CNBD that is not gated by cyclic nucleotides . In general , CNBDs might represent sensor domains that can relay information on ligands other than cyclic nucleotides . Second , signalling pathways that control sperm motility are located to the flagellum: The GC receptor for chemoattractant binding in sea urchin ( Bönigk et al . , 2009; Pichlo et al . , 2014 ) , the CatSper channel in humans , mice , and sea urchin ( Chung et al . , 2014; Kirichok et al . , 2006; Seifert et al . , 2015 ) , the Slo3 K+ channel in mice and humans ( Brenker et al . , 2014; Navarro et al . , 2007 ) , and the HCN and the CNGK channel in sea urchin ( Bönigk et al . , 2009; Gauss et al . , 1998 ) . In contrast , the DrCNGK channel is located in the head rather than the flagellum . What might be the functional significance of such a peculiar location ? The CNGK channel probably serves two related functions . In seminal fluid , sperm of freshwater fish are immotile due to a high [K+] and high osmolarity . Upon release into hypoosmotic freshwater , sperm become motile for a few minutes ( Morisawa et al . , 1983; Takai and Morisawa , 1995; Wilson-Leedy et al . , 2009 ) . The osmolarity-induced activation hyperpolarizes sperm and induces a Ca2+ signal ( Krasznai et al . , 2000 ) . We propose that the CNGK triggers Ca2+ signalling events upon spawning: In the high-K+ seminal fluid , partially open CNGK channels keep sperm depolarized . When exposed to low-K+ hypoosmolar conditions , sperm hyperpolarize and , ultimately , Ca2+ is entering the cell and activates general motility ( Figure 7 ) . Moreover , during the search for the micropyle on the egg surface , the sense of direction might be provided by haptic interaction with tethered molecules that line the opening or the funnel of the micropyle ( Iwamatsu et al . , 1997; Ohta and Iwamatsu , 1983; Yanagimachi et al . , 2013 ) . The haptic interactions could directly control CNGK activity in the head . For example , near or inside the micropyle , the CNGK might become further activated by alkaline pH and initiate the Ca2+-dependent ‘drilling’ behaviour . On a final note , the study of zebrafish sperm provides insight into adaptive mechanisms of sperm evolution . Boundary conditions might constrain sperm to develop different signalling strategies for similar functions . One obvious constraint is the ionic milieu , which strongly affects ion channel function . In freshwater , ion concentrations are low and opening of Na+- , K+- , and non-selective cation channels would hyperpolarize rather than depolarize cells . We speculate that , in freshwater fish , a depolarization-activated Ca2+ channel like CatSper may not work and has been replaced by another Cav channel . Furthermore , the Thr/Val difference in ApCNGK versus DrCNGK , which determines Na+ blockage , probably represents an adaptation to the respective ionic milieu . Na+ blockage of sea urchin CNGK resists hyperpolarization in seawater and , thereby , facilitates the opening of depolarization-activated CatSper channels . The observation that CNGK channels from seawater organisms carry this Thr residue indicates a specific evolutionary pressure on this pore residue . Why is this Thr residue lost in CNGK channels of freshwater organisms ? We speculate that Na+ blockage disappeared along with the loss of CatSper genes and that Ca2+ ions enter fish sperm through a Ca2+ channel that is activated by hyperpolarization rather than depolarization . Future work needs to identify this Ca2+ channel , its mechanism of activation , and its role for fertilization of teleost fish . In summary , we identify a zebrafish CNGK channel that is activated at alkaline pH , and is set apart from its cousins of sea urchins that are activated by cGMP . Orthologues of CNGK also exist in the choanoflagellate S . rosetta , suggesting that this channel sub-family is phylogenetically ancient . Interestingly , this protozoon has a sexual life cycle: during anisogamous mating , small flagellated cells fuse with large cells ( Levin and King , 2013 ) . This mating behaviour represents the ancestor of sexual reproduction in animals ( Levin and King , 2013; Umen and Heitman , 2013 ) . The role of the S . rosetta CNGK channel for sexual reproduction without sperm will be interesting to study . Chemicals were purchased from AppliChem ( Darmstadt , Germany ) , Biozym ( Hessisch Oldendorf , Germany ) , Carl Roth ( Karlsruhe , Germany ) , GE Healthcare Life Sciences ( Munich , Germany ) , Life Technologies ( Carlsbad , CA ) , Merck KgAa ( Darmstadt , Germany ) , Merck Millipore ( Billerica , MA ) , PolyScience ( Warrington , PA ) , Qiagen ( Hilden , Germany ) , Serva ( Heidelberg , Germany ) , Sigma-Aldrich ( Steinheim , Germany ) , and Thermo Scientifica ( Waltham , MA ) . Enzymes and corresponding buffer solutions were ordered from Ambion ( Austin , TX ) , MBI Fermentas ( Vilnius , Lithuania ) , New England Biolabs ( Frankfurt on the Main , Germany ) , and Roche ( Basel , Switzerland ) . Primers were synthesized from Eurofins Genomics ( Ebersberg , Germany ) . Chemicals for mammalian cell culture were ordered from Carl Roth , Life Technologies , and Biochrom ( Berlin , Germany ) . CHOK1 and HEK293 cells were obtained from the American Type Culture Collection ( ATCC , Manassas , VA ) . We identified two putative annotated sequences in the database: Eb934551 and XM_0013354 . Eb934551 contained the putative N-terminal region and XM_0013354 contained the putative repeat 3 , repeat 4 , and parts of repeat 2 . Using four sets of primer pairs , we did nested PCR reactions on testis cDNA to obtain the full-length sequence: the primer pairs #4811/#4812 and #4813/#4814 were used for XM_0013354 and the primer pairs C0274/C0275 and C0276/C0277 for Eb934551 . The primer sequences were: TATTTCAAGTAGCTGTTACCG ( #4811 ) , ACATTCCCTTATAATAATGTCC ( #4812 ) , AAAAAAGCTAAGCTTTTCAGAAACACAG ( #4813 ) , AAAATCTGACAGGTACCCTGCAGAATGC ( #4814 ) , CATACAGGATGCATGACCCC ( C0274 ) , CCAGGAATGTATGTGTAGGTC ( C0275 ) , GAGGAATTCATGCATGACCCCAGAGAAATGAAG ( C0276 ) and CTCGGATCCGTATGTGTAGGTCTTTAATTTCAGGG ( C0277 ) . Due to failure of expression , we used a codon-optimized version ( human codon usage ) of the DrCNGK gene separated into three modules ( Eurofins Genomics ) . Each module was flanked by restriction sites . The first module contained bases 1 to 2 , 108 and was flanked on the 5’ end with BamHI and on the 3’ end with XbaI; the second module contained bases 2 , 109 to 4 , 655 and was flanked on the 5’ end with XbaI and on the 3’ end with EcoRI; the third module contained bases 4 , 656 to 6 , 360 and was flanked on the 5’ end with EcoRI and on the 3’ end with NotI . At the 3’ end , the coding sequence for the hemagglutinin tag ( HA-tag ) was added . The construct was cloned into the pcDNA3 . 1 vector ( Life technologies ) ( DrCNGK ) . To enhance expression levels , we added a QBI SP163 sequence ( Stein et al . , 1998 ) in front of the start codon ( QBI-DrCNGK ) . Moreover , we added the coding sequence for a flag-tag at the 5’ end of the DrCNGK gene . We performed two PCR reactions with primer pairs C0991/C0962 and C0417/C0990 and a recombinant PCR reaction on the resulting PCR products with primer pair C0417/C0962 . Primer sequences were: CCCGGACGGCCTCCGAAACCATGGACTACAAGGACGACGACGACAAGC ( C0991 ) , TTCAGACCGGCATTCCAAGCCC ( C0962 ) , CGCGGATCCAGCGCAGAGGCTTGGGGCAGC ( C0417 ) , GTCGTCGTCGTCCTTGTAGTCCATGGTTTCGGAGGCCGTCCGGG ( C0990 ) . ApCNGK pore mutants: for the pore mutant ApCNGK-4V , the following amino-acid substitutions in the ApCNGK wild-type channel ( Bönigk et al . , 2009 ) were produced: T252V , T801V and T1986V . Three PCR reactions were required for each mutation: two with the primers containing the point mutation and one recombinant PCR reaction . For the amino-acid exchange T252V , the primer pairs #4433/C0531 and C0530/#4409 were used and for the recombinant PCR , the primer pair #4433/#4409 . The PCR product was cloned into the ApCNGK gene with restriction enzymes BamHI and XhoI . For the amino-acid exchange T802V , primer pairs #4436/C0533 and C0532/#4439 were used and for the recombinant PCR the primer pair #4436/#4439 . The PCR product was cloned into the ApCNGK gene with restriction enzymes XhoI and XbaI . For the amino-acid exchange T1986V , the primer pairs #4412/C0535 and C0534/#4447 were used and for the recombinant PCR , the primer pair #4412/#4447 . The PCR product was cloned into the ApCNGK gene with restriction enzymes BamHI and XbaI . The primer sequences were: GGTTCTGCTCGAGATTCTGTAGG ( #4409 ) , CAACACCGGATCCGGTGAGAGCAGTG ( #4412 ) , AAAGTTGGGATCCAATACAGCG ( #4433 ) , TACAGAATCTCGAGCAGAACC ( #4436 ) , AAGTCTAGACGGTAGACTGATCGCCTGG ( #4439 ) , AAATCTAGATTAGGCATAATCGGGCACATCATAGGGATACACCACCGTTTGTCTCAGCG ( #4447 ) , GCCACCTCTGTAGGCTACGGAGAC ( C0530 ) , GTCTCCGTAGCCTACAGAGGTGGC ( C0531 ) , ATGACATCCGTGGGCTACGGAGAC ( C0532 ) , GTCTCCGTAGCCCACGGATGTCAT ( C0533 ) , CTGACCTCCGTTGGCTACGGTGACATC ( C0534 ) , GTCACCGTAGCCAACGGAGGTCAGAG ( C0535 ) . For expression in X . laevis oocytes , the DrCNGK and QBI-DrCNGK constructs were cloned into a modified version of the expression vector pGEMHEnew ( Liman et al . , 1992 ) ; because cloning of DrCNGK was only possible using BamHI and NotI , a NotI restriction site present in pGEMHEnew was removed and a new one was introduced into the multiple-cloning site . This new vector has been named pGEMHEnew-NotI . In vitro transcription to generate cRNA was performed using the mMESSAGEmMACHINEKit ( Ambion ) ; the plasmid was linearized with SpeI . For in situ hybridization , a short fragment of the DrCNGK gene coding for amino acids 1 , 085-1 , 219 was cloned into the pBluescript vector using PstI and HindIII . Nested PCR reactions were performed using for the 1st reaction the primer pair #4791/#4792 and for the 2nd reaction the primer pair #4793/#4794 . Primer sequences were: ATTTTGCCGTGGAGTCCATGG ( #4791 ) , AAGTCAATATTAAACGTTGCATCC ( #4792 ) , GGAAGCTTTCCGAAGCATTACAGCCG ( #4793 ) , GTTGGATCCAAGTGTGTCACCCATGAC ( #4794 ) . For the antisense probe , the plasmid was linearized with HindIII and transcribed by T7 RNA polymerase . RNA was labeled with Digoxigenin ( DIG RNA Labelling Mix , Roche ) . Animals were sacrificed according to the “Guidelines for housing and care , transport , and euthanasia of laboratory fishes” , ( “Empfehlung für die Haltung , den Transport und das tierschutzgerechte Töten von Versuchsfischen” , published by the Tierärztliche Vereinigung für Tierschutz e . V . January 2010 ) . To obtain intact sperm , zebrafish male were anesthetized with MS-222 ( 0 . 5 mM , 3 min ) . After a brief wash with fresh water , the head was quickly separated from the body . The body of the fish was ventrally opened and two testis strands were removed and transferred into ES buffer ( see Electrophysiology ) or phosphate-buffered saline ( PBS ) containing ( in mM: NaCl 137 , KCl 2 . 7 , Na2HPO4 6 . 5 , KH2PO4 1 . 5 , pH 7 . 4 ) , additionally containing 1 . 3 mM EDTA , mPIC protease inhibitory cocktail ( Sigma-Aldrich ) , and 1 mM DTT . Sperm were collected with a pipette tip . After 15 min , 80% of the supernatant was transferred into a fresh reaction tube . To separate heads from flagella , the sperm suspension was sheared 30-40 times on ice with a 24 gauge needle ( Braun , Bethlehem , PA ) . The sheared suspension was centrifuged for 10 min ( 800xg , 4°C ) to sediment intact sperm and sperm heads . This procedure was repeated twice . The purity of flagella preparations was assessed using dark-field microscopy ( Figure 1—figure supplement 2 ) . For testis sections , males were ventrally sliced and kept overnight in 4% paraformaldehyde . After 24 hr , testis strands were removed and embedded in paraffin . Sections ( 8 µm ) were made using a microtome ( Leica , Wetzlar , Germany ) . A rabbit polyclonal antibody produced by Peptide Specialty Laboratories ( PSL , Heidelberg , Germany ) was directed against the cytosolic loop C-terminal of the CNBD of the first repeat ( anti-repeat1 , amino acids 483 - 497 ) . Antibodies were purified with a peptide affinity column provided by PSL . Rat monoclonal antibody YENT1E2 ( anti-repeat3 ) was directed against the extracellular loop between S5 and the pore region of the third repeat ( amino acids 1 , 254 - 1 , 269 ) . Anti-α-tubulin ( mouse , B-5-1-2 , Sigma-Aldrich ) , anti-β-actin ( mouse , abcam , Cambridge , United Kingdom ) , anti-HA ( rat , 3F10 , Roche ) , anti-calnexin ( rabbit , abcam ) , and anti-flag-tag ( mouse , M2 , Sigma-Aldrich ) antibodies were used as controls . Sperm were immobilized on SuperFrost Plus microscope slides ( Thermo Fisher Scientific , Waltham , MA ) and fixed for 5 min with 4% paraformaldehyde . After preincubation with 0 . 5% Triton X-100 and 5% chemiblocker ( Merck Millipore ) in PBS , sperm were incubated for 1 hr with antibodies YENT1E2 ( 1:10 ) or anti-repeat1 ( 1:500 ) diluted in 5% chemiblocker ( Merck Millipore ) and 0 . 5% TritonX-100 in PBS ( pH 7 . 4 ) . Sperm were visualized with Cy3-conjugated secondary antibody ( Jackson ImmunoResearch Laboratories , West Grove , PA ) . For in situ hybridization , tissue was permeabilized with protein kinase K ( 1 µg/ml in 0 . 1 M Tris/HCl , pH 8 . 0 ) and hybridized using the DrCNGK-3 antisense probe . After washing , the antibody staining was performed using an anti-Digoxigenin antibody ( 1:500 , Roche ) conjugated with alkaline phosphatase . RNA was visualized with a mixture of nitro-blue tetrazolium chloride ( 500 µg ) and 5-bromo-4-chloro-3'-indolyphosphate p-toluidine salt ( 188 µg , Roche ) . Cross sections were covered with a glass slip . For antibody staining of the in situ hybridization sections , cover slips were removed keeping the slides 5 min in xylene and briefly in PBS . This step was repeated; the fixative was removed from the sections . Afterwards , sections were stained as described for sperm immunocytochemistry . Proteins were probed with antibodies: anti-repeat1 ( 1:500 ) or anti-repeat3 ( 1:10 ) and visualized with Cy3-conjugated secondary antibody ( Jackson ImmunoResearch Laboratories ) . For Western blotting , zebrafish tissue or cells heterologously expressing the DrCNGK channel were resuspended in PBS buffer containing 1 . 3 mM EDTA , mPIC protease inhibitor cocktail ( Sigma-Aldrich ) , and 1 mM DTT . For lysis , cells were triturated 20x with a cannula ( 24G , Braun ) and sonicated three times for 15 s . After a clearing spin ( 25 , 000xg , 30 min , 4°C ) , the pellet was resuspended and sonicated two times for 10 s in 200 mM NaCl , 50 mM Hepes ( pH 7 . 5 ) , mPIC , and 1 mM DTT . Triton X-100 was added to a final concentration of 1% . Proteins were solubilized for 1–2 hr at 4°C . A final clearing spin ( 10 , 000xg , 20 min , 4°C ) was performed . For Western blot analysis , proteins were separated using 4-12% NuPAGE gradient gels ( Life Technologies ) and transferred overnight ( 4°C , 12–15 V ) onto PVDF membranes ( Immobilion FL , Merck Millipore ) , using a Xcell SureLock minigel chamber ( Life Technologies ) . Membranes were incubated with Odyssey blocking buffer ( LI-COR Biosciences , Lincoln , NE ) . Proteins were probed with the following antibodies: anti-repeat1 ( 1:1 , 000 ) , anti-repeat3 ( 1:10 ) , anti-α-tubulin ( 1:2 , 000 ) , anti-β-actin ( 1:1 , 000 ) , anti-HA ( 1:1 , 000 ) , anti-calnexin ( 1:5 , 000 ) , and anti-flag-tag ( 1:200 ) . Proteins were visualized using IRDye800CW-conjugated secondary antibodies ( 1:10 , 000 , LI-COR Biosciences ) , IRDye680-conjugated secondary antibodies ( 1:10 , 000 , LI-COR Biosciences ) , or horseradish peroxidase-conjugated secondary antibodies ( 1:5 , 000 , Jackson ImmunoResearch Laboratories ) . Visualization took place either with a chemiluminescence detection system ( LAS-3000 Luminescent Image Analyzer , FUJIFILM , Life Science , Stamford , CT ) or with fluorescent secondary antibodies ( Odyssey infrared imaging system , Li-Cor Bioscience ) . The Novex Sharp pre-stained protein standard ( Life Technologies ) was used as molecular mass standard . We electrically recorded from intact zebrafish sperm and from isolated sperm heads using the patch-clamp technique in the whole-cell configuration . Recordings were accomplished within 4 hr after preparation . Seals between pipette and sperm were formed at the neck region in standard extracellular solution ( ES ) . The following pipette solutions were used: standard intracellular solution ( IS ) ( in mM ) : NaCl 10 , K+ aspartate 130 , MgCl2 2 , EGTA 1 , Na2ATP 2 , and Hepes 10 at pH 8 . 4 , 7 . 9 , 7 . 4 , 6 . 9 , or 6 . 4 adjusted with KOH; Cl--based IS ( in mM ) : NaCl 10 , KCl 130 , MgCl2 2 , EGTA 1 , Na2ATP 2 , and Hepes 10 at pH 7 . 4 adjusted with KOH . The following bath solutions were used: standard ES ( in mM ) : NaCl 140 , KCl 5 . 4 , MgCl2 1 , CaCl2 1 . 8 , glucose 10 , and Hepes 5 at pH 7 . 4 adjusted with NaOH; for K+-based ES solutions , the equivalent amount of Na+ was replaced by K+ ( concentrations are indicated in the Figure legends ) . Calculations of the free Ca2+ concentrations were carried out using the Maxchelator program ( http://maxchelator . stanford . edu/webmaxc/webmaxcE . htm ) assuming a residual Ca2+ concentration in water of 1 µM . At pH 6 . 4 , [Ca2+]i was 7 . 7 nM and at pH 7 . 4 , it was 91 pM . To obtain an intracellular solution with 1 µM free Ca2+ , 1 mM CaCl2 was added to the IS solution at pH 7 . 4 . Caged compounds ( 100 µM BCMACM-caged cAMP or 100 µM BCMACM-caged cGMP ) were added to the IS . The final concentration of DMSO was 0 . 1% . The compounds were photolyzed by a ~1 ms flash of ultraviolet light from a Xenon flash lamp ( JML-C2; Rapp OptoElectronic , Wedel , Germany ) . The flash was passed through a BP295-395 nm filter ( Rapp OptoElectronic ) and delivered to the patch-clamp chamber in the microscope by a liquid light guide . Pipette resistance in IS/ES was between 11 . 5 and 15 . 0 MΩ . Voltages were corrected for liquid junction potentials . For functional studies in X . laevis oocytes , 50 nl DrCNGK RNA ( 0 . 3 , 0 . 4 , and 0 . 6 µg/µl ) per oocyte were injected . Oocytes were purchased from EcoCyte Bioscience ( Castrop-Rauxel , Germany ) or prepared from dissected animals . Briefly , frogs were anesthetized with MS-222 ( 0 . 5% , 10-20 min ) , follicles were removed , opened with forceps and washed several times with ND96 solution . For defolliculation , oocytes were transferred for 1–2 hr ( RT ) into Ca2+-free OR-2 solution containing 3 mg/ml collagenase type IV ( Worthingthon Biochemical Corp . , Lakewood , NJ ) . Defolliculated oocytes were stored in ND96 solution containing ( in mM ) : NaCl 96 , KCl 2 , MgCl2 1 , CaCl2 1 . 8 , Hepes 10 at pH 7 . 6 , pyruvate 2 . 5 , and gentamycin 1 . The OR-2 solution contained ( in mM ) : NaCl 82 . 5 , KCl 2 . 5 , MgCl2 1 , Hepes 5 at pH 7 . 6 . We recorded in the Two-Electrode Voltage-Clamp configuration . Most data were recorded with a Dagan Clampator One ( CA-1B , Dagan , Minneapolis , MN ) amplifier and digitized with Digidata 1320A ( Axon Instruments , Molecular Devices , Sunnyvale , CA ) . Analogue signals were sampled at 2 kHz . The holding potential was -80 mV . Pipette solution: 3 M KCl . Bath solutions were ND96-7K ( in mM ) : NaCl 96 , KCl 7 , MgCl2 1 , CaCl2 1 . 8 , Hepes 10 at pH 7 . 4 adjusted with NaOH; K+- based solution K96-7Na ( in mM ) : NaCl 7 , KCl 96 , MgCl2 1 , CaCl2 1 . 8 , Hepes 10 at pH 7 . 4 adjusted with KOH . Recordings with reduced osmolarity were carried out in ND48-7K solution ( in mM ) : NaCl 48 , KCl 7 , CaCl2 1 . 8 , MgCl2 1 , HEPES 10 at pH 7 . 4 adjusted with NaOH . Pipette resistance of voltage electrodes ranged between 1 . 5 and 3 . 0 MΩ and of current electrodes between 0 . 5 and 1 . 5 MΩ . Different analogues of cyclic nucleotides were added to the bath solution as indicated . Oocytes recordings with bicarbonate-based solutions were performed at Stanford University . Data were recorded with an OC-725C amplifier ( Warner Instruments , Hamden , CT ) using Patchmaster ( HEKA Elektronik , Lambrecht , Germany ) as acquisition software . Analogue signals were sampled at 1 kHz . The holding potential was -60 mV . Pipette solutions and pipette resistance as described above . Bath solutions: K+ bicarbonate-based solution ( in mM ) : NaCl 7 , K-bicarbonate 96 , MgCl2 1 , CaCl2 1 . 8 , Hepes 5 at pH 7 . 65 . Solution was made fresh on each day of recording; K+ gluconate-based solution ( in mM ) : NaCl 7 , K-gluconate 96 , MgCl2 1 , CaCl2 1 . 8 , Hepes 5 at pH 7 . 65 adjusted with KOH . NH4Cl was dissolved in K+gluconate-based solution . We recorded ApCNGK and mutant ApCNGK currents from transfected ( Lipofectamine 2000 , Life technologies ) HEK293 cells with the patch-clamp technique in the whole-cell configuration . A HEK293 cell line stably expressing the ApCNGK channel was used for inside-out recordings . The pipette solution for whole-cell recordings was standard IS . Channels were activated with 100 µM cGMP . The pipette solution for inside-out recordings was standard ES . The following bath solutions were used for inside-out recordings: IS-30 NMDG-0Na+ solution ( in mM ) : NaCl 0 , NMDG 30 , KCl 110 , EGTA 0 . 1 , Hepes 10 at pH 7 . 4 adjusted with KOH; IS-NMDG-30Na+ solution ( in mM ) : NaCl 30 , KCl 110 , EGTA 0 . 1 , Hepes 10 at pH 7 . 4 adjusted with KOH . 30 NMDG-0Na+ and 0 NMDG-30Na+ solutions were mixed to obtain the desired Na+ concentrations . 100 µM Na+-cGMP was added to the bath solution . For the solution with 0 mM Na+ , we used 100 µM Na+-free cGMP . Pipette resistance in IS/ES was between 4 . 0 and 7 . 0 MΩ . We measured changes in [Ca2+]i , and pHi in a rapid-mixing device ( SFM-4000; BioLogic , Claix , France ) in the stopped-flow mode using the Ca2+ indicator Cal-520-AM ( AAT Bioquest , Sunnyvale , CA ) or the pH indicator BCECF-AM ( Life Technologies ) . All sperm from a zebrafish male were diluted into 100 µl of ES solution and incubated with either 10 µM Cal-520-AM and 0 . 5% Pluronic for 120–180 min or 10 µM BCECF-AM for 10 min . Sperm were washed once , diluted 1:20 into ES solution , and loaded into the stopped-flow device . The sperm suspension was rapidly mixed 1:1 ( vol/vol ) with control ES solution or with ES solution containing NH4Cl to obtain final concentrations of 10 mM and 30 mM after mixing . Fluorescence was excited by a SpectraX Light Engine ( Lumencor , Beaverton , OR ) . Cal-520 was excited with a 494/20 nm ( Semrock , Rochester , NY ) , BCECF with a 452/45 nm ( Semrock ) excitation filter . Emission was recorded by photomultiplier modules ( H9656-20; Hamamatsu Photonics ) . Fluorescence of Cal-520 was recorded using a 536/40 nm ( Semrock ) emission filter and normalized ( without background subtraction ) to the value before stimulation . BCECF fluorescence was recorded in the dual emission mode using a 494/20 nm ( Semrock ) and a 549/15 nm ( Semrock ) emission filter . The pHi signals represent the ratio of F494/549 and were normalized ( without background subtraction ) to the value before stimulation . All stopped-flow traces represent the average of 3–6 recordings . The signals were normalized to the first 5-10 data points before the onset of the signal to yield △F/F and △R/R , respectively . Proteins of whole sperm , isolated heads , or flagella were resuspended in an SDS sample buffer and loaded on a SDS gel; after proteins had migrated approximately 1 cm into the separation gel , the gel was stained with Coomassie . The single gel band was excised for every sample , and proteins were in-gel digested with trypsin ( Promega , Sunnyvale , CA ) ; peptides were separated by RP-LC ( 180 min gradient 2–85% acetonitrile , ( Thermo Fisher Scientific ) ) using a nanoAcquity LC System ( Waters , Milford , MA ) equipped with a HSS T3 analytical column ( 1 . 8 µm particle , 75 µm x 150 mm ) ( Waters ) and analyzed twice by ESI-LC-MS/MS , using an LTQ Orbitrap Elite mass spectrometer ( Thermo Fisher Scientific ) with a 300-2 , 000 m/z survey scan at 240 , 000 resolution , and parallel CID of the 20 most intense precursors from most to least intense ( top20 ) and from least to most intense ( bottom20 ) with 60 s dynamic exclusion . All database searches were performed using SEQUEST and MS Amanda algorithm ( Dorfer et al . , 2014 ) , embedded in Proteome Discoverer ( Rev . 1 . 4 , Thermo Electron 2008-2011 , Thermo Fisher Scientific ) , with both a NCBI ( 26 , 623 entries , accessed December 20th , 2010 ) and a Uniprot ( 40 , 895 entries , accessed April 24th , 2014 ) zebrafish sequence protein database , both supplemented with the DrCNGK protein sequence ( Figure 1—figure supplement 2 ) . Only peptides originating from protein cleavage after lysine and arginine with up to two missed cleavages were accepted . Oxidation of methionine was permitted as variable modification . The mass tolerance for precursor ions was set to 8 ppm; the mass tolerance for fragment ions was set to 0 . 6 amu . For filtering of search results and identification of DrCNGK , a peptide FDR threshold of 0 . 01 ( q-value ) according to Percolator ( Käll et al . , 2007 ) two unique peptides per protein and peptides with search result rank 1 were required . Alignments for the calculation of the phylogenetic tree were done with ClustalOmega . Tree was depicted with Tree view ( Page , 1996 ) . The following ion channel sequences were used for the phylogenetic tree: CNGK channels from zebrafish ( Danio rerio , DrCNGK , XP_001335499 . 5 ) ; rainbow trout ( Oncorhynchus mykiss , OmCNGK , CDQ79437 . 1 ) ; spotted gar ( Lepisosteus oculatus , LoCNGK , W5MTF2 ) ; West Indian ocean coelacanth ( Latimeria chalumnae , LcCNGK , H3BE11 ) ; sea urchin ( Arbacia punctulata , ApCNGK ) ; acorn worm ( Saccoglossus kowalevskii , SkCNGK , XP_002731383 . 1 ) ; amphioxus ( Branchiostoma floridae , BfCNGK , XP_002592428 . 1 ) ; starlet sea anemone ( Nematostella vectensis , NvCNGK , XP_001627832 ) ; vasa tunicate ( Ciona intestinalis , CiCNGK , XP_002123955 ) ; sponge ( Amphimedon queenslandica , AqCNGK , I1G982 ) ; choanoflagellate ( Salpingoeca rosetta , SrCNGK , XP_004992545 . 1 ) ; murine HCN channel 1 ( Mus musculus , mHCN1 , NP_034538 ) , 2 ( mHCN2 , NP_032252 ) , 3 ( mHCN3 , NP_032253 . 1 ) , and 4 ( mHCN4 , NP_001074661 ) , and the HCN channel from sea urchin ( Strongylocentrotus purpuratus , SpHCN1 , NP_999729 ) ; rat cyclic nucleotide-gated channels CNGA1 ( Rattus rattus , rCNGA1 , NP_445949 ) , A2 ( rCNGA2 , NP_037060 ) , A3 ( rCNGA3 , NP_445947 . 1 ) , and A4 ( rCNGA4 , Q64359 ) ; the KCNH channels from fruit fly ( Drosophila melanogaster , DmEAG , AAA28495 ) and human ( Homo sapiens , hERG , BAA37096 . 1 ) ; murine voltage-gated Nav channels ( Mus musculus , mNav 1 . 1 , NP_061203 and mNav 1 . 6 , NM_001077499 . 2 ) ; murine voltage-gated Cav channels ( Mus musculus , mCav1 . 1 , NP_055008 , mCav2 . 3 , NP_033912 . 2 , and mCav3 . 1 , NP_033913 . 2 ) ; and voltage-gated Kv channels from fruit fly ( Drosophila melanogaster , DmShaker , CAA29917 . 1 ) and mouse ( Mus musculus , mKv3 . 1 , NM_001112739 . 1 ) . Sperm were loaded for 45 min at room temperature with either 30 µM DEACM-caged cAMP , 30 µM DEACM-caged cGMP , or 40 µM caged Ca2+ ( NP-EGTA , Life Technologies ) . A UV light-emitting diode ( 365-nm LED; M365L2-C , Thorlabs , Newton , NJ ) was used for photolysis of caged compounds . Experiments using caged Ca2+ and DEACM-caged nucleotides were carried out using a UV power of 25 and 22 mW , respectively . Flash duration was 300 ms . Pluronic ( 0 . 5% ) was added to the incubation solution . Sperm were kept quiescent during incubation in ES solution ( 292 mOsm x L-1 ) . Swimming was initiated by a hypoosmotic shock diluting sperm 1:20 in an activation solution containing ( in mM ) : NaCl 70 , KCl 5 . 4 , MgCl2 1 , CaCl2 1 . 8 , glucose 10 , and Hepes 5 at pH 7 . 4 adjusted with NaOH ( 167 mOsm x L-1 ) . Swimming behaviour was observed with a dark-field condenser in an inverse microscope ( IX71 , Olympus , Tokio , Japan ) with 20x magnification ( UPLSAPO , NA 0 . 75 ) . Movies were recorded at 30 Hz using a back-illuminated electron-multiplying charge-coupled device camera ( DU-897D; Andor Technology , Belfast , United Kingdom ) . Sperm trajectories were tracked using custom-made software written in MATLAB ( Mathworks ) . The software can be made available upon request . The average swimming path ( ASP ) was calculated by filtering the tracked coordinates with a second degree Savitzky-Golay filter with a 200 ms span . The curvature ( κ ) of the swimming path was calculated using the formula:bbb κ=x˙y¨−y˙x¨x˙2+y˙23/2 , where x and y are the coordinates of the ASP . To assess cAMP loading and release , we recorded the fluorescence increase due to DEACM-OH release after photolysis of DEACM-caged cAMP ( Bönigk et al . 2009 ) . Release and fluorescence excitation was achieved simultaneously using the same UV LED ( power 1 . 75 mW ) . Light was coupled to the microscope with a dichroic mirror ( 455DRLP , XF2034 , Omega Optical , Brattleboro , VT ) and fluorescence was long-pass filtered ( 460ALP; XF309; Omega Optical ) . Single cells were recorded at 50 Hz . Data analysis: Statistical analysis and fitting of data were performed , unless otherwise stated , using Sigma Plot 11 . 0 , GraphPadPrism 5 , or Clampfit 10 . 2 ( Molecular Devices ) . All data are given as mean ± standard deviation ( number of experiments ) . Note: All cell lines used in this study will be sent for STR profiling . Mycoplasma testing was performed using the Promokine Mycoplasma Test Kit 1/C ( PromoCell GmbH , Heidelberg , Germany ) . Results of this test can be supplied upon request .	Mammalian sperm cells fertilize egg cells inside the female’s body; while for fish and other marine animals it is common for fertilization to take place outside in the environment . In general , sperm cells become attracted to egg cells by various chemical or physical signals . Sperm detect these cues and generate electrical signals that control their own movements and eventually guide sperm to the egg . In 2009 , researchers identified a potassium ion channel , called CNGK , that starts the electrical signal in the sperm cells of sea urchins . This channel is activated by signalling molecules inside cells , called ‘cyclic nucleotides’ , and its activity ultimately leads to calcium ions flowing into the sperm cell’s tail . This influx of calcium ions in turn controls the beating of the tail and , thereby , steers the sperm cell towards the egg . It was previously thought that this CNGK channel is found only in animals without a backbone ( i . e . in invertebrates ) . However , Fechner et al . – including some of the researchers involved in the 2009 work – now report that the CNGK channel also exists in the sperm cells of a freshwater fish , the zebrafish . Unexpectedly , the CNGK channel is located in the heads of this fish’s sperm cells rather than in the tails . Electrophysiological experiments then revealed that the fish version of CNGK is not activated by cyclic nucleotides , but is activated when the inside of the cell becomes more alkaline . In zebrafish sperm , a more alkaline pH inside the cell causes calcium ions to flow in and this influx of calcium ions triggers a unique spinning-like swimming movement that is different from the swimming of other sperm from other species . Fechner et al . suspect that this unusual swimming behaviour guides sperm through a small hole in the protective coating of fish eggs , which eventually leads to fertilization . These findings suggest that while channel proteins found in sperm cells from different species look similar and serve similar roles , they are activated in ways that can be very different .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	A K+-selective CNG channel orchestrates Ca2+ signalling in zebrafish sperm
Mechanoelectrical transduction is a cellular signalling pathway where physical stimuli are converted into electro-chemical signals by mechanically activated ion channels . We describe here the presence of mechanically activated currents in melanoma cells that are dependent on TMEM87a , which we have renamed Elkin1 . Heterologous expression of this protein in PIEZO1-deficient cells , that exhibit no baseline mechanosensitivity , is sufficient to reconstitute mechanically activated currents . Melanoma cells lacking functional Elkin1 exhibit defective mechanoelectrical transduction , decreased motility and increased dissociation from organotypic spheroids . By analysing cell adhesion properties , we demonstrate that Elkin1 deletion is associated with increased cell-substrate adhesion and decreased homotypic cell-cell adhesion strength . We therefore conclude that Elkin1 supports a PIEZO1-independent mechanoelectrical transduction pathway and modulates cellular adhesions and regulates melanoma cell migration and cell-cell interactions . Cells sense and respond to their physical surroundings by converting mechanical inputs into biochemical signals , a process referred to as mechanotransduction . The most rapid mode of mechanotransduction is mediated by mechanically activated ( MA ) ion channels that are activated within milliseconds of a stimulus , resulting in a localised flow of ions across the plasma membrane , thus converting physical stimuli into electrical signals . The discovery of the mammalian MA channels , PIEZO1 and PIEZO2 ( Coste et al . , 2012; Coste et al . , 2010 ) , has highlighted the diverse array of cells and tissues that express such ionotropic force sensors ( Albuisson et al . , 2013; Florez-Paz et al . , 2016; Hung et al . , 2016; Maksimovic et al . , 2014; Martins et al . , 2016; Miyamoto et al . , 2014; Ranade et al . , 2014; Rocio Servin-Vences et al . , 2017; Yang et al . , 2016 ) . The activation of these channels by externally applied mechanical stimuli underpins our senses of touch , proprioception and hearing ( Florez-Paz et al . , 2016; Maksimovic et al . , 2014 ) , and contributes to the physiological function of mechanoresponsive tissues such as the vasculature ( Evans et al . , 2018; Li et al . , 2014 ) , the urothelium ( Martins et al . , 2016; Miyamoto et al . , 2014 ) and the cartilage ( Rocio Servin-Vences et al . , 2017 ) . Recent evidence has suggested that PIEZO1 is not only activated by exogenously applied forces but can also be activated by cell-generated forces ( Blumenthal et al . , 2014; Ellefsen et al . , 2019; Nourse and Pathak , 2017 ) . Thus , MA channels may be involved in inside-out mechanical signalling and contribute to the ability of cells to probe the physical nature of their microenvironment . Because PIEZO channels do not account for all mammalian MA channel activity , identifying additional MA channels and characterising their function represent important outstanding challenges in the field ( Dubin et al . , 2017 ) . Inside-out mechanical signalling has been implicated in the development and metastasis of cancers . The stiffening of the local microenvironment ( Kai et al . , 2016 ) , changes in cell packing and cell-cell adhesions ( Cichon et al . , 2015 ) and the tumour topology ( Lee et al . , 2016a ) have all been shown to impact tumourigenicity or invasiveness . In the case of melanoma , cancer progression/transformation is directly linked to mechanical changes at a cellular level: cells become more compliant ( Jonas et al . , 2011 ) , exhibit increased contractility ( Paszek et al . , 2005; Sanz-Moreno et al . , 2011 ) and display altered morphology during transformation ( Poole and Müller , 2005 ) . During the process of metastasis , cells must break away from the primary tumour and navigate microenvironments with diverse physical properties including: confined pores in the ECM , the planar interface of the basement membrane , or tracks and channels formed from ECM fibres or the earlier passage of cancerous cells ( Wolf et al . , 2009 ) . Given the mechanical changes during melanoma development and metastasis it is important to characterise mechanotransduction molecules in these cells , including MA ion channels . We report the identification of MA ion channel activity at the cell-substrate interface in WM266-4 metastatic melanoma cells , using elastomeric pillar arrays to apply fine mechanical stimuli to cells via their connections to the subjacent matrix ( Poole et al . , 2014; Sianati et al . , 2019 ) , thus simulating deflections arising from cell-generated forces . Simultaneous recordings using whole-cell patch-clamp electrophysiology allowed us to directly measure the resulting MA currents . We have identified a crucial molecule ( TMEM87a ) required for this channel activity and have renamed this polypeptide Elkin1 , from the Greek word Elko , meaning ‘to pull’ . The deletion of Elkin1 resulted in altered cell migration and increased interaction forces between melanoma cells and laminin 511 ( LM511 ) , a functionally important extracellular matrix ( ECM ) molecule . In addition , Elkin1 deletion modulated cell-cell interactions , leading to facilitated dissociation of Elkin1-KO cells from organotypic spheroids . To establish whether melanoma cells exhibit MA channel activity , metastatic WM266-4 melanoma cells ( originally isolated from a secondary tumour ) were cultured on uncoated pillar arrays made of polydimethylsiloxane ( PDMS ) . Mechanical stimuli were applied directly to cell-substrate contact points by physically deflecting a single pilus subjacent to the cell ( Figure 1A , B ) and the electrical response of the cell was monitored using whole-cell patch-clamp . Deflection-activated currents were measured in all WM266-4 cells ( 10/10 ) and the current amplitude increased with increasing stimulus size ( Figure 1C ) . Variable inactivation kinetics were measured ( Figure 1—figure supplement 1 , Figure 1—source data 1 ) , as also previously demonstrated for PIEZO1-mediated currents activated by substrate deflection ( Poole et al . , 2014; Sianati et al . , 2019 ) . The reversal potential , as determined from a current-voltage relationship for the peak MA current , was +6 . 6 mV , indicating that the underlying current was passed by a non-selective cation channel ( Figure 1D ) . Stimulus-response plots were generated by calculating the precise pillar deflection for each applied stimulus ( Figure 1E ) . We additionally tested whether MA currents were activated in WM115 melanoma cells ( isolated from the primary tumour from the same patient as WM266-4 ) . Larger deflections were required to activate currents in WM115 cells , compared to WM266-4 ( Figure 1—figure supplement 1 ) . These data demonstrate that displacements at the interface between melanoma cells and their substrate evoke MA currents . To examine whether MA channel activity is correlated to cell migration speeds , we performed pillar-array experiments using LM511 , a substrate that supports the migration of metastatic melanoma cells ( Oikawa et al . , 2011 ) . We first confirmed that LM511 promotes increased migration of WM266-4 cells using transwell assays , compared to other LM isoforms ( Figure 1F ) . We then repeated the analysis of MA channel activity in WM266-4 cells cultured on pillar arrays coated with LM511 and noted that current kinetics were unchanged ( Figure 1—source data 1 ) . However , MA currents were more sensitive when evoked in WM266-4 cells attached to LM511 ( Figure 1E ) . Under these conditions , current saturation occurred within the stimulus range , allowing us to use a Boltzmann sigmoidal fit to determine the MA current sensitivity . Half-maximal activation of MA currents was seen with approximately 18 nm of substrate deflection ( Effective deflection ED50; standard error = 20 . 5 nm ) . These data indicate a correlation between migratory properties and the MA current sensitivity to deflections applied at cell-substrate contact points . The robust MA current activation observed in cells cultured on LM511 also provided an excellent system to investigate the molecules required for this mechanoelectrical transduction . PIEZO1 is an obvious candidate for mediating this activity , in particular because the biophysical characteristics of the observed deflection-activated currents are consistent with those previously described for this channel ( Poole et al . , 2014 ) . To directly test if PIEZO1 mediates the MA current in WM266-4 cells , we knocked down PIEZO1 expression and examined whether these currents were still detectable . In PIEZO1 knock-down cells cultured on LM511-coated pillar arrays , MA currents were activated in response to pillar deflection in 10/10 cells measured and the resulting stimulus-response curves were similar to controls ( Figure 1—figure supplement 2 ) . Similarly , treatment with Ruthenium Red ( RR , a channel blocker that inhibits PIEZO1 and TRP channels ) did not inhibit these MA currents ( Figure 1—figure supplement 2 ) . From these data we conclude that PIEZO1 is unlikely to be the MA channel responsible for the current measured in WM266-4 cells . To identify novel MA channels in WM266-4 cells , we undertook a proteomic-based strategy . Given that both intracellular ( Poole et al . , 2014; Zhang et al . , 2017 ) and extracellular proteins ( Chiang et al . , 2011 ) can tune the sensitivity of MA channels , we analysed the proteome of WM266-4 cells rather than taking a comparative proteomic approach ( Supplementary file 1 ) . Two known non-selective cation channels were identified , PIEZO1 and TRPV2 . However , RR is a channel blocker of both PIEZO1 ( Coste et al . , 2012 ) and TRPV2 ( Caterina et al . , 1999 ) , indicating that neither likely mediates the deflection-evoked currents in WM266-4 cells ( Figure 1—figure supplement 2 ) . We then examined the proteomics data for proteins of unknown function with four or more predicted transmembrane ( TM ) domains . We prioritised the investigation of Elkin1 due to its expression in melanoma cells but not healthy melanocytes , its expression in additional mechanosensitive cells ( Alveolar Type II cells ) and its upregulation in additional human cancers ( Human Protein Atlas [Uhlén et al . , 2005] available from www . proteinatlas . org ) . We generated miRNA constructs targeting Elkin1 and found that knockdown of Elkin1 transcript resulted in a dramatic reduction in MA currents to deflections up to 1000 nm ( Figure 2A , B ) . These data suggested that Elkin1 contributes to MA currents in melanoma cells . Three human isoforms ( representing splice variants ) of Elkin1 have been identified: isoforms 1 and 3 ( 555 and 494 aa respectively ) , contain six predicted TM domains ( Figure 2C ) . Isoform 2 ( 181 aa ) does not contain any predicted TM domains and was not examined in this study . We cloned hsElkin1-iso1 and hsElkin1-iso3 from WM266-4 cDNA and generated C-terminal GFP fusion constructs . We confirmed the plasma membrane localisation of these two isoforms in transiently transfected HEK-293T cells using cell-surface biotinylation followed by Western blot analysis ( Figure 2D , Figure 2—figure supplement 1 ) . Both isoforms were present at the plasma membrane , as well as the intracellular fraction . To further examine the subcellular distribution of hsElkin1 , we transiently expressed these plasmids in WM266-4 cells and imaged the cells using total internal reflection fluorescence ( TIRF ) microscopy , which limits fluorescence excitation to molecules near the cell-substrate interface . hsElkin1-iso1-GFP localised to both the membrane and a dispersed population of dynamic foci , while hsElkin1-iso3-GFP assembled into structures that were co-labelled by Lifeact-mCherry . ( Figure 2E–J; Figure 2—videos 1 and 2 ) . Laser-scanning confocal imaging to visualise hsElkin1-iso1-GFP throughout the cell revealed that this protein also localised to the Golgi apparatus ( as previously described [Hirata et al . , 2015] ) : in contrast , hsElkin1-iso3-GFP was enriched in actin-based ruffles that were present at the cell periphery ( Figure 2—figure supplement 2 ) . The presence of Elkin1 in the plasma membrane , as described here , is consistent with Elkin1 forming an integral component of a mechanoelectrical transduction pathway . To test the hypothesis that Elkin1 contributes to mechanoelectrical transduction , we examined whether its expression in a heterologous cell system is sufficient to reconstitute MA currents . Un-tagged Elkin1 isoforms were overexpressed in the HEK-293T P1KO cell line ( Lukacs et al . , 2015 ) , which lacks functional PIEZO1 . The control cells exhibited no current activation in response to stimuli within our deflection range of 1–1000 nm ( 0/8 cells ) . However , mechanically activated currents were detected in response to similar stimuli in HEK-293T P1KO cells expressing either hsElkin1-iso1 or hsElkin1-iso3 ( 8/8 and 9/9 respectively ) ( Figure 3A , B ) . The observed current kinetics were consistent with direct mechanical activation ( Figure 3—source data 1 ) and the mechanically evoked currents measured in WM266-4 cells ( Figure 1—source data 1 ) . We confirmed that we could mechanically evoke currents in a second cell line lacking PIEZO1 , N2a Piezo1-/- ( Moroni et al . , 2018 ) , where 6/8 cells expressing hsElkin1-iso1 responded to pillar deflection compared to 4/10 control cells ( Figure 3—figure supplement 1 ) . To investigate whether cells expressing Elkin1 were sensitive to alternative modes of mechanical stimuli , we used cellular indentation and high-speed pressure-clamp ( HSPC ) . The HEK-293T P1KO cell line is unresponsive to indentation , ( 0/6 cells responding ) as previously described ( Dubin et al . , 2017; Figure 3C , D ) ; in contrast , MA currents were evoked in cells expressing either human Elkin1 isoform ( hsElkin1-iso1 6/6 cells , hsElkin1-iso3 5/5 cells responding ) ( Figure 3C , D ) . Indentation-evoked currents were only measured in a small fraction of N2a Piezo1-/- cells expressing hsElkin1-ios1 or hsElkin1-iso3 ( 1/10 and 1/12 responding respectively compared to 0/8 control cells , Figure 3—figure supplement 1 ) . When membrane stretch was applied using high-speed pressure-clamp ( HSPC ) , no currents were evoked in negative controls or HEK-293T cells overexpressing human Elkin1 ( negative pressure: hsElkin1-iso1 0/11 , hsElkin1-iso3 0/13 , positive pressure: hsElkin1-iso1 0/5 , hsElkin1-iso3 0/5 responding ) ( Figure 3E ) . In contrast , both positive and negative pressure evoked currents in all positive controls where PIEZO1 was expressed ( Figure 3—figure supplement 2 ) . From these data we conclude that the expression of Elkin1 is sufficient to confer MA channel activity to substrate deflection and cell indentation stimuli in a heterologous system lacking endogenous PIEZO channels , but not to membrane stretch . The Mus musculus homolog of Elkin1-iso1 shares 93% sequence identity at the protein level with the human variant . However , we did not observe robust MA currents in HEK-293T P1KO cells following expression of mouse mmElkin1-iso1 ( 5/8 cells responding , Figure 4A , B ) , indicating a difference in sensitivity between these homologous proteins . These data were confirmed in the N2a Piezo1-/- background ( Figure 3—figure supplement 1 ) , with 11/26 cells responding . Alignments between the protein sequences ( Corpet , 1988 ) revealed that the divergence in amino acid sequences primarily resides in the N-terminus of the proteins ( Figure 4—figure supplement 1 ) . However , expression of N-terminal truncation mutants of hsElkin1 conferred similar MA currents to those of the wild type ( WT ) protein ( Figure 4—figure supplement 1 ) , suggesting that this region is not required for MA currents . To further characterise the differences between the mouse and human proteins , we focused on two non-conservative changes within the region containing the six predicted TM domains . Introducing the two human residues into the mouse polypeptide ( mmElkin1-iso1-F271L-N292G ) , led to robust MA channel activity , in contrast with WT mmElkin1-iso1 . Conversely , introducing either of the mouse residues into the human polypeptide ( hsElkin1-iso3-L210F or hsElkin1-iso3-G231N ) led to a reduction in the amplitude of the MA currents measured ( Figure 4A , B ) , despite the fact that these variants were still present in the plasma membrane ( Figure 4—figure supplement 2 ) . These data indicate that two residues can account for the differences in activity between the mouse and human proteins , and that these two residues are required for the activation of robust Elkin1-dependent currents by substrate deflection . Having demonstrated that human Elkin1 is involved in mechanoelectrical transduction in melanoma cells , we sought to investigate whether the presence of this protein influences cell migration . To address this question , we used CRISPR/Cas9 to gene-edit cells such that no functional Elkin1-iso1/3 were expressed ( Figure 5—figure supplement 1 ) . Duplicate clonal populations of wild-type ( WT ) and Elkin1-knockout ( KO ) cells were isolated . Using pillar arrays , we observed reduced deflection-activated currents in the Elkin1-KO clones compared to the WT ( Figure 5A , B ) . A residual current at large deflections was still present in some KO cells; we hypothesise that this effect is due to minor compensation from PIEZO1 ( Supplementary file 1 and as noted previously [Rocio Servin-Vences et al . , 2017] ) or the activity of an as-yet-unidentified MA channel . Nevertheless , these data confirm that Elkin1 is required for sensitive MA channel activity in WM266-4 cells . We first tested whether Elkin1-dependent mechanotransduction regulates cell migration using a transwell assay . While transmigration of the WT clones was not different from control populations , both Elkin1-KO clones exhibited reduced transmigration ( Figure 5C ) . This phenotype could be rescued by the transient overexpression of wild-type hsElkin1-iso3 . In contrast , overexpression of hsElkin1-iso3 L210F ( Figure 5D ) , a variant associated with significantly reduced MA currents ( Figure 4B ) , did not rescue the transmigration defect . We additionally generated an Elkin1-KO clone from the A375 melanoma cell line ( which has previously been shown to express functional PIEZO1 [Hung et al . , 2016] ) . The A375 Elkin1-KO cells exhibited reduced transmigration , compared to the A375 WT cells ( Figure 5E ) , indicating that the link between Elkin1 activity and migration is not restricted to WM266-4 cells and can be measured in a cell line where PIEZO1 is functionally active . We therefore conclude that disruption of Elkin1 not only impairs mechanosensitivity , but also cell migration . We expanded our analysis to investigate how Elkin1 regulates different modes of migration . To examine how Elkin1 mediates the unconfined migration of individual cells on a 2D surface , we plated cells on dishes coated with LM511 , and observed their movement using phase-contrast and epi-fluorescent microscopy . A representation of the migration tracks shows that the WT clones migrated further over the course of the experiment ( Figure 6A ) . To test whether this effect was exclusive to migration on LM511-coated substrates , we repeated the experiments on dishes coated with poly-L-lysine ( PLL ) ( Figure 6B ) . The Elkin1-KO clones exhibited reduced track mean speed on both substrates ( Figure 6C ) and a reduced distance migrated from origin ( calculated as the Euclidean distance , Figure 6D ) . The percent reduction in the mean of the measured track mean speed was similar for experiments conducted on LM511 ( Elkin1-KO cells 29% reduction in the mean ) to PLL ( Elkin1-KO cells 32% reduction ) . These data demonstrate that Elkin1 modulates cell migration in unconfined environments . To investigate the impact of Elkin1 on confined migration , stripes of LM511 were printed onto glass coverslips and the unprinted regions were passivated to block cell adhesion , thus restricting attachment to the printed regions ( maximum width of 5 µm ) ( Figure 6E ) . During the experiment cells would occasionally detach and then reattach; our analysis only included periods when cells were attached and elongated on the patterned region . The mean migration speed was lower in the Elkin1-KO clones compared to WT ( Figure 6F ) ( 22% reduction in the mean Elkin1-KO track mean speed ) , indicating that disrupting Elkin1 inhibits the ability of cells to undertake both unconfined and confined migration on hard surfaces . Given that cell dissociation from the primary tumour is required for tumour metastasis , we investigated the effect of Elkin1 deletion using in vitro organotypic spheroids . We selected representative WT and Elkin1-KO clones and labelled these cells using a viral vector expressing GFP . Spheroids formed from GFP-expressing cells were implanted in 3D collagen gels and imaged at 24 , 48 and 72 hr post-implantation . At every time point , more Elkin1-KO cells had dissociated from the spheroid compared to WT ( Figure 7A–G ) and at 48 and 72 hr the Elkin1-KO cells were further away from the edge of the spheroid ( Figure 7H , Figure 7—figure supplement 1 ) . Imaging the first 12 hr after spheroid implantation highlighted the fact that Elkin1-KO cells dissociated more readily from the spheroid mass ( Figure 7—videos 1 and 2 ) . We investigated whether isolated Elkin1-KO cells embedded in 3D collagen gels exhibited a change in migratory properties to determine if an increase in migration speed could account for these data . In this 3D environment , the mean track speed was not different between WT and Elkin1-KO clones , however the track straightness was reduced in the Elkin1-KO clones ( Figure 7—figure supplement 2 ) . We additionally noted that the Elkin1-KO cells were less spherical as they broke away from the spheroid ( Figure 7I , Figure 7—figure supplement 1 ) as they were in the 3D collagen gels ( Figure 7—figure supplement 2 ) ) . However , these effects on sphericity and track straightness were moderate , with much of the data set overlapping . These data thus suggest that the increased distance of the Elkin1-KO cells from the spheroid is due to facilitated dissociation , not due to increased migration speed . One model to account for the increased dissociation of cells from Elkin-1-KO spheroids as well as their altered migration properties is that deletion of Elkin1 changes physical cellular interactions . To test this idea we used atomic force microscopy ( AFM ) to measure unbinding forces after short-term contact between cells and LM511 or homotypic cell-cell contacts ( Hofschröer et al . , 2017; Puech et al . , 2006 ) . The unbinding of LM511 from Elkin1-KO cells required a larger force than the WT clone ( 37% increase in the mean of the maximum unbinding force ) ( Figure 8A , B ) . In contrast , lower unbinding force was measured after homotypic contact between Elkin1-KO cells , in comparison with WT cells ( 27% decrease in the mean of the maximum unbinding force ) ( Figure 8C , D ) . To determine if deletion of Elkin1 influenced cell-cell organisation over longer time scales , we created chimeric spheroids between Elkin1-KO and WT cells by mixing equal numbers of a GFP-labelled and an unlabelled clone . Regardless of which clone expressed the GFP , the spheroid was organised such that Elkin1-KO cells were found in the outer layer of the spheroid and the WT cells in the core ( triplicate experiments , 23 spheroids total ) ( Figure 8E , F ) . To test whether the channel activity of Elkin1 is linked to cell partitioning within spheroids , we stably transfected Elkin1-KO cells with either hsElkin1-iso3 or hsElkin1-iso3-L210F . In chimeric spheroids the partitioning of cells was partially rescued in cells expressing hsElkin1-iso3 but not hsElkin1-iso3-L210F , where cells were evenly distributed throughout the resulting spheroids ( duplicate experiments , 6 and 9 spheroids , respectively ) ( Figure 8—figure supplement 1 ) . Taken together these data indicate that deletion of Elkin1 has a differential effect on cell-substrate versus cell-cell binding and regulates cell-cell associations in organotypic spheroids . We have identified a novel mechanoelectrical transduction pathway in melanoma cells , activated by mechanical stimuli applied at the cell-substrate interface . The sensitivity of MA currents found in melanoma cells grown on LM511 was comparable to those found in the sensitive mechanoreceptors of the dorsal root ganglia required for fine touch ( Poole et al . , 2014 ) . Furthermore , these molecular-scale pillar movements lie within the range of matrix displacements that arise due to cells pulling on their surroundings ( Legant et al . , 2013; Legant et al . , 2010 ) . In WM266-4 melanoma cells these deflection-activated currents were dependent on the Elkin1 protein . Elkin1 contains a LUSTR ( Lung Seven Transmembrane ) domain , which defines a family of proteins of as yet unknown function found in plants and animals , but not bacteria , archaea or viruses ( Edgar , 2007 ) . This LUSTR family also includes TMEM87b , a protein that shares 48% homology with hsElkin1-isoform1 . There is limited information regarding the physiological function of Elkin1 , however the gene is expressed in diverse human tissues including organs where mechanical feedback is important , such as the lungs and the bladder ( GTEx portal ) . Elkin1 was previously suggested to be a Golgi-associated protein that , when overexpressed , partially rescued a VPS54-KO endosome to trans-Golgi network ( TGN ) retrograde transport phenotype ( Hirata et al . , 2015 ) . In contrast , we found that a fraction of both hsElkin1-iso1 and -iso3 is localised to the plasma membrane , with hsElkin1-iso1 also present within the Golgi . Elkin1 has also been demonstrated to be amongst the fraction of cell surface proteins that undergo N-linked glycosylation ( Park et al . , 2018 ) , further supporting our data . Given that Elkin1 knockdown did not inhibit endosome to TGN retrograde signalling ( Hirata et al . , 2015 ) , but did ablate MA channel activity ( shown here ) , we propose that Elkin1 is an essential component of a novel mechanoelectrical transduction pathway . In support of this model , heterologous expression of Elkin1 was associated with the appearance of MA currents activated by substrate deflections in two cell lines in which Piezo1 was deleted . The latency between stimulus and response ( <2 ms ) and the activation time constant ( <1 ms ) were similar to those reported for PIEZO1-mediated currents ( Poole et al . , 2014; Rocio Servin-Vences et al . , 2017 ) and sufficiently rapid to suggest that the Elkin1-dependent current is directly activated by the mechanical stimulus ( Christensen and Corey , 2007 ) . In addition , none of the other MA non-selective cation channels ( PIEZO2 , TRPV4 ) were detected in WM266-4 cells , suggesting that Elkin1 either mediates these MA currents or modulates an unidentified MA channel . These data definitively demonstrate that Elkin1 is required for a novel , PIEZO1-independent mechanoelectrical transduction pathway . Mutation of a single residue in a predicted TM helix and of a single residue in a predicted intracellular loop led to a marked reduction in MA channel activity , without a concomitant reduction in surface localisation of the protein . Taken together these data indicate that Elkin1 exhibits several hallmarks of a novel MA channel ( Christensen and Corey , 2007 ) . However , it is possible that Elkin1 is an accessory molecule that modulates the activation of an , as yet uncharacterised , MA channel or an ion channel that requires additional proteinaceous tethers in order to be activated by mechanical inputs . Future studies with purified protein will be required to definitively test the hypothesis that Elkin1 is an ion channel activated by mechanical inputs . In HEK-293T P1KO cells , human Elkin1 isoform1 and 3 were associated with a current activated by substrate deflection . Expression of mouse Elkin1 in the HEK293T P1KO cells was also associated with the de novo appearance of deflection activated currents , though these currents were of smaller amplitude . Channel activity could also be evoked by indentation ( however , maximal currents were smaller than reported for the PIEZOs [Coste et al . , 2010] ) but not by membrane stretch applied using HSPC . This mechanical response profile is distinct to other MA ion channels: PIEZO1 and PIEZO2 respond to membrane stretch , cell indentation and substrate deflection ( Coste et al . , 2010; Poole et al . , 2014; Rocio Servin-Vences et al . , 2017 ) ( though PIEZO2 is less sensitive to membrane stretch than PIEZO1 [Wang et al . , 2019] ) and TRPV4 responds only to substrate deflections in mammalian cells ( Rocio Servin-Vences et al . , 2017 ) . These data suggest that the expression of distinct MA channels allows cells to distinguish between stimuli arising from cell-generated forces at cell-substrate contacts versus exogenous stimuli that stretch the membrane or compress the cell . Such differential activation profiles may reflect multiple roles for distinct MA channels during tumour development and metastasis . During metastasis , cells first dissociate from the primary tumour and then , as they migrate through the body , the cells encounter environments of differing dimensionality and degree of confinement ( Paul et al . , 2017; van Helvert et al . , 2018 ) . Disruption of Elkin1 expression altered the confined and unconfined migration of cells in quasi-1D and 2D environments . Our results stand in contrast with the effect of PIEZO1 on melanoma cell migration: PIEZO1 knockdown reduced the speed of cell migration in confined environments , but had no effect on unconfined cell migration ( Hung et al . , 2016 ) . In other tumours , PIEZO2 knockdown in a breast cancer cell line ( BrM2 ) , inhibited the cells’ ability to enter confined spaces , but did not influence the speed of migration of the confined cells ( Pardo-Pastor et al . , 2018 ) . These data suggest that distinct MA channels mediate separate mechanical transduction pathways regulating different aspects of tumour cell migration . In addition to the Elkin1-dependent changes in migration , we report that Elkin1-KO cells exhibited increased dissociation from organotypic spheroids embedded in collagen gels . The process of cell dissociation from a spheroid mass is defined by a complex interplay between the strength of cell-cell adhesions , the contractility of the cells at the edge of the spheroid and the physical characteristics of the surrounding matrix ( Ahmadzadeh et al . , 2017 ) . Our data demonstrate that Elkin1 expression regulates cell-substrate and cell-cell binding forces . In addition , the partitioning of cells within the organotypic spheroids was dependent on Elkin1 . We therefore propose that Elkin1-dependent expression regulates the balance between cell-cell and cell-matrix adhesion; ablation of the protein increases invasion of surrounding matrix ( by reducing cell-cell adhesions and facilitating cellular dissociation ) and slows migration ( by increasing cell-substrate binding ) . Previous studies of the role of TRPV4 in breast cancer development have demonstrated that overexpression of TRPV4 can lead to a decrease in E-cadherin expression , driving EMT in this cancer type ( Lee et al . , 2017 ) . This TRPV4-dependent switch in E-cadherin expression may also lead to increased dissociation from the tumour mass . However , in contrast to Elkin1 , increased levels of TRPV4 lead to an increase , rather than a decrease in cell migration ( Lee et al . , 2016b; Lee et al . , 2017 ) . A similar phenomenon ( decreased cell-cell binding forces , facilitated dissociation from organotypic spheroids and decreased cell migration ) has been reported in melanoma cells overexpressing the Na+/H+ exchanger , NheI ( Hofschröer et al . , 2017 ) . This exchanger locally modulates the pH of the cellular environment and is not known to be mechanically responsive . Given that Elkin1 is required for MA currents that are separable from the PIEZO channels and that Elkin1-dependent currents are activated at the cell-substrate interface , these channels may transduce distinct mechanical inputs . The PIEZOs may be acting to sense confinement ( Hung et al . , 2016; Pardo-Pastor et al . , 2018 ) and the overall mechanical status of the cells ( Rocio Servin-Vences et al . , 2017 ) , while Elkin1-dependent mechanoelectrical transduction may encode information about the mechanical nature of the cells’ microenvironment thus facilitating modulation of cell-cell versus cell-substrate binding . Such integration of multiple mechanoelectrical transduction pathways could engender cells with a tuneable and diverse repertoire of mechanical sensing . All melanoma cell lines ( WM266-4 , WM115 , and A375 ) were obtained from the ATCC and cultured in complete Minimum Essential Medium Eagle ( MEME ) or High Glucose DMEM supplemented with 10% FBS and 1% Penicillin/Streptomycin . MEME was additionally supplemented with 1% l-glutamine . HEK-293T and HEK-293T P1KO ( Lukacs et al . , 2015 ) ( a gift from A . Patapoutian ) were cultured in DMEM medium , 10% FBS and 1% Penicillin/Streptomycin . All cultures were maintained at 37°C , 5% CO2 . Fugene HD ( Promega ) or polyethyleneimine ( PEI , Sigma , MW 600–800 kDa ) was used to transfect cells at the following ratios ( w/w ) : DNA:Fugene at 1:3 and DNA:PEI at 1:4 . WM266-4 cells were transduced with pRRLSIN . cPPT . PGK-GFP . WPRE lentivirus ( a gift from Ian Alexander , originally from Didier Trono/Inder Verma , Addgene plasmid #12252 ) , in which a human PGK promoter drives EGFP expression . The virus was produced in HEK-293T cells following PEI transfection of the lentiviral backbone together with packaging plasmid ( psPAX2 ) and the vesicular stomatitis virus ( VSV-G ) envelope . To create cell lines with stably-integrated DNA for spheroid rescue experiments , plasmids encoding Elkin1 and GFP were linearised using ScaI . Linearised DNA was transfected into WM266-4 3C6 ( Elkin1 KO clone ) using Fugene ( as above ) . Genetically modified cells ( virally transduced or stably expressing Elkin1 ) were isolated in FACS buffer ( 2 mM EDTA , 2% FBS , 2% Penicillin/Streptomycin in PBS ) using the BD FACS Jazz ( low pressure sorting with a 100 μm nozzle at 17 psi , 4°C ) . For all experiments to be conducted within 6 hr of cell collection , cells were released from culture flasks using enzyme-free cell dissociation buffer ( Sigma-Aldrich ) . Cell lines were authenticated using STR profiling and checked for mycoplasma contamination by CellBank Australia . To prepare samples for mass spectrometry , peptides were prepared from cultured WM266-4 cells using the Pierce Mass Spec Sample Prep Kit for Cultured Cells as per manufacturer’s instructions ( ThermoFisher ) . LysC and Trypsin were used to generate peptides from 1 mg of total protein . Peptides were separated using isoelectric focussing ( IEF ) in immobilized pH gradient ( IPG ) gel strips into six separate fractions , as previously described ( Eravci et al . , 2014 ) . Desalted peptides of these fractions were then separated on an 8–60% acetonitrile gradient ( 240 min ) with 0 . 1% formic acid at a flow rate of 200 nL/min using the EASY-nLC II system ( Thermo Fisher Scientific ) on in-house manufactured silica microcolumns packed with the ReproSil-Pur C18-AQ 3 µm resin . A Q Exactive plus mass spectrometer ( Thermo Fisher Scientific ) was operated in the data dependent mode with a full scan in the Orbitrap followed by top 10 MS/MS scans using higher-energy collision dissociation ( HCD ) . Analysis of MS and MS/MS spectra was performed using MaxQuant software ( version 1 . 5 . 1 . 2 ) and proteins were identified by searching against the human reference proteome database UP000005640 from Uniprot . The sequences of all primers used for this study are listed in Supplementary file 2 . mRNA was isolated from WM266-4 cells using the RNEasy kit as per manufacturer’s instructions ( Qiagen ) . First strand cDNA synthesis was carried out using 200 ng of isolated mRNA , random primer mix ( New England Biolabs ) and M-MuLV reverse transcriptase ( New England Biolabs ) . The resulting samples were used for qPCR analysis ( see below ) and as a template to amplify hsElkin1-iso1/3 cDNA . For electrophysiology experiments , these sequences were cloned into pRK5 with a cistronic eGFP driven by an internal IRES sequence . For the purposes of live cell imaging , hsElkin1-iso1/iso3 sequences were fused to mGFP using the same vector . miRNA knockdown reagents were generated using the BLOCK-iT Pol II miR RNAi Expression Vector , as per manufacturer’s instructions ( Invitrogen ) . Briefly , complementary ssDNA sequences ( Eurofins , Belgium ) that encode the miRNA of interest were annealed and cloned into the pcDNA 6 . 2GW-EmGFP vector . Three distinct miRNA sequences were generated for each target gene and assembled in series within a single plasmid to generate the final construct for knock-down experiments . The Golgi network was detected using CellLight Golgi-RFP , BacMam 2 . 0 ( ThermoFisher ) . Sequences were analysed using the T-Coffee multiple sequence alignment server ( coffee . crg . cat ) and formatted using Boxshade ( https://embnet . vital-it . ch/software/BOX_form . html ) . To create plasmid constructs for gene editing , guide RNAs ( gRNAs ) were designed using the online CRISPR design tool ( Zhang Lab , MIT - http://crispr . mit . edu/ ) . Single stranded DNA oligos were obtained from Eurofins ( Belgium ) , annealed and cloned into the pSpCas9n ( BB ) −2A-GFP plasmid ( a gift from Feng Zhang , Addgene plasmid # 48140 ) ( Ran et al . , 2013 ) . Four plasmids were generated , each with gRNAs that targeted sequences in intron 7 or exon 9 of the Elkin1 gene . These constructs were transfected into either WM266-4 or A375 cells . Isolation of clonal populations was a two-step process . First , a population of transfected cells was initially isolated based on GFP expression: 24–48 hr post-transfection cells were resuspended in FACS buffer ( 2 mM EDTA , 2% FBS , 2% Penicillin/Streptomycin in PBS ) and cells expressing GFP were collected using the BD FACS Jazz low pressure sorting with a 100 μm nozzle at 17 psi , 4°C . Following expansion of this population , cells were sorted a second time ( as above ) as single cells in individual wells of a 96 well plate . Single-cell clones were initially cultured in a media comprising of 50% complete media/50% conditioned media for 7–14 days until cells had started to divide and form colonies . Edited clones were screened for genomic deletions covering the Elkin1 gene using PCR . Primers for screening CRISPR edited clones were designed using Primer-BLAST ( NCBI ) to span the ~2 . 5 kb deleted region of the gene . Genomic DNA ( gDNA ) was extracted from cells using the Illustra Genomic Prep Mini Spin kit ( GE Life Sciences ) , as per manufacturer’s instructions . Genomic PCR was conducted using 10 ng of template DNA and Hot Start Taq DNA polymerase ( New England Biolabs ) . Primers and probes to analyse Elkin1 and PIEZO1 transcript levels were designed using PrimerQuest Design Tool ( IDT , Singapore ) and a predesigned assay was used to detect levels of HPRT1 ( housekeeping gene ) ( Integrated DNA Technologies , Singapore ) . The reaction underwent 40 cycles . Using the difference in cycle threshold ( ΔCt ) , the fold change in expression ( 2-ΔΔCt ) was calculated and compared to a control sample . Positive masters and pillar array casting were described previously ( Poole et al . , 2014 ) . Briefly , positive silicon masters were silanised using vapour phase Trichloro ( 1H , 1H , 2H , 2H-perfluorooctyl ) silane ( Sigma-Alrich ) for 16 hr . Negative masters were cast from this substrate in polydimethylsiloxane ( PDMS ) ( Sylgard 184 , Dow Corning ) , mixed at a ratio of 1:10 and cured at 110°C for 15 min . Negative masters were silanised as above and used to cast pillar arrays . Arrays were coated with degassed PDMS ( 1:10 ) and left for 30 min . A thickness two coverslip activated with oxygen plasma generated using a low pressure Zepto plasma system ( Diener , Germany ) was placed over the still liquid PDMS . Pillar arrays were cured for 1 hr at 110°C . Pillar arrays were activated using the oxygen plasma system and either cells were directly seeded onto this activated surface or arrays were first functionalised by coating with 10 μg/ml LM-511 ( BioLamina , Sweden ) for one hour at 37°C . Cells were seeded at a concentration of 2 × 104 cells/mL in complete media and incubated overnight . Whole-cell patch pipettes were prepared from thick-walled filamented glass ( Harvard Apparatus , USA ) using a pipette puller fitted with a box filament ( P-1000 , Sutter Instruments , USA ) . Pipettes were heat-polished with a homemade micro-forge to give a final resistance of 3 MΩ – 6 MΩ . Pipettes were filled with a solution containing 110 mM KCl , 10 mM NaCl , 1 mM MgCl2 , 1 mM EGTA and 10 mM HEPES ( pH 7 . 3 ) . Extracellular solutions contained 140 mM NaCl , 4 mM KCl , 2 mM CaCl2 , 1 mM MgCl2 , 4 mM glucose and 10 mM HEPES ( pH 7 . 4 ) . Whole-cell patch-clamp data was obtained on either a Zeiss 200 inverted microscope and an EPC-10 amplifier in combination with Patch-master software or a Nikon Ti-E inverted microscope and an Axopatch 200B with pClamp 10 software . Data were analysed using either Fitmaster software ( HEKA Electronik GmbH , Germany ) or Clampfit software ( Molecular Devices , USA ) . Pipette and membrane capacitance were compensated and to minimise voltage errors , series resistance was compensated by at least 60% . Mechanically-activated currents were recorded at a holding potential of −60 mV . Mechanical stimuli were applied by serially deflecting an individual pilus using a blunt , heat-polished pipette ( tip diameter approx . 2 μm ) driven by a MM3A-LS nanomanipulator ( Kleindiek Nanotechnik , Germany ) . Multiple stimuli ranging between 1 nm–1 µm were applied with a delay between each stimulus of at least 10 s . To quantify the stimulus , bright-field images of the deflected pilus were taken before and during stimulation using a 40x/0 . 6 NA objective . The centre of each pilus was calculated off line by applying a 2D-gaussian fit of intensity values ( Igor , Wavemetrics , USA ) and the magnitude of pillar deflection calculated from successive images by comparing the difference in the calculated centre point . Transwell assays were conducted using the HTS FluoroBlok Multiwell Insert System ( Corning ) with an 8 . 0 µm pore size . The bottom surface of the membrane was coated with laminin ( 10 µg/ml ) for 3 hr at 37°C . Membranes were then blocked for 1 hr using 2% polyvinylpyrolidone . The bottom well was filled with 500 µl media and 1 × 104 cells were added to the top well . Cells were allowed to transmigrate for 16 hr before the bottom side of the membrane was fixed with 4% PFA . In order to quantify the number of cells that had transmigrated , the fixed sample was permeabilised using 0 . 1% TX-100 for 5 min at room temperature and nuclei were stained using Hoechst at a concentration of 0 . 1 µg/ml . Cells were imaged using epifluorescence imaging on an inverted microscope ( Zeiss Axiovert 200 or Leica DMIL ) fitted with a 10x/0 . 22 NA objective and standard filters; 4 regions of each filter were imaged . The ‘Analyze Particles’ function of ImageJ was used to count cells within each field . Data from each independent experiment were normalized to the average number cells that were counted on the underside of each filter for triplicate controls . µ-Slide 8-well dishes ( ibidi , Germany ) were coated with either 10 µg/ml LM-511 ( BioLamina , Sweden ) , or 0 . 1% ( w/v ) poly-L-lysine ( PLL ) ( Sigma-Aldrich , USA ) . Approximately 2500 cells were added to each well , and after 2 hr , nuclei were labelled by adding Hoechst dye at a concentration of 0 . 1 µg/ml . Cell migration was monitored for 16 hr using live-cell microscopy ( see below ) . Data were obtained from three experiments . Data were confirmed by repeating this assay with cells expressing cytoplasmic GFP , using the same imaging conditions . Microcontact-printing ( von Philipsborn et al . , 2006 ) was used to pattern substrates for quasi-1D experiments . Microcontact-printing stamps ( Chiang et al . , 2011 ) were incubated with 10 µg/ml LM511 and 1 µg/ml of goat anti-mouse IgG antibody labelled with AlexaFluor 594 ( to visualize patterns ) for 30 min at 37 °C . Stamps were rinsed with ultrapure water , dried with inert gas , then stamped on clean coverslips activated using oxygen plasma . Coverslips were incubated with 1 mg/mL PLL-g-PEG ( SuSoS , Switzerland ) in PBS , at room temperature for 30 min , to block unprinted regions . Cells ( 3 × 104 cells/mL ) were seeded on each substrate and nuclei were labelled by adding 0 . 1 µg/ml Hoechst . Cells were monitored using live-cell microscopy for 12 hr and data were obtained from at least 50 cells across three experiments . We only included in our data the periods of movement that correspond to when the cell was attached and elongated on the patterned region . To determine whether cells were confined/unconfined we analysed the cell shape . Cells with an aspect ratio of 2:1 or more were classed as confined and the number of confined vs non-confined cells was determined by sampling still images at five points during each experiment . To assay individual cells in collagen gels , a cell suspension ( 1 . 25 × 106 cells/mL in complete media ) was mixed with 1 . 5 mg/mL rat tail Collagen I ( total volume 100 µL ) and gelated at 37°C , 5% CO2 for 15 min . Isolated cells in collagen gels were imaged using a Leica SP8 DLS fitted with a 10x/0 . 4 NA objective every 10 min for 14 hr . All cultures were maintained at 37°C and 5% CO2 . Both WT and Elkin1-KO clones were virally transduced with a construct encoding eGFP to enable visualisation of cells . For spheroid formation , 1 × 103 cells were seeded onto Ultra-low attachment 96-well plates ( Corning , USA ) in complete MEME media and formed spheroids following 72 hr of incubation . Composite spheroids were created by mixing 5 × 102 of each cell-type to give 1 × 103 cells total . Collagen gel mixtures were made from 1 . 5 mg/mL – 2 mg/mL rat tail Collagen I ( Corning , USA ) in 10 mM NaOH , 1 × PBS and complete media on ice . A base collagen gel was formed in each well of a glass-bottom 96-well plate ( Greiner Bio-one , Austria ) by incubating 30 µL of the collagen mixture at 37°C , 5% CO2 for 6 min . The collagen gel mixture containing the spheroid was then applied to the base gel and underwent gelation at 37°C , 5% CO2 for 15 min . Cells in collagen gels were cultured under 200 µL of complete media . Spheroids and invasive cells were imaged at 24 , 48 , 72 hr using a Leica SP8 DLS fitted with a 10x/0 . 4 NA objective . Imaris 9 . 1 . 2 software ( Bitplane AG , Zurich , Switzerland ) was used to segment cells by creating surfaces with a filter of 200 μm3 to discard cell debris . For organotypic spheroid assay analysis , cell surfaces were filtered by centre of image mass versus z depth , where the z value was determined by orientating the 3D image stacks to the plane of view showing the cells on the glass and cells attached to the glass were excluded from analysis . Cells from the 3D dissociated cell migration assay were tracked using autoregressive motion , applying a threshold of 1500s to filter track duration . Intensity , morphological and tracking data were then exported and further analysed using GraphPad Prism software ( La Jolla , CA , USA ) . To detect Elkin1 variants at the plasma membrane , the cell surface fraction was biotinylated and subsequently isolated following established protocols ( Tarradas et al . , 2013 ) . Briefly , HEK-293T P1KO cells attached to PLL-coated culture dishes were transiently transfected with plasmids encoding GFP-tagged Elkin1 variants . After 24 hr the cell surface fraction was labelled , on ice , with freshly prepared 2 . 5 mg/ml EZ-link Sulfo-NHS-LC-LC-biotin ( 21338 , ThermoFisher Scientific ) in DPBS containing Ca++ . After quenching with 100 mM glycine , cells were lysed as above using RIPA buffer containing protease inhibitor cocktail . A portion of this lysed sample was reserved as the ‘input’ sample . The biotinylated fraction was then isolated using NeutrAvidin Ultralink Resin ( 53150 , ThermoFisher Scientific ) . After recovery from the NeutrAvidin beads , samples were prepared as for gel electrophoresis by mixing with Bolt LDS sample buffer and Bolt reducing agent ( B0007 and B0009 respectively , ThermoFisher Scientific ) . Samples were separated on a 10% Bolt Bis-Tris Plus gel ( NW00102BOX , ThermoFisher Scientific ) , transferred to a PVDF membrane and subjected to standard antibody detection . GFP-fusion proteins were detected using rabbit polyclonal anti-GFP ( SAB4301138 , Sigma-Aldrich , 1:1000 ) and HRP-linked anti-rabbit IgG ( 7074 , Cell Signaling Technologies , 1:1000 ) . AFM analysis of cell-binding forces was conducted using a JPK Nanowizard , fitted with a CellHesion module ( JPK Instruments AG ) . To measure cell-LM511 interaction forces , cantilevers ( MLCT-O10 , Bruker ) were first activated using oxygen plasma generated using a low-pressure Zepto plasma system ( Diener , Germany ) and then incubated in a droplet of LM511 ( 20 µg/ml ) for 1 hr . To measure cell-cell interaction forces , cantilevers were coated with wheat germ agglutinin ( L4895 , Sigma Aldrich ) . Before collecting data , the sensitivity and spring constant of each cantilever was determined using in-built routines in the JPK software . To measure cell-LM511 binding forces , the following settings were used: speed 2 µm/sec , set point 500 pN , with contact times of 2 s held at constant height . At least five force-distance curves were collected from at least 10 cells per condition ( multiple force-distance curves were not collected successively on each cell to minimise adaptive changes ) . To measure cell-cell binding forces the CellHesion module was used to move the stage through an extended 100 µm pulling range . To attach a cell to the cantilever , cells in suspension were added to a sample dish containing adherent cells , the calibrated cantilever was positioned over an unattached cell and carefully brought into contact . After 10 s attachment time the cantilever was slowly withdrawn from the surface , with a single cell attached . To collect data the cantilever was positioned over an adherent cell and data collected with the following parameters: speed 5 µm/sec , set point 250 pN , contact time 2 s held at a constant height . At least five force-distance curves were collected from at least 10 different cell-cell interaction pairs . Force-distance curves were analysed to determine the maximum unbinding force during cantilever retraction . All data were analysed using Prism seven or Prism 8 ( GraphPad ) . Normality was determined with D’Agostino-Pearson omnibus normality test . Normally distributed data were analysed using parametric statistical tests . All t-tests were two-tailed and when there were significant differences in the variance , Welch’s correction was used . Data from qPCR experiments and migration assays were analysed using a Kruskal-Wallis one-way ANOVA test with Dunn’s multiple comparisons . Stimulus-response plots were generated by binning data by stimulus size and averaging within bins for each cell , then across cells . Ordinary two-way ANOVA with Sidak’s multiple comparisons was used to analyse stimulus-response plots . The transwell migration assay data were normalised against the parental clone and analysed using a parametric one-way ANOVA test with Tukey’s multiple comparisons . Data from organotypic spheroid experiments were averaged for each spheroid and then compared with one-way ANOVA with Tukey’s multiple comparisons .	When cells receive signals about their surrounding environment , this initiates a chain of signals which generate a response . Some of these signalling pathways allow cells to sense physical and mechanical forces via a process called mechanotransduction . There are different types of mechanotransduction . In one pathway , mechanical forces open up specialized channels on the cell surface which allow charged particles to move across the membrane and create an electrical current . Mechanoelectrical transduction plays an important role in the spread of cancer: as cancer cells move away from a tumour they use these signalling pathways to find their way between cells and move into other parts of the body . Understanding these pathways could reveal ways to stop cancer from spreading , making it easier to treat . However , it remains unclear which molecules regulate mechanoelectrical transduction in cancer cells . Now , Patkunarajah , Stear et al . have studied whether mechanoelectrical transduction is involved in the migration of skin cancer cells . To study mechanoelectrical transduction , a fine mechanical input was applied to the skin cancer cells whilst measuring the flow of charged molecules moving across the membrane . This experiment revealed that a previously unknown protein named Elkin1 is required to convert mechanical forces into electrical currents . Deleting this newly found protein caused skin cancer cells to move more slowly and dissociate more easily from tumour-like clusters of cells . These findings suggest that Elkin1 is part of a newly identified mechanotransduction pathway that allows cells to sense mechanical forces from their surrounding environment . More work is needed to determine what role Elkin1 plays in mechanoelectrical transduction and whether other proteins are also involved . This could lead to new approaches that prevent cancer cells from dissociating from tumours and spreading to other body parts .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "cancer", "biology" ]	2020	TMEM87a/Elkin1, a component of a novel mechanoelectrical transduction pathway, modulates melanoma adhesion and migration
In the mouse , Period-2 ( Per2 ) expression in tissues peripheral to the suprachiasmatic nuclei ( SCN ) increases during sleep deprivation and at times of the day when animals are predominantly awake spontaneously , suggesting that the circadian sleep-wake distribution directly contributes to the daily rhythms in Per2 . We found support for this hypothesis by recording sleep-wake state alongside PER2 bioluminescence in freely behaving mice , demonstrating that PER2 bioluminescence increases during spontaneous waking and decreases during sleep . The temporary reinstatement of PER2-bioluminescence rhythmicity in behaviorally arrhythmic SCN-lesioned mice submitted to daily recurring sleep deprivations substantiates our hypothesis . Mathematical modeling revealed that PER2 dynamics can be described by a damped harmonic oscillator driven by two forces: a sleep-wake-dependent force and an SCN-independent circadian force . Our work underscores the notion that in peripheral tissues the clock gene circuitry integrates sleep-wake information and could thereby contribute to behavioral adaptability to respond to homeostatic requirements . The sleep-wake distribution is coordinated by the interaction of a circadian and a homeostatic process ( Daan et al . , 1984 ) . The biological substrates underlying the circadian process are relatively well understood: circadian rhythms in overt behavior of mammals are generated by the suprachiasmatic nuclei ( SCN ) located in the hypothalamus ( Hastings et al . , 2018 ) . At the molecular level , so-called ‘clock genes’ interact through negative transcriptional/translational feedback loops ( TTFLs ) , where the CLOCK/NPAS2:ARNTL ( BMAL1 ) heterodimers drive the transcription of their target genes , among them the Period ( Per1 , 2 ) and Cryptochrome ( Cry1 , 2 ) genes . Subsequently , PER and CRY proteins assemble into repressor complexes that inhibit CLOCK/NPAS2:ARNTL-mediated transcription , including their own . The resulting reduction of the repressor complex allows a new cycle to start . This feedback loop , present in almost each cell of the mammalian body , interacts with other molecular pathways , together ensuring a period of ~24 hr ( Hastings et al . , 2018 ) . The SCN synchronizes peripheral clock gene expression rhythms through its rhythmic behavioral , electrical , and humoral output generated across the day ( Schibler et al . , 2015 ) . Accumulating evidence suggests that , perhaps surprisingly , clock genes are also involved in the homeostatic aspect of sleep regulation ( Franken , 2013 ) . This is illustrated by the sleep deprivation ( SD ) -induced increase in the expression of the clock gene Per2 in tissues peripheral to the SCN , including the cerebral cortex , liver , and kidney ( Curie et al . , 2013; Curie et al . , 2015; Franken et al . , 2007; Maret et al . , 2007 ) . Moreover , the highest level of peripheral Per2 expression is reached after the time of day mice were awake most , suggesting that also during spontaneous periods of waking Per2 expression accumulates . Accordingly , lesioning of the SCN , which eliminates the circadian sleep-wake distribution , attenuates the circadian amplitude of clock gene transcripts and proteins in peripheral tissues ( Akhtar et al . , 2002; Curie et al . , 2015; Tahara et al . , 2012; Saini et al . , 2013; Sinturel et al . , 2021 ) . Together , these studies suggest that sleeping and waking are important contributors to clock gene expression , but dissecting the contribution of the sleep-wake distribution and circadian time is challenging because the two change in parallel . By simultaneously recording electroencephalogram ( EEG ) , electromyogram ( EMG ) , locomotor activity ( LMA ) , and PER2-dependent bioluminescence signals from cortex and kidney in freely behaving mice , we established that the circadian sleep-wake distribution importantly contributes to the daily rhythmic changes in central and peripheral PER2 levels . To further test this hypothesis , we predicted that ( i ) in behaviorally arrhythmic SCN-lesioned ( SCNx ) mice , daily recurring SDs mimicking a circadian sleep-wake distribution will temporarily reinstate high-amplitude PER2 bioluminescence rhythms , and ( ii ) in intact rhythmic animals , reducing the amplitude of the circadian sleep-wake distribution will result in a reduced amplitude of PER2 rhythms . While daily SDs indeed enhanced the amplitude of peripheral PER2 rhythms in SCNx mice , the protocol used to reduce the amplitude of the sleep-wake distribution did not reduce PER2 amplitude in all mice . To reconcile the sleep-wake-driven and circadian aspects of PER2 dynamics , we implemented a mathematical model in which waking represents a force that sets in motion a harmonic oscillator describing PER2 dynamics and found that the sleep-wake distribution , also under undisturbed conditions , is an important contributor to the daily changes in PER2 bioluminescence . Moreover , we discovered a second , SCN and sleep-wake independent force with a circadian period that underlay the residual circadian PER2 rhythms in SCNx mice , and that the phase relationship between these two forces is important for predicting the amplitude response in PER2 rhythms to sleep-wake perturbations . It now has been well documented that enforced wakefulness affects Per2 mRNA and protein levels in various tissues and mammalian species ( Franken , 2013; Hoekstra et al . , 2019; Möller-Levet et al . , 2013; Vassalli and Franken , 2017 ) , but it is not known whether circadian rhythms in spontaneous sleep-wake behavior contribute to the daily changes in PER2 levels . To address this question , we equipped Per2Luc KI B6 and SKH1 mice ( n = 6 and 5 , respectively ) with wireless EEG/EMG recorders ( i . e . , NeuroLoggers ) and monitored simultaneously sleep-wake state , PER2 bioluminescence , and LMA under constant darkness ( DD; see Figure 1—figure supplement 1 ) . Time asleep in mice kept in the RT-Biolumicorder quantified with the NeuroLoggers under DD was similar to that quantified under LD conditions using a tethered EEG acquisition system ( 48 hr baseline in DD vs . LD in C57BL/6J mice: NREM sleep: 42 . 2% ± 1 . 6% , REM sleep 6 . 1% ± 0 . 3% of total recording time; n = 6; compared to 41 . 3 ± 1 . 1 and 5 . 2% ± 0 . 2% , respectively; n = 12 , data taken from Diessler et al . , 2018 ) . Sleep and EEG have not been recorded in SKH1 mice previously . Also in these mice , similar sleep durations were obtained under the two conditions ( 48 hr baseline in DD vs . LD: NREM sleep: 40 . 4% ± 2 . 1% , REM sleep 5 . 8% ± 0 . 9%; n = 5; compared to 40 . 0% ± 1 . 0% and 6 . 9% ± 0 . 2% , respectively; n = 8; see Figure 1—figure supplement 3 ) . Next , we reproduced the circadian changes in PER2 bioluminescence as well as the response to a 6 hr SD ( Figure 1A ) as described previously ( Curie et al . , 2015 ) in both peripheral bioluminescence , mainly of renal origin , and central bioluminescence , emitted by the cortex ( see Materials and methods for explanation of the SD procedure ) . Similar to what was observed in that publication , SD elicited a tissue-specific response , with an immediate increase in central PER2 bioluminescence after SD , while in the periphery the response was delayed and PER2-bioluminescence increases were observed in the second and fifth hour of recovery ( Figure 1A ) . In addition to these direct effects on PER2 bioluminescence within the first 6 hr of recovery , the SD also caused a long-term reduction of rhythm amplitude during both recovery days; that is , the first ( REC1 ) and last ( REC2 ) 24 hr period of recovery . In the periphery , this decrease amounted to ca . 30% on both days ( REC1: 17 . 5–41 . 1%; REC2: 18 . 7–42 . 4% , 95% confidence intervals [95% CI] ) . In the central recordings , PER2-bioluminescence amplitude decreased significantly during REC1 ( 32% [10 . 2–54 . 5%] ) , whereas the decrease during REC2 ( 22% [0 . 06–44 . 8%] ) no longer reached significance levels ( linear mixed model with fixed conditional effect [‘BSL , ’ ‘REC1 , ’ ‘REC2’] and random intercept effect [‘Mouse’]; periphery: BSL vs . REC1 p=0 . 0014; vs . REC2 p=0 . 0011; central: BSL vs . REC1 p=0 . 022; vs . REC2 p=0 . 082; sinewave-fitted baseline amplitudes: periphery 0 . 258 [0 . 15–0 . 36]; central 0 . 193 [0 . 08–0 . 31 a . u . ] ) . This long-term reduction of PER2-biolumnescence is reminiscent of the long-term SD effects on rhythm amplitude we observed for Per2 expression and for other clock genes in the cortex ( Hor et al . , 2019 ) . Although a circadian modulation of PER2 bioluminescence of both peripheral and central origin is evident ( see Figure 1B and Figure 1—figure supplement 4 ) , we observed additional changes in bioluminescence that occurred simultaneously with changes in sleep-wake state , indicating that PER2 bioluminescence increases during wake-dominated periods and decreases during sleep-dominated periods in both tissues . To quantify this observation , transitions from sleep ( irrespective of sleep state ) to wake and from wake to sleep were selected ( see Materials and methods for selection criteria ) . Examples of the selected transitions are indicated in Figure 1B as a hypnogram . A similar number of sleep-to-wake and wake-to-sleep transitions passed selection criteria during the two-and-a-half baseline days ( periphery: 31 . 2 ± 3 . 9 and 29 . 0 ± 3 . 9; central: 25 . 6 ± 1 . 4 and 24 . 6 ± 1 . 4 , respectively; mean ± SEM ) . Transitions obtained in mice in which central bioluminescence was recorded were shorter ( longest common wake period after sleep-to-wake transitions: 39 vs . 57 min; longest common sleep period after wake-to-sleep transitions: 63 vs . 75 min , for central and periphery , respectively ) . Although PER2 bioluminescence increased during wakefulness and decreased during sleep in both tissues , tissue-specific differences were observed ( Figure 1C ) . At sleep-to-wake transitions , tissue differences concerned an initial decrease in peripheral PER2 bioluminescence after wake onset , followed by a steep increase saturating at 125% . In contrast , central levels of PER2 bioluminescence increased from the start and followed a linear time course throughout the waking period . Despite these different dynamics , similar bioluminescence levels were reached in both tissues . After wake-to-sleep transitions , peripheral PER2 bioluminescence initially increased before quickly decreasing and then leveling out at around 84% ( Figure 1C ) . In contrast , the central signal decreased linearly throughout the sleep period reaching levels of 91% at the end . Changes in sleep-wake state are accompanied by physiological alterations , such as changes in body and brain temperature ( Sela et al . , 2020; Vishwakarma et al . , 2021 ) , which could influence bioluminescence by changing substrate levels and/or the rate of the enzymatic reaction . Although the circadian rhythms of subcutaneous temperature and PER2 bioluminescence are ca . 4 hr out of phase ( Figure 1—figure supplement 6 ) , this does not exclude the possibility that fast changes in physiology associated with sleep-wake transitions contribute to luciferin availability . Therefore , we assessed sleep-wake-related changes in mice carrying two other luciferase reporter constructs . Besides the CAG-Luc mice we used to decide on the route of luciferin administration ( Figure 1—figure supplement 2B ) , we also had access to Pkg1-Luc mice in which bioluminescence is under the control of the promotor of the housekeeping gene Pkg1 ( see Materials and methods ) . As expected from a housekeeping gene , bioluminescence in Pkg1-Luc mice was relatively constant and did not increase at sleep-wake transitions ( Figure 1—figure supplement 5 ) . In contrast , in CAG-Luc mice increases at transitions were larger than in Per2Luc mice , consistent with the rapidly activated CMV-immediate-early enhancer contained within the CAG synthetic promotor ( Brightwell et al . , 1997; Collaco and Geusz , 2003 ) . These results show that the bioluminescence changes at sleep-wake transitions are specific to the luciferase reporter construct and therefore not an artifact of sleep-wake-related changes in physiology . Taken together , three observations indicate that the sleep-wake distribution importantly contributes to both central and peripheral PER2-bioluminescence dynamics: ( 1 ) enforced wakefulness has both acute and long-term effects on PER2 bioluminescence; ( 2 ) the changes in PER2 bioluminescence at sleep-wake transitions demonstrate that also spontaneous waking is associated with an increase , and sleep with a decrease , in PER2 bioluminescence; and ( 3 ) that PER2 bioluminescence is high when mice are spontaneously awake more . The central PER bioluminescence signals tended to be noisier than peripheral signals ( signal-to-noise ratio: –1 . 15 ± 1 . 0 and 0 . 37 ± 1 . 0 dB , respectively; estimated on 3 min values in baseline of the 6 hr SD experiments according to Leise et al . , 2012 ) . Moreover , the peripheral signal was easier to obtain and minimally invasive to the animal . We therefore decided for the next experiments to focus on peripheral PER2 bioluminescence using the Per2Luc SKH1 mice . Because waking correlates highly with LMA ( Figure 1—figure supplement 7 ) , we recorded LMA as proxy for wakefulness to avoid invasive EEG/EMG surgery and mice having to adapt to a ca . 2 . 7 g recorder mounted on the head . The SCN is the main driver of the circadian sleep-wake distribution because animals in which the SCN is lesioned ( SCNx ) lack a circadian organization of sleep-wake behavior under constant conditions ( Baker et al . , 2005; Edgar et al . , 1993 ) . Under these conditions , the amplitude of clock gene expression in peripheral organs is significantly reduced , but not eliminated ( Akhtar et al . , 2002; Curie et al . , 2015; Tahara et al . , 2012; Sinturel et al . , 2021 ) . Given our results in freely behaving mice , we expect that imposing a rhythmic sleep-wake distribution in SCNx mice will reinstate high-amplitude rhythmicity in PER2 bioluminescence . Conversely , reducing the amplitude of the circadian sleep-wake distribution in SCN-intact mice is expected to reduce the amplitude of PER2 bioluminescence . We tested these predictions in two complementary experiments ( see experimental design in Figure 1—figure supplement 1 ) . In the first experiment , we enforced a daily sleep-wake rhythm in arrhythmic SCNx mice by sleep depriving them for 4 hr at 24 hr intervals during four subsequent days . In the second experiment , we aimed to acutely reduce the circadian distribution of sleeping and waking in intact mice according to a ‘2hOnOff’ protocol , comprising 12 2-h SDs each followed by a 2 h sleep opportunity ‘window’ ( SOW ) , previously utilized in the rat ( Yasenkov and Deboer , 2010 ) . Lesioning the SCN rendered LMA arrhythmic under DD conditions ( Figure 2—figure supplement 1 ) . In arrhythmic SCNx mice , we confirmed that rhythms in PER2 bioluminescence were strongly reduced , but not completely abolished ( Figure 2A–C ) . During the second baseline measurement after SCNx ( BSL2 in Figure 2B ) , we observed in two of the four mice erratic high values ( see also Figure 1—figure supplement 2 ) that we cannot readily explain especially because in the subsequent four SD days the variance in the PER2 bioluminescence signal among mice was again as small as in the earlier recordings ( Figure 2A vs . B ) . The erratic values in these two mice led to a >10-fold increase in the residual sum of squares ( RSS ) , indicating a poorer fit , and to a higher amplitude of the fitted sinewaves in the second compared to the first baseline recording ( Figure 2C , BSL1 vs . BSL2 ) . The repeated SDs induced a robust oscillation in the PER2-bioluminescence signal , restoring amplitude to the levels observed prior to SCN lesioning and higher than those observed in the baseline recordings ( Figure 2C ) . The latter observation shows that the increased amplitude during the SD did not result from an SD-mediated alignment of different individual phases in baseline . Importantly , the oscillation continued after the end of the SDs while decreasing in amplitude ( Figure 2C ) . In the next experiment , we aimed at reducing the amplitude of the circadian sleep-wake distribution in intact mice using the 2-day 2hOnOff protocol . We measured sleep-wake state and PER2 bioluminescence during baseline conditions , the 2hOnOff procedure , and the two recovery days in two cohorts of mice . In the first cohort , PER2 bioluminescence was monitored , and mice were taken out of the RT-Biolumicorder for the 2 hr SDs . Bioluminescence was therefore quantified only during the SOWs , thus biasing the read-out of PER2 bioluminescence during the 2hOnOff protocol towards levels reached during sleep . A second cohort was implanted with tethered EEG/EMG electrodes to determine the efficacy of the 2hOnOff protocol in reducing the sleep-wake distribution amplitude . As expected , mice exhibited a circadian PER2-bioluminescence rhythm under baseline conditions ( Figure 3A ) . Contrary to expectation , the 2hOnOff protocol did not significantly decrease the ongoing circadian PER2 oscillation when analyzed at the group level ( Figure 3B ) . When inspecting the individual responses , four out of the six mice did show a consistent 27% reduction ( range 23–31% ) in PER2-biolumininescence amplitude during the 2-day 2hOnOff protocol , while in the remaining two mice amplitude increased by 40% ( Figure 3C ) . In all six mice , PER2-bioluminescence amplitude reverted to baseline values during recovery , irrespective of whether it was increased or decreased during the 2hOnOff protocol . The distinct bimodal , opposing response in amplitude observed among individual mice and the subsequent reverting back to baseline suggest that the 2hOnOff protocol did affect the ongoing PER2-bioluminescence rhythm . This phenomenon might relate to factors not accounted for in our experimental design . In the modeling section at the end of the Results section , we explore this observation further and find that variation in circadian phase could underlie the individual differential response ( for details , see ‘Modeling PER2 dynamics’ ) . We used the second cohort of mice to determine the efficacy with which the 2hOnOff protocol reduced the circadian sleep-wake distribution . In contrast to bioluminescence , the amplitude of the circadian sleep-wake distribution did decrease consistently in all mice ( Figure 3D and E ) . The circadian rhythm in the sleep-wake distribution was , however , not eliminated because the sleep obtained during the 2 hr SDs as well as during the 2 hr SOWs both varied as a function of time of day ( sleep during SDs: average: 12 . 4% , min-max: 4 . 9–23 . 0%; during SOWs: average: 54 . 3% , min-max: 35 . 7–76 . 4% ) . This was especially evident at the beginning of the subjective light phase of day 2 when the average time spent asleep during the SD reached 23% . The sleep obtained during the SDs at this time could be due to a substantial sleep pressure because of lost sleep during day 1 of the 2hOnOff protocol ( total time spent asleep/day , mean ± SEM; baseline: 11 . 1 ± 0 . 1 hr; 2hOnOff: day 1: 7 . 3 ± 0 . 3 hr , day 2: 8 . 6 ± 0 . 7 hr , paired two-tailed t-test , BSL vs . 2hOnOff-day 1: t ( 7 ) = 8 . 86 , p<0 . 001; BSL vs . 2hOnOff-day 2: t ( 7 ) = 4 . 18 , p=0 . 004; 2hOnOff-day 1 vs . -day 2 , t ( 7 ) = 1 . 97 , p=0 . 09 ) , combined with the difficulty for the experimenters to visually detect and prevent sleep in pinkish hairless mice under dim-red light conditions . During recovery from the 2hOnOff protocol , baseline amplitudes were re-established ( Figure 3E ) . The results above demonstrate that both circadian and sleep-wake-dependent factors need to be considered when studying PER2 dynamics . To understand how these two factors collectively generate the variance in PER2 bioluminescence , we put our experimental data into a theoretical framework and modeled changes in peripheral PER2 bioluminescence according to a driven harmonic oscillator ( Curie et al . , 2013 ) . This type of oscillator is not self-sustained but depends on rhythmic forces to set in motion and maintain the oscillation . In the model , we assume the sleep-wake distribution to represent one such force . In the absence of rhythmic forces , the oscillator loses energy , resulting in a gradual reduction of its amplitude according to the rate set by its damping constant ( γ ) . Besides amplitude and timing ( i . e . , phase ) of the recurring forces and the damping constant , the oscillatory system is further defined by a string constant ( ω02 ) with the natural angular frequency ( ω0 ) defining the intrinsic period . Because waking was not quantified alongside bioluminescence in the SCNx and the 2hOnOff experiments , we used LMA as a proxy for the driving force provided by wakefulness ( F→WAKE ) . The 6 hr SD experiment showed that LMA and time spent awake are strongly correlated and that their hourly dynamics closely changed in parallel ( Figure 1—figure supplement 7 ) . Moreover , applying the model to this data set ( see below ) using either wakefulness or LMA as F→WAKE yielded similar fits ( Figure 4—figure supplement 1 ) , demonstrating that LMA is an appropriate proxy for wakefulness for the purpose of our model . We performed the analyses at the group level; that is , the mean LMA level of all mice was used to reflect F→WAKE , and the free parameters in the model were optimized by fitting the motion of the oscillator to the mean PER2-bioluminescence levels ( see Materials and methods ) . Besides F→WAKE , we implemented a circadian force to account for the residual PER2-bioluminescence rhythm observed in the SCNx mice ( Figure 2 ) . We assumed that this additional SCN and sleep-wake-independent force reflects a peripheral circadian process ( F→PERI ) present in both intact and SCNx mice ( Sinturel et al . , 2021 ) . This force was modeled as a sinewave with amplitude and phase as free parameters in both experiments while period was set to the period estimated from the baseline PER2-bioluminescence rhythm observed in the SCNx mice ( 23 . 7 ± 0 . 6 hr; n = 4 ) . A schematic overview of the influence of F→PERI and F→WAKE on PER2 bioluminescence is presented in Figure 4A . We first optimized the parameters of the model describing PER2-bioluminescence dynamics in the SCNx experiment ( data from Figure 2 ) , and then predicted PER2 bioluminescence under the 2hOnOff experiment . F→PERI ’s amplitude and phase again required optimization as both differed between the two experiments ( nonoverlapping 95% CI for both in Table 1 ) . With the parameters listed in Table 1 , the model captured the dynamic changes in PER2 bioluminescence in the SCNx experiment with high precision including the residual rhythmicity in baseline , the reinstated pronounced rhythmicity during the four SDs , and its subsequent dampening thereafter ( Figure 4B , black line , Table 2 ) . The model also accurately captured average bioluminescence dynamics in the 2hOnOff experiment ( Figure 4C , black line , Table 2 ) . It furthermore predicted an 18% reduction of PER2-bioluminescence amplitude during the 2hOnOff protocol compared to baseline ( relative amplitudes estimated by the model: 0 . 31 vs . 0 . 38 [a . u . ]; Figure 4C ) , consistent with our hypotheses and the 27% reduction in PER2-bioluminescence amplitude observed in the four mice in which amplitude did decrease ( see Figure 3C ) . To further test the performance of the model , we predicted the peripheral PER2-bioluminescence data obtained in the 6 hr SD experiment using the parameters optimized for the 2hOnOff experiment ( Figure 1A , Table 1 ) . Also , the results of this experiment could be accurately predicted with the model , including the higher PER2-bioluminescence levels reached over the initial recovery hours after SD and the subsequent longer-term reduction in amplitude ( Figures 1A and 4D , Table 2 ) . The model could not reliably predict the central PER2-bioluminescence dynamics mainly due to an earlier phase of the central compared to the peripheral signal ( Figure 4—figure supplement 2D ) . Moreover , the increase in PER2 levels immediately following the SD ( Figure 1A ) was missed by the model , further underscoring the tissue-specific relationship between time-spent-awake and PER2 requiring the model to be optimized according to the tissue under study . The model accurately captured peripheral PER2-bioluminescence dynamics under three different experimental conditions . The model’s robust performance prompted us to assess in silico whether ( 1 ) both forces are required to predict PER2-bioluminescence dynamics , ( 2 ) lesioning the SCN affects the forces exerted on PER2 bioluminescence under undisturbed , DD conditions , and ( 3 ) differences in the circadian phase could predict the opposing response in PER2-bioluminescence amplitude among mice in the 2hOnOff experiment . To evaluate if F→PERI and F→WAKE combined are required to predict PER2 dynamics , we compared the performance of the model by dropping these two modeling terms; that is , removing either F→PERI or F→WAKE . Removing F→PERI from the model did not change the coefficient obtained for F→WAKE ( 8 . 25e-5 , which is still within the 95% CI estimated in the full model; see Table 1 ) . However , this simpler model ( only four free parameters to estimate compared to six in the full model ) could not reliably predict the bioluminescence data of the SCNx and 2hOnOff experiments ( Figure 4B and C , blue lines ) and fit statistics indicated a poorer fit ( i . e . , a higher Bayesian information criterion [BIC] score , see Table 2 ) . For instance , without F→PERI the model is unable to capture the residual PER2 rhythmicity in the baseline of the SCNx experiment ( Figure 4B , ‘model 1’ until 72 hr ) , and it took longer for the PER2 bioluminescence to reach high-amplitude levels compared to the full model because of F→PERI being in phase with the timing of the SDs in the full model . With the same strategy , we evaluated the model’s performance when dropping F→WAKE ( Figure 4B and C , green lines ) . As the sleep-wake distribution no longer affects the model , F→PERI dynamics are unperturbed throughout the experiment and can therefore be solved as a sinewave in the differential equation . Fit statistics for the SCNx experiment were poor compared to the full model as the effects of SDs on PER2 amplitude could not be captured ( Figure 4B , Table 2 ) . Removing F→WAKE when modeling the 2hOnOff experiment resulted in BIC scores similar to those obtained with the full model ( Table 2 ) , although the amplitude reduction during the 2hOnOff protocol could not be captured ( Figure 4C ) . Also , the PER2-bioluminescence dynamics in the 6 hr SD experiment were captured best when using both F→WAKE and F→PERI ( Figure 4D , Table 2 ) . Together , these results demonstrate that both forces are required to accurately predict PER2 dynamics under the tested experimental conditions . To answer the second question , we compared the forces driving PER2 bioluminescence in the baseline recordings of the 2hOnOff ( intact mice ) and the SCNx experiments . The model estimated that the average absolute force ( F→WAKE , F→PERI , F→γ , and F→ω02 combined ) exerted on the relative PER2-bioluminescense levels was more than twice as high in intact mice compared to SCNx mice ( 2 . 11e-2 vs . 0 . 90e-2 h-2 ) . This large difference was not due to a difference in F→WAKE , which was somewhat lower in intact mice ( 2hOnOff vs . SCNx; 2 . 01e-3 vs . 2 . 22e-3 h-2 ) . Although F→PERI was higher by 32% ( 2 . 45e-3 vs . 1 . 85e-3 h-2 ) , this difference did not substantially contribute to the higher amplitude of the PER2 rhythm in intact mice . To illustrate this , we substituted the value of F→PERI obtained in the 2hOnOff baseline with the lower value obtained in the SCNx baseline , which resulted in a 17% amplitude reduction of the PER2 rhythm . The presence of a circadian sleep-wake distribution impacted PER2 amplitude to a larger extent: running the simulation with the parameters estimated for the 2hOnOff experiment but with the sleep-wake distribution of the SCNx mice resulted in a 32% reduction . Therefore , the circadian sleep-wake organization is an important contributor to high-amplitude oscillations in PER2 . In the 2hOnOff experiment , the larger PER2 momentum resulted in a four times higher string and damping forces ( 2hOnOff vs . SCNx; F→ω02 : 19 . 9e-3 vs . 4 . 4e-3 h-2; F→γ: 9 . 74e-4 vs . 2 . 52e-4 h-2 ) that together with the larger F→PERI underlie the larger average absolute force . Moreover , this analysis demonstrated that the direct effects of F→WAKE and F→PERI on PER2 bioluminescence do not depend on an intact SCN and that the magnitude of the two forces is comparable . Although the model predicted the expected decrease in PER2-bioluminescense rhythm amplitude in the 2hOnOff experiment , this decrease was not observed in the mean bioluminescence data . The individual data showed , however , that in four mice this intervention did decrease PER2-bioluminescence amplitude , while in the remaining two amplitude increased ( Figure 3C ) . With the model , we addressed the third question: Do circadian phase differences predict the opposite effects of the 2hOnOff intervention on amplitude ? Surprisingly , when systematically varying the phase of F→PERI during the baseline prior to the 2hOnOff intervention ( Figure 4—figure supplement 2C ) , we found that at phase advances larger than 2 . 5 hr , the model predicted an increase of PER2 amplitude during the subsequent 2hOnOff protocol instead of a decrease ( illustrated in Figure 4—figure supplement 2A for a 6 hr phase advance ) . Circadian phase differences among animals might relate to individual differences in period length of their free-running rhythms that accumulate over time . We did not find evidence for a difference in PER2 bioluminescence or LMA phase at the end of the baseline recording ( 2 . 5 days under DD ) between animals that showed a decrease in PER2-bioluminescence amplitude compared to animals that showed an increase . However , during the first day of recovery following the 2hOnOff protocol ( 5 . 5 days under DD ) , we observed a ca . 2 hr phase advance in bioluminescence and 1 hr phase advance in LMA in the two mice that increased their amplitude during the preceding 2hOnOff protocol ( Figure 4—figure supplement 2B , left ) compared to the four mice for which we observed the anticipated decrease in PER2 amplitude ( Figure 4—figure supplement 2B , right ) . Whether these differences in phase contributed to the opposite response cannot be answered with the current data set . Nevertheless , the model yielded a perhaps counterintuitive but testable hypothesis by demonstrating that phase angle between the sleep-wake distribution and peripheral circadian clock-gene rhythms is an important variable in predicting outcome and emphasizes the importance of carefully controlling the initial conditions . Per2 transcription can be initiated from its cognate E-boxes by CLOCK/NPAS2:ARNTL . This transcriptional activation is at the core of the TTFL and drives the circadian changes in PER2 . Enforced wakefulness not only affects Per2 levels but also modulates the expression of other components of the TTLF ( Mang and Franken , 2015; Hor et al . , 2019 ) . The SD-evoked increase in Per2 expression could therefore be mediated through other clock genes in the circuitry , as was demonstrated by the differential SD-evoked response in Per2 levels in mice lacking the core clock genes Npas2 and both Cry1 and –2 genes ( Franken et al . , 2006; Wisor et al . , 2002 ) . Apart from a TTFL-mediated activation , Per2 transcription can be induced by other signaling molecules directly acting on elements within the Per2 promotor ( Schibler et al . , 2015 ) . For example , ligand-bound glucocorticoid receptors can induce Per2 transcription by binding to their glucocorticoid response elements ( Cheon et al . , 2013; So et al . , 2009 ) . Similarly , cAMP response element ( CRE ) -binding protein ( CREB ) , heat-shock factor 1 ( HSF1 ) , and serum-response factor ( SRF ) can directly activate Per2 transcription through CREs , heat-shock elements , and CArG-boxes , respectively , present in the Per2 gene ( Gerber et al . , 2013; Saini et al . , 2012; Tamaru et al . , 2011; Travnickova-Bendova et al . , 2002 ) . Through these pathways , Per2 responds to stress , light , temperature , blood-borne systemic cues , and cellular activation as an immediate early gene ( IEG ) . Because of this , Per2 can appear rhythmic even in the absence of a functional TTFL , provided these signaling pathways fluctuate cyclically ( Kornmann et al . , 2007 ) . In behaviorally arrhythmic SCNx animals , the residual PER2 rhythms we observed might similarly result from SCN-independent corticosterone ( Andrews , 1968 ) or body temperature ( Satinoff and Prosser , 1988 ) rhythms , or , alternatively , might be TTFL-driven locally ( Sinturel et al . , 2021 ) . Important for the current study is that several of the pathways known to directly influence Per2 expression are activated by either spontaneous and/or enforced waking ( e . g . , corticosterone [Mongrain et al . , 2010] , temperature [Hoekstra et al . , 2019] , Hsf1 and Srf [Hor et al . , 2019] , pCREB [Cirelli and Tononi , 2000] ) and are therefore good candidates linking sleep-wake state to changes in PER2 . The observed changes in PER2 bioluminescence were rapid and suggest that increases in protein can occur within an hour of spontaneous wakefulness . Other studies document that PER genes can indeed be rapidly transcribed and translated . For instance , a light pulse given at CT14 leads within an hour to a significant increase in Per2 transcript in the SCN ( Yan and Silver , 2002 ) . One study reported a large increase in hepatic Per2 transcript levels within 1 hr after food presentation in fasted rats ( Wu et al . , 2010 ) , underscoring the ability of this transcript to rapidly adapt to homeostatic need . Per2 translation is not solely dependent on de novo transcription and , for example , in the SCN , light was shown to promote Per2 translation ( Cao et al . , 2015 ) , suggesting that transcription would not be necessary to increase PER2 protein levels . Such mechanism could also underlie the very fast ( <30 min ) 1 . 5-fold increase in PER2 protein observed in fibroblasts after serum shock ( Cao et al . , 2015 ) . Finally , as the PER2 protein levels measured are the net result of translation and degradation , also sleep-wake-dependent changes in PER2-degradation rate may contribute both to its increase during wakefulness and its decrease during sleep . PER2 degradation is crucial in setting TTFL period and the timing and stability of the circadian sleep-wake distribution ( Chong et al . , 2012; D’Alessandro et al . , 2017 ) . One established pathway leading to PER2 degradation involves Casein kinase 1 ( Csnk1 ) -mediated phosphorylation ( Eide et al . , 2005 ) followed by the recruitment of the ubiquitin ligase β-transducin repeat-containing proteins ( Btrc ) ( Masuda et al . , 2020; Ohsaki et al . , 2008; Reischl et al . , 2007 ) . Other kinases , such as Salt-inducible kinase 3 ( Sik3 ) ( Hayasaka et al . , 2017 ) , and phosphorylation-independent ubiquitin ligases , such as Transformed mouse 3T3 cell double minute 2 ( Mdm2 ) ( Liu et al . , 2018 , p . 2 ) , also target PER2 for degradation . Using a modeling approach to estimate the role of Btrc in circadian period length , Reischl et al . , 2007 estimated a linear decay rate of 0 . 18/hr for PER2 degradation , which is not inconsistent with the approximately 0 . 10/hr net decay rate we observed for the PER2 bioluminescence during sleep . However , the dynamics of PER2 degradation have been assessed in a circadian context exclusively , and effects of sleep-wake state have not been quantified previously . The obvious next step is to determine which pathway ( s ) contributes to the wake-driven changes in PER2 protein . We already established that the SD-incurred increase in Per2 in the forebrain partly depended on glucocorticoids ( Mongrain et al . , 2010 ) . Along those lines , restoration of daily glucocorticoid rhythms in adrenalectomized rats reinstates PER2 rhythms in several extra-SCN brain areas ( Segall and Amir , 2010 ) . To determine the contribution of the aforementioned wake-driven factors , a genetic screen could be deployed where one-by-one the regulatory elements in the Per2 promoter are mutated , and the effect of sleep-wake driven Per2 changes is assessed . This approach has already been taken for the GRE and CRE elements in the Per2 promoter to test their respective roles in circadian phase resetting and integrating light information ( Cheon et al . , 2013; So et al . , 2009; Travnickova-Bendova et al . , 2002 ) . The model accurately captured the main features of peripheral PER2 dynamics observed in all three experiments , thereby giving further credence to the notion that sleep-wake state importantly contributes to the changes in PER2 observed in the periphery . Moreover , it demonstrated that the large amplitude of the circadian PER2 rhythm in intact mice is likely the result of the momentum gained in the harmonic oscillator through the daily recurring sleep-wake distribution . This is conceptually different from a currently accepted scenario , in which direct and indirect outputs from the SCN assure phase coherence of locally generated self-sustained circadian rhythms ( Schibler et al . , 2015 ) . According to this model , loss of amplitude observed at the tissue level in SCNx mice is caused by phase dispersion of the continuing rhythms in individual cells . Among the SCN outputs thought to convey phase coherence are feeding , LMA , and changes in temperature . Because these outputs all require or are associated with the animal being awake , it can be argued that in both models the circadian sleep-wake distribution is key in keeping peripheral organs rhythmic . The harmonic oscillator model further showed that , although important , the sleep-wake force alone was not sufficient to predict PER2 dynamics . In addition to account for the residual PER2 rhythms observed in undisturbed SCNx mice , the SCN-independent and sleep-wake-independent circadian force greatly improved the performance of the model . The synergistic effect of both forces ( F→PERI and F→WAKE ) was needed to explain the rapid response to the SDs observed in the SCNx experiment and also in maintaining robust circadian rhythms during the SDs in the 2hOnOff experiment as illustrated in Figure 4B and C , respectively . Furthermore , this synergistic effect greatly depended on the relative phase of the two forces as we could illustrate in silico for the 2hOnOff experiment: a relative subtle change in the phase of F→PERI might underlie the increase ( instead of the predicted decrease ) in PER2 amplitude in two of the six mice recorded . Which pathways set the phase of F→PERI and whether it is truly independent of F→WAKE , as assumed in the model , our current results cannot answer . In an earlier modeling effort using a similar approach in SCN-intact , light-dark entrained mice , we also required a second force to correctly predict the phase of the observed rhythm in brain Per2 expression ( Curie et al . , 2013 ) . In that publication , we based the second force on the pattern of corticosterone production sharply peaking at ZT11 just prior to the light-dark transition . The phase of F→PERI in the 2hOnOff experiment , which followed a more gradual , sinewave function of which values became positive shortly after ZT11 ( extrapolated from the preceding LD cycle ) , seems consistent with this . In the SCNx experiment , the phase of F→PERI was positioned ~7 hr earlier with a positive driving force starting at the end of each of the SDs . As SD is accompanied by an increase in corticosterone ( Mongrain et al . , 2010 ) , the phase of F→PERI could be associated with corticosterone signaling also in the SCNx experiment . Thus in intact mice , the SCN output would dictate the phase of corticosterone production in the adrenals ( and thus that of F→PERI ) , while in SCNx mice the phase of the corticosterone rhythm can be reset by stressors such as SD . As PER2 and Per2 levels in the SCN seem insensitive to SD ( Curie et al . , 2015; Zhang et al . , 2016 ) , this could explain why the phase of F→PERI is maintained in sleep-deprived SCN-intact mice . Moreover , ex vivo experiments demonstrated that the adrenal gland can generate bona fide circadian rhythms in corticosterone release independent of the SCN ( Andrews , 1968; Engeland et al . , 2018; Kofuji et al . , 2016 ) , even though SCNx is generally thought to abolish rhythms in circulating corticosterone levels ( Moore and Eichler , 1972 ) . While rhythmic corticosterone release represents a plausible candidate contributing to F→PERI , especially considering its role in synchronizing peripheral clocks ( Balsalobre et al . , 2000; Cuesta et al . , 2015; Dickmeis , 2009; Le Minh et al . , 2001 ) , our current results cannot rule out other sources underlying or contributing to F→PERI . Above we argued that also wakefulness ( i . e . , F→WAKE in the model ) could influence peripheral PER2 through corticosterone signaling , further complicating the issue . Indeed , adrenalectomy was found to reduce ( but not abolish ) the SD-induced increase in Per2 expression in the forebrain ( Mongrain et al . , 2010 ) . However , enforced but not spontaneous wakefulness is accompanied by increases in corticosterone and the model could predict PER2 dynamics without having to distinguish between the two types of waking . Therefore , other candidate signals among those listed above must be considered to understand the biological basis of F→PERI and F→WAKE . Model optimization yielded an unexpected short 21 . 8 hr period for the natural frequency of the PER2 oscillator , which , in addition , differed from the 23 . 7 hr period we set for F→PERI . While in the intact mice of the 2hOnOff experiment it is difficult to independently determine F→PERI ’s period , we estimated a 23 . 7 hr period length for the residual PER2 rhythmicity observed during baseline in SCNx mice , which we assume is driven by F→PERI . In intact mice kept under constant conditions , SCN output drives behavioral sleep-wake rhythms and synchronizes peripheral clock-gene rhythms forcing the entire system to oscillate at the intrinsic period of the SCN . Similarly , in behaviorally arrhythmic SCNx mice , we assume that the period of the observed residual PER2 rhythm reflects that of the only remaining driver , F→PERI , as F→WAKE is no longer rhythmic and direct effects of the SCN are absent . Ex vivo experiments showed that periods vary among tissues and do not depend on whether tissues were obtained from an intact or SCNx mouse ( Cederroth et al . , 2019; Yoo et al . , 2004 ) , pointing to tissue-specific TTFLs , which we assume to underlie the intrinsic rhythmicity of both F→PERI and PER2 bioluminescence in our experiments . The difference in period length of F→PERI and that of the intrinsic PER2 oscillator therefore suggests that F→PERI is not of renal origin; that is , the tissue that contributed most to the bioluminescence signal we recorded . Our data demonstrated that both central and peripheral PER2-bioluminescence dynamics are affected by sleep-wake state , not only after SD but also after spontaneous periods of wakefulness . Despite this general observation , we found clear tissue-specific differences: the 6 hr SD elicited an immediate increase in the central PER2 signal , while in the periphery this increase occurred several hours later , confirming our earlier findings in brain versus liver and kidney ( Curie et al . , 2015 ) . Tissue specificity was also observed after spontaneous bouts of wakefulness: in the brain , PER2 bioluminescence immediately increased after the animal woke up and continued to do so until the end of the waking bout , whereas in the kidney the increase in PER2 bioluminescence became apparent only 5–10 min after the transition . Similar differences in PER2 dynamics were observed after falling asleep , albeit in opposite direction . Also , the model suggested a tissue specificity as central dynamics of the 6 hr SD experiment could not be accurately predicted with the parameters optimized on peripheral PER2-bioluminescence data . In its current form , the model describes the global effects of external forces ( the sleep-wake distribution and the SCN-independent circadian force ) on the behavior of the oscillator , but not their acute effects . Translated into molecular terms , the model only makes predictions on the TTFL aspect of PER2 regulation , not on PER2 as an IEG . Accordingly , the model cannot capture the fast changes at the transitions . Similarly , the short-lasting high levels of PER2 observed in the brain immediately after the 6 hr SD might reflect an IEG response rather than the state of the TTFL oscillator , which would explain why the model could not predict it . As sleep-wake states are brain states , it stands to reason that changes in brain PER2 levels capture more of the acute IEG effects than in the periphery . One could test this hypothesis by quantifying the PER2 response after activating a peripheral tissue , provided this can be achieved without affecting sleep-wake state as well . The method we implemented to quantify PER2 protein levels presents advantages over previous methods used . It enabled us to acquire data at a time resolution needed to link changes in PER2 to sleep-wake state transitions in individual mice . Moreover , because of the within-subject experimental design , there is a substantial reduction in data variability , and , as illustrated with the effects of individual phase on PER2 amplitude in the 2hOnOff experiment , we could assess the presence of individual differences in the initial conditions that might influence experimental outcome . Finally , the number of mice needed for these experiments has been greatly reduced while obtaining better quality data . A limitation of this method is the assumption that changes in bioluminescence reflect changes in PER2 protein levels . Using western blot , we previously validated that changes in bioluminescence during baseline and after a 6 hr SD indeed reflect changes in PER2 protein ( Curie et al . , 2015 ) . Nevertheless , substrate availability can importantly contribute to the signal as demonstrated in the experiment in which we delivered luciferin in the drinking water . Even the use of osmotic mini-pumps does not guarantee constant delivery as its release rate is temperature dependent ( Alzet , manufacturer’s notes ) and bioluminescence’s increase during wakefulness might therefore result from the accompanying increase in temperature during this state . Arguments against a possible temperature effect on bioluminescence changes come from a study in which , using the same osmotic mini-pumps , the expression of two clock genes known to oscillate in anti-phase could be confirmed ( Ono et al . , 2015 ) , which would not be possible if temperature was the main determinant of bioluminescence . In the current data , the circadian rhythm in wakefulness in the CAG-Luc and Pkg1-Luc mice was not accompanied by changes in bioluminescence . Moreover , we observed that the circadian rhythms of subcutaneous temperature and PER2 bioluminescence are ca . 4 hr out of phase ( Figure 1—figure supplement 6 ) , supporting that the large circadian changes in bioluminescence are not driven by changes in luciferin availability . Moreover , the lack of an increase in bioluminescence during wake bouts in Pkg1-Luc mice demonstrates that changes in rate-limiting availability of luciferin did not contribute to the fast sleep-wake-evoked changes in PER2 . In this study , we used a unique combination of methods allowing us to collect high-resolution data of sleep-wake state in conjunction with PER2 levels and found that the sleep-wake distribution profoundly affects PER2 bioluminescence both short and long term . Such behavior-dependent plasticity of the time-keeping machinery in tissues peripheral to the SCN enables the organism to respond to challenges as time-restricted feeding experiments have demonstrated ( Damiola et al . , 2000; Saini et al . , 2013 ) . Besides its importance in regulating feeding and energy homeostasis ( Bass and Takahashi , 2010 ) , the clock circuitry also plays a prominent role in sleep homeostasis ( Franken , 2013 ) . PER2 seems perfectly suited as an integrator of sleep-wake state and circadian time because it is sensitive to a variety of sleep-wake-driven signals . Our model suggests that having a large amplitude rhythm protects from acute disturbances of sleep as observed in the 2hOnOff experiment , while sleep-depriving arrhythmic SCNx mice had immediate and large effects on PER2 . These rapid effects could only be achieved through the synergistic effect of a second force that we found to be independent of the SCN and the sleep-wake distribution . The coordination of the sleep-wake force and this second force in the model was critical in predicting the effects of sleep-wake interventions on PER2 . Research on the nature of this second force would therefore be important to facilitate phase resetting and normalize disrupted clock gene rhythms under conditions of jet lag and shift work , complementing strategies aimed at altering the timing of the central pacemaker . To measure peripheral PER2-bioluminescence levels , we made use of the Per2Luc KI construct ( Yoo et al . , 2004 ) . The KI construct was originally generated on a C57BL/6J-129 mixed background and subsequently brought onto a C57BL/6J ( B6 ) background by backcrossing for at least 11 generations . These mice were then crossed with outbred SKH1 mice ( Crl:SKH1-Hrhr; Charles River ) to create hairless Per2Luc KI mice . We used male homozygous Per2Luc KI B6 and hairless heterozygous Per2Luc KI SKH1-B6 hybrid ( here referred to as SKH1 mice ) mice . Mice were kept under a 12 hr light/12 hr dark cycle with light- and dark onset referred to as Zeitgeber time ( ZT ) -0 and -12 , respectively . Age at time of recording varied between 12 and 24 weeks . Food and water was available ad libitum , and after surgery mice were singly housed . All experiments were approved by the Ethical Committee of the State of Vaud Veterinary Office Switzerland under license VD2743 , 3201 , and 3402 . Because Per2Luc KI mice express ubiquitously luciferase and the RT-Biolumicorder cannot discriminate between different sources of bioluminescence , we assessed which peripheral organ ( s ) was/were the major source of bioluminescence . Two male heterozygous Per2Luc SKH1 mice were implanted with an osmotic mini-pump ( model 1002; 35 mg/mL luciferin ) and 5 days later lightly anesthetized with 2 . 5% isoflurane and imaged for 60 s ( Xenogen IVIS Lumina II ) around ZT6 . The main source of dorsal bioluminescence overlapped with the expected location of the kidney , whereas ventrally almost no bioluminescence was detected ( see Figure 1—figure supplement 2A ) . Most bioluminescence quantified during the experiment is of dorsal origin due to the orientation of the mouse relative to the PMT , suggesting that the kidneys are the main source of peripheral bioluminescence in Per2Luc SKH1 mice . To confirm our previous report that the main source of central bioluminescence was the brain ( Curie et al . , 2015 ) , we imaged one male B6 PER2::LUC mouse at ZT3 again using the IVIS system ( Figure 1—figure supplement 2 ) . When luciferin is infused centrally and the skull is thinned and a glass cone mounted , all emitted bioluminescence originates from the head of the mouse , with no detectable signal from the periphery . For details on the surgery see ‘Surgical procedures and experimental design . ’ . In a second pilot experiment , we investigated the optimal route of luciferin administration . Although luciferin administration via drinking water has been used before to measure bioluminescence ( Saini et al . , 2013 , Iwano et al . , 2018 , Hall et al . , 2018 , Sinturel et al . , 2021 ) , we were concerned that this route could limit luciferin availability in a circadian fashion because drinking behavior has a strong circadian rhythm ( Bainier et al . , 2017 ) . To address these concerns , we made use of mice expressing constitutively luciferase under control of the synthetic CAG promotor ( Cao et al . , 2004 , Jackson catalog number 008450 ) , thus allowing for testing of circadian fluctuating levels of luciferin . Mice received luciferin via the drinking water or via an osmotic mini-pump and served as their own control . Four male CAG-Luc mice were housed for two subsequent experiments in constant darkness in the RT-Biolumicorder . During the first experiment , 0 . 5 mg/mL luciferin was dissolved in the drinking water . At the end of this experiment , mice received subcutaneously an osmotic mini-pump ( Alzet , model 1002 ) under light anesthesia ( isoflurane; 2–4% mixed with O2 ) containing 70 mg/mL of luciferin and could recover for 2 days before bioluminescence and activity was monitored for the second experiment in the RT-Biolumicorder . Experimental design of the three main experiments is depicted in Figure 1—figure supplement 1 with the upper panel ( Experiment 1 ) illustrating the central and peripheral recordings of PER2 bioluminescence alongside EEG/EMG and LMA before , during , and after a 6 hr SD , and the middle and bottom panels ( Experiment 2 and 3 ) the SCNx and the 2hOnOff experiments , respectively . In the latter two experiments , peripheral PER2 bioluminescence and LMA were recorded . Mice were implanted with EEG and EMG electrodes under deep ketamine/xylazine anesthesia . Three gold-plated screws ( frontal , parietal , and cerebellar ) were screwed into the skull over the right cerebral hemisphere , where the cerebellar screw served as a reference for the other two electrodes . Two additional screws were used as anchor screws . For the EMG , a gold wire was inserted into the neck musculature along the back of the skull . For brain delivery of D-luciferin , a cannula ( Brain Infusion Kit1 , Alzet ) was introduced stereotaxically into the right lateral ventricle ( 1 mm lateral , 0 . 3 mm posterior to bregma , and 2 . 2 mm deep ) under deep anesthesia ( ketamine/xylazine; intraperitoneally , 75 and 10 mg/kg , respectively ) , and connected to the mini-pump . A depression ( diameter 2 mm ) was made in ( but not through ) the skull in a region of the left frontal cortex ( approximate coordinates 2 mm lateral to midline , 2 mm anterior to bregma ) , in which a glass cylinder ( length 4 . 0 mm; diameter 2 . 0 mm ) was positioned and fixed with dental cement . The EMG and three EEG electrodes were subsequently soldered to a connector and cemented to the skull . The cerebellar screw served as a reference for the parietal and frontal screw and the EMG . After the first recovery day , mice were habituated to the weight of the wireless EEG recording system by attaching a dummy of same size and weight to their connector . Two days before habituation to the RT-Biolumicorder , mice were implanted with the osmotic mini-pump ( model 1002 , Alzet; luciferin 35 mg/mL ) under light anesthesia . 8–10 days post-surgery , mice were placed in the RT-Biolumicorder at the end of the light phase ( ~ZT10-ZT12 ) for 2 days in LD to habituate to the novel environment . At the end of the second habituation day , the dummy was replaced with a wireless EEG ( NeuroLogger , TSE Systems GmbH ) . After two-and-a-half days of baseline recording in constant darkness , mice were sleep deprived for 6 hr at a time they were expected to rest ( ZT0 under LD conditions ) by gentle handling as described ( Mang and Franken , 2012 ) . In short , mice are left undisturbed as long as they do not show signs of sleep . Sleep is prevented by introducing and removing paper tissue , changing the litter , bringing a pipet in the animal’s proximity , or gentle tapping of the cage . As opposed to what the term might suggest , mice are not handled . After SD , mice were placed back into the RT-Biolumicorder for the subsequent two recovery days . Four SKH1 mice were recorded over the course of the experiment and served as their own control . Briefly , their PER2-bioluminescence rhythm was monitored before SCNx ( once ) , under undisturbed conditions post-SCNx ( twice ) , and after the second measure under SCNx conditions , the mice were subjected to the repeated 4 hr SDs . Bilateral lesion of the two SCNs was performed stereotaxically ( Kopf Instruments , 963LS , Miami Lakes , FL ) under ketamine/xylazine anesthesia ( intraperitoneal injection , 75 and 10 mg/kg , at a volume of 8 mL/kg ) . Two electrodes ( 0 . 3 mm in diameter ) were introduced bilaterally at the following coordinates ( position of the frontal electrode: anteroposterior using bregma as reference: ± 0 . 2 mm lateral , + 0 . 5 mm bregma , depth: –5 . 9 mm; the second electrode was positioned 0 . 7 mm posterior to the frontal one ) . Electrolytic lesions ( 1 mA , 5 s ) were made using a direct current ( DC ) lesion device ( 3500 , Ugo Basile , Comerio , Italy ) . After lesion , mice were housed in constant dark ( DD ) conditions for at least 10 days to verify absence of circadian organization of overt behavior . Activity was quantified using passive infrared sensors ( Visonic SPY 4/RTEA , Riverside , CA ) . ClockLab software ( Actimetrics , Wilmette , IL ) was used for data acquisition and analyses . SKH1 mice ( n = 8 ) were implanted with EEG and EMG electrodes as described previously ( Mang and Franken , 2012 ) to determine sleep-wake state . The surgery took place under deep xylazine/ketamine anesthesia . Briefly , six gold-plated screws ( diameter 1 . 1 mm ) were screwed bilaterally into the skull over the frontal and parietal cortices . Two screws served as EEG electrodes , and the remaining four screws anchored the electrode connector assembly . As EMG electrodes , two gold wires were inserted into the neck musculature . The EEG and EMG electrodes were soldered to a connector and cemented to the skull . Mice recovered from surgery during several days before they were connected to the recording cables in their home cage for habituation to the cable and their environment , which was at least 6 days prior to the experiment . The habituation to the room and the recovery from the 2-day SD procedure took place under LD 12:12 conditions . During the baseline recording and SD days , red light at very low intensity was present to allow the experimenters to observe the mice . Mice were sleep deprived for 2 hr according to the ‘gentle handling’ method . SKH1 mice ( n = 6 ) were implanted with an osmotic mini-pump ( Alzet , 1002 , luciferin concentration: 35 mg/mL; blue flow moderator ) 2 days before the habituation . At the end of the light phase ( ~ZT10-ZT12 ) , mice were moved from their cage to the RT-Biolumicorder for 2–3 days of habituation in LD . They were housed for 2 . 5 days in DD , after which the 2hOnOff protocol was initiated at light onset ( ZT0 ) under the preceding LD conditions . At the start of each SD , mice were moved from the RT-Biolumicorder and placed into a novel cage that was in the same room as the EEG-implanted mice . Fifteen minutes before the end of each SD , mice were brought back to their RT-Biolumicorder cage . The adenoviral vector used in the experiment was constructed by cloning the cassette flanked by PacI sites from vector prLV1 ( Du et al . , 2014 ) into vector pCV100 , an E1/E3-deleted replication-incompetent first-generation adenovirus vector based on human adenovirus serotype 5 ( a variant of pGS66 [Schiedner et al . , 2000] with an additional deletion of Ad5 nt 28133–30818 ) . In this construct , firefly luciferase is expressed from a bidirectionally active , minimal Pgk1 promoter ( note: the other side of the promoter carries Renilla luciferase cDNA , which was not used/measured in the framework of this study ) . See Supplementary file 1 for the construct sequence . Liver cells were transduced with the pCV100 vector ( 2 . 1e10 to 4 . 5e11 adenoviral particles ) via tail vein injection using an illuminated restrainer according to Saini et al . , 2013 and Sinturel et al . , 2021 . Male C57BL/6J mice kept under LD 12:12 were injected 2–5 days prior to pump implantation , and pump was implanted 3 days prior to commencing experiment in DD in the RT-Biolumicorder . To allow passage of photons emitted from liver , mid-section of mice’ backs was shaved . The temporal dynamic of PER2 bioluminescence was modeled according to the equation of motion describing a driven damped harmonic oscillator:d2xdt2+γdxdt+ω02x=F where x is the displacement of the oscillator , γ is the damping constant of the model , and ω02 is the string constant defining the natural frequency of the model . The driving forces used in this model are the LMA representing waking F→WAKE , and a circadian force F→PERI , represented as a sinewave . The momentary force exerted on the oscillator is represented as the sum of these two forces ( see Figure 4A ) :Ft=βLMAt+Asinωt+φ where β , A , and φ are respectively the coefficient applied on LMA amount , the amplitude of the circadian sinewave force , and the phase of the circadian sinewave force . These coefficients were the free parameters to be optimized in the model . ω is the angular velocity of the sinewave and was set to 2*pi/23 . 7 , based on the residual PER2-bioluminescence rhythm present in the baseline of SCNx mice . To solve this equation and optimize for parameters that best describe the observed PER2 bioluminescence , we proceeded as follows: the relative bioluminescence data from both experiments were averaged across mice using 30 min bins . LMA was averaged across mice using 6 min bins . LMA could not be measured directly during SDs and was estimated by assessing the increase in LMA during SDs measured in EEG-implanted mice , which was found to be 2 . 4 times higher compared to average baseline levels . Thus , the SD effect was estimated using 2 . 4 times the mean activity observed during baseline ( i . e . , 181 . 9 and 180 . 2 for the SCNx and 2hOnOff , respectively ) . To solve the second-order ordinary differential equation ( ODE ) of the driven harmonic oscillator , we transformed it into the following system of two first-order ODEs describing the change of position and speed of our oscillator:x1′=x2x2′=F−γx2−ω02x1 We then used the fourth-order Runge–Kutta ( RK4 ) numerical method to approximate the solution using a fixed time step of 0 . 1 hr . In the SCNx experiment , initial values of speed ( x20 ) and position ( x10 ) were set to 0 . For the 2hOnOff experiment , the model was generated for 20 days prior experiment to reach a steady state using replication of LMA observed in baseline ( T0–T24 ) . The position and speed of the oscillator at the end of the 20 days were taken as initial values for the fitting . We optimized the model for the following parameters: intercept ( equilibrium position of the oscillator ) , natural frequency , damping constant , and coefficient for the force exerted by LMA , amplitude of circadian force and phase of circadian force . We optimized the fitting by minimizing the RSS between predicted position of our model and the observed PER2-bioluminescence level . The box-constrained PORT routines method ( nlminb ) implemented in the optimx/R package ( Nash and Varadhan , 2011 ) was used to minimize the RSS . The goodness of fit of the model was assessed as follows . We assumed that the model errors follow a normal distribution and computed a BIC value for the model according toBIC=nln ( RSSn ) +kln ( n ) where n is the number of observations , and k is the number of optimized parameters +1 of the model . We approximated the Bayes factor ( BF ) between our model and a flat model ( linear model with intercept only ) as follows:BF≈exp ( −12 ( BICflat−BICmodel ) ) To compute confidence interval of our model parameters , we used 500 Monte Carlo simulations and calculated a confidence interval for our parameters based on 95% empirical quantiles ( 95% CI ) . The method was adapted from the code of Marc Lavielle ( Inria Saclay [Xpop] and Ecole Polytechnique [CMAP] , Université Paris-Saclay , France ) for nonlinear models , available here: http://sia . webpopix . org/nonlinearRegression . html . Statistics were performed in R ( version 4 . 0 . 0 ) , SAS ( version 9 . 4 ) , and Prism ( version 7 . 0 ) , with the threshold of significance set at α = 0 . 05 . Performed statistical tests are mentioned in the text and figure legends .	Circadian rhythms are daily cycles in behavior and physiology which repeat approximately every 24 hours . The master regulator of these rhythms is located in a small part of the brain called the supra-chiasmatic nucleus . This brain structure regulates the timing of sleep and wakefulness and is also thought to control the daily rhythms of cells throughout the body on a molecular level . It does this by synchronizing the activity of a set of genes called clock genes . Under normal conditions , the levels of proteins coded for by clock genes change throughout the day following a rhythm that matches sleep-wake patterns . However , keeping animals and humans awake at their preferred sleeping times affects the protein levels of clock genes in many tissues of the body . This suggests that , in addition to the supra-chiasmatic nucleus , sleep-wake cycles may also influence clock-gene rhythms throughout the body . To test this theory , Hoekstra , Jan et al . measured the levels of PERIOD-2 , a protein coded for by the clock gene Period-2 , while tracking sleep-wake states in mice . They did this by imaging a bioluminescent version of the PERIOD-2 protein in the brain and the kidneys , at the same time as they recorded the brain activity , movement and muscle response of animals . Results showed that PERIOD-2 increased on waking and decreased when mice fell asleep . Additionally , in mice lacking a circadian rhythm in sleep-wake behavior – whose changes in PERIOD-2 levels with respect to time were greatly reduced – imposing a regular sleep-wake cycle restored normal PERIOD-2 rhythmicity . Next , Hoekstra , Jan et al . developed a mathematical model to understand how sleep-wake cycles together with circadian rhythms affect clock-gene activity in the brain and kidneys . Computer simulations suggested that sleep-wake cycles and circadian factors act as forces of comparable strength driving clock-gene dynamics . Both need to act in concert to keep clock-genes rhythmic . The model also predicted the large and immediate effects of sleep deprivation on PERIOD-2 levels , giving further credence to the idea that waking accelerated clock-gene rhythms while sleeping slowed them down . Modelling also suggested that having regular clock-gene rhythms protects against sleep disturbances . In summary , this work shows how sleep patterns contribute to the daily rhythms in clock genes in the brain and body . The findings support the idea that well-timed sleep-wake schedules could help people to adjust to new time zones . It might also be useful to inform other strategies to reduce the health impacts of shift work .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology", "neuroscience" ]	2021	The sleep-wake distribution contributes to the peripheral rhythms in PERIOD-2

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.